Ryzen 7 3700X – Part 2

Testing and Benchmark!

Read Part 1 Here!

Test Setup

zen 2 test system
zen test system
coffee lake test system
  • All applications, games and processors are tested with the drivers and hardware listed above—no performance results were recycled between test systems.
  • All games and applications are tested using the same game version.
  • All games are set to their highest quality setting unless indicated otherwise.

Super Pi

SuperPi is one of the most popular benchmarks with overclockers and tweakers. It has been used in world-record competitions since forever. It is a purely single-threaded CPU test that calculates Pi to a large number of digits—32 million for our testing. Released in 1995, it only supports x86 floating point instructions and thus makes for a good test for single-threaded legacy application performance.

The Ryzen 7 3700X falls behind Ryzen 9 3900X and ahead of Ryzen 7 2700X in this test Intel is the leader here leading Ryzen 9 3900X with the core i9 3600K and best performing CPU here is core i7 8700K which leads the core i9 9900K

wPrime

While SuperPi focuses on calculating Pi, wPrime tackles another mathematical problem: finding prime numbers. It uses Newton’s Method for that. One of the design goals for wPrime was to engineer it so that it can make the best use of all cores and threads available on a processor.

Ryzen thrashes Intel here with the Ryzen 7 3700X leading intel core i9 9900K and behind Ryzen 9 3900X.

Rendering — Cinebench

Cinebench is one of the most popular modern CPU benchmarks because it is built around the renderer of Maxon’s Cinema 4D software. Both AMD and Intel have been showing this performance test at various public events, making it almost an industry standard. Using Cinebench R20, we test both single-threaded and multi-threaded performance.

Once again, Ryzen leads here thrashing intel with Ryzen 7 3700X leading core i9 9900K and behind Ryzen 9 3900X.

Rendering — Blender

Blender is one of the few professional-grade rendering programs out there that’s both free and open source. That fact alone helped build a strong community around the software, making it a highly popular benchmark program as well due to its ease of use. For our testing, we we’re using the Blender “BMW 27” benchmark scene.

Here too, Ryzen keeps the same results from the Cinebench Render.

Rendering — Corona

Corona Renderer is a modern photorealistic renderer that’s available for Autodesk 3ds Max and Cinema 4D. It delivers physically plausible and predictable output due to its realistic lighting algorithm, global illumination, and beautiful materials. Corona does not support GPU rendering, so CPU performance is very important for all its users.

Once again, the results are same as the Cinebench render.

Rendering — KeyShot

The standalone KeyShot rendering software features fast and efficient workflows that help you get high-quality realistic product shots in the shortest-possible time frame. Real-time raytracing, multi-core photon mapping, adaptive material sampling, and a dynamic lighting core provide high-quality images that update instantly even when interactively working on the scene. KeyShot is optimized for usage on CPUs only, which lets them use more complex algorithms than a GPU-based renderer. Unlike our other rendering tests, we record “frames per second” in KeyShot while rendering, so higher numbers are better.

In all four including keyshot, we get the same results from the Cinebench render.

To be continued…

Ryzen 9 3900X – Part 2

Of course, one of the biggest architectural improvements to the 3rd generation Ryzen is the process node shrink or die shrink as it is more commonly known.

Read Part 1 of this here!

Energy Efficiency and Power Consumption

Of course, one of the biggest architectural improvements to the 3rd generation Ryzen is the process node shrink or die shrink as it is more commonly known. Not only does this increase the transistor budget and allow for more cache and cores in the package, it also potentially reduces power consumption and heat. This is what AMD calls its: “7nm Energy Efficiency Leadership.”

Here are some points on this taken directly from AMD:

  • The higher natural performance of the “Zen 2” architecture dovetails beautifully with the superior performance, power, and density characteristics of the 7nm process. Some key metrics specifically enabled by the process change include:
  • 29% smaller CCX size vs. 12nm (~31mm² vs. ~44mm²), enabling new area for the “Zen 2” revisions
  • Up to 75% higher perf/W compared to 2ndGen AMD Ryzen™ Processors with the 12LP process
  • Up to 58% higher perf/W compared to 9th Gen Intel Core processors with the 14nm++process
  • 2X L3 cache size (32MB)vs. 2nd Gen AMD Ryzen™ Processors, not previously possible in the area with 12nm
  • Up to 2X more cores in the same package (AMD Ryzen™ 9 3950X vs. AMD Ryzen™ 7 2700X)
  • Up to +350MHz of core frequency at the same voltage vs. 12LP Altogether, these highly desirable characteristics position the 3rd Gen AMD Ryzen™ processor as the leader in desktop energy efficiency in 2019.
  • You can see this proven out for yourself in the below data, where AMD commands full-stack leadership over 9th Gen Intel Core in performance per watt.
ryzen 9 3900x power consumption

Our Ryzen 9 3900X actually idles with a slightly higher power draw than our Intel system. However, under load things switch around. It’s important to note that the AMD system has a motherboard with far more integrated features, and its possible that some of that difference is easily due to the motherboards added complexity.

While overclocked, our Ryzen 9 3900X once again pulled a bit more power than the Intel test system. However, AMD’s claims are not about raw power draw but rather “performance per watt.” By that metric, AMD is more efficient as it’s offering has far greater core density and performance in multi-threaded workloads. I’m not seeing math that tracks with what AMD’s claims are, but as I said this was a test based on our limited selection of motherboards. At the time of this test, I had only one X570 motherboard on hand and it was the MSI MEG X570 GODLIKE which has features like a tiny LCD readout that shows a cartoon of a dragon fighting ninja’s or something.

Features like this create additional power draw. In contrast, the Intel system is using an ASUS ROG Maximus XI APEX, which is by comparison a stripped down no frills and no nonsense motherboard. The APEX motherboard has beefy VRM’s sure, but we aren’t pulling the kind of power they are capable of delivering under more extreme circumstances. It’s probable that using the right motherboards for comparison that AMD’s math would be a bit closer to its claims. I do agree that AMD offers more performance per watt, but your mileage may vary as to just how much more performance per watt you get in a specific combination. I long for the days when AMD and Intel used the same sockets and chipsets. It made comparisons like this much easier.

Performance and Testing Methodology

Finally, we come down to the performance data. However, I must preface this by saying the Ryzen 9 3900X was the only Ryzen 3000 series CPU I had time to evaluate prior to the embargo date. For general performance testing, we used virtually the same software we use in our motherboard testing and evaluation. However, I have added a couple of additional benchmarks into the mix. I had planned to add some additional games into the lineup but, ran out of time for that. Gaming performance is something we’ll follow up on and we will also review additional processors in the 3rd generation Ryzen family. So, keep that in mind. We are far from done on this topic. This information is concerning the Ryzen 9 3900X only.

Due to the schedular improvements to Windows 10 build 1903, we decided to use build 1903 for all system testing. While many of the applications here have been used before in our motherboard reviews, even with the same configurations, the numbers are not comparable. Those were all done on builds 1803 and 1809 using the drivers that were current then. All systems were freshly formatted and all the latest drivers and OS patches were used. All of the systems were updated to their latest BIOS revisions. Finally, for the Intel system, I did install the CPU microcode updates relevant to that CPU. Its important to note that build 1903 does contain improved mitigations for several security flaws on Intel processors. However, I did not go out of my way to download any additional or optional mitigation patches. Hyperthreading also remained enabled for all testing.

We also followed AMD’s recommendations for using CPPC2 which is enabled by using AMD’s balanced power plan. Essentially, we created a “best case” scenario for each system outside of the hardware configurations. For the hardware, it was impossible to use the same memory modules on all of the test systems due to the nature of memory compatibility on different motherboards. That said, we were able to use common frequencies and keep the timings relatively close for the most part. This was not the case with the Threadripper system, which would not cooperate regarding running tighter timings. It simply could not complete all the tests at the same timings used by other systems. The timings in the table are the RAM’s default timings at maximum speeds. They were all run at 16,18,18,36,1T @ DDR4 3200MHz.

Finally, all systems were run at stock and overclocked values. The “stock” settings are their automatic or default base/boost clocks. The boost clocks are shown in the graphs for reference. An overclocked value for each CPU was used as well, running all cores at a fixed speed. In the case of the Ryzen 9 3900X, an all core frequency of 4.3GHz was the highest we could achieve with full stability. This is the value that you will see. I am not convinced this is the best way to overclock a Ryzen 9 3900X, but I’ll talk more on that point later.

Application & Synthetic Testing

Sandra Memory Bandwidth

Note: All systems were run in dual channel mode, excluding our Threadripper system which used quad-channel mode. Again, all RAM was run at DDR4 3200MHz. The timings on all systems were 16,18,18,36,1T, excluding the Threadripper system which uses the default 18,19,19,39,1T values.

All of our Ryzen systems score essentially the same result. I was actually surprised that the Ryzen 9 3900X didn’t do better. However, it is capable of much greater speeds which is a topic for another day.

Sandra CPU Dhrystone

While the memory results weren’t anything special, its fair to say that the Ryzen 9 3900X is nothing short of a monster. The Ryzen 9 3900X is considerably faster than the Threadripper 2920X in this test. You would need a very expensive HEDT CPU to beat these numbers.

WinRAR – Multithread

WinRAR is a weird test. While it does scale with additional threads, its still very much influenced by single-threaded performance. I don’t think it stretches that far into additional cores either. This is evidenced by the fact that the Core i9 9900K easily keeps up and slightly edges out the 3900X. It does show the strength of AMD’s.

WinRAR – Single Thread

It’s not really a surprise that Intel’s Core i9 9900K is the king here. However, if you look at the results, once again we see near IPC parity between AMD’s 3rd generation Ryzen and Intel’s Coffee Lake architecture. The boost clocks of the Ryzen 9 3900K are relatively close to Intel’s Core i9 9900K. The latter actually loses to the Ryzen 9 3900X using their stock boost clocks. Its only overclocked to 5.0GHz that the Intel pulls ahead. One important note, while the Core i9 9900K can boost to 5.0GHz for single threaded applications, I almost never saw this actually happen. It ran at 4.8 or 4.9GHz most of the time.

wPrime v2.10

This this test we once again see very similar results between our Core i9 9900K and the Ryzen 9 3900X. At their normal boost clocks, AMD is actually faster here. The Intel Core i9 9900K gets close but, has to be overclocked to 5.0GHz on all cores to achieve this.

POV-Ray

This is a test that clearly likes having access to more CPU cores. This is evidenced by the fact that the AMD Ryzen 9 3900X destroys all the other test systems save for the Threadripper 2920X which has the same core and thread count. Despite the memory bandwidth advantage of the Threadripper, it fails to match the 3rd generation Ryzen. This comes down to all the various architectural improvements and reduced latencies.

V-Ray

The Ryzen 9 3900X is an absolute monster here. The 100MHz advantage when all of the Ryzen 9 3900X’s cores are overclocked vs. the Threadripper probably has very little to do with the massive difference we see between those two platforms. Otherwise, we would see different deltas across the other CPU’s if clock speed mattered that much.

Cinebench R20 – Multithread

As I said, the single threaded results do indicate an advantage for AMD’s architecture here. It is a best-case scenario as AMD is only 8 points off Intel’s Core i9 9900K despite the latter having a clock speed advantage.

Adobe After Effects CC – Puget Systems Benchmark

This is the first time we’ve used this test. This was suggested by one of our forum members, and I have to say, I’ve been rather impressed with it. However, it is worth pointing out that there is some GPU acceleration here as I’ve tried this with various GPU’s and it can radically alter the results. Adjustments to the CPU’s clock speeds and core counts also makes a substantial impact, which is why I left the test in the lineup. We used the same graphics cards for all of the comparisons, making the GPU a constant variable.

Blender

Blender is a mixed bag for me. On one hand, it’s a nice test but if there is a remote hint of instability, this test shows it. Essentially, if I could run Blender, that configuration would pass any test I threw at it. This is the test that actually highlighted how unstable our Ryzen 9 3900X was at 4.4GHz or beyond. That said, it’s a good multithreaded test which shows a massive improvement as cores are added, but it also highlights the difference between our Threadripper 2920X and the Ryzen 9 3900X. Again, the Ryzen 9 3900X has no equal here. It’s worth noting that the fastest result came from running all the cores at 4.3GHz.

Gooseberry

In this test we see more of AMD’s claims of content creation dominance in action. At default settings and via a manual overclock, the Ryzen 9 3900X simply crushes the competition.

Gaming Performance

I think it was essentially a foregone conclusion that the Ryzen 9 3900X would be faster in most content creation and productivity applications due to its higher core and thread count. The clocks and IPC should be close enough for that to be the case in most scenarios. However, Intel has dominated on the gaming front. This was especially clear at 1920×1080 where the 1st and 2nd generation Ryzen CPU’s fell way behind in benchmarks. This wasn’t as much of a problem at higher resolutions indicating that 1920×1080 is at this point, rather CPU limited. As a result, all of our gaming tests were conducted at 1920×1080, but at low settings to try and isolate the CPU performance and avoid using the GPU too much. The same GPU was used for all our game tests, which was a GIGABYTE RTX 2080 Ti Aorus Xtreme 11G. It was running its factory overclock.

Tests at these settings are in no way representational of what you should expect to see in actual gaming unless you really use a $1,299.99 video card to game at 1920×1080 with potato mode levels of detail in games. In game benchmarks were used in each of the tests. Finally, we wanted to add more games into the mix. I had added Destiny 2 to the list. It was actually tested on the other configurations but would not launch on our Ryzen 9 3900X and MSI MEG X570 GODLIKE test system. On some of the other systems, I ran into problems using FRAPS or MSI Afterburner for benchmarking the game and eventually ran out of time for testing. We will certainly be evaluating CPU performance on more games in the future, but for now we are going to use the standard tests used in our motherboard reviews. Again, these results are all done using systems with the latest build of Windows 10. The OS is at the latest patch level and we are using the latest drivers available at the time of this writing. This includes the pre-release AMD chipset drivers for X570 for all three AMD test systems as it’s a unified driver.

Heaven

This is the only totally synthetic test we employ here. More than any of the other tests, I’ve used, it seems to show the greatest delta between different processors which is why I used it here. As you can see, there is allot we can take from these results. First and foremost, there is no benefit to having more than 8c/16t here. Secondly, clock speeds do matter, and AMD and Intel are closer in terms of IPC as far as this test goes. The latency issues we talked about earlier regarding Zen and Zen+ can be seen in the delta between the Ryzen 7 2700X and the Threadripper 2920X.  

The Division 2

In this test, Intel’s Core I9 9900K is clearly the fastest in the lineup. However, AMD’s Ryzen 9 3900X isn’t too far behind. The curious thing is that the 3rd generation Ryzen is only a bit faster than the previous generation CPU. By no means is the difference anywhere near what AMD suggested in its product brief and all the talk about “GameCache.”

Shadow of the Tomb Raider

SoTTR is a fantastic looking game that embraces the latest technologies such as ray tracing. Obviously, we aren’t concerned with that here, but it’s a very modern game technologically speaking. However, once again, Intel shows a still fairly substantial lead here. It also benefits quite a bit from being overclocked. The AMD systems didn’t, which is curious. We also see a fairly decent performance increase going from the Ryzen 7 2700X to the Ryzen 9 3900X.

Hitman 2

It should come as no surprise that Intel remains dominant here. We can see another substantial gain relative to a minor clock speed increase. We also see a fairly large gap between Intel and AMD here. It is a gap that’s relatively similar to what we saw in the Division 2. The gap is equally as pronounced comparing the Ryzen 7 2700X to the Ryzen 9 3900X.

Direct IPC Comparison

Naturally, when a new CPU comes out, we want to know precisely how its IPC stacks up against competing CPU’s as well as the previous iteration. For this test, we had to select a common clock frequency that was achievable by all the CPU’s used in the comparison. Since my test Ryzen 7 2700X maxes out at 4.2GHz on all cores, that’s the clock speed we went with. Unfortunately, I didn’t have an 8c/16t Ryzen on hand for this comparison. As a result, I simulated it using the 3900X. I disabled four of the cores. Unfortunately, due to the layout of the CPU, Ryzen Master required a symmetrical layout which meant that I had to disable two cores in each CCX within the same CCD.

This has the unfortunate result of leaving our simulated eight core Ryzen with more L3 cache than it would ordinarily have. Therefore, the topology of the simulated eight core Ryzen 3000 series CPU, isn’t quite right. I’m fairly certain I’m going to get flak on this, and we are aware that this isn’t a perfect simulation. Having said that, our previous tests show that the cache difference didn’t create a huge delta between the Ryzen 7 2700X and the Ryzen 9 3900X and the latter still had four more cores and higher clock speeds to boot.

More importantly, we were able to compare a 2nd and 3rd generation Ryzen CPU using the same memory frequencies and timings as an Intel Core i9 9900K. This is as “apples to apples” as we can get without having more Ryzen 3000 series chips on hand. We will potentially revisit this topic at a later date, but this gives us some idea of how these CPU’s compare at the same frequency. Naturally, this isn’t a realistic scenario as I can clock RAM on the Ryzen 9 3900X to DDR4 3600MHz and I’ve got 4,000MHz RAM for the Core i9 9900K. Using PBO or manual overclocking, the Ryzen 9 and Core i9 9900K would also be at dramatically different clock speeds. This comparison is largely academic, and potentially entertaining. One thing we can potentially learn is what would happen if AMD closed the clock speed gap completely. How would Intel fair in gaming then? Well, let’s find out.

We didn’t run through our entire suite of tests. We disregarded things that wouldn’t be impacted by these settings such as memory bandwidth. I lost the data I collected from the Sandra CPU test, so that was omitted here. Lastly, I also omitted testing on the Threadripper system. Mainly this is because I didn’t think it was very useful for a direct IPC comparison. The platform is very different than X470, X570 and Z390. The CPU’s internal layout and memory configuration would put the CPU behind the Ryzen 9 3900X. I thought about using it as it has the same core and thread count, but again, the platforms are so different, that I didn’t think the comparison was nearly as useful. In a sense, we already have that comparison as the “all core” overclocks are fairly close in the other tests and the Threadripper 2920X lost every single one of those tests aside from the Sandra memory test.

wPrime

Right out of the gate, I was surprised to see that AMD’s 3rd generation Ryzen came out on top here.

V-Ray

Once again, the victory goes to AMD. We can also see the fairly substantial gains compared to Zen+.

POV-Ray

Here we see a massive win in favor of AMD. The Ryzen 7 2700X is also not that far behind Intel’s Core i9 9900K. This too was quite surprising.

Cinebench R20 – Multithread

Again, as Intel puts it, this is a “best-case” test for AMD’s Ryzen CPUs. Despite losing a third of its cores and threads, the 3rd generation Ryzen still comes out on top by a wide margin.

Cinebench R20 – Single Thread

Here we see the same trend of AMD dominance.

Adobe After Effects CC – Puget Systems Benchmark

It’s worth noting that this is also the only test where the Threadripper 2920X beat the Ryzen 9 3900X. If you look back at the earlier benchmarks, the Threadripper 2920X scored a 970 in this test compared to 930 on the Ryzen 9 3900X. Undoubtedly, this must have something to do with the memory size, and or memory bandwidth that X399 and Threadripper brings to the table as it’s the only factors which favors the Threadripper system. It used the same video card, a slower SSD, and a slower processor in every other test. Clearly, the latency issues going across CCX complexes didn’t have an impact either. If it did, it was offset by the memory. To be fair, Adobe products like this are designed with workstations in mind. Adobe programs also really like memory and memory bandwidth.

Of course, another takeaway from this is that both AMD systems were faster in this case. There are potential architectural reasons for this. Of course, another likely variable has to do with the fact that the latest versions of Windows 10 contain some built-in mitigations for Intel’s security flaws. Additionally, I did al the microcode and firmware updates that applied as well. These factors almost certainly play into these results.

Blender

This is another case which favored both AMD systems. This shows us that AMD’s 2nd and 3rd generation Ryzen CPUs can absolutely dominate in content creation without having to lead through having a greater core count. This isn’t as surprising as one might initially believe. If you think about it, Ryzen was still relatively competitive in these areas even after Intel caught up with AMD on core and thread counts. Intel still generally won where core and thread counts were equal, but we can see this may often be simply due to clock speed. Do keep in mind that there are some mitigations in place for Intel’s security issues here, so that is a factor as well. We knew that those can cause that IPC gap to shrink and that’s what we are seeing here.

Gooseberry

This is a test that plays out almost exactly as one might expect. Intel’s Core i9 9900K beats out AMD’s Ryzen 7 2700X, but fails to match the newer Ryzen 9 3900X, even with a third of its cores disabled. Again, security mitigations built into the OS, updated microcode and mitigations in firmware certainly contribute to this.

Heaven

This is another test with a surprising result. AMD’s Ryzen 9 3900X manages another win at this clock speed.

The Division 2

Again, we see the 3rd generation Ryzen edge out the Intel Core i9 9900K. At their normal clocks, we see quite the opposite, showing that this game seems to be sensitive to clock speeds.

Shadow of the Tomb Raider

This is a test where we see what we expected to see. Intel’s performance is far greater than that of its rival. In a way, this is precisely what AMD said would happen in that its often older titles that benefit from the added cache. Hitman and Shadow of the Tomb Raider are fairly new games built on up to date game engines. I’m convinced the cache has allot to do with the improved performance relative to the Ryzen 7 2700X, but its not enough to close the gap with Intel’s Core i9 9900K.

Hitman 2

This is another title where AMD absolutely gets beaten down by Intel’s Core i9 9900K. It proves to be a beast in this game. Again, this is somewhat odd given the clock speed parity. As AMD said, it wins some and loses some. That’s certainly the case in our testing.

Overclocking

3rd generation Ryzen CPU’s share several things in common with their predecessors. In fact, overclocking is largely the same in most respects. The Ryzen 3000 series supports overclocking in one of three ways. Manual, Precision Boost 2, and Precision Boost Overdrive. We will talk a little about each of these methods and how they differ, as well as explore the potential of our test CPU and what you might expect overclocking one of your own.

Precision Boost 2

Precision Boost 2 or “PB2” is basically a form of automatic overclocking. It’s an opportunistic algorithm which tries to increase the clock speeds of loaded cores within specific limits. It does not overclock a single core or all cores specifically. AMD’s literature on the matter is quite clear on the subject. However, this is more or less how it looks unless you watch everything in real time using something like the Ryzen Master software application which we will talk about as well. Essentially, the only real limitations are imposed by four things. PPT, TDC, EDC and the OEM boost clock. The OEM in this case being AMD. The processor for example may have a preprogrammed limit of 95A TDC and a 140A EDC. It will therefore modulate its clock speeds to stay within those ranges. Its sort of like a rev limiter for your CPU. If it hits the limit, it retards the clockspeed to stay within its default thresholds and do so until the situation changes. It analyses these conditions once every millisecond.

To understand what PPT, TDC and EDC are, it’s just easier to cut and paste from AMD’s literature:

1.Package Power Tracking (“PPT”):The PPT threshold is the allowed socket power consumption permitted across the voltage rails supplying the socket. Applications with high thread counts, and/or“heavy” threads, can encounter PPT limits that can be alleviated with a raised PPT limit.

            a.Default for Socket AM4 is at least142Won motherboards rated for 105W TDP processors.

            b.Default for Socket AM4 is at least 88W on motherboards rated for 65W TDP processors.

2.Thermal Design Current (“TDC”):The maximum current (amps) that can be delivered by a specific motherboard’s voltage regulator configuration in thermally-constrained scenarios.

            a.Default for Socket AM4 is at least 95Aon motherboards rated for 105W TDP processors.

            b.Default for Socket AM4 is at least 60A on motherboards rated for 65W TDP processors.

3.Electrical Design Current (“EDC”):The maximum current (amps) that can be delivered by a specific motherboard’s voltage regulator configuration in a peak (“spike”) condition for a short period of time.

            a.Default for Socket AM4 is 140Aon motherboards rated for 105W TDP processors.

            b.Default for Socket AM4 is 90A on motherboards rated for 65W TDP processors.

Essentially, AMD’s Precision Boost 2 automatically adjusts clock speeds to stay within the above limits. Setting your system up to use Precision Boost 2 is perfectly safe as the CPU’s guidelines are relatively conservative.

Precision Boost Overdrive

Precision Boost Overdrive or “PBO” works the same as Precision Boost 2. However, instead of using the CPU’s established default limits for PPT, TDC and EDC, it uses those values as defined by the motherboard. When PBO is enabled, the same algorithm is applied, but operates within the limits set by the motherboard’s firmware. Once enabled, it’s just like PB2 in that you can essentially set it and forget it. The processor will adjust its clock speeds up or down as needed. Motherboard VRM implementations, firmware configuration and thermal solutions vary greatly. As a result, PBO results can vary quite a bit. PBO still doesn’t raise the limits of any core beyond the specified boost clock limits. So, if you have a processor like the Ryzen 9 3900X with a maximum boost clock of 4.6GHz, no core will ever go beyond that value.

However, 3rd generation Ryzen CPU’s have the ability to use a maximum boost clock offset which is adjustable in 25MHz increments up to 200MHz. This means that a boost clock of 4.6GHz can now be extended to 4.8GHz under the right conditions. It is a separate feature from PBO specifically, called Auto-OC by AMD. However, it works on conjunction with PBO. The limits still apply as it will not reach those extended frequencies unless it can do so within the established PPT, TDC and EDC limits of the motherboard. Essentially, conditions have to be ideal to allow for the CPU to achieve the desired 4.8GHz boost clock in this scenario. Cooling and environmental conditions come into play with all of that so your mileage may vary. However, compared to PB2, PBO+Offset would be considerably more aggressive and the chances you’ll see those higher boost clocks are much greater. Speaking from personal experience as well as from the testing, PBO works very well to increase performance in lightly threaded applications.

There is data out there from the 1st generation and 2nd generation Ryzen launches showing PB2 and PBO giving greater performance benefits than manual overclocking in some scenarios. Basically, any application that can’t use all the CPU’s cores and threads will potentially be faster using PBO or PB2 than manual overclocking.

Manually Overclocking the Ryzen 9 3900X

Unfortunately, I had no idea what to expect from the CPU in terms of manual overclocking. When you cover these types of launches, you aren’t always told what to expect in terms of overclocking. At least, that’s been my experience. Given the PBO+offset feature and previous Ryzen overclocking experience, I was hoping to at least achieve an all core manual overclock equal to the maximum boost clock, or at least get close to it. As always, overclocking with a new processor and platform always creates a learning curve. However, overclocking with the 3rd generation Ryzen, is similar to the 1st and 2nd generation CPUs in most respects. Mainly, overclocking all cores manually works the same as it did previously. All you really need to do is adjust your load-line calibration and set your CPU voltage. After that, it’s simply a matter of adjusting your CPU voltage until you achieve stability at a given frequency.

However, there are significantly more options for tuning the UEFI BIOS than I’ve seen on previous socket AM4 motherboards. X570 is also somewhat immature, although already worlds ahead of where X370 was at its launch. In any case, I do need to spend more time with this specific motherboard and CPU combination. That is something I’ll be doing in the future as I still need to review the motherboard.

I found that the system could POST at speeds up to 4.6GHz, but at that speed it was unable to reach the Windows desktop. No amount of voltage would change this. I even went as far as 1.55v with no success. All the load-line calibration settings and increasing the SoC voltage didn’t help either. A word of warning, the UEFI BIOS of the MSI MEG X570 GODLIKE warned that PCI-Express 4.0 signaling would be dropped and negotiated to PCIe Gen 3.0 speeds if the SoC voltage went past its default 1.1v value. I continued my testing and found that I could get the system into Windows at 4.5GHz, but it was far from stable. Essentially, I couldn’t actually run any applications that put significant strain on the CPU. Some of the game benchmarks would run, but none of the processor intensive applications would. Attempting to do so would cause the system to randomly reboot on me.

Dropping the clocks down to 4.4GHz @ 1.35-1.4v and beyond bore no fruit either. While I could run many of the benchmarks the system was ultimately unstable. It couldn’t complete any of the major CPU intensive applications. Game benchmarks like Heaven would simply crash to desktop. This is where I feel the CPU’s maximum capability probably lies. Its stable enough to give me some hope that the right combination of settings might work to make it stable. One thing I noticed was that the CPU temperature never went higher than 79c. That’s not exactly cool, but its far from throttling at that temperature. Most of the time the crashes occurred while it still read within the 65c-70c range. This tells me that it isn’t temperature that’s the issue here. It may simply be a setting, or it might actually be past the CPU’s maximum clock speed for all its cores.

At 4.3GHz and 1.35v, I could leave almost everything at its default values and the system was rock solid and ran any test I threw at it without issue. This is the value I settled on for my maximum “all core” overclock. As I said, I feel like manual mode shortchanges you when you can pick up single threaded performance through considerably higher boost clocks using PBO or PBO+Offset. I did try these modes, but I need to spend more time with them. PBO+Offset has no guarantee of granting you the extra clock speed and in my case it never did. However, I saw PBO boost into the 4.5GHz range fairly often in single threaded tasks. However, PBO and PB2 rarely achieved more than 4.2GHz on heavily multi-threaded workloads. In the UEFI, you do have control over the PPT, TDC and EDC values. Ryzen Master gives you access to these as well.

Our testing did show a benefit to the manual overclock in some cases. However, if your buying this primarily for a gaming system, I think the added heat, power consumption and effort probably isn’t worthwhile. This is something I saw back with the 2ndgeneration Ryzen CPU’s. On my own personal system, I’ve often found PBO to be the way to go, although you do pick up more performance in some tasks using an all core manual overclock. This obviously depends on your usage case.

To be Continued….

I had given up on AMD… Until Yesterday! Ryzen 9 3900X! Part – 1

Today marks AMD’s official launch of its new Ryzen 5, 7 and 9 “3000 series” CPU’s.

Today marks AMD’s official launch of its new Ryzen 5, 7 and 9 “3000 series” CPU’s. AMD hasn’t really kept many secrets about what Ryzen 3000 series parts are. About the only things that have remained secret are the general performance and overclocking capabilities of its new CPU’s. Even that, AMD hasn’t kept completely in the dark. As is often the case with such early statements concerning performance, its often a best-case scenario which is showcased to generate hype about the product.

AMD has made many bold claims about its 3rd generation offerings and in the past, AMD has often paid the price for over hyping its products. I’ll preface this article by saying that much of the hype is deserved, but that some of the statements AMD has made have been done through a certain lens. As usual, we’ll separate fact from fiction, reality vs. marketing and tell you just how these CPU’s fit in today’s landscape.

We will see how these new processors do against 2nd generation offerings as well as Intel’s mainstream gaming champion, the Core i9 9900K. AMD has made claims that these CPU’s are upwards of 21% faster than the previous generation and that Zen 2 has a 15% IPC improvement over its previous architecture. We will see if this is true or not.

ryzen 9 3900x event

Background Info

First, some background. AMD launched its Ryzen CPU’s in early 2017. Ryzen was well received, but not because it was dominant in terms of performance. Ryzen was successful because AMD launched a product that had twice the core and thread density of its rival, while still offering reasonably good instructions per clock cycle (IPC) performance. When Ryzen hit the market, Intel was still peddling quad cores just as it had done for roughly ten years prior in the mainstream. While it had offered higher densities in the HEDT market, such systems are prohibitively expensive for many people. I’ve personally been using Intel’s HEDT platform since X58, so I am all too familiar with its pricing. Since the launch of its Core 2 Duo, AMD has had a massive IPC deficit that it is only now able to deal with. AMD’s performance advantage in some applications was simply due to having double the core density of its rival.

When AMD’s Ryzen CPU’s hit the market, AMD had a performance advantage in multi-threaded workloads. This combined with good enough IPC for single threaded applications, along with attractive pricing is what enthusiasts responded to. However, it soon became apparent that Ryzen had a substantial weakness: Gaming. This is primarily at 1080P where the resolution is still largely CPU limited, although GPU’s do have a significant impact. Intel was considerably faster in gaming, and given most users are probably at 1080P, Ryzen was a little less attractive. People still went for it despite this weakness. However, it wasn’t just games where Ryzen sometimes suffered, it was still at an IPC deficit compared to Intel’s offerings and because of that, single threaded workloads weren’t exactly Ryzen’s strong suit. Many found AMD a viable option due to price and platform longevity. AMD promised socket compatibility through at least 2020. So far, all subsequent generations of Ryzen CPU have been a drop-in upgrade to the early X370 and B350 chipset motherboards.

Despite being a budget CPU maker for most of the last three decades, AMD has never wanted to play second fiddle to Intel. It’s done so primarily because it had no choice. When AMD dominated in most benchmarks with its Athlon 64, it charged as much if not more than Intel did. The stars have aligned as it were, and some circumstances have allowed AMD to catch Intel and compete with it on nearly equal footing. Not only has Intel struggled with its manufacturing process, but recently discovered security flaws required patches which mitigate both security holes and Intel’s performance advantages. Today marks AMD’s return to glory as it will no longer be considered a budget alternative to Intel, as it tries to lead the market rather than play catch up to Intel.

In order to do so, AMD had to address Ryzen’s weaknesses and improve performance in single-threaded applications. Not only that, AMD felt the need to push core density in the mainstream market ahead of Intel once again. This is in order to dominate in multi-threaded workloads. Achieving this required a multi-pronged effort. It required a die shrink, architectural improvements, driver work, firmware changes and of course some help from Microsoft.

Meet the Ryzen 3000 Series

There are five new models to start with another making its way into the lineup later in September. Ryzen CPU’s based on this 3rd generation architecture will have 3000 series model numbers. The available models are: The Ryzen 5 3600 and 3600X, Ryzen 7 3700X and 3800X and lastly, the Ryzen 9 3900X. There is an upcoming Ryzen 9 3950X that’s already known to the public, but this CPU isn’t available today, so I won’t spend a lot of time talking about it. Its specifications are listed below, and it will be available in September, with the rest of the 3rd generation Ryzen family being available immediately.

ryzen 3000 price

Here is the 3rd generation AMD Ryzen desktop processor product stack. This table includes the cooler information as well as L2 and L3 cache sizes.

ryzen 3000 specs

As you can see, AMD has Ryzen CPU’s competing directly with Intel at virtually every price point. AMD claims more performance per watt and or more cores at any given price point. There is a typo in the slide as the Ryzen 5 3600 non-X should appear at the bottom. AMD claims stronger content creation at each price point and at least comparable gaming. AMD is rather up front in its briefings that it doesn’t dominate Intel when it comes to gaming but, gets close enough “winning some and losing some” here and there. Even at a glance, there are a few standout items of note in the specs. Of course, the Ryzen 9 series will appeal to enthusiasts with a slightly larger budget and a concern for performance above all else. Another noteworthy item is the 65watt TDP parts that trade a little clock speed and 3MB of L3 cache for a lower TDP. These parts come in a little bit cheaper but, should offer very similar performance to their higher TDP brethren at stock speeds. If your only concern is gaming, you might want to look towards the larger cache models with the higher clock speeds.

The AMD Ryzen 9 3900X

Your new Ryzen CPU will most likely come in a retail package similar to this one. It’s no 12 sided plastic dice thing like Intel gives you, but the box protects the contents and you don’t need an engineering degree to open it.

ryzen 9 3900x package

Inside the box you’ll find a CPU cooler a case badge, and naturally a CPU. This example is an AMD Ryzen 9 3900X. It is a 12 core and 24 thread part with a TDP of 105w. As a result, this comes with the big daddy of the included CPU coolers, the Wraith Prism.

Naturally, with every new product comes a slew of marketing material. In this case, a large slide deck from AMD. The slides go over many of AMD’s marketing claims as well as product changes, and architectural information. We will go over most of these slides, in this article. So strap in, this is going to be a long one.

ryzen 9 3900x performance

Right out of the gate, you can see AMD is making bold claims about its processors and their performance improvements. Using the Ryzen 2700X as a baseline, it shows the lowly Ryzen 5 3600 offering 9% more single-threaded performance in Cinebench R20. As you scale in price, clocks and core count, this gap widens to 21% for the Ryzen 9 3900X. The Ryzen 5 series are all 6c/12t parts. The Ryzen 7 series are both 8c/16t and the Ryzen 9 series only has one entry today consisting of 12 cores and 24 threads.

Windows 10 Optmizations (Build 1903)

From time to time, the software has to be updated to take advantage of newer hardware features. AMD has worked with Microsoft to do just that.

ryzen 3000 windows 10
ryzen 3000 windows 10

One of the drawbacks of Ryzen 1000 and 2000 series CPU’s are latency issues between CCX complexes. One reason for these issues comes down to Windows not knowing anything about the architectural layout of Ryzen CPU’s and assigning tasks to different cores across different CCX complexes with no regard for the topology of the processor. Naturally, crossing CCX complexes in the older CPU’s incurred a latency penalty. Working with Microsoft, build 1903 of Windows 10 now has “Topology Awareness” for all Ryzen derived CPU’s. This includes the 1000 and 2000 series, as well as all current and past Threadripper CPU’s. As of the May 2019 update, the Windows 10 schedular is now aware of the CCX topology and attempts to localize tasks to cores and threads within a single CCX complex as often as possible to improve performance. In a sense, it fills up a CCX complex before moving to another. Taking advantage of this requires nothing on your part aside from installing the Windows updates and chipset drivers.

However, newer 3rd generation CPU’s have a new feature called “CPPC2”. This is for 3rdgeneration CPU’s and requires the latest firmware and chipset drivers along-side Windows 10 build 1903 to make use of it. This feature acronym stands for “Collaborative Power and Performance Control.”

Here is what AMD has to say about CPPC2: “CPPC2, or Collaborative Power and Performance Control, is a newer method of clock speed selection in the UEFI standard that deprecates the classic pState model in favor of CPU firmware-controlled clockspeeds. Whereas pState clockspeed selection may take as much as 30ms, handing control over to the processor’s own firmware speeds this up by 20-30X into the range of 1-2ms. This change is particularly beneficial to the brief and bursty workloads like webpage rendering and application launches.” AMD further acknowledges the fact that the gains from this and the topology awareness scheduling are modest at best. It’s documentation notes that the improvements are typically in line with what one might see in major graphics card driver releases.

ryzen 3000 dram latency

One weakness of the Zen and Zen+ architectures is gaming performance. At 1920×1080, you are largely still CPU limited and this is the resolution probably about half the PC gaming community uses for one reason or another. This was a point AMD felt the need to address directly, and it does so in two significant ways. Improved DRAM latency and increased L3 cache sizes.

Each CCD contains 32MB of L3 cache. Models with more than one CCD show as having twice as much cache in the spec sheet, but that is because each CCD gets its own cache. Most models will have a single CCD complex. Even so, any Ryzen 3000 series CPU has double the cache that was in the original Zen and Zen+ CPU’s. As AMD puts it: “This change was directly enabled by the 7nm process, allowing for a larger transistor budget inside the CCD.” There is a definite correlation between cache and game performance. This was true back when I got into building PC’s over two decades ago and it’s just as true today. As a result, AMD’s increased cache has been branded “AMD GameCache” as you can see above. AMD’s materials on the subject specifically state that it wishes the L3 cache to be referenced as “GameCache” and that it will do so on its website and future materials on the matter. AMD claims that the increased cache shows upwards of a 21% increase in the average framerates at 1920×1080. It’s materials on the subject even show that the doubled L3 cache is more than twice as effective as increasing the memory clock speeds to DDR4 3600MHz speeds and beyond.

AMD’s slide deck goes on to show general specifications for each model and how it stacks up against the competition at each price point in gaming, content creation, and power consumption. As you can see, AMD shows a massive lead in content creation, and shows relative parity with Intel on gaming. Again, AMD freely admits that gaming remains its competitor’s strong suit. However, it shows substantial improvement compared to the previous generations of Ryzen processors. I’ve included the information for reference.

65w TDP Models

As I mentioned earlier, one of the more interesting standouts in the mix are the two available 65w models. AMD highlights its Ryzen 7 3700X as its most efficient CPU in terms of performance, heat and power consumption. The slide deck shows a FLIR image of similar heat sinks on the Core i7 9700K compared to its Ryzen 7 3700X. We will talk more about Ryzen’s efficiency later, but I wanted to point out this specific slide as there are two 65w TDP models available. The Ryzen 7 3700X also represents a potential sweet spot for price and performance.

The X570 Chipset / Updated Socket AM4 Platform

As always, a new CPU brings with it a new platform. At the very least CPU releases always bring motherboard refreshes which may use an existing chipset, but, have VRM’s and other optimizations to better leverage the CPU’s potential via overclocking or it may include newer I/O technologies. In this case, AMD is releasing its X570 platform along side the Ryzen 3000 series CPU’s. Naturally, there is a lot to talk about concerning a new platform, but unfortunately, we did not have the time necessary to do a proper motherboard review, or even a deep dive into the platform itself as well as cover the CPU’s. As a result, this will be a short overview of the changes.

ryzen 3000 am4

The Ryzen 3000 series retains its socket AM4 pinout and compatibility. You can use the 3rd generation of Ryzen CPU’s with any X470 motherboard with a BIOS update. Even some X370 motherboards have or will receive BIOS updates to allow installation of 3rdgeneration Ryzens into existing systems. As many of you know, the X570’s claim to fame is supporting PCI-Express generation 4.0. To be perfectly honest, this is damn near more trouble than its worth at this stage. AMD repeatedly hypes this technology for its Navi cards which are also releasing at the same time as its 3rd generation Ryzen processor family.

AMD has stated that there are certain cases where specific visual effects and operations are sensitive to PCIe bandwidth. Specifically, it highlights DaVinci Resolve as one application that showcases this, and it goes on to talk about the impact PCI-Express 4.0 can have on 8k mastering. I’m not an expert on such matters, and that’s a topic for another time. I will say that at present, I don’t think Navi will be fast enough to make serious use of PCI-Express 4.0, and that right now, its biggest advantages have nothing to do with graphics cards. PCIe 4.0 certainly has its place, and eventually, I’m sure we won’t be able to imagine going backwards. However, right now it’s one of the main selling points of X570. Even AMD has admitted that performance with Ryzen 3000 series CPU’s will be virtually identical on X470.

The biggest advantage of going to PCI-Express 4.0 is for NVMe storage. Previous solutions limit the bandwidth an NVMe SSD can use. There are drives commonly available with read speeds that exceed 3500MB/s. A Gen 2.0 or 3.0 interconnect between the chipset and CPU can easily be saturated with one or more NVMe devices depending on their specifications and the PCI-Express generation in question. Certainly, even a Gen 3.0 interconnect, or even Intel’s extremely efficient DMI 3.0 bus runs into problems with saturation when high end NVMe SSD’s are configured in two or three drive arrays. Therefore, PCIe Gen 4.0’s most exciting benefit has nothing to do with graphics cards and everything to do with storage, or at least the chipset downlink.  

The X570 chipset is based off the IO die of the Ryzen 3000 series CPU’s. It is a 12nm part produced by GlobalFoundries. This part has a 16w TDP, which is considerably greater than that of its predecessor. AMD chose to build an entirely new chipset based off the IO die of Ryzen itself rather than upgrade the basic design of the X470 that preceded it. While this seems odd, the new platform gains flexibility and additional I/O. Gone are the legacy PCIe 2.0 slots and limited lanes of the previous chipsets. The one unfortunate side effect of this is that the chipset consumes considerably more power than its predecessor, and most companies have seen the need to go with an active fan in the chipset heat sink. The MSI MEG X570 GODLIKE motherboard used for our testing here has such a fan. I can tell you this fan is rather annoying on startup. Fortunately for me, this fan didn’t run very often.

ryzen 3000 x570

As you can see by the block diagram, the X570 chipset offers an increased PCI-Express lane count as well as the option for PCIe Gen4.0 compatibility and throughput. Again, the most exciting aspect of this is that the chipset downlink gets its bandwidth doubled using the same number of lanes.

ryzen 3000 overclock

As an enthusiast, the BIOS changes are something that elevates X570 to a more premium status over X470. Having spent time with this, I can tell you that there is a sea of settings and I didn’t have time to explore them all prior to finishing this article.

ryzen 3000 connection speed

The chart above came from AMD’s reviewers guide. This showcases the amount of I/O that is available to Ryzen systems based on X570. However, this chart doesn’t really tell you the whole story. For example, I’d almost bet real money that you will never see 14 SATA ports on any motherboard sold in North America. If you look at the block diagram it shows several tables that say: “pick one.” Essentially, the motherboard manufacturers gave a set number of lanes per “block” to use and can allocate them however they wish. Most of those SATA ports would come at the cost of the lanes used for the M.2 slots. Given the popularity of NVMe based M.2 SSD’s, I do not see anyone going beyond the 4 to 6 ports normally allocated for this purpose on existing X370 and X470 motherboards.

One of the biggest advantages of the 3rd generation Ryzen processor family is its improved memory controller. We will talk more about that in the architectural portion of this article. However, it does relate to the platform. The new memory controller, and therefore the X570 motherboards support up to 128GB of DDR4 RAM using 4x32GB DIMMs. ECC memory is supported by the processor’s memory controller. However, AMD states that it is up to the motherboard manufacturers to enable the feature on its models. Official transfer rates vary based on the DIMM configuration. This table was taken from the AMD review guide and should help make this clear.

ryzen 3000 memory

As you can see, only speeds up to DDR4 3200MHz are supported officially. Using four modules, RAM speeds drop to DDR4 2933 and DDR4 2667MHz officially. Having said, that, your mileage may vary. I have had some luck getting some modules to work at speeds up to DDR4 3200MHz in groups of four on X470 and X399 systems. Whether or not that will work with X570 motherboards remains to be seen. I haven’t had time to try that yet. As we saw with the original X370 and X399 motherboards, there is potential to improve memory clocks in four DIMM configurations over time.

However, one of the biggest advantages of the 3rd generation Ryzen CPU’s is the ability to run higher clocked RAM kits than it could before. While some people had better luck than others on earlier Ryzen systems, many people couldn’t go beyond DDR4 3200MHz in many cases. AMD states that speeds above DDR 4 4200MHz that are essentially plug and play are not uncommon. MSI specifically announced that it was able to reach speeds upwards of DDR4 5133MHz.

One of the most important things to understand is that the memory clock, memory controller clock, and Infinity Fabric clocks are all at a fixed 1:1:1 ratio until you reach DDR4 3600MHz speeds. That means that all of these things run at those speeds. Once you pass the DDR4 3600MHz mark, dividers have to be used. This will enable a 2:1 ratio for the memory and memory controller clocks. The Infinity Fabric will automatically be set to 1800MHz in 2:1 mode. Doing this creates a memory latency penalty of approximately 9ns. AMD states that this deficit can potentially be overcome with even higher clocked RAM, higher CPU frequencies or sub-timing adjustments.

Unfortunately, its time to address the elephant in the room. And that’s price. PCI-Express Gen 4.0 signaling is a bit tricky it seems. As a result, motherboard manufacturers are left with two choices. Either use thicker PCB’s made from conventional, less expensive materials. Or, they can turn to newer and more advanced materials for PCB construction to maintain signal integrity. Even the MSI MEG X570 GODLIKE used for our testing here is still an 8-layer PCB and it’s made from fancier materials. This is one of the driving forces behind the price increases you will see over the older X470 motherboards. Another major factor is that AMD wants to position itself as a premier brand rather than the “other guy” peddling cheap alternatives to the products people really want but, can’t afford.

To that end, you should expect to see beefier VRM’s, thicker PCBs, and a ton of premium features. The example we used for this review is the MSI MEG X570 GODLIKE. It comes with a $699.99 price tag. Now, its an extreme example and somewhat of an outlier. Its onboard features help elevate the price, but be warned, if you want a Ryzen 3000 series CPU on a budget, you might want to look at existing X470 solutions which will co-exist with X570 for the time being.

Socket Compatibility Isn’t 100%

While AMD prides itself on its deep socket compatibility, such compatibility is a bit of a tricky proposition. I’ve argued that AMD tends to maintain socket compatibility far longer than it should. AMD’s product briefs do talk in depth about just how hard it was to make its 3rd generation processors maintain the same socket AM4 pinout and aside from having made the promise earlier, I’d argue it wasn’t worth the headache. That said, they did succeed although compatibility isn’t guaranteed. That is, you may not ever be able to install a 3rd generation Ryzen CPU into an X370 or B350 chipset-based motherboard. It is still up to the manufacturers to update the BIOS. AMD provided a chart for reference on this issue.

ryzen 3000 socket compatibility

For the most part, compatibility is broad, but it doesn’t appear that you will be able to install 1st generation Ryzen CPU’s into X570 motherboards. Similarly, it will be hit and miss with 3rd generation Ryzen CPUs and X370 or B350 motherboards. A320 motherboards are listed as a no go for 3rd generation Ryzen’s. However, AMD did not that this Is still up to the motherboard makers and A320 support for 3rd generation Ryzen’s is possible, and it’s been done at least once. In short, check with your motherboard manufacturer and flash your BIOS before getting rid of your 1st or 2ndgeneration Ryzen CPU’s and before you buy a 3rd generation chip.

Read Part 2 Here!

Here’s a Big Deal! AMD Ryzen 7 3700X Part – 1!

AMD today released its 3rd generation Ryzen desktop processor family based on its “Zen 2” microarchitecture.

amd logo

AMD today released its 3rd generation Ryzen desktop processor family based on its “Zen 2” microarchitecture. The company surprised everyone with its “Zen” desktop family because expectations from AMD on the processor front had faded due to a decade of Intel’s unchallenged market dominance and its eventual stagnation in per-core performance growth over the past few years. Something as simple as a 4% IPC uplift from AMD for its 2nd generation “Zen+” processors was met with cheers as Intel began waking up to the reality of a resurgent AMD. Wafer supply woes causing price hikes shielded AMD from Intel’s 8th generation Core. Intel fought back with the 9th generation Core processors, but pricing and supply issues in the desktop retail channel pushed sales to AMD. Fast forward to mid-2019 and AMD is in the thick of things.

Today, AMD is not only launching its 3rd generation Ryzen processors, but also its Radeon RX 5700 “Navi” graphics cards. What’s common between the two is the 7 nanometer silicon fabrication process they’re built on, which is significantly more advanced than the 14 nm process Intel has been stuck with. With Intel’s upcoming 10 nm “Ice Lake” processors arriving no sooner than 2020 for the desktop platform, AMD is eyeing a free rein on the market for a good three quarters by releasing “Zen 2” with the idea of toppling Intel’s 9th generation Core processors at every price point.

ryzen 7 3700x

At the heart of AMD’s effort is the “Zen 2” microarchitecture, which sets out to match or exceed the IPC of Intel’s latest “Coffee Lake” microarchitecture. This would be the first time in over 15 years that AMD beat Intel at IPC. While Intel led AMD at IPC, AMD led Intel at CPU core count. Intel responded to previous generations of Ryzen processors by increasing core counts of its mainstream-desktop processors for the first time in a decade. With the 9th generation Core, Intel achieved core-count parity. AMD’s response is not only matching the 9th generation Core at IPC, but also restoring AMD’s competitiveness by increasing core counts, at least in the high-end. Intel gave its coveted Core i9 brand-extension to its 8-core LGA1151 processor. AMD created the new Ryzen 9 series to match the Core i9 LGA1151 at price and IPC, while beating it at core counts. We hence have the Ryzen 9 3900X and the upcoming 3950X.

The Ryzen 9 3900X is a 12-core/24-thread processor, a 50% increase in core-counts right off the bat against the Core i9-9900K. The Ryzen 9 3950X, which will join the product stack this fall, is a 16-core/32-thread monstrosity priced at $750, while retaining its mainstream desktop credentials. Such high core counts are possibly not only due to the switch to 7 nm, but also because AMD has taken the multi-chip module (MCM) approach to building these processors, which are both similar and dissimilar to the Ryzen Threadripper. They’re similar in that the CPU cores are spread across two separate dies. They’re dissimilar in that there’s a second kind of die, the I/O controller.

With its first EPYC and Ryzen Threadripper processors, particularly the high core count WX models, AMD ran into several structural problems with memory bandwidth sharing between the CPU cores. The company fixed these with its 2nd generation EPYC processors, in which all dies with CPU cores talk to a centralized I/O controller die that has a monolithic memory controller, thereby making it possible for a CPU core to have the full bus width of the memory interface. With its 3rd generation Ryzen processors, AMD takes a similar approach. Two 8-core CPU complex dies talk to an I/O controller die over Infinity Fabric, which has the processor’s dual-channel memory interface and PCI-Express root complex.

The reasons for not building a monolithic 16-core die on 7 nm are economic. AMD is contracting TSMC to build its 7 nm wares, and it would want to minimize its silicon design to the smallest indivisible unit, an 8-core “Zen 2” chiplet. The company can build socket AM4 Ryzen processors with one or two of these chiplets to achieve up to 16 cores or drop up to eight of these on an SP3r2/TR4 package to achieve up to 64 cores. To minimize redundant components like with MCMs that use “Zeppelin” dies, AMD disintegrated the memory controller, PCIe root complex, and integrated southbridge on to the I/O controller die. This die has components that aren’t as power critical as CPU cores, so AMD could build it on the existing 12LPP (12 nm) process at GlobalFoundries. The Ryzen 9 3900X is an MCM with two 7 nm CPU core chiplets, and the I/O controller die. Models that have 8 CPU cores or less, such as the Ryzen 7 3700X or the Ryzen 5 3600X, only have one 7 nm chiplet besides the I/O controller die. This way, AMD makes the most out of its limited allocation at TSMC, which is building 7 nm chips for a dozen other companies.

In this review, we have with us the Ryzen 7 3700X, an 8-core/16-thread processor launched at the same $329 price as the 2700X and over $50 cheaper than the Core i7-9700K. Besides a higher core count and similar IPC, these processors offer the latest PCI-Express gen 4.0 bus, which doubles bandwidth for graphics cards and SSDs that support it.

amd vs intel price
Ryzen 7 3700X Market Segment Analysis

A Closer Look

The Ryzen 7 3700X ships in a large cubical paperboard box with carbon-fiber texture on some of its faces. There are clear markings on the front that tell you this is a 3rd generation Ryzen processor, which has PCI-Express gen 4.0 support. The back also mentions “Zen 2”. There are also some “VR ready” and NVMe logos on the box.

AMD includes a Wraith Prism RGB cooling solution with this processor, capable of thermal loads of up to 140 W.

cooler cables

Besides 4-pin PWM for its fan, the cooler includes two additional cables, an addressable 3-pin RGB cable to control the lighting and a USB cable that plugs into one of your motherboard’s USB 2.0/1.1 headers.

The Ryzen 7 3700X processor looks like any conventional AMD processor with a large IHS dominating the top, and a 1,331-pin micro-PGA in the bottom. You see national-origin markings for three places—USA, China, and Taiwan. The 7 nm “Zen 2” CPU chiplets are made in TSMC, Taiwan. The 12 nm I/O controller die is made in the US at GlobalFoundries. The two dies are packaged into the MCM at a facility in China.

Architecture

AMD’s 3rd generation Ryzen processors use the “Zen 2” microarchitecture. The 2nd generation Ryzen chips use an enhanced first-generation “Zen” derivative called “Zen+”, which has process and boost algorithm improvements eke out roughly a 4% IPC uplift. With “Zen 2”, AMD’s key design goal is to finally beat Intel in the IPC game. IPC, or instructions per clock, is loosely used to denote a CPU core’s performance at a given clock speed. For the past 15 or so years, Intel dominated AMD at IPC, while AMD attempted to make their processors competitive by cramming in more CPU cores than Intel at any given price point for competitive multi-threaded performance. Today’s software environment is increasingly multi-threaded, as are games. With “Zen 2”, AMD set itself an ambitious double-digit-percentage IPC uplift target to catch up or overtake Intel’s latest “Coffee Lake” microarchitecture at IPC. AMD didn’t stop there and even increased core counts for the platform at higher price points. The 3rd generation Ryzen family even includes a 16-core processor, which is a tremendous core count for the mainstream-desktop platform.

ryzen 3000 architechture

AMD’s 3rd generation Ryzen processors use the “Zen 2” microarchitecture. The 2nd generation Ryzen chips use an enhanced first-generation “Zen” derivative called “Zen+”, which has process and boost algorithm improvements eke out roughly a 4% IPC uplift. With “Zen 2”, AMD’s key design goal is to finally beat Intel in the IPC game. IPC, or instructions per clock, is loosely used to denote a CPU core’s performance at a given clock speed. For the past 15 or so years, Intel dominated AMD at IPC, while AMD attempted to make their processors competitive by cramming in more CPU cores than Intel at any given price point for competitive multi-threaded performance. Today’s software environment is increasingly multi-threaded, as are games. With “Zen 2”, AMD set itself an ambitious double-digit-percentage IPC uplift target to catch up or overtake Intel’s latest “Coffee Lake” microarchitecture at IPC. AMD didn’t stop there and even increased core counts for the platform at higher price points. The 3rd generation Ryzen family even includes a 16-core processor, which is a tremendous core count for the mainstream-desktop platform.

Before we get into the interesting and quirky way AMD crammed 16 cores into this chip, let’s talk about the “Zen 2” CPU core. After the colossal failure that was “Bulldozer,” AMD set out to once again build strong and monolithic CPU cores that share nothing except L3 cache with other cores. It achieved this desired result with “Zen”, which posted a mammoth 40%–50% IPC increase over “Bulldozer”, catapulting AMD back into competitiveness. “Zen” cores IPC sits somewhere between “Haswell” and “Skylake/Coffee Lake”, which was enough for AMD as it backed the IPC increase with higher core counts compared to Intel. Over the 8th and 9th generations of Core processors that retained the same IPC as “Skylake”, Intel shored up core counts to match AMD. Wanting to set up a definitive edge over Intel, AMD not only worked to increase IPC, but also core counts.

The “Zen 2” CPU core has essentially the same component layout and hierarchy as “Zen”, but with major changes and broadening of key components. As with “Zen” (or most x86 CPU cores), the “Zen 2” core is made up of five key components: Fetch, Decode, Integer, Floating-point, and load-store. Fetch and Decode tell the CPU core what needs to be done and what data or instructions are needed; Integer and Floating-Point Unit execute a mathematical model of what needs to be done depending on the data type and nature of the instruction; Load/Store are the I/O of the CPU core. At various levels, there are tiny buffers, registers that store instructions, and larger caches that cushion data-transfers between various components.

AMD updated the Fetch and Decode units, which contribute to IPC, by making the CPU work “smarter”. The updated Integer and FPU make the CPU work “harder”, the Load/Store unit’s job is to make sure the other components aren’t starved of things to do. The Fetch unit is updated with a TAGE branch predictor. Invented in 2006, TAGE is considered to be the best branch-prediction technique by the IEEE. AMD broadened the BTB (branch target buffers) at L1 and L2 by doubling the L1 entries to 512k, and L2 entries to 7,000 from 4,000. The ITA (indirect target array) has also been expanded. The design goal for updating the Fetch unit is to lower “mispredictions” (bad guesses) that wasted load/storage operations by 30 percent. The 32 KB L1 instruction cache has also been improved. The Decode unit has two improvements to the Op cache: improved instruction fusion and the ability to push up to 4,000 fused instructions per clock cycle.

We now move on to the two components that contribute the most to the IPC, the Integer and Floating-point Units. The Integer unit receives incremental updates in the form of a broader integer scheduler that handles 92 entries (up from 84), with four 16-entry ALU queues and one 28-entry AGU queue. The general-purpose physical register file has now been expanded to 180 entries, up from 168. The issue-per-cycle has been broadened to 7 from 6, which now includes 4 ALUs and 3 AGUs. The reorder-buffer (ROB) has been broadened to 224 entries, up from 192. The SMT (simultaneous multi-threading) logic has been tweaked to better share the ALUs and AGUs among the logical processors. The FPU has the bulk of the innovation with “Zen 2”. The load/store bandwidth of the FPU has been doubled to 256-bit, up from 128-bit on “Zen”. 

The core now also supports a sort of AVX-256: AVX/AVX2-flagged instructions with 256-bit registers. There are many applications for this, such as physics simulation, audio-stack execution, and memory-copy performance improvement. Multiplication operation latency has been improved by 33 percent.

ryzen 3000 architecture

Lastly, we move on to the Load/Store unit with a similar round of generational enhancements. The entry store queue is expanded to 48 entries, up from 44. The L2 TLB (translation lookaside buffer) has been expanded by 33% to 2,000 entries, and its latency improved. The 32 KB L1 Data cache has two 256-bit read paths and one 256-bit write path, with 64-byte load and 32-byte store alignment boundaries. The load/store bandwidth to L2 has been doubled to 32 bytes per clock.

ryzen 3000 architecture

We now move on to the cache-hierachy, which is essentially the same as “Zen.” Notwithstanding the technical changes described above, the “Zen 2” core still has a 32 KB 8-way L1I cache, a 32 KB 8-way L1D cache, and a dedicated 512 KB 8-way L2 cache. AMD doubled the shared L3 cache size to 16 MB. Every CCX (quad-core compute complex) on a “Zen 2” processor now has 16 MB of shared L3 cache. The doubling in L3 cache size was necessitated not just because Intel shares larger amounts of L3 cache among individual cores on the “Coffee Lake Refresh” silicon (16 MB shared among all 8 cores); but also because the larger L3 cache on a “Zen 2” CCX cushions data-transfers with the I/O controller die.

ryzen 3000 architecture

This brings us to the interesting and quirky way AMD achieved 16 cores. The Ryzen 9 3900X and Ryzen 7 3700X processor packages are codenamed “Matisse”. This is a multi-chip module (MCM) of one or two 7 nm 8-core “Zen 2” CPU chiplets and one I/O controller die built on the 12 nm process. AMD made sure only those components that tangibly benefit from the shrink to 7 nm—namely, the CPU cores—are built on the new process, while those components that don’t benefit from 7 nm stay on the existing 12 nm process, on the I/O controller die.

ryzen 3000 infinity fabric

These components include the processor’s dual-channel DDR4 memory controller; a 24-lane PCI-Express gen 4.0 root-complex, and an integrated southbridge that puts out some platform connectivity directly from the AM4 socket, such as SATA 6 Gbps and USB 3.1 ports. Infinity Fabric is the interconnect that binds the three dies by providing a 100 GB/s data path between each CPU chiplet and the I/O controller. The memory clock is now practically de-coupled from the Infinity Fabric clock, which should improve memory overclocking headroom. AMD also claims to have put in a lot of work to improving memory module compatibility across brands, especially since Samsung stopped mass-production of the expensive B-die DRAM chip that favored AMD processors. The memory scaling article talks a little more about this.

AMD “Valhalla” X570 Desktop Platform

ryzen 3000 x570

AMD delivered on its promise of 3rd generation Ryzen “Matisse” processors being backwards compatible with older socket AM4 motherboards, going all the way back to the AMD 300-series chipset, with a simple BIOS update. To make the most out of Ryzen “Matisse”—namely, PCI-Express gen 4.0 connectivity and increased CPU/memory overclocking headroom, you’re expected to use one of the latest motherboards that use the AMD X570 chipset. The X570 is an entirely different chip from the X470 and X370. The older chipsets were supplied by ASMedia, and were rather slim in their downstream connectivity. 

The X470 only puts out 8 PCIe gen 2.0 downstream lanes, for example. The X570 modernizes all I/O by putting out up to 16 PCIe gen 4.0 downstream lanes. This enables additional M.2 PCIe gen 4 slots on your motherboards for the latest SSDs featuring PCIe gen 4 support and creates room for many new bandwidth-hungry onboard devices, such as 10 GbE adapters, next-generation Thunderbolt, 802.11ax controllers, etc. Along with the “Matisse” SoC, the X570 also puts out a number of 10 Gbps USB 3.1 gen 2 ports. Motherboards based on X570 also implement modern network connectivity options, such as 2.5 GbE and 802.11ax WLAN.

Read Part 2 Here

Shots fired as AMD cut price of Navi GPUs just before launch to combat Nvidia’s RTX Super cards

AMD’s Navi cards are just over 24 hours away from release right now, but after having been gazumped by Nvidia’s RTX Super graphics cards earlier in the week, AMD have just announced a little gazumping of their own in the form of an unexpected price cut for both the RX 5700 and RX 5700 XT.

navi

AMD’s Navi cards are just over 24 hours away from release right now, but after having been gazumped by Nvidia’s RTX Super graphics cards earlier in the week, AMD have just announced a little gazumping of their own in the form of an unexpected price cut for both the RX 5700 and RX 5700 XT. How about that for a nice bit of friendly competition, eh?

The new AMD Navi prices are as follows:

  • AMD Radeon RX 5700: Originally $379, now $349
  • AMD Radeon RX 5700 XT: Originally $449, now $399
  • AMD Radeon RX 5700 XT 50th Anniversary Edition: Originally $499, now $449

That’s a $50 cut for both RX 5700 XT cards (the 50th Anniversary edition is ever so slightly faster than the standard version), and a $30 saving on the non-XT model. UK pricing is still TBC at time of writing, but all three price drops are pretty significant, especially when you consider where they sit among Nvidia’s RTX cards.

While I can’t say what their actual performance is like until Sunday, you may remember that throughout the run-up to Navi’s arrival, AMD have maintained that the RX 5700 is set to be a little bit faster than Nvidia’s RTX 2060, while the RX 5700 XT is supposedly a little bit quicker than the RTX 2070. Now that the RTX Super cards have arrived on the scene, however, I reckon a more appropriate pair of competitors for them are, you guessed it, Nvidia’s RTX 2060 Super and RTX 2070 Super.

Take a look at Nvidia’s respective US RRP prices, however, and you’ll find they’re now quite a bit more expensive:

  • Nvidia GeForce RTX 2060: $349
  • Nvidia GeForce RTX 2060 Super: $399
  • Nvidia GeForce RTX 2070: $499
  • Nvidia GeForce RTX 2070 Super: $499

If AMD’s comparisons hold true, then that’s quite the saving indeed – and a very loud shot fired against Nvidia’s pack of RTX Super cards. I should point out, of course, that Nvidia’s cards also let you take advantage of real-time ray tracing and their upscaling performance boosting DLSS tech, neither of which are present on the Radeon RX 5000 cards, and that you get two free games (Control and Wolfenstein: YoungBlood) when you opt for an RTX Super card. Regular RTX cards, meanwhile, just come with a free copy of Wolfenstein.

That adds a little bit of extra value to Nvidia’s proposition, but AMD have their own bundle deal as well: a three-month subscription to Xbox Game Pass for PC, which gives you access to over 100 games for absolutely nothing. It’s also worth noting that the same deal applies to more or less any new AMD graphics card bought between July 1st 2019 and March 10th 2020, starting from the RX 560 all the way up to the Radeon 7, as well as any new Ryzen 3000 CPU or one of AMD’s existing Ryzen 2nd Gen CPUs.

The proof, of course, will be in the benchmark-flavoured pudding as to whether it’s really worth jumping ship to AMD or not, but if you’re in the market for a 1440p graphics card, then the competition has just got a heck of a lot hotter. Will this price cut help AMD’s Navi cards win the fight for best graphics card? All will be revealed on Sunday.

Nokia 6.1 Price in India Cut, Now Starts at Rs. 6,999

Notably, Nokia 6.1 was launched last year with a starting price of Rs. 16,999.

The revised Nokia 6.1 price is now reflecting on the Nokia Online Store

HIGHLIGHTS

  • Nokia 6.1 4GB + 64GB variant price slashed to Rs. 9,999
  • Flipkart and Amazon.in are yet to show the updated pricing
  • The Nokia phone was launched in India back in April last year

Nokia 6.1 price in India has been dropped as low as Rs. 6,999. The price cut is currently reflecting on the Nokia India Online Store. To recall, the Nokia 6.1 aka Nokia 6 (2018) was launched back in April last year in two different variants — 3GB + 32GB and 4GB + 64GB. The mobile phone is based on Google’s Android One programme and sports a full-HD display. The Nokia 6.1 also has an octa-core Qualcomm Snapdragon 630 SoC onboard.

Nokia 6.1 price in India

According to the listing on the Nokia India Online Store, the Nokia 6.1 ₹8,949 price in India now starts Rs. 6,999 for the base 3GB + 32GB variant. The 4GB + 64GB option of the Nokia 6.1, on the other hand, is available with a revised price tag of Rs. 9,999.

The latest price revision is yet to be reflected on e-commerce sites such as Amazon.in and Flipkart. Similarly, offline retailers may still offer the handset with its previous pricing.

The Nokia 6.1 was launched in India last year with a starting price of Rs. 16,999, though most recently it was seen on the Nokia Online Store with a price of Rs. 8,999 for the 3GB + 32GB variant, whereas its 4GB + 64GB model was available at Rs. 10,999.

We’ve reached out to HMD Global for clarity on the latest price cut and will update this space when we hear back.

Nokia 6.1 specifications

The Nokia 6.1 runs Android Pie and features a 5.5-inch full-HD (1080×1920 pixels) IPS display with a 16:9 aspect ratio. It is powered by the Snapdragon 630 SoC, paired with up to 4GB RAM.

For photos and videos, the Nokia 6.1 has a single 16-megapixel image sensor at the back — along with an f2.0 lens and an LED flash. An 8-megapixel sensor is available at the front with an f/2.0 fixed focus lens.

The Nokia 6.1 has 32GB and 64GB of onboard storage options that both are expandable via microSD card (up to 128GB). Connectivity options on the phone include 4G VoLTE, Wi-Fi 802.11ac, Bluetooth v5.0, GPS/ A-GPS, FM radio, USB Type-C (v2.0), and a 3.5mm headphone jack. There is Nokia’s spatial audio technology. Also, the phone has a fingerprint sensor at the back.

HMD Global has provided a 3,000mAh battery on the Nokia 6.1 that is claimed to deliver up to 16 hours of talk time on a single charge.

PUBG LITE Beta quick review: Feels like heaven for PUBG loyalists on low-end PCs

PUBG LITE Beta quick review: Feels like heaven for PUBG loyalists on low-end PCs

pubg lite preparation time

PUBG LITE is now available in India under the beta testing program. This free-to-play game brings the best of PUBG on PC as well as PUBG MOBILE to deliver the best PUBG experience.

If you like playing battle royale games and adore PUBG’s interpretation of the genre, you get enough choices to enjoy the experience anyplace at any time. If you want to settle in for serious battlecraft with the most realistic simulation, PUBG on PC is the best way to get that experience. If you want to enjoy quick sessions of PUBG to massage your gamer ego by beating a few noobs on the move, PUBG MOBILE is at your service. However, there’s a massive gap between these two versions of the same game and it seems that the studio has finally filled in the gap with PUBG LITE.

Okay, the PUBG team may be promoting the game as a more accessible version of the hardcore version on PC but in a lot of ways, this game feels more than just being accessible. And the best part about PUBG LITE is that it costs nothing to download and play – similar to the PUBG MOBILE. Despite coming with many compromises, the PUBG team promises the authentic PUBG experience and even in its beta stage, that claim seems to be true.

The game has been available since yesterday and I spent a few hours with it on my regular non-gaming laptop to see whether it manages to paste a massive smile.

Where does PUBG LITE exactly sit in delivering the PUBG experience?

As I said, the PUBG team is claiming to offer the authentic PUBG experience with this game. However, in the time I spent with the game, I think the game delivers an experience that surpasses the authentic PUBG experience.

The gameplay experience is a mix between PUBG for PC and PUBG MOBILE. You get the same physics system as PUBG for PC but knitted with all the conveniences of PUBG MOBILE. This, in my opinion, is enough to compel players to try out the game.

I spent a lot of hours in PUBG and even after dedicating a lot of time to the game, it’s a struggle for a casual gamer like me. The almost realistic physics and simulation require real fighting instincts to manage to reach in the top standings. On the other hand, PUBG MOBILE is joyous with is easier gameplay style but the physics and watered down graphics at times make players want the same experience in a better package. PUBG LITE hits the sweet spot in between.

With PUBG LITE, you are getting the same PUBG physics, which is a bonus for gamers who like a little more challenge. At the same time, the game helps you in its own sweet ways to be more competitive. Every time I was being shot at by an enemy, the game notified me about the direction from where the firing was coming – just like PUBG MOBILE. Even when an enemy approaches with vehicles, the mini-map shows the positioning of the enemy thus giving casual players like me a chance to fight back against the pro-level players. The game also keeps the player updated on how much points he/she is earning while surviving each round.

The in-game physics are quite impressive. The way the player moves or crouches behind a wall while under attack seems realistic. The vehicle dynamics also encourage players to use them in order to move inside the world. Like all versions of PUBG, you can dress up your character with clothes earned from in-game crates. To add to the realism, players need to pick up weapons, medication or other items from the floor manually. What’s even more challenging is that players need to reload the weapon manually when the magazine is empty.

The best part, however, is that all players are right now learning the game and hence, you aren’t facing pro-level players right now. I was able to reach the top 5 most of the time with very simple battle skills.

I also found the controls to be fairly easy. You get simple controls schematics that you may have witnessed in gaming franchises such as Grand Theft Auto and Battlefield. I tried it with a controller but the game is best played with a mouse and keyboard setup.

PUBG LITE nails it in the graphics department

The fully loaded version of PUBG impresses so much with its graphics that often I find myself getting distracted (and eventually shot by enemies). With the PUBg LITE, you can get almost the same graphics, albeit with noticeable compromises. I played this game on a laptop specced with an Intel Core i7-8750H processor and an NVIDIA GeForce 1050Ti Max-Q GPU and in the maxed-out settings, the overall visual effects look good but not as good as PUBG. PUBG LITE does miss out on the amazing light effects and some high-resolution textures of few objects. The player details are a notch below than the full version of PUBG but it’s still not bad in any way — fro someone coming from PUBg MOBILE, it is a massive upgrade. The reflections are good but not great. In short, the graphics feel like they are from the early years of the PS4 era. But given that this game is meant to run on low-end laptops, I can’t complain.

I also played the game in the lowest graphical settings and the just looked like a scaled-up version of PUBG MOBILE in its lowest settings. It doesn’t look pretty in these settings but if you are restricted by your hardware, this is the best experience you are going to get.

First impressions: A must-play game for battle royale enthusiasts

As a gamer accustomed to fighting titles and open-world shooters on PC, PUBG LITE impressed me highly with what it offers. The team has just nailed the sweet spot in delivering the perfect gameplay experience that many casual PUBG players have been longing for since a year now. With good graphics and amazing physics, you can’t get a better battle royale experience than PUBG LITE. And especially when you consider that this is free to play, it becomes a must-have for all gamers.