Techgage logo

Workstation Powerhouses: AMD Ryzen Threadripper 16-core 2950X & 32-core 2990WX Review

AMD Ryzen Threadripper 2nd-gen Packaging

Date: August 13, 2018
Author(s): Rob Williams

Just as we were starting to believe that 16-core CPUs were amazing, AMD has come along and dropped a 32-core bombshell. It’s called Ryzen Threadripper 2990WX, a workstation chip that cuts through renderers like butter. We’re putting it up against the forthcoming 2950X, and six other unsuspecting processors.



AMD Ryzen Threadripper 16-core 2950X & 32-core 2990WX Review

AMD’s newest Ryzen Threadripper processors haven’t been kept much of a secret, and if you’ve followed the progression of Ryzen’s second-generation chips, you pretty much know what to expect from the new top-end 2000-series SKUs. In a nutshell, we have 12nm dies, better clocks, greater efficiency, and… thirty-two freaking cores!

Let’s rehash some history. Ahead of summer 2017, the most powerful desktop processor someone could buy was the 10-core Intel Core i7-6950X, priced at $1,750. The following summer, AMD released its Ryzen Threadripper 1950X, boasting 16 cores for $999. At the time, and even right now, 16 cores is impressive. But here we have a chip with twice the number of cores and threads. I didn’t think AMD would do it, but here we are. This current era of AMD is beyond aggressive, and it’s exciting.

Consider this. Intel’s first six-core CPU was the i7-980X, released back in 2010. It took four years from that point to see a core increase. It came to us in the form of the eight-core i7-5960X Extreme Edition. Things began to accelerate from that point on, giving us the 10-core i7-6950X in 2016. AMD gave us its 16 cores in 2017 (Intel gave us 18, to give credit there), and now, AMD is giving us 32. I’m still in relative awe here.

The question must be raised: “What about EPYC?” EPYC was the reason I thought this 32-core chip wouldn’t happen, at least so soon. But, EPYC is different, as is all enterprise-level gear. It adheres to many different certifications and processes, and not to mention supports an eight-channel memory controller. You can also combine two chips in the same server unit, something Threadripper can’t do (it wouldn’t surprise me if AMD changed that with TR3, because I’m done thinking ‘not happening’ to everything that sounds too good to be true).

“But isn’t $1,799 a bit low for a market-leading processor?” I’ve mulled this a bit, and do admit I thought AMD should have charged more for this chip than it has. I’ve come to the conclusion, though, that AMD is actually just doing the market at large a favor by forcing a long-overdue CPU pricing recalibration. A 10-core chip shouldn’t have cost people $1,750 a mere two years ago.

Before moving too much further, here’s a look at AMD’s product stack as it currently stands… but please note that the 2970X and 2920X are due in October. The 2950X, meanwhile, is due at the end of the month, whereas the biggest chip, 2990WX, is available right this moment.

AMD’s Current-gen Ryzen Processor Lineup
CoresClock (Turbo)L2+L3MemoryTDPPrice
Threadripper WX-series
2990WX32 (64T)3.0 GHz (4.2)16+64MBQuad250W$1799
2970WX24 (48T)3.0 GHz (4.2)12+64MBQuad250W$1299
Threadripper X-series
2950X16 (32T)3.5 GHz (4.4)8+32MBQuad180W$899
2920X12 (24T)3.5 GHz (4.3)6+32MBQuad180W$649
Ryzen 7
R7 2700X8 (16T)3.7 GHz (4.3)4+16MBDual105W$329
R7 27008 (16T)3.2 GHz (4.1)4+16MBDual95W$299
Ryzen 5
R5 2600X6 (12T)3.6 GHz (4.2)3+16MBDual95W$219
R5 26006 (12T)3.4 GHz (3.9)3+16MBDual65W$189
R5 1600X6 (12T)3.6 GHz (4.0)3+16MBDual95W$219
R5 16006 (12T)3.2 GHz (3.6)3+16MBDual65W$189
R5 1500X4 (8T)3.5 GHz (3.7)2+16MBDual65W$174
R5 14004 (8T)3.2 GHz (3.4)2+8MBDual65W$169
Ryzen 3
R3 1300X4 (4T)3.5 GHz (3.7)2+8MBDual65W$129
R3 12004 (4T)3.1 GHz (3.4)2+8MBDual65W$109
Ryzen w/ Radeon Vega Graphics
R5 2400G4 (8T)3.6 GHz (3.9)0.5+4MBDual65W$169
R3 2200G4 (4T)3.5 GHz (3.7)0.5+4MBDual65W$99

Due to the fact that I spent about 80 hours the past week performance testing, and have a SIGGRAPH conference looming, I am not going to get into the nitty-gritty here as much as I’d like, but that doesn’t apply to the test results. There are some 35 graphs in this review, tackling many different scenarios.

SIGGRAPH is a professional visualization conference that brings in the world’s brightest workstation users to the west coast to talk shop and promote their wares. AMD will be doing just that at the event, because this 32-core chip is perfectly suited for that audience. To be absolutely clear, the 2970WX and 2990WX are not for gamers; they’re for workstation users (or anyone who needs cores more than IPC). We’re testing tens of thousands of dollars worth of workstation software here, not $60 games.

In the reviewer kit AMD sent press, the company included Cooler Master’s new Wraithripper cooler, one that was designed specifically for this product. Interestingly, Cooler Master had begun design work on the cooler before there was ever an intention to make it an official cooler. When the company talked to AMD about its creation, the idea was hatched, and the rest is history.

It’s hard to believe that an air cooler is marketed towards a CPU series that offers 32-core SKUs, and it’s a testament to the fact that AMD has actually made an efficient CPU architecture with Zen (and subsequently, this Zen+). I am being extremely serious when I say that when I ran all of my benchmarks, I was never concerned about temperatures when using this air cooler. I mean sure, it’s massive, but I use water (NZXT Kraken) to cool the top-end Core X-series chips, and I actually need more cooling ability than I have. If you need an example, the PC crashed during the AVX-512-powered Sandra test with the i9-7960X, entirely because of heat. After using NZXT’s software to manually crank the fan and pump to max, the issue disappeared. Meanwhile, I’m using this Wraithripper air cooler to handle the 32-core PC beside it.

I am not a cooling genius, so I’ll let those who actually know what they’re doing relay their thoughts. As far as I am concerned, though, the Wraithripper is more than capable for this job. Even during stress, the fins don’t get too hot to touch.

And with that, the next page is going to go over our testing methodology, and the five systems used for testing eight processors. After returning from AMD’s Italy press event a couple of weeks ago, I overhauled our test suite, and chose the eight most relevant processors for this kind of performance look. So… enough said; let’s just get a move on:

Test Methodology & Systems

Benchmarking a CPU may sound like a simple enough task, but in order to deliver accurate, repeatable results, and not to mention results that don’t favor one vendor over another, strict guidelines need to be adhered to. That in turn makes for rigorous, time-consuming testing, but we feel that the effort is worth it.

This page exists so that we can be open about how we test, and give those who care about testing procedures an opportunity to review our methodology before flaming us in the comments. Here, you can see a breakdown of all of our test machines, specifics about the tests themselves, and other general information that might be useful.

Let’s start with a look at the test platforms, for AMD’s TR4 (MSI’s MEG X399 Creation and Aorus’ X399 GAMING 7) and AM4 (ASUS’ CROSSHAIR VII HERO), along with Intel’s LGA2011-v3 (ASUS’ ROG STRIX X299-E GAMING), and LGA1151 (EVGA Z370 FTW).

To prevent unexpected performance results, the “Multi-Core Enhancement” optimizations (effectively overclocking all cores to max turbo, instead of just two cores) offered by ASUS on its respective motherboards is disabled. There is, however, an exception. For Ryzen 2 testing, I had to use the DOCP profile on the motherboard to take proper advantage of the memory kit (though it doesn’t appear to overclock anything with the basic settings).

All platforms were run with DDR4-3200 speeds, and 14-14-14 timings. This is the first time in the history of the site that we’ve been able to benchmark multiple platforms against each other with identical memory configurations (all thanks to the fact that I could benchmark two rigs at once since AMD included a perfectly spec’d memory kit to complement the other – cheers to whoever decided that one). Host and scratch disk SSDs were also identical between all systems, so this is as true apples-to-apples as we can get.

Here’s the full breakdown of the test rigs:

Techgage’s CPU Testing Platforms

AMD TR4 Test Platform #1
ProcessorAMD Ryzen Threadripper 2990WX (3.0GHz, 32C/64T)
MotherboardMSI MEG X399 Creation
CPU tested with BIOS 1.0EN43 (Jul 27, 2018)
MemoryG.SKILL TridentZ (F4-3200C14-8GTZ) 8GB x 4
Operates at DDR4-3200 14-14-14 (1.35V)
GraphicsNVIDIA TITAN Xp (12GB; GeForce 398.82)
StorageHost OS: WD Blue 3D NAND 1TB (SATA 6Gbps)
Scratch Drive: Samsung 970 EVO 512GB (NVMe)
Power SupplyCooler Master Silent Pro Hybrid (1300W)
ChassisCooler Master MasterCase H500P Mesh
CoolingCooler Master Wraithripper Tower Cooler
Et ceteraWindows 10 Pro (Build 17134), Ubuntu 18.04 (4.15 kernel)
AMD TR4 Test Platform #2
ProcessorAMD Ryzen Threadripper 2950X (3.5GHz, 16C/32T)
AMD Ryzen Threadripper 1950X (3.4GHz, 16C/32T)
MotherboardGIGABYTE X399 Aorus Gaming 7
CPU tested with BIOS F10 (July 10, 2018)
MemoryG.SKILL Flare X (F4-3200C14-8GFX) 8GB x 4
Operates at DDR4-3200 14-14-14 (1.35V)
GraphicsNVIDIA TITAN Xp (12GB; GeForce 398.82)
StorageHost OS: WD Blue 3D NAND 1TB (SATA 6Gbps)
Scratch Drive: Samsung 970 EVO 512GB (NVMe)
Power SupplyEnermax RevoBron 80+ Bronze (600W)
ChassisEnermax Equilence
CoolingEnermax Liqtech TR4 AIO (240mm)
Et ceteraWindows 10 Pro (Build 17134), Ubuntu 18.04 (4.15 kernel)
NVIDIA GeForce GTX 1080 Ti - SLI Configuration NVIDIA GeForce GTX 1080 Ti - SLI Configuration
As tested configuration: AMD Ryzen Threadripper 2990WX
NVIDIA GeForce GTX 1080 Ti - SLI Configuration
Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
As tested configuration: AMD Ryzen Threadripper 2950X
Sony Ps4 Pro Angled View
Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
As tested configuration: AMD Ryzen Threadripper 1950X
Sony Ps4 Pro Angled View

There’s not too much to say here outside of the fact that since I wanted to keep the 2990WX to its own machine for further testing, the 2950X was tested in the same motherboard as the 1950X. That gave me a chance to test a new Threadripper in a year-old motherboard, and I can report it was flawless (I flashed the latest EFI with the 1950X still in the board, which made things a touch easier).


AMD AM4 Test Platform
ProcessorsAMD Ryzen 7 2700X (3.7GHz, 8C/16T)
MotherboardASUS Crosshair VII HERO Wi-Fi
CPU tested with BIOS 0804 (July 9, 2018)
MemoryG.SKILL Flare X (F4-3200C14-8GFX) 8GB x 4
Operates at DDR4-3200 14-14-14 (1.35V)
GraphicsNVIDIA TITAN Xp (12GB; GeForce 398.82)
StorageHost OS: WD Blue 3D NAND 1TB (SATA 6Gbps)
Scratch Drive: Samsung 970 EVO 512GB (NVMe)
Power SupplyEVGA Bronze 600B1 (600W)
ChassisFractal Design Define C
CoolingNoctua NH-U12S SE-AM4 (1x120mm)
Et ceteraWindows 10 Pro (Build 17134), Ubuntu 18.04 (4.15 kernel)
NVIDIA GeForce GTX 1080 Ti - SLI Configuration NVIDIA GeForce GTX 1080 Ti - SLI Configuration
As tested configuration: AMD Ryzen 7 2700X
NVIDIA GeForce GTX 1080 Ti - SLI Configuration

Nothing special had to be done on the CROSSHAIR VII HERO to get up and running quickly. As covered above, I had to run with the DOCP Standard profile to get the memory to run at optimum settings. Again, in testing, I couldn’t see how that improves performance outside of applying the proper memory settings, so I believe the setting requires further input for an actual boost to performance (eg: overclocking).


Intel LGA2011-3 Test Platform
ProcessorsIntel Core i9-7980XE (2.6GHz, 18C/36T)
Intel Core i9-7960X (2.8GHz, 16C/32T)
Intel Core i9-7900X (3.3GHz, 10C/20T)
MotherboardASUS ROG STRIX X299-E GAMING
CPU tested with BIOS 1004 (Nov 14, 2017)
MemoryG.SKILL TridentZ (F4-3200C14-8GTZ) 8GB x 4
Operates at DDR4-3200 14-14-14 (1.35V)
GraphicsNVIDIA TITAN Xp (12GB; GeForce 398.82)
StorageHost OS: WD Blue 3D NAND 1TB (SATA 6Gbps)
Scratch Drive: Samsung 970 EVO 512GB (NVMe)
Power SupplyCorsair Professional Series Gold AX1200 (1200W)
ChassisCorsair Carbide 600C
CoolingNZXT Kraken X62 AIO (280mm)
Et ceteraWindows 10 Pro (Build 17134), Ubuntu 18.04 (4.15 kernel)
CyberPowerPC AMD VR Gaming PC - Keyboard Switches CyberPowerPC AMD VR Gaming PC - Keyboard Switches
As tested configuration: Intel Core i9-7980XE
CyberPowerPC AMD VR Gaming PC - Keyboard Switches
CyberPowerPC AMD VR Gaming PC - Keyboard Switches CyberPowerPC AMD VR Gaming PC - Keyboard Switches
As tested configuration: Intel Core i9-7960X
CyberPowerPC AMD VR Gaming PC - Keyboard Switches
CyberPowerPC AMD VR Gaming PC - Keyboard Switches CyberPowerPC AMD VR Gaming PC - Keyboard Switches
As tested configuration: Intel Core i9-7900X
CyberPowerPC AMD VR Gaming PC - Keyboard Switches

This ASUS motherboard tries to trick us into accepting core boost optimizations after selecting the XMP profile. That’s disregarded, to keep things as apples-to-apples as possible.


Intel LGA1151 Test Platform
ProcessorsIntel Core i7-8700K (3.7GHz, 6C/12T)
MotherboardEVGA Z370 FTW
CPU tested with BIOS 1.08 (March 20, 2018)
MemoryG.SKILL Flare X (F4-3200C14-8GFX) 8GB x 4
Operates at DDR4-3200 14-14-14 (1.35V)
GraphicsNVIDIA TITAN Xp (12GB; GeForce 398.82)
StorageHost OS: WD Blue 3D NAND 1TB (SATA 6Gbps)
Scratch Drive: Samsung 970 EVO 512GB (NVMe)
Power SupplyCorsair RM650X (1200W)
ChassisNZXT S340 Elite Mid-tower
CoolingCorsair Hydro H100i V2 AIO Liquid Cooler (240mm)
Et ceteraWindows 10 Pro (Build 17134), Ubuntu 18.04 (4.15 kernel)
Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
As tested configuration: Intel Core i7-8700K
Sony Ps4 Pro Angled View

There’s nothing to report here at all. EVGA doesn’t have weird boosts built-in, which makes it a great platform for benchmarking with. Its biggest offense is that the EFI is garish.

Testing Considerations

For the bulk of our testing, we use Windows 10 build 17134 with full updates as the base. Our basic guidelines are:


Encoding Tests

Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
Rendering: Adobe Dimension CC
Sony Ps4 Pro Angled View
Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
Rendering: Arnold in Maya 2018
Sony Ps4 Pro Angled View
Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
Rendering: Blender
Sony Ps4 Pro Angled View
Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
Rendering: Cinebench
Sony Ps4 Pro Angled View
Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
Rendering: Cinema 4D
Sony Ps4 Pro Angled View
Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
Rendering: Corona in 3ds Max 2019
Sony Ps4 Pro Angled View
Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
Rendering: POV-Ray
Sony Ps4 Pro Angled View
Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
Rendering: Radeon ProRender in Maya 2018
Sony Ps4 Pro Angled View
dBpoweramp - Convert FLAC to MP3 dBpoweramp - Convert FLAC to MP3
Rendering: V-Ray Next in 3ds Max 2019
dBpoweramp - Convert FLAC to MP3
Intel Core i7-6700K (CPU-Z & GPU-Z) Intel Core i7-6700K (CPU-Z & GPU-Z)
Rendering: V-Ray Benchmark
Intel Core i7-6700K (CPU-Z & GPU-Z)

Adobe Premiere Pro
Agisoft PhotoScan
DaVinci Resolve
HandBrake
LameXP
MAGIX Vegas

(You can click each name to go straight to that result.)


Rendering Tests

Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
Encoding: Adobe Premiere Pro (Media Encoder)
Sony Ps4 Pro Angled View
Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
Encoding: Agisoft PhotoScan
Sony Ps4 Pro Angled View
Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
Encoding: Capture One
Sony Ps4 Pro Angled View
Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
Encoding: DaVinci Resolve
Sony Ps4 Pro Angled View
Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
Encoding: HandBrake
Sony Ps4 Pro Angled View
Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
Encoding: LameXP
Sony Ps4 Pro Angled View

Adobe Dimension
Arnold (Maya 2018) (Also relevant to: 3ds Max, C4D, Houdini, Katana, Softimage)
Blender
Cinebench
Cinema 4D
Corona (3ds Max 2019) (Also relevant to: C4D)
KeyShot
POV-Ray
Radeon ProRender (Maya 2018) (Also relevant to: 3ds Max, Blender, C4D, SolidWorks)
V-Ray Next (3ds Max 2019) (Also relevant to: C4D, Houdini, Maya, Rhino, SketchUp)
V-Ray Benchmark
SiSoftware Sandra 2018

(You can click each name to go straight to that result.)


Gaming Tests

UL 3DMark

(You can click each name to go straight to that result.)

Linux Tests

Ubuntu 18.04 is the OS of choice for our Linux testbed, as it’s both simple to set up, and so de facto that everyone reading the results should feel at home. The OS is left as stock as possible, with minor software added, and everything updated.

Before testing begins, we take the Phoronix Test Suite suggestion of enabling the “performance” power profile. That command is (as sudo):

echo performance | tee /sys/devices/system/cpu/cpu*/cpufreq/scaling_governor

Blender - Pavillon (Linux)

Compiling
Compression
Encryption
Scientific Analysis
HandBrake
Blender
Ray Tracing

(You can click each name to go straight to that result.)


If you think there’s some information lacking on this page, or you simply want clarification on anything in particular, don’t hesitate to leave a comment.

Encoding: Premiere Pro, Vegas, PhotoScan, Resolve & LameXP

We’re going to kick off this Threadripper performance look with a handful of encoding tests. Encoding is one of those scenarios that can be extremely hit-or-miss when it comes to taking good advantage of big CPUs. Sometimes, applications will give the impression that they’re making proper use of the CPU, but I’ve found more than once that some applications actually just use the entire CPU very poorly.

Fortunately, the situation is getting a lot better over time. As an example, Adobe Lightroom, for most of its life, didn’t use more than a few cores and threads. Today, the application can use most of whatever CPU you can hand it, although I am not sure it uses it entirely effectively (based on the fact that I planned to include it here, but irregularities across multiple runs told me to drop it).

The same could be said for Capture One, a reader-requested application that was going to make a debut here. It could cap CPUs at 100% usage just as well, but the performance was the same on the 2990WX as it was the 1950X. And 7900X. That said, I plan to test both applications more down-the-road, to see if I can’t eke more reliable benchmarks out of them.

Those applications didn’t make the cut, but Adobe’s Premiere Pro, MAGIX’s Vegas, Agisoft’s PhotoScan, Blackmagic Design’s DaVinci Resolve, and an application most reading this haven’t heard of, LameXP (music converter), help fill out this page.

Adobe Premiere Pro CC

AMD Ryzen Threadripper 2950X & 2990WX Performance in Adobe Premiere Pro (Software AVC Encodes)
AMD Ryzen Threadripper 2950X & 2990WX Performance in Adobe Premiere Pro (Software HEVC Encode)

Right off the bat, we had to start with some results that need explaining. Clearly, AMD’s 32-core wonder does very well in video encoding, but not always. Defying simple logic, the 32-core chip fell to the back of the pack in the Techgage video test, which is an encode of this YouTube video. Considering the fact that the 2950X and other Ryzens perform much better in that test, it’s safe to say that the threads and architecture as a whole are not being used effectively enough.

Those tests above represented straight CPU encodes, but things fare better (in terms of your patience) when the GPU gets involved:

AMD Ryzen Threadripper 2950X & 2990WX Performance in Adobe Premiere Pro (CUDA AVC Encodes)
AMD Ryzen Threadripper 2950X & 2990WX Performance in Adobe Premiere Pro (CUDA HEVC Encode)

With the GPU, every encode time dramatically drops, and while the 2990WX still falls back in that particular Techgage project, it leads the pack in the other straight encodes.

MAGIX Vegas

This application is a recent entrant to our suite, having debuted a few weeks ago as a GPU test. Like Premiere Pro, Vegas also uses the CPU to good effect, so for a test here, I ran one of our projects though a CPU-focused AVC encoder, converting 4K/60 to 1080/60.

AMD Ryzen Threadripper 2950X & 2990WX Performance in MAGIX Vegas

So far, the 2990WX is exercising its muscle pretty well. But, based on the difference between the 16- and 32-core chips, it does seem like more optimization could take place. Conversely, it could also be that different projects would give more interesting results. While I haven’t talked much about it up to this point, the 2950X is delivering strong performance as well, consistently (and this continues) edging ahead of the 1950X it’s replacing.

Agisoft PhotoScan

PhotoScan is a powerful 3D model and mapping application that allows you to generate detailed information and models from a set of images. In our particular test, 80 or so images or product packaging is used, and is riddled with guideline markers to help tear apart the object from the scene. Have an action figure you want to turn into a 3D asset? PhotoScan is a tool that could do that.

AMD Ryzen Threadripper 2950X & 2990WX Performance in Agisoft PhotoScan

This application takes great advantage of big CPUs, but for whatever reason, it behaves better with SMT disabled, on both AMD and Intel. Intel fares better with SMT on, but on the 2990WX in particular, it should definitely be turned off. It could be that in time, the application will take better advantage of the mammoth chip. Other scenarios show that it can be done. I have provided Agisoft with detailed information about this, so here’s to hoping an update tackling the issue is in our future. We will follow up SMT testing further in the future.

DaVinci Resolve

This is another application that’s been recommended to us a few times, and coincidentally, a brand-new version is en route, giving us a chance to check it out ahead of its release (it’s in a public beta). I am ridiculously inexperienced with Resolve, so the project here is a simple 4K/60 video converted to a QuickTime YouTube upload using default settings. For good measure, anti-camera shake is applied to the video.

AMD Ryzen Threadripper 2950X & 2990WX Performance in DaVinci Resolve

Clearly, the CPU is only relied-upon for so much with this application, or perhaps project. Resolve is best known as a GPU encoding suite, so it’ll probably be better-suited for our future workstation GPU content. What you as an end-user needs to know, though, is that the application can definitely take advantage of 10 and more cores. Just not much more than that (at least in our case).

LameXP

For many years, I used dBpoweramp for music conversion, but while the application supports up to 16 threads, it doesn’t go beyond that. LameXP does, but only to 32 threads. That’s enough to top out a SATA SSD’s realistic write speed of about 300MB/s, so for even more threads, NVMe would come in handy (one was used as the project / scratch disk for all tests on this page).

AMD Ryzen Threadripper 2950X & 2990WX Performance in LameXP

This test seems to be sensitive to both IPC and cores, so if you have the perfect blend, you’ll rule the roost. The top four CPUs here perform extremely well in comparison to the bottom few.

We’ll make note here that we will be testing the performance impact of SMT/multithreading in future articles, since time constraints limited what we could delve into here. However, it’s clear from initial testing that software optimization for more than 32 cores, real or virtual, is mixed at best.

Rendering: Arnold, Blender, KeyShot, Radeon ProRender & V-Ray Next

There are few things I find quite as satisfying as rendering: seeing a bunch of assets thrown into a viewport that manage to turn into a beautiful scene (well, none that I touch). Rendering also happens to be one of the best possible examples of what can take advantage of as much computer hardware as you can throw at it. This is true both for CPUs and GPUs.

On this page and next, we’re tackling many different renderers, because not all renderers behave the same. That will be proven in a few cases. If you don’t see a renderer that applies to you, it could to some degree in the future, should you decide to make a move – either to a different design suite, or different renderer plugin. An example: V-Ray supports more than just Autodesk’s 3ds Max that I test with; it also supports Cinema 4D, Maya, Rhino, SketchUp, and announced only just this week, Houdini.

Another cross-suite renderer? Only the best-named one: Arnold.

SolidAngle Arnold

AMD Ryzen Threadripper 2950X & 2990WX Performance with SolidAngle's Arnold

We’re off to a great start here. The projects used here, are ones I found online for free, so they may not be the most representative of more complex projects, but I think the point of scalability gets across. AMD’s 32-core wonder dominates in both renders here. The 2950X continues to deliver just a bit more performance than its older brother.

Blender

AMD Ryzen Threadripper 2950X & 2990WX Performance in Blender

With Arnold, both tested projects showed a 15 second delta between them on the 2990WX, but with Blender, an even greater lead can be seen in the BMW scene. What’s a little funny about that is just how simple in design that project is in comparison to Agent 327, which was designed specifically to show off what Blender can do. Big or small, the projects will scale extremely well on beefier CPUs.

KeyShot

AMD Ryzen Threadripper 2950X & 2990WX Performance in KeyShot

KeyShot is one of the coolest rendering applications I’ve ever touched, but it does give us some odd results with some chips. The Porsche model rendered pretty much as expected straight up the stack, but the interior render, which has many reflective surfaces, didn’t render very well on the 2990WX. Again, this is a hint that optimizations to the software could be made to take better advantage of these chips. This is highlighted by the fact that it isn’t an SMT issue, as the 16-core chips with SMT are faster than the 32-core 2990WX.

Really though. Download the KeyShot demo, a bunch of free scenes, and have some fun.

Radeon ProRender

AMD Ryzen Threadripper 2950X & 2990WX Performance with AMD's Radeon ProRender

For some reason, AMD’s own Radeon ProRender doesn’t scale well with the new 32-core chip. I have tested multiple beta versions of the plugin (to be released this week), and nothing changed. I am not sure at this point if the issue is related to the Maya plugin, or the projects I’m using, but it’s unfortunate. Of all the renderers I expected to scale brilliantly on the 2990WX, it was this one. Again, highlighting the need for optimization going forward (32 cores is new territory outside of Xeon Phi CPUs).

Chaos Group V-Ray Next

AMD Ryzen Threadripper 2950X & 2990WX Performance with Chaos Group's V-Ray Next

I took a look at V-Ray’s GPU Next renderer a few months ago, so it was nice to be able to test out the CPU version for this review, since it’s typically what most people will be using for production renders (GPU renderers are more commonly used during the design phase of a project for quick and dirty iterations).

Interestingly, V-Ray is the first renderer to not show much of a strength on the 2990WX, which is a bit of a shame as I was hoping for a lot more. What’s strange about this is that with the standalone benchmark, seen below, the gap is widened far more than it is above.

AMD Ryzen Threadripper 2950X & 2990WX Performance in V-Ray Benchmark

I am not sure why there’s such a discrepancy here, but with SIGGRAPH up next week, I think I have some questions for some of the excellent folk at Chaos Group.

Rendering: Adobe Dimension, Cinebench, Cinema 4D, Corona, POV-Ray

I hope you’re not sick of rendering, because there’s a lot more of it on this page. On the previous page, many big-name renderers were tackled, but on this page, some others are as well, along with synthetic tests that will allow you to test your own PC and compare its results to ours.

Adobe Dimension

At AMD’s Ryzen Threadripper second-gen event, held in Modena, Italy a couple of weeks ago, the company had some demo PCs showing off what the new chip could do. I was intrigued by an application I had never seen before, one that turned out to be Adobe Dimension. Here I thought After Effects was the go-to for Adobe rendering. Nope, Dimension is another, and it’s an interesting one.

This is a renderer for those who essentially want to take a real-life photo and add rendered objects to it. This would mostly be used in marketing, but as an end-user just experimenting, it’s quite satisfying to use. Our render here involves a countertop in a beautiful kitchen that has only two rendered objects: a gelato container, and a glass cup to hold it in. Lots of rays, refractions, and all that good stuff.

AMD Ryzen Threadripper 2950X & 2990WX Performance in Adobe Dimension CC

As you can probably guess without much effort, this application hasn’t cooperated with me so much. The project I chose to test with, won’t open on the 2990WX when SMT is left enabled, hence the reason for it being disabled. That’s also the reason I included a second test (in orange), using a project provided by AMD. The fact that some projects have issue on this CPU isn’t a secret to the company (thankfully), and it’s actively working with Adobe to get it fixed. Further testing will take place in due time.

Cinema 4D & Cinebench

In my look at AMD’s and Intel’s 16-core processors a few weeks ago, I mentioned that testing Cinebench at this point in time seemed pointless, since it was four generations out-of-date. Coincidentally (or not, since SIGGRAPH is next week), MAXON announced C4D R20 last week, making Cinebench five generations out of date. I hope the company decides to turn R20 into a benchmark, but even if not, Cinema 4D becomes a stature in our testing going forward.

Let’s start with what’s familiar:

AMD Ryzen Threadripper 2950X & 2990WX Performance in Cinebench Multi-thread

No surprises here. During the press event in Modena, AMD had overclockers on hand to push the 2990WX with liquid nitrogen, and before long, a world record was broken: 7600+. That’s an impressive number, but even at stock speeds, the 2990WX breaks the 5,000 mark, which is quite something considering the 7980XE sits at 3,400.

One great thing about Cinebench is that it also gauges single-threaded performance, and that gives us just about what we’d expect to see:

AMD Ryzen Threadripper 2950X & 2990WX Performance in Cinebench Single-thread

Intel’s super-fast i7-8700K sits comfortably at the top, with all of the CPUs below it largely being ranked by our expectations of each chip’s IPC. The 1950X is an odd result at the bottom, though, as I had definitely expected the 2990WX to be there. This is consistent with POV-Ray, coming up soon, but it comes down to the improved cache system in the second-gen Zen+ architecture. Clocks speeds also play a part as well.

AMD Ryzen Threadripper 2950X & 2990WX Performance in Cinema 4D

In real-world tests, Cinema 4D emulates V-Ray to a certain extent in that real-world performance doesn’t align with the synthetic testing, but the gains are much more favorable to the 2990WX here than they were in V-Ray. Here, AMD clearly reigns supreme. If only Intel had that 28-core Core i9 shown-off at Computex here today; we’d see some interesting results. If AMD sent us Wraithripper for the 2990WX, it’d be appropriate for Intel to send us LN2 with that future chip.

Corona Renderer

AMD Ryzen Threadripper 2950X & 2990WX Performance with Corona Renderer

Corona Renderer is not one I’m very familiar with, but given the strengthening that was being done to our test suite, I had to give it a shot. And… it’s impressive, generating gorgeous results very quickly. For my testing, I plugged it into 3ds Max, and the results are once again kind to AMD, putting the 2990WX at the top. If only the 8700K was an eight-core chip, then the 2700X would have better competition!

POV-Ray

AMD Ryzen Threadripper 2950X & 2990WX Performance in POV-Ray Multi-thread
AMD Ryzen Threadripper 2950X & 2990WX Performance in POV-Ray Single-thread

POV-Ray is a classic renderer that’s become a de facto benchmark over the years, thanks to its amazing ability to scale. And by classic, I’m not kidding. I benchmarked version 3.7 eleven years ago with my Intel Core 2 Extreme QX6850 review. That’s still the current version, but as the graphs above suggest, 3.8 is in beta, and has been opted for here, as it does actually deliver higher results.

Ultimately, like Cinebench, POV-Ray paints a great picture of the 2990WX, although not to the same extent. The single-threaded performance on the 2990WX still proves to be better than the 2950X, which defies logic (due to the lower clock speed), but on a different day, the roles could have been reversed. They are extremely close as it is.

Linux: Rendering, Compiling, Compression, & Cryptography

On the previous pages, there were many encoders and renderers tackled, while on this page, Linux is going to help us tackle those same scenarios, among others. To do most of the testing here, I rely on Phoronix’s excellent Test Suite (PTS), which makes it super easy to test a wide-range of scenarios – not just for CPU or GPU, but also I/O and memory.

The OS acting as a base for testing is Ubuntu’s 18.04, which is left as stock as possible. The performance power profile is enforced before testing is undertaken, although on the 2990WX machine, the terminal kicked back that the file I usually echo to doesn’t exist. This is the only platform I’ve had this happen on, so it’s definitely related to the design of the chip, and I’d suspect EPYC would have generated the same kind of result. Nonetheless, it proved to not be important, based on HandBrake testing between the OSes (which were for the most part exact).

Compiling

AMD Ryzen Threadripper 2950X and 2990WX Performance in Compiling (Linux)

I’d have to be a real meathead to not start this page off with a compile benchmark, a test almost as de facto as SuperPi used to be (is it still?). As expected, AMD’s many cores helped it reach the top of the chart here, though Intel gives the chip a real run for its money with the ImageMagick compile. The kernel compiling in nearly 30 seconds really makes me want to haul out my overclocking hat.

Blender

AMD Ryzen Threadripper 2950X and 2990WX Performance in Blender (Linux)

The use of this test might be a little redundant since I also use it for Windows testing, but PTS tests the same scene I use in a different, more thorough way, so it’s nice to have a second opinion. And, once again, the 2990WX really shines here.

HandBrake

AMD Ryzen Threadripper 2950X and 2990WX Performance in HandBrake (Linux)

As I found out the hard way during testing and retesting, HandBrake 1.1.1 does not love AMD that much, as the encode times were much worse than Intel’s in Windows. Yet, I didn’t see that in Linux. After investigation, it turns out that 1.1.1 somehow introduced a change that dramatically impacted AMD performance. As an example, in Linux, the 2700X took 736 seconds for the x265 test, whereas in Windows, it took 1700.

1.1.0 does not have this issue, and that’s the version used for the testing above. Ubuntu’s repository doesn’t have 1.1.1, and that worked to my favor here, since it gave me actually usable results. Hopefully the next version of the application gets rid of this issue (please, devs?).

Ray Tracing

AMD Ryzen Threadripper 2950X and 2990WX Performance in Ray Tracing (Linux)

A benchmark that concludes in under 2 seconds might seem dumb, but is it reaaally when it still scales? A render is a render, after all, and they all scale pretty much infinitely if you want them to. In all three tests, the 2990WX’s raw horsepower is put to get use, and because Tachyon favors AMD somehow, the 2990WX is actually separated quite a bit from the 7980XE. A full 0.9s!!!

If you’ve never heard of these renderers, the information is still relevant to other ray tracing renderers, like AMD’s Radeon ProRender or Blender’s Cycles.

Rodinia

AMD Ryzen Threadripper 2950X and 2990WX Performance in Rodinia (Linux)

These tests represent some of the most grueling calculations, so it’s a true stress test, and one that fared well for AMD. The 2990WX performed well in the solver test, but that’s easy to ignore with the kinds of gains seen in the LavaMD one (it’s not a volcano health website; I looked).

Encryption

AMD Ryzen Threadripper 2950X and 2990WX Performance in Encryption (Linux)

Continuing its reign at the top, the 2990WX pushes well ahead of the competition in the encryption test, one that’s normally Intel’s domain (as you can see by how far ahead the 7960X pulls ahead of the 2950X).

Compression

AMD Ryzen Threadripper 2950X and 2990WX Performance in Compression (Linux)

While the 2990WX killed it in the encryption test, it has to get by in the compression test by simply beating Intel by a few percentage points. That aside, it’s worth noting just how well compression can scale with processor cores. Most of us don’t spend our days compressing files, but it sure is nice to have it be as fast as it can be.

Memory

AMD Ryzen Threadripper 2950X & 2990WX Memory Performance (Linux)

I debated whether or not it was important to include this set of results here, since it’s also covered on the following page with Sandra, but it actually does prove useful to show that not all benchmarks are built alike. Here, AMD’s 2990WX sits comfortably at the top, but as you’ll see on the following page, it’s actually Intel which does with Sandra’s take (probably thanks to the use of AVX-512).

Ultimately, memory is a very hard thing to gauge, and the results may not even affect your personal workload as much as you think it does. I’ve been tempted to do RAM scaling testing for a while, so I feel like this might be as good a reason to plan for it than any other.

System: SiSoftware Sandra 2018

For the sake of time, I didn’t run as many Sandra tests as I normally would, especially since I wanted to focus on as many real-world scenarios as possible. Still, it’s not a bad idea to turn to synthetic benchmarks to get a true sense of scaling in certain areas, because some, more than others, are unexpected.

Case-in-point: despite the fact that the 2990WX is more powerful overall than the 7980XE, that chip slaughtered this one with almost twice as many cores. This will come as no surprise to anyone who’s been following the site, or CPUs in general, for a while. Intel’s held a dominating position in multimedia in general for quite a while, and its latest use of the AVX-512 instruction set helps out a lot with that. With the value AMD is offering consumers right now, software developers are going to be forced to notice before long, and begin improving the performance on Zen. It’s unfortunate to require 32 AMD cores to beat out 10 of Intel’s here. But again, the encode tests on the third page of this article prove that these synthetic results do not always translate to real-world scenarios. They’re still useful to gauge overall potential, however.

AMD Ryzen Threadripper 2950X & 2990WX Performance in SiSoftware Sandra 2018 Multimedia

The multimedia test didn’t fare super-well for AMD, but the mathematics test does. Here, the 2990WX leaps well ahead of the 7980XE, skewing the rest of the results by hogging so much space up top.

AMD Ryzen Threadripper 2950X & 2990WX Performance in SiSoftware Sandra 2018 Arithmetic

The crypto test on the Linux page exhibited great results for the 2990WX, so it’s no surprise that the trend continues here. AMD has actually been super-strong in crypto on both the CPU and GPU sides of the fence, marking one of the greatest gains in performance with recent architectures that I haven’t noticed since Intel introduced the AES-NI instruction set way back in the day.

AMD Ryzen Threadripper 2950X & 2990WX Performance in SiSoftware Sandra 2018 Cryptography

Finally, as mentioned on the previous page, different memory benchmarks allegedly give different test results, as the Stream-based benchmark in our Linux testing put AMD cleanly at the top, whereas Sandra paints a bit of a different picture thanks to its use of AVX-512, which is exclusive to Intel’s Core X-series chips (on the desktop).

AMD Ryzen Threadripper 2950X & 2990WX Performance in SiSoftware Sandra 2018 Memory Bandwidth

All of the platforms that use quad-channel memory controllers easily hit well over 50GB/s at DDR4-3200 speeds, and if memory matters, you clearly want better than dual-channel.

Gaming & Final Thoughts

With its WX branding, AMD has made it clear that the top-end Threadrippers are not your ordinary CPUs. They’re definitely not meant for gaming, or, at least, that’s not their focus. If you want ultimate gaming, you want high IPC and clock speeds more than lots of cores, so the i7-8700K or 2700X should be considerations.

The unfortunate caveat with many-core chips is that the more cores you add, the lower your overall clocks are going to be. That means that slower multi-core chips will likewise be slower in general usage tasks (not to say you’d notice all of the impact; much of this would require benchmarks to appreciate the differences), whereas high clock chips, like the i7-8700K, will give you the snappiest performance possible.

I will say it’s with a bit of surprise, then, that the 2950X has performed so well in 3DMark, using its effective performance to help drag up its overall score to again lead the pack:

AMD Ryzen Threadripper 2950X & 2990WX Performance in UL 3DMark Fire Strike (1080p)
AMD Ryzen Threadripper 2950X & 2990WX Performance in UL 3DMark Fire Strike (4K)
AMD Ryzen Threadripper 2950X & 2990WX Performance in UL 3DMark Time Spy (1440p)

Can you see why the 2990WX isn’t targeting gamers? At 1080p, the detriment is very noticeable, but given just how far behind the top chip’s CPU scores are behind everything else, this again seems like a case of the application not taking advantage of the hardware on offer, a problem I’m hoping will start to disappear more as time goes on, and developers begin to consider these architectures in their code. The problem exists on the gaming side just as well as the software side.

Final Thoughts

It’s not going to be difficult summing this one up. Without question, AMD is offering an unbelievable chip with the Ryzen Threadripper 2990WX. I’m still surprised that it even exists, more than most, it seems. I just couldn’t see AMD literally doubling the core count after a single year, but here we are.

I’ve talked lots about core counts, but in actuality, it’s not the core counts that really matter at the end of the day. Do you buy your Radeons and GeForces based on their core counts? Probably not. What you want is the overall performance. The same principle should be applied here. On one hand, it’s unfortunate that AMD needs 32 cores to only just beat Intel’s 18-core in select tests, but AMD’s charging about the same ($1799) for its 2990WX that Intel is still charging for its 7980XE ($1850).

But it’s not those select tests that are going to matter to most people who are reading this review to learn how the chip can impact their workload. Like any workstation scenario, not everyone’s needs are the same. One chip might excel in one area, but if it’s not the area you deal with, then the point is moot.

AMD Ryzen Threadripper 2990WX Specs Sticker

If you use Arnold, for example, you stand to see incredible gains in performance with a chip like the 2990WX. The same could be said for pretty much every other renderer tested. Encoding scenarios are really hit or miss, even within the same suite. AMD performed great in certain tests, but fell behind in others – often attributed to applications that wouldn’t use the CPU properly.

Having tested the 16-core and 18-core chips before this, I felt like I was really quite prepared for this performance look. Admittedly, though, I was thrown some curve-balls, but it’s really to be expected. You can’t just double the cores and expect every bit of software to pretend like it doesn’t notice. HandBrake in Linux used literally half of the CPU, for example. Then there were applications like PhotoScan which did not use AMD’s many-core chips too well – or Intel for that matter, since SMT off proved better for both of them.

AMD Ryzen Threadripper 2990WX In Its Protective Case

Fortunately, scaling worked well overall, just to different levels. If you are a render addict, it’s almost guaranteed that you’re going to see enormous gains, and I’d suspect those gains will only get better as time goes on, as developers design for these architectures and core counts. I have suspicions that AMD will be demoing Threadripper with a partner at next week’s SIGGRAPH, which will be a testament to that.

Even with some 35 graphs, it’s hard to wrap this up and feel like I covered enough ground. There’s much more that could have been tackled, but that’s probably best left for follow-up content. Gaming will never be a focus of mine with these biggest Threadrippers, but I will definitely be following-up on testing once in a while to see how the ecosystem has improved to better cater to these big chips. I’ve been in contact with a few ISVs already, and hope to talk to more at SIGGRAPH next week to gauge their thoughts.

If there’s anything I didn’t cover that you think I should have, please feel free to jump into the comment section and let me know. If you also have ideas for follow-up testing, that’d be welcomed, as well. New GPUs are coming soon, so time is strapped, but I am eager to get more testing done with the 2990WX soon. Having all that power on tap is kind of addicting.

Cooler Master Wraithripper Logo

I don’t really think about awarding Editor’s Choices that often; it really has to impress me to earn it. And the Ryzen Threadripper 2990WX definitely does earn it. It’s rough around some edges, partly due to AMD’s obsession with releasing more cores before developers realize they need to code for them, but it’s an incredible performer; a true “ultimate” workstation processor. It might not be for your workload, but if it is, the gains will be simply sublime. And, you don’t need exotic cooling for it.

If it hasn’t been stated enough already though, software optimization is required here, especially with SMT involved. While some software may have been exposed to multi-socket systems with very high-end servers and core-counts, for workstations it’s a different matter. The tell-tale signs are good this early in the release based on synthetic tests, but there is considerable room for improvement in some live software.

I know I didn’t touch on the 2950X as much as I should have, but I feel like we honestly knew what to expect from that to begin with, especially given the second-gen Ryzen series, and also what we knew from the 1950X (which was performance tackled just a few weeks ago). Fortunately, the 2950X was in fact faster than the 1950X in every test, sometimes by a greater margin than I expected. A nice iteration, but of course not a reason to jump up from the 1950X. For those looking to build a new rig, and are tempted by a 16-core processor priced at $899, it’s an excellent option.

ASUS ZenFone 2 Smartphone - Techgage Editor's Choice
AMD Ryzen Threadripper 2990WX

Copyright © 2005-2019 Techgage Networks Inc. - All Rights Reserved.