Comprehensive Review AMD Ryzen Threadripper PRO 5995WX, Performance Verified!

• The High Performance "Zen 3" Architecture. The fastest cores in the world for PC gamers.
• "Zen 3" cores are designed for the world's best gaming processors with a historic 19% higher instructions per cycle, lower latency and power efficiency leadership.
• The combination of a state-of-the-art architecture and industry-leading 7nm process give the AMD Ryzen™ 5000 Series a +24% generational improvement in energy efficiency, and a stunning 2.8X lead over competing architectures.
• Total Redesign with Lower Latency

The Review AMD Ryzen Threadripper PRO 5995WX. On August 12, 2022, the ban on the sale of individual CPUs from the Ryzen Threadripper PRO 5000 WX Series, the top CPU for HEDT (High-End Desktop), was lifted. The 5995WX was announced in March, but until now it has only been available as a factory workstation. And now the ban on buying single CPUs has been lifted, so it can be used as an option for self-built PCs.

The Ryzen Threadripper PRO 5000WX series is available in all five models, but the model sold in stores is higher than the 24-core/48-thread Ryzen Threadripper PRO 5965WX. The reason why there is no model below is allegedly due to the same number of cores as the current Ryzen 9.

And it costs USD 7251 including tax for the top model, the Ryzen Threadripper PRO 5995WX. Compared to the previous generation, Ryzen Threadripper PRO 3000WX, the number of cores has not changed, but the price has increased by more than USD 2100. The impact of currency depreciation is also considered strong, but I think the biggest reason is that AMD stands alone in the HEDT CPU market.

I was lucky enough to have the opportunity to test the Ryzen Threadripper PRO 5995WX. Verification was very difficult as we encountered various difficulties such as stability, timing, and equipment constraints, but we wanted to verify through various tests how “powerful” it is compared to mainstream CPUs.

CPU Monster is dedicated to WRX80 motherboards.

First, let’s briefly compare the specifications. The following table compares the Ryzen Threadripper PRO 5995WX with comparable models + α. As mentioned above, the number of cores in the Ryzen Threadripper PRO 5995WX has not increased from the previous generation. However, the architecture has been changed to Zen 3 and the maximum boost clock has been slightly raised. However, the TDP remains the same at 280W.

In addition, the “WRX80” chipset adopted from the Ryzen Threadripper PRO 3000WX series will continue to be used as a motherboard. The “TRX40” motherboards used in the Ryzen Threadripper 3000X series have significantly changed the memory bus width and PCI Express Lane counts specifications, so a WRX80 motherboard is a must. Therefore, switching from the Ryzen Threadripper PRO 3000WX series is only possible with a BIOS update.

To be honest, the differences between the Ryzen Threadripper PRO 3000WX and the 5000WX series are minor. Apart from the fact that the architecture has changed from Zen 2 to Zen 3, only security-related features (AMD Shadow Stack) have been added. The number of PCI Express lanes and memory channels has increased dramatically compared to the non-PRO Ryzen Threadripper 3000X series, so switching to the Ryzen Threadripper PRO 5000WX series will depend on whether or not these differences matter.

What about the verification environment?

Although the Ryzen Threadripper PRO 5995WX (hereinafter referred to as 5995WX) was acquired this time, the Ryzen Threadripper PRO 3995WX (hereinafter referred to as 3995WX) could not be obtained at that time, which was one generation earlier. Therefore, we will compare it to the Ryzen Threadripper 3990X (hereinafter referred to as 3990X), which has the largest core in the non-PRO. The company is also preparing Ryzen 9 5950X (hereinafter referred to as 5950X) and Core i9-12900KF (hereinafter referred to as 12900KF) from mainstream CPUs. It would be great if you could see it from the point of view of the strongest HEDT CPU performance from the point of view of the mainstream CPU.

1

Ryzen Threadripper system memory is a total of 64GB of the 8GB SanMax DDR4-3200 module borrowed as verification equipment for the 5995WX, while Ryzen and Core i9 are 32GB. However, this verification is designed to work well in a 32GB environment.

The OS enables Windows 11 Pro, core isolation (VRS), and Secure Boot for all features. Drivers for Radeon are the most challenging part of AMD’s environment, but they chose AMD Software: PRO Edition, which is tuned for creative applications (with wider different release intervals).

Multi-threaded performance is well improved

CPUs like Threadripper shine best for CG rendering. First of all, we will verify with the “CINEBENCH R23” standard. It’s interesting to see how far the single-threaded performance has extended by the increase in clocks.

The 5995WX and 3990X have the same number of cores but have advantages over the 5995WX in terms of architecture, clocks, and memory bus width. This is why the 5995WX is 15% higher on multithreaded scores. In addition, the single-threaded score also increased by more than 15% from 3990X. If you think the horsepower of the 3990X is still not enough, the 5995WX could be a good place to go.

Compared to the mainstream 5950X and 12900KF CPUs, the single-threaded CPU is lower, but the multi-threaded score is about three times higher. Considering the number of cores ratio (1:4), the growth rate is slow, but it is commendable that this performance can still be produced in a single socket.

Besides CINEBENCH R23, I wanted to try some bench CG rendering. First, is the official benchmark for Blender. The Blender version is the latest “3.2.1” at the time of verification.

In each scene, 5995WX scores 18-21% higher than 3990X. It also only scored about three times as much as the 5950X, which has fewer cores. THIS IS GENERAL FOR CINEBENCH R23.

Next is verification on the V-Ray Benchmark. This bench can also be raised using CUDA or OptiX, but this time it’s a Radeon environment, so only “CPU V-Ray” uses only the CPU used.

The trend is the same here. The 5995WX is not only faster than CINEBENCH, but it can be judged that it is stable and fast on the CG rendering bench, and can expect up to 20% performance over the 3990X and more than 3 times the performance of the 5950X.

What about when CPU load is low

Next, let’s try the “Standard Test” of the PCMark 10 benchmark to see the overall performance. Unlike CINEBENCH R23, the CPU load is very light, so CPUs like the Threadripper series don’t score. A mainstream CPU like the 5950X should score better. However, with the increase in single-thread performance, it is expected to get a score higher than 3990X.

In this verification environment, PCMark 10 and other Adobe-affiliated applications often encounter problems such as errors appearing in scenes using OpenCL. I decided to compare it to the score at the end of the race, but in the end, I couldn’t get enough results to compare in the 3990X environment (I stopped at the Productivity test). Unfortunately, these results only show that the 5995WX scores are lower than mainstream CPUs.

I also tried UL Procyon, which is from the same UL company in “PCMark 10”. The first is “Photo Editing Benchmark” which runs Photoshop and Lightroom Classic. We also had an issue that stopped midway through testing, but as a result of changing the driver to AMD Software: PRO Edition, the situation has improved a lot.

The overall score (blue bar) gives the 5995WX a lower rating than mainstream, high-clocked CPUs like the 12900KF. However, if we focus on Batch Processing (gray bars) here, we would like to note that although the 5995WX is higher than the 5950X, the Image Retouching (orange bars) is lower at the 5995WX than the 5950X.

In other words, from these results, it can be said that the 5995WX can perform better than the 5950X in the workflow that only ends with Lightroom Classic, but it can be said that it is inferior to the 5950X in terms of workflows involving Photoshop. The reason why the 5995WX can surpass the 5950X is thought to be due to the wide memory bus. Given that the 5995WX can carry up to 2TB of memory, the 5995WX has the potential to shine in scenes associated with developing ultra-high resolution RAW.

Also, try the UL Procyon Office Productivity Benchmark. This is a bench that runs Office 365 (Word/Excel/PowerPoint/Outlook).

In this test, the 5995WX did not have a strong advantage over mainstream CPUs. However, it has been shown to have some performance improvements over the 3990X.

How do creative apps perform?

Next, let’s verify it with a creative app other than CG rendering. First of all, try it with the free video encoder “HandBrake”. The time it takes to encode a 4K@60fps video (approximately 3 minutes) to the full MP4 or MKV HD@30fps format on the HandBrake has been measured. For image quality and other settings, the “Super HQ 1080p30 Surround” and “H.265 MKV 1080p30” presets are used.

Since all the processing here is done on the CPU, the Threadripper series has the advantage of having a very large number of cores, but note that the difference between the 3990X and 5950X is almost non-existent in H.265. H.264 (Super HQ 1080p30 Surround) is more effective than the 3990X because it finishes in less time.

And the 5995WX has a speed of 3990X or higher even at H.265. If you’re cost-effective enough to get there with the 5950X, but you’re more willing to put every second of the minute aside than money, investing in the TP PRO 5000WX series isn’t a bad idea.

Next, we compare the time it takes to encode a 4K video (approximately 3 minutes) edited in “Premiere Pro 2022” to MP4 format in “Media Encoder 2022”. The encoding is software (CPU), VBR 50Mbps one-pass encoding, and frame interpolation using optical flow.

This test failed to complete the H.265 encoding in the Threadripper series. The 5950X and 12900KF were able to finish the race without difficulty, so there is a possibility of failure because the CPU has too many cores. Experience has shown that H.264, where the results are comparable, has a very high percentage of GPU processing, and as a result, there is little difference between any of the CPUs (or rather, GPU performance can be significantly reduced).

I’d also like to try DaVinci Resolve 18, which relies heavily on the GPU. Create 8K videos with four 4K videos lined up and add simple processing like color adjustments, and output to 8K MP4 (H.265) format. Since the bitrate is set automatically and the encoder is “Native”, processing will not complete no matter how long it takes, so AMD using GPU encoding is selected.

GPU (especially VRAM capacity) is more important than CPU, and the Threadripper series doesn’t offer any advantages. Depending on how the timeline is created, CPU performance may be more critical, but even if you say video encoding, the Threadripper series is not unconditionally strong.

Next up is the After Effects 2022 test, which measures the time it takes to apply “content-based fill” to a 10-second Full HD video. The cache is cleared before it finishes.

It is seen that the 5995WX is superior to the 5950X but still far from the 12900KF. The advantages of the 5995WX, which can be installed in large numbers in situations where a large amount of main memory is used come into play, but with materials and verification like this, it doesn’t appear to have reached mainstream CPUs.

Another thing is to measure the time to be analyzed with “3D camera tracking” for 4K video with a playback time of 40 seconds.

Again, the 3990X couldn’t finish the race, so we can’t see any difference in performance between the 5995WX and the 3990X. However, because the 12900KF completes processing in a very short time, it is difficult to find the advantages of the Threadripper series.

I’d also like to try Topazlabs’ DeNoise AI. Denoise processing was performed on 100 JPEG images measuring 6000 × 4000 pixels and the time to output to JPEG was measured. The learning model during the denoise process uses “Severe Noise”.

Since any environment takes about 2 to 3 seconds per sheet, it can be said that the difference due to CPU output is relatively easy to understand as a result of processing 100 sheets at a time. The fastest is the 12900KF, but the 5995WX is also very fast. Since the CPU core is underutilized, it can be concluded that the speed of the 5995WX is related to the thickness of the memory bandwidth.

Finally, take a look at light baking times (computing previous exposure information and “baking” into textures) using the Unreal Engine. Testing using “Infiltrator” and lighting quality in “Production” settings, click the “Build” button to compare processing times. Unreal Engine uses 4.27.

The 3990X, which has been lackluster so far, tops the list, but as found in previous tests, the behavior around the GPU drivers is a bit suspicious, so it’s too soon to assume that the 3990X is faster. However, it can be said to be faster than mainstream CPUs due to the large number of cores.

What about power consumption?

Let’s check the power consumption using actual app bench verification here. Stability values after 10 minutes of system startup and maximum values in the above benchmark with HandBrake were measured with the Ratock RS-WFWATTCH1 system, but also the power flowing to the two EPS12V systems during HandBrake encoding using Elmorlabs’ PMD (Power Measurement Tool).

The 3990X consumes more power in terms of the overall system, but when you look at the energy flowing through the EPS12V, the 5995WX consumes more power instantly. Given the massive power consumption of 12900KF, it can be said that the CPU has a very high wattage performance for CG rendering and HandBrake processing.

And the graph below tracks the state of the CPU during coding with this HandBrake with “HWiNFO PRO”. Here’s a limited comparison of both the 5995WX and 3990WX.

The graph above will solve the mystery of power consumption. Looking at the average value of 64 cores only, the 5995WX is lower than the 3990X, but the 5995WX is higher even though the situation is the same as the maximum clock. Since the average and maximum values of the 3990X are almost the same, it can be said that the motherboard for the 3990X I tried this time is tuned to boost almost all cores in much the same way, while the 5995WX is tuned to preferentially increase the highly demanded core clock (though it is assumed that the load on the CPU core is about the same in both cases).

The following graph shows the transition between “CPU Packet Power” and “12V EPS power consumption by PMD” during encoding.

First, in the CPU Package Power, it’s 3990X longer at higher scenes. The reason for the 3990X higher overall system power consumption is thought to be because it also increases the core clock, which is not very expensive.

And when we look at the 12V EPS power consumption, we can confirm that the 5995WX jumped instantly, but overall, there doesn’t seem to be any difference between the 5995WX and the 3990X. If you look at the average values for the entire period (15 minutes) in the graph, the 5995WX is 193W while the 3990X is 190W, which is hardly different. While no material can be said to be superior to the 5995WX from these data alone, it can be said that at least the “WRX80 Creator” used for this verification managed to put out the 5995WX while reducing power consumption.

Summary: Will there be developments in the market without a benchmark?

This ends the Ryzen Threadripper PRO 5995WX verification. Since the core count remains the same at 64, we didn’t see any surprising results, but we can confirm that it can deliver performance that surpasses mainstream CPUs depending on the application. There are many scenes where creative applications lose out to mainstream CPUs, but the fact that the memory capacity can be expanded up to 2 TB is commendable.

However, it has to be said that the price of the core 5995WX unit is high. Prices are influenced by currency depreciation and can be said to be expensive because it is a product for the chosen professional more than anything else, but I feel that is also due to the absence of the Intel Core X and Xeon series, which could be rivals. There are currently no Intel CPUs that can compete directly with the Threadripper series in terms of core count, process rules, and performance.

If Sapphire Rapids, the line of hope, is also delayed until 2023, you should opt for the Threadripper series if you want a CPU with multiple cores for HEDT. There is no progress without the pressure of racing. What’s wrong with writing this when it’s a Threadripper verification article, but if Intel doesn’t try hard to keep up with HEDT, the next generation Threadripper could be a mediocre update with just the Zen 4 architecture.

Check out our store if you are interested in buying this product Review AMD Ryzen Threadripper PRO 5995WX with great deals and other parts and accessories there. You can also click the Shop Now button below to buy it from external vendors.

Do you think you have other thoughts about the Comprehensive Review AMD Ryzen Threadripper PRO 5995WX, Performance Verified? You can comment below or discuss more related to “Comprehensive Review AMD Ryzen Threadripper PRO 5995WX, Performance Verified!” in the CnwinTech Forum. Also, read more articles about Razer or other interesting tech tips and tricks at CnwinTech.