Techgage logo

A Real HEDT CPU: AMD Ryzen Threadripper 3960X & 3970X Workstation Performance Review

AMD Ryzen Threadripper Chip Shot Thumbnail

Date: November 29, 2019
Author(s): Rob Williams

We’ve already put AMD’s newest Ryzen Threadripper chips through our Linux test gauntlet, so it’s now time to shift our focus back to Windows, where we have an even more daunting number of tests to cover. It won’t take long to notice a trend: AMD’s 24-core 3960X and 32-core 3970X are ridiculously fast processors.



Introduction

We might have mentioned it already, but this month has been intense for benchmarking. We kicked it off with a look at Intel’s Core i9-9900KS, and that already feels like forever ago now. After that, we rolled out two updated test suites at once, kicking them off with our look at the Ryzen 9 3950X in both Windows and Linux.

Intel then launched its new Core X-series, and we rolled out more Windows performance for that, and then hopped in Linux to greet the new Threadrippers. Oh – let’s not forget that we also took a look at AMD’s new Radeon Pro W5700, and NVIDIA’s GeForce GTX 1660 SUPER. Takes deep breath.

AMD Ryzen Threadripper Third-gen Out Of Its Box

Fortunately, this review is going to be an easy one to write, largely because we’ve already tackled these new Threadrippers before. And, due to this being written ahead of travel in the AM, we’re going to get through this front page a bit quicker than normal.

Let’s at least tackle the basics. For this Threadripper launch, AMD is likewise launching a new socket, TRX40, which means that old Threadrippers will not work in these new boards, and the new Threadrippers won’t work in the old ones. We all know that AMD truly loves backwards-compatibility, so we forgive it for this shift. We just hope you’re happy to splurge, because opting into this platform is not cheap. Somehow, it’s actually refreshing to see a different face behind the desktop premiums for once.

Before we go too much further, here’s AMD’s current Ryzen lineup:

AMD’s Ryzen & Ryzen Threadripper Lineup
CoresClock (Turbo)L2+L3MemoryTDPPrice
Threadripper WX-series
3970X32 (64T)3.7 GHz (4.5)144MBQuad280W$1999
3960X24 (48T)3.8 GHz (4.5)140MBQuad280W$1399
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 9
R9 3950X16 (32T)3.5 GHz (4.7)72MBDual105W$749
R9 3900X12 (24T)3.8 GHz (4.6)70MBDual105W$499
Ryzen 7
R7 3800X8 (16T)3.9 GHz (4.5)36MBDual95W$399
R7 3700X8 (16T)3.6 GHz (4.4)36MBDual65W$329
Ryzen 5
R5 3600X6 (12T)3.8 GHz (4.4)35MBDual95W$249
R5 36006 (12T)3.6 GHz (4.2)35MBDual65W$199
Ryzen w/ Radeon Vega Graphics
R5 3400G4 (8T)3.7 GHz (4.2)0.5+4MBDual65W$149
R3 3200G4 (4T)3.6 GHz (4.0)0.5+4MBDual65W$99

AMD has decided to bump the price on its 24- and 32-core Threadrippers this generation, and as our performance results are going to show, it’s rather hard to complain about. After all, not everyone needs the biggest chips. The Ryzen 9 3950X in itself is a fantastic chip that is found on a more affordable platform. If you need as many cores as Threadripper can offer, you know it already.

The reason for the premium is clearly tied to the vast improvements made. In years’ past, we’ve regularly talked about single-digit percentage gains from one generation to the next, for either AMD or Intel. With Zen 2, performance got a jolt in the arm, and with all design quirks of the last-gen Threadripper gone, these new models are much easier to recommend to anyone who needs a top-performing CPU, regardless of their workload. There are of course some exceptions, though they are rarer than usual here.

AMD Ryzen Threadripper Third-gen CPU Holder

The new Threadripper packaging is fairly interesting, not quite as grandiose as the previous generations, and that’s probably a good thing. After opening the box, you can lift out its CPU holder, which can have its plastic cover lifted upwards off of it. A little lever can be pulled down and removed, at which point the CPU can be slid out.

A downside with this new packaging is that AMD did not include convenient CPU holders for storing the chips for later use. This is definitely a #BenchmarkerProblem, but we haven’t had any issue storing our CPUs up to this point. For the regular user, the packaging is juuuuust fine.

Inside the box, AMD includes its Threadripper Torx screwdriver (with torque limiter), a sticker, and a bracket for a CPU cooler, if it’s required (it was with our NZXT Kraken X62).

AMD Ryzen Threadripper Third-gen Package Contents

We received two motherboards for potential inclusion, and we chose ASUS’ ROG Zenith II Extreme to run with first, since in terms of dealing with quirks, ASUS has been kindest to us. When diving into testing, we didn’t pay any attention to the motherboards, simply due to time. We didn’t realize until writing this that the boards available are pretty pricey, with models most commonly starting at ~$500.

ASUS ROG Zenith II Extreme Motherboard

We can’t speak too much to the quality of this particular board yet, but as it retails for $849, we’d sure as heck hope it’d be built to stand the test of time. By far, our favorite feature so far is the new BIOS OLED screen built-in, which actually lets you see what hardware is being engaged, or what part of the boot process it’s at. This is a lot better than seeing a two-digit readout that Google can barely help with. It also doubles as a temperature readout during normal use:

ASUS Zenith II Extreme Temperature Readout

As mentioned above, we’re going to speed through the intro here more than we’d like, but we wanted to make sure this review got live before we wound up out-of-pocket for a week. There’s no shortage of benchmarks to be had, though gaming has been skipped over for now (not that it’s a particular focus of these Threadrippers to begin with, but we still wanted to explore).

On the next page, you can peruse our test systems and methodologies; or jump straight to the tests on page three.

Test Methodology & Systems

Benchmarking a CPU may sound like a simple enough task, but in order to deliver accurate, repeatable results, strict guidelines need to be adhered to. This 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 MEG X399 Creation), TRX40 (ASUS Zenith II Extreme), and AM4 (Aorus X570 MASTER), along with Intel’s LGA2011-v3 (ASUS’ ROG STRIX X299-E GAMING), and LGA1151 (ASUS’ ROG STRIX Z390-E GAMING).

On our Intel platforms that use ASUS motherboards, we disable the “MultiCore Enhancement” feature, which effectively operates the CPU beyond stock speeds. The same feature doesn’t exist on our AMD platforms. We validated our configuration with AMD, Intel, and ASUS before settling on it.

On the mitigation front, nothing is explicitly done outside of having the most up-to-date EFI and chipset driver installed on every motherboard. Systems are effectively default, and whichever security mitigations are applied will be automatic ones applied by the motherboard firmware or driver vendor. All platforms are run with 32GB DDR4-3200 (14-14-14) memory configurations.

Note that AMD sent a 64GB kit of memory along with our Threadripper sample, and using those does improve performance in places. For the sake of keeping things apples-to-apples, though, we’ve stuck to the same DDR4-3200 kit as the other test machines. We plan on doing memory scaling testing in the near-future.

Here’s the full breakdown of the test rigs:

Techgage’s CPU Testing Platforms

AMD AM4 Test Platform
ProcessorsAMD Ryzen 9 3950X (3.5GHz, 16C/32T)
AMD Ryzen 9 3900X (3.8GHz, 12C/24T)
AMD Ryzen 7 3700X (3.6GHz, 8C/16T)
AMD Ryzen 5 3600X (3.8GHz, 6C/12T)
MotherboardAorus X570 MASTER
CPUs tested with BIOS F10c (November 8, 2019)
MemoryG.SKILL Flare X (F4-3200C14-8GFX) 8GB x 4
Operates at DDR4-3200 14-14-14 (1.35V)
GraphicsNVIDIA RTX 2080 Ti (12GB; GeForce 441.12)
StorageWD Blue 3D NAND 1TB (SATA 6Gbps)
Power SupplyEVGA Bronze 600B1 (600W)
ChassisFractal Design Define C
CoolingCorsair Hydro H115i PRO RGB (280mm)
Et ceteraWindows 10 Pro (1903, Build 18362)

AMD TR4  / TRX40 Test Platform
ProcessorAMD Ryzen Threadripper 3970X (3.7GHz, 24C/48T)
AMD Ryzen Threadripper 3960X (3.8GHz, 32C/64T)
AMD Ryzen Threadripper 2990WX (3.0GHz, 32C/64T)
AMD Ryzen Threadripper 2950X (3.5 GHz, 16C/32T)
MotherboardTRX40: ASUS Zenith II Extreme
^ CPUs tested with BIOS 0601 (November 4, 2019)
TR4: MSI MEG X399 Creation
^ CPUs tested with BIOS 7B92v13 (April 11, 2019)
MemoryG.SKILL Flare X (F4-3200C14-8GFX) 8GB x 4
Operates at DDR4-3200 14-14-14 (1.35V)
GraphicsNVIDIA RTX 2080 Ti (12GB; GeForce 441.12)
StorageWD Blue 3D NAND 1TB (SATA 6Gbps)
Power SupplyCooler Master Silent Pro Hybrid (1300W)
ChassisCooler Master MasterCase H500P Mesh
CoolingTRX40: NZXT Kraken X62 AIO (280mm)
TR4: Enermax LIQTECH TR4 (240mm)
Et ceteraWindows 10 Pro (1903, Build 18362)

Intel LGA1151 Test Platform
ProcessorsIntel Core i9-9900KS (4.0GHz, 8C/16T)
Intel Core i9-9900K (3.60GHz, 8C/16T)
Intel Core i7-8700K (3.70GHz, 8C/16T) (Only Some Tests)
MotherboardASUS ROG STRIX Z390-E GAMING
CPUs tested with BIOS 1302 (September 20, 2019)
MemoryG.SKILL Flare X (F4-3200C14-8GFX) 8GB x 4
Operates at DDR4-3200 14-14-14 (1.35V)
GraphicsNVIDIA RTX 2080 Ti (12GB; GeForce 441.12)
StorageWD Blue 3D NAND 1TB (SATA 6Gbps)
Power SupplyCorsair RM650X (1200W)
ChassisNZXT S340 Elite Mid-tower
CoolingCorsair Hydro H100i V2 AIO Liquid Cooler (240mm)
Et ceteraWindows 10 Pro (1903, Build 18362)

Intel LGA2011-3 Test Platform
ProcessorsIntel Core i9-10980XE (3.0GHz, 18C/36T)
Intel Core i9-9980XE (3.0GHz, 18C/36T)
MotherboardASUS ROG STRIX X299-E GAMING
CPU tested with BIOS 2002 (October 4, 2019)
MemoryG.SKILL Flare X (F4-3200C14-8GFX) 8GB x 4
Operates at DDR4-3200 14-14-14 (1.35V)
GraphicsNVIDIA RTX 2080 Ti (12GB; GeForce 441.12)
StorageWD Blue 3D NAND 1TB (SATA 6Gbps)
Power SupplyCorsair Professional Series Gold AX1200 (1200W)
ChassisCorsair Carbide 600C
CoolingNZXT Kraken X62 AIO (280mm)
Et ceteraWindows 10 Pro (1903, Build 18362)

Testing Considerations

For our testing, we use Windows 10 build 18362 (1903) with full updates. Version 1909 was released in the middle of our benchmarking, so we’ve stuck to 1903 here. For the sake of sanity checking, we cloned one of our test platform SSDs to upgrade to 1909, and couldn’t find any change to performance across ten or so benchmarks. In the grand scheme of Windows releases, 1909 is minor, but we’ll still upgrade next time we need to start from scratch.

Here are some basic guidelines we follow:


Encoding Tests

Intel Core i7-6700K (CPU-Z & GPU-Z) Intel Core i7-6700K (CPU-Z & GPU-Z)
Encoding: Adobe Lightroom
Intel Core i7-6700K (CPU-Z & GPU-Z)
Intel Core i7-6700K (CPU-Z & GPU-Z) Intel Core i7-6700K (CPU-Z & GPU-Z)
Encoding: Adobe Premiere Pro
Intel Core i7-6700K (CPU-Z & GPU-Z)
Intel Core i7-6700K (CPU-Z & GPU-Z) Intel Core i7-6700K (CPU-Z & GPU-Z)
Encoding: Blackmagic RAW Speed Test
Intel Core i7-6700K (CPU-Z & GPU-Z)
Intel Core i7-6700K (CPU-Z & GPU-Z) Intel Core i7-6700K (CPU-Z & GPU-Z)
Photogrammetry: Agisoft Metashape
Intel Core i7-6700K (CPU-Z & GPU-Z)
Intel Core i7-6700K (CPU-Z & GPU-Z) Intel Core i7-6700K (CPU-Z & GPU-Z)
Music Encoding: LameXP
Intel Core i7-6700K (CPU-Z & GPU-Z)

Adobe Lightroom Classic
Adobe Premiere Pro
Agisoft Metashape
Blackmagic RAW Speed Test
HandBrake
LameXP

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


Rendering Tests

Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
Rendering: Autodesk Arnold in Maya
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: Chaos Czech Corona Renderer in 3ds Max
Sony Ps4 Pro Angled View
dBpoweramp - Convert FLAC to MP3 dBpoweramp - Convert FLAC to MP3
Rendering: Chaos Group V-Ray Next in 3ds Max
dBpoweramp - Convert FLAC to MP3
Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
Rendering: MAXON Cinebench
Sony Ps4 Pro Angled View
Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
Rendering: MAXON Cinema 4D
Sony Ps4 Pro Angled View
Sony Ps4 Pro Angled View Sony Ps4 Pro Angled View
Rendering: POV-Ray
Sony Ps4 Pro Angled View
Intel Core i7-6700K (CPU-Z & GPU-Z) Intel Core i7-6700K (CPU-Z & GPU-Z)
Rendering: V-Ray Next Benchmark
Intel Core i7-6700K (CPU-Z & GPU-Z)


Arnold (Maya 2019) (Also relevant to: 3ds Max, C4D, Houdini, Katana, Softimage)
Blender
Cinebench
Cinema 4D
Corona (3ds Max 2019) (Also relevant to: C4D)
KeyShot (Also relevant to: 3ds Max, Maya, Creo, SketchUp, SolidWorks, NX, Rhino)
POV-Ray
V-Ray Next (3ds Max 2019) (Also relevant to: C4D, Houdini, Maya, Rhino, SketchUp)
V-Ray Benchmark
SiSoftware Sandra 2020

(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 Pro & Agisoft Metashape

We’re going to kick off this performance look with a handful of encode 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 we’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, for most of its life, Adobe’s Lightroom didn’t use more than a few cores and threads. Today, the application can use most of whatever CPU you can hand it.

The performance look on this page is going to tackle Adobe’s ever-popular Premiere Pro, as well as MAGIX’s Vegas Pro. That duo takes care of video encoding for this page, while Agisoft’s Metashape will help with a photogrammetry scenario.

Adobe Premiere Pro CC – CPU

Adobe Premiere Pro 2020
Adobe Premiere Pro 2020 - 1080p YouTube CPU Encode (AVC) Performance (AMD Ryzen Threadripper 3970X & 3960X)
Adobe Premiere Pro 2020 - 4K YouTube CPU Encode (AVC) Performance (AMD Ryzen Threadripper 3970X & 3960X)

From the get-go, AMD’s new Threadrippers impress. There isn’t too much to say that isn’t obvious when the products we’re looking at are glued to the top, but it’s important to at least pay attention to the generational gains in performance between the new and old 32-core Threadrippers. The last-gen 2990WX (and siblings) had a design that didn’t jive with certain different workloads, but especially encoding.

This is the first bit of proof we can see that the new Threadrippers really do fix the problem with NUMA modes that plagued the last generation. This is the kind of performance scaling we hoped for, but weren’t sure we’d get. How about some codec tests where the GPU makes an entrance?

Adobe Premiere Pro CC – Codec Comparisons

Adobe Premiere Pro 2020 - 4K60 AVC to 1080p HEVC Encode Performance (AMD Ryzen Threadripper 3970X & 3960X)
Adobe Premiere Pro 2020 - 4K60 AVC to 1080p HEVC (CUDA) Encode Performance (AMD Ryzen Threadripper 3970X & 3960X)

We’re going to see some variations with the results below, so let’s start here with the AVC to HEVC tests above. In both of these, the scaling from our initial project tests carries over here, with both of the new Threadrippers keeping pegged to the top. That continues once the GPU is involved, resulting in dramatically improved overall encode times, regardless of the CPU involved.

Things shake up a wee bit with RED:

Adobe Premiere Pro 2020 - 8K24 RED to 1080p HEVC Encode Performance (AMD Ryzen Threadripper 3970X & 3960X)
Adobe Premiere Pro 2020 - 8K24 RED to 1080p HEVC (CUDA) Encode Performance (AMD Ryzen Threadripper 3970X & 3960X)

In this RED transcode, the CPU only encode mode bodes (sorry) well for the new 24- and 32-core. With the GPU added in, scaling changes up a fair bit, although it’s hard for us to get a read on what makes for the most perfect blend of CPU processing + GPU processing. The 16-core 3950X soared to the top here, taking seconds off of the previous encode time. It could be that RED prefers high clocks and a bunch of cores rather than more modest clocks with many more cores. Either way, this is what we go. How does ProRes change things up?

Adobe Premiere Pro 2020 - 8K24 ProRes 422 to 1080p HEVC Encode Performance (AMD Ryzen Threadripper 3970X & 3960X)
Adobe Premiere Pro 2020 - 8K24 ProRes 422 to 1080p HEVC (CUDA) Encode Performance (AMD Ryzen Threadripper 3970X & 3960X)

With these ProRes results, we can see that the 3950X domination when joined by a GPU carries over to this ProRes test. Interestingly, the new Threadrippers fall a fair bit back of Intel here (with the GPU involved), something we hope to see improved in the future, especially as software developers begin paying more attention to not only AMD, but properly threading for so many cores.

MAGIX Vegas

MAGIX Vegas Pro 17 - Median FX Test
MAGIX Vegas Pro 17 - Median FX CPU Encode Performance - (AMD Ryzen Threadripper 3970X & 3960X)
MAGIX Vegas Pro 17 - Median FX NVENC Encode Performance - (AMD Ryzen Threadripper 3970X & 3960X)

In Premiere Pro, we saw that the new Threadrippers could fall behind some other competition once the GPU is introduced, and (perhaps not so) surprisingly, we’re seeing similar things here. The 3950X yet again places on top of the GPU test, though it’s not as though it’s running that far ahead of the Threadrippers behind it. Since we do see good scaling between 3900X and 3950X, it could be that this is another case of where optimization would improve scaling – or our test project just isn’t demanding enough (but we think it is, as the Median FX filter is seriously computationally intensive).

What we haven’t tested though, is the impact of lower-end GPUs with encoding tasks, since not everyone will have an RTX 2080 Ti to throw into a workstation. Under such workloads, the core counts of the CPU may factor in more. However, at this point we haven’t had the time to fully explore this, and may be something we get a chance to look at later.

Agisoft Metashape

Agisoft Metashape
Agisoft Metashape Photogrammetry Performance - Build Dense Cloud (AMD Ryzen Threadripper 3970X & 3960X)
Agisoft Metashape Photogrammetry Performance - Build Depth Maps (AMD Ryzen Threadripper 3970X & 3960X)
Agisoft Metashape Photogrammetry Performance - Build Mesh (AMD Ryzen Threadripper 3970X & 3960X)
Agisoft Metashape Photogrammetry Performance - Build Texture (AMD Ryzen Threadripper 3970X & 3960X)

Metashape is a fun application to test with, but its performance results are rather unpredictable at times. And, because a singular process actually encompasses multiple different workloads, the CPU and GPU will be used differently throughout. That’s why a breakdown like this is important, to see which tasks each CPU is good at.

Ultimately, for everything but the Build Texture phase, Metashape really seems to love processors that sport big clock speeds, and a healthy amount of cores. Adding more cores to your Metashape problem isn’t the solution, but for the sake of erring on the side of caution, your Metashape workload may scale a bit differently than ours.

We also have an aerial project we’ve been contemplating adding to our suite, but the amount of time it takes to run is rather prohibitive at the moment. Our test suite already takes 10 straight hours to get through on the fastest chip we have – so we need to begin optimizing before adding more! That said, we have been working on adding RealityCapture, as well, but still have to perfect our test scripts.

Encoding: Adobe Lightroom, BRAW Speed Test, HandBrake & LameXP

On this page, we’re going to be tackling a few additional encoding-type projects. Since the beginning of its life, we’ve benchmarked with Adobe’s Lightroom, but dropped it for about a year or two because it wouldn’t reliably scale. Over time, things changed, and now the application seems pretty efficient on multi-core CPUs.

In addition to Lightroom, we’ve also added Blackmagic RAW Speed Test, which acts as a simple way to see how a CPU can handle playback of BRAW footage at different compression levels. In time, we’ll be adding a much fuller Resolve test to the suite, but this BRAW test fills in for now. Finally, we’re also testing with LameXP, an open-source music encoder that can take advantage of many-core CPUs, as well as the super-popular HandBrake transcoder.

Adobe Lightroom Classic

Adobe Lightroom Classic
Adobe Lightroom Classic - RAW to JPEG Export Performance (AMD Ryzen Threadripper 3970X & 3960X)

Months ago, we discovered just how much Adobe Lightroom loves Zen 2, because we saw the 12-core 3900X beat out every other chip in our collection – and by a seriously healthy margin. When the 3950X released, we encountered an oddity that had that chip fall behind the 3900X, something we can’t quite explain. Given that problem, we had worried that these bigger Threadrippers would suffer the same fate, but… not so.

That said, there are still some oddities to speak of. First, the 3970X didn’t manage to take the cake here, across multiple retestings to validate data. The result doesn’t surprise us, given what we saw with the 3950X, but it does highlight that it seems like further optimization in LR could be put to good use.

The main takeaway is that the 3960X becomes the fastest CPU we’ve ever tested in Lightroom. We should also mention that during our testing, we found that Lightroom can prove to be very sensitive to memory latencies and bandwidth. From testing we did eleven years ago, we had a suspicion of that anyway, but at least based on initial testing, we can say that density definitely matters, not just frequency. We don’t want to talk about it too much until we can devote more time to actually testing various configurations and generating some numbers. This is something to look forward to in the future.

What is worth pointing out is the huge difference in performance from one generation to the next with 2990WX to the 3970X, nearly 3.5x the performance for the same 32-cores. Those NUMA node issues of the 2nd gen chips are clearly highlighted here.

Blackmagic RAW Speed Test

Blackmagic RAW Speed Test
Blackmagic RAW Speed Test (AMD Ryzen Threadripper 3970X & 3960X)

The original many-core Threadripper chips (2970WX, 2990WX) didn’t scale too well with Blackmagic’s RAW, likely due to the same memory design that impacted many other video workloads. The new Threadripper has no problem rising to the absolute top, going toe-to-toe with each other. Since both perform so similarly, it could be that we’re running into a bottleneck of some sort – something future testing will include a look at.

HandBrake

HandBrake AVC Encode Performance - (AMD Ryzen Threadripper 3970X & 3960X)
HandBrake HEVC Encode Performance - (AMD Ryzen Threadripper 3970X & 3960X)

HandBrake is an excellent transcode utility for many reasons, but a great one is that it makes for easy testing, and it’s also representative of more users than most of the other tools tested here. At the second-gen Threadripper launch, we encountered an issue with scaling in HandBrake, something that was fixed with the release of 1.2.2.

Fast-forward to today, and we are really excited to see that third-gen Threadripper is serious about scaling. These results are fantastic. Even the $749 3950X looks like a great value from where it sits.

LameXP

LameXP
LameXP - FLAC to MP3 Encode Performance - (AMD Ryzen Threadripper 3970X & 3960X)

Given the type of workload LameXP is, we weren’t sure if the new Threadripper could really shine, but behold: both chips still scale, but hit a plateau long before the others. Because of how many tracks are being encoded at once, we’re going to experiment with faster storage with our next round of testing and see if it impacts the scaling at all. In the past, I/O was never a bottleneck, but it could become one with such powerful processors at the top.

Rendering: Arnold, Blender, KeyShot, V-Ray Next

There are few things we find quite as satisfying as rendering: seeing a bunch of assets thrown into a viewport that turn into a beautiful scene. Rendering also happens to be one of the best possible examples of what can take advantage of as much PC 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 way. 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 to a different design suite or renderer. An example: V-Ray supports more than just 3ds Max; it also supports Cinema 4D, Maya, Rhino, SketchUp, and Houdini.

Autodesk Arnold

Autodesk Arnold in Maya 2019
Autodesk Arnold CPU Render Performance - Jaguar E-Type Scene 1 (AMD Ryzen Threadripper 3970X & 3960X)
Autodesk Arnold CPU Render Performance - Sophie Scene (AMD Ryzen Threadripper 3970X & 3960X)

We’ve mentioned our uncertainties with what to expect from the new Threadrippers a few times, and it’s only because after spending so much time testing the original and last-gen Threadrippers, we were cautiously optimistic that we’d actually see the kinds of gains AMD was promising. With rendering, we truly didn’t expect too much, since rendering was the exact workload a chip like the 2990WX was best-suited for. But yet, we do in fact see massive generation-over-generation performance gains.

Meanwhile, we’re seeing immediate proof that if rendering is important to you, you don’t want to skimp on your processor. The traditionally fast 8-cores listed here prove slow in this heavy workload.

Blender – CPU

Blender 2.8
Blender 2.80 Cycles CPU Render Performance - BMW (AMD Ryzen Threadripper 3970X & 3960X)
Blender 2.80 Cycles CPU Render Performance - Classroom (AMD Ryzen Threadripper 3970X & 3960X)

If you hate to see similar results, then the rest of this article might appear boring, because by all appearances, AMD has come to kick ass and take names. Both Threadrippers again soar to the top in Blender, but we know how to shake things up: by adding a GPU.

Blender – CPU + GPU

Blender 2.80 Cycles CPU+GPU Render Performance - BMW (AMD Ryzen Threadripper 3970X & 3960X)
Blender 2.80 Cycles CPU+GPU Render Performance - Classroom (AMD Ryzen Threadripper 3970X & 3960X)

In some of our video encode tests, we saw that the 3970X and 3960X might fall a tad behind after adding a GPU, pointing to possible optimizations that could be done (from some or multiple sources). With Blender, no such detriment was witnessed, with the scaling left in tact. The astute among you may have noticed that the CPU+GPU tests are far quicker than the CPU only renders – so clearly, the GPU is exceptionally important here.

For more Blender performance, you can refer to our in-depth look at Blender 2.80 from a few months ago. It does not include the latest CPUs, but the results above can fill in, since it was all tested with the same version. We tested 2.81 on half of the CPUs listed here, and found no performance differences between the two sets of results.

KeyShot

KeyShot 9
Luxion KeyShot 9 - Character Render Performance (AMD Ryzen Threadripper 3970X & 3960X)
Luxion KeyShot 9 - Room Render Performance (AMD Ryzen Threadripper 3970X & 3960X)

With as much testing as we’ve done with KeyShot, it’s hard to believe that we only introduced it into our test suite last fall, with the launch of the second-gen Threadrippers. Ahead of that launch, it became obvious quick that we needed more tests which took advantage of such big processors, and we’re glad we did, since it’s an awesome tool to play with.

On that note, the last-gen Threadripper had a bit of a regression problem with one of our KeyShot projects, specifically an interior scene which used a lot of glass. This was a problem that ended up being ironed out by the time KeyShot 8 rolled around, and continues to be a non-existent problem in KeyShot 9, as even the 2990WX performs exceptionally well here. Yet, even that result can’t hold a candle to the new 32-core. Just look at those performance gains – from 645 to 427 seconds in a single generation. Is it any wonder AMD has decided to charge a bit of a premium on this year’s 32-core?

Chaos Group V-Ray Next – CPU

Chaos Group V-Ray in Autodesk 3ds Max 2019
Chaos Group V-Ray - Flowers CPU Render Performance (AMD Ryzen Threadripper 3970X & 3960X)
Chaos Group V-Ray - Teaset CPU Render Performance (AMD Ryzen Threadripper 3970X & 3960X)

V-Ray typically performs strongly on Intel CPUs, but when AMD has far higher core counts to boast, that Intel edge no longer matters. Both the 24- and 32-cores continue to perform so well in our workloads, but we’ve seen hiccups with GPUs before; will it happen again?

Chaos Group V-Ray Next – CPU + GPU

Chaos Group V-Ray - Flowers CPU+GPU Render Performance (AMD Ryzen Threadripper 3970X & 3960X)
Chaos Group V-Ray - Teaset CPU+GPU Render Performance (AMD Ryzen Threadripper 3970X & 3960X)

The answer is a definite “Nope”. V-Ray is one solution that happens to thread just fine in heterogeneous rendering, and with the new Threadrippers, some seriously attractive performance can be seen. Again, you don’t want to skimp on a GPU – it’s likely just as important as the CPU, but that of course assumes you are not locked to the CPU-only renderer.

V-Ray Next Benchmark

Chaos Group V-Ray Next Benchmark - CPU Render Score (AMD Ryzen Threadripper 3970X & 3960X)
Chaos Group V-Ray Next Benchmark - CPU+GPU Render Score (AMD Ryzen Threadripper 3970X & 3960X)

Scaling with V-Ray’s standalone benchmark seems to agree pretty well with our real testing. The new Threadrippers once again rule on top, especially so on the CPU only side. From 30K with the 9980XE to 43K with the 3970X is quite a jump.

Rendering: Cinebench, Cinema 4D, Corona, LuxMark, POV-Ray

We covered a handful of major renderers on the previous page, but we’re not done yet. On this page, we’re going to take a look at a few more, including some industry mainstays and newbies. That includes Corona Renderer, which we recently upgraded to version 5. We’re foregoing Adobe Dimension performance for this review, since we haven’t seen realistic scaling with the new version 3.0, and have not yet been able to investigate (typical Adobe new-release teething problems).

To give you an opportunity to test your own hardware against ours, we’re also including the ever-popular Cinebench standalone benchmark, which represents current R20 performance. This test, along with the latest version of POV-Ray, act as our only single-threaded angles in the article. For good measure, the performance on this page will be capped off the real Cinema 4D, to see how it agrees with CB.

Cinema 4D R21

Maxon Cinema 4D R21
Maxon Cinema 4D R21 - Candies Render Performance (AMD Ryzen Threadripper 3970X & 3960X)
Maxon Cinema 4D R21 - Interior Render Performance (AMD Ryzen Threadripper 3970X & 3960X)

We’re going to be overhauling our Cinema 4D tests in the near-future, both with better projects, and if all goes well, a viewport test. For now, we’re sticking to the same projects we’ve been using for the past couple of years. Now, if you think back to a “couple of years” ago, the best any enthusiast could buy was a 10-core. Today, we have 32 cores available, and with these results, we can see that Cinebench can take advantage of every single one of them.

Somehow, the 16- and 24-core Zen 2 chips scored the same in the interior render, even after retesting, and despite the scaling being just fine all-around in the (simpler) Candies render. At least we’re not seeing some actually detrimental loss in performance, as we have seen with the last-gen 24+ core Threadrippers.

Cinebench R20

Maxon Cinebench R20 - Multi-threaded Score (AMD Ryzen Threadripper 3970X & 3960X)
Maxon Cinebench R20 - Single-threaded Score (AMD Ryzen Threadripper 3970X & 3960X)

It’s hard to be surprised with the results here, given not just what we’ve seen of the previous Zen 2 chips, but also what we saw with the real-world Cinema 4D above. The project Maxon chose for this benchmark seems pretty neutral overall, given how things scale – at least on the multi-thread side of things. For peak single-thread, AMD owns this benchmark, with the 9900KS not trailing far behind.

Corona Renderer

Corona 5 in 3ds Max
Chaos Czech Corona Renderer 5 Performance - Livingroom Scene (AMD Ryzen Threadripper 3970X & 3960X)
Chaos Czech Corona Renderer 5 Performance - Sales Gallery Scene (AMD Ryzen Threadripper 3970X & 3960X)

We’ve gone over many renderers by this point already, and like the rest, Corona shows that the new Threadrippers can seriously tear up a rendering workload. It’s worth comparing the new 32-core to the last-gen 32-core once again. When is the last time you saw these kinds of generational gains on a desktop processor?

LuxMark

LuxMark v4
LuxMark Food (C++) Render Performance (AMD Ryzen Threadripper 3970X & 3960X)
LuxMark Hall Bench (C++) Render Performance (AMD Ryzen Threadripper 3970X & 3960X)

With the C++ renderer in LuxMark using Intel’s Embree ray tracing kernels, it’d be easy to assume that Intel simply has an edge by default. And well, the results here imply that too. In LuxMark, Intel CPUs definitely enjoy a speed-up, allowing the 18-core chips to fly past every AMD chip in the Hall Bench render. We also happen to test Embree in our Linux tests, and there, AMD’s new Threadripper led the pack. Just when you thought you could predict a workload!

POV-Ray

POV-Ray
POV-Ray 3.8 Multi-threaded Score (AMD Ryzen Threadripper 3970X & 3960X)
POV-Ray 3.8 Single-threaded Score (AMD Ryzen Threadripper 3970X & 3960X)

We wrap up our rendering tests with one of the classics, POV-Ray. Cinebench really does seem to get the lion’s share of the standalone benchmark attention, but we’ve been using POV-Ray for about just as long. We featured both in our Core 2 Extreme QX6850 review in 2007. Twelve years ago, the best desktop processor had four cores and cost $1,000. How far we’ve come.

This is another test that Intel has an edge on, especially with the single-threaded performance. Both of the 9900K chips keep well ahead of the others before it. For multi-thread, the winners have become rather predictable at this point.

System: SiSoftware Sandra

While this article has no lack of synthetic benchmarks, SiSoftware’s Sandra makes it very easy to get reliable performance information on key metrics, such as arithmetic, multimedia, cryptography, and memory. Sandra is designed in such a way that it takes the best advantage of any architecture it’s given, so each CPU always has its best chance to shine.

That means a couple of things. This is definitely the “best” possible performance outlook for any chip, and doesn’t necessary correlate with real-world performance in other tests. It’s best used as a gauge of what’s possible, and to see where one architecture obviously differs from another.

SiSoftware Sandra 2020

Multimedia

SiSoftware Sandra 2020 - Multi-media Performance (AMD Ryzen Threadripper 3970X & 3960X)

Core count matters a lot with all of the tests on this page, but the result with multimedia is pretty impressive for AMD’s new Threadrippers. This test can take advantage of AVX-512 acceleration, which is why the 18-cores keep so far ahead of the 16-core 3950X. With all of their might, the new 24- and 32-core chips negate their lack of AVX-512 pretty easily. Now just imagine if Intel could churn out a 32-core Core X real quick, then things would get interesting again.

Arithmetic

SiSoftware Sandra 2020 - Arithmetic Performance (AMD Ryzen Threadripper 3970X & 3960X)

Oh. My. GOPS. The 3970X manages to become the first CPU we’ve tested to hit the 1,000 GOPS (1 TOPS) milestone in the arithmetic test. This is one test that actually bodes really well for the last-gen 32-core chip, as well.

Cryptography

SiSoftware Sandra 2020 - Cryptography (High) Performance (AMD Ryzen Threadripper 3970X & 3960X)
SiSoftware Sandra 2020 - Cryptography (Higher) Performance (AMD Ryzen Threadripper 3970X & 3960X)

We see some shake-ups with the cryptography tests, with the first one being the fact that the new 24-core managed to beat out the 32-core, something we can’t actually recollect if we were able to sanity check the 24-core, but did the 32. In the higher security test, the 32-core redeemed itself, delivering more expected scaling. Meanwhile, AVX-512 comes to Intel’s rescue here, allowing it to charge past AMD.

Memory Bandwidth

SiSoftware Sandra 2020 - Memory Bandwidth (AMD Ryzen Threadripper 3970X & 3960X)

Memory bandwidth tests can vary a fair bit depending on how they were designed, and we plan to evaluate adding AIDA64’s to our suite in the future, as a bit of a second opinion. In Sandra’s take, which uses Stream as the base of the test, Intel exhibited some super-strong performance – a result not tied to AVX-512. AMD’s new Threadrippers placed just behind Intel’s 18-cores, with the 24-core somehow delivering a better result than the 32-core here.

Final Thoughts

This is one of those articles where words are not even needed, because the results really do speak for themselves. We often talk about how important it is to “know your workload”, but if you equip yourself with one of these new Threadrippers, you can be pretty much assured that it’d be hard to hit a workload that isn’t going to run great on it.

The last-gen Threadripper knocked our socks off at the time. The 24- and 32-core models were simply jaw-dropping at the time of their release, especially as Intel was (and still is) locked to 18 cores for its Core X line. The biggest caveat, though, was that those chips had a memory design which didn’t jive well with all workloads, so whether they were worth purchasing really required understanding your workload.

After seeing what was possible with the first Zen 2 chips, especially the 3950X, we remained reserved about what we’d actually see from these new Threadripper chips. Would the odd threading issues finally be gone? Based on our testing, we can definitely say that they seem to be. There are occasions where behavior remains unexpected, but we usually attribute those to the software itself, since the overwhelming number of tests scale fine.

AMD Ryzen Threadripper Third-gen Close-up

With rendering, the last-gen top-end Threadrippers were fantastic, but encoding tasks could choke the CPUs up, allowing technically slower CPUs to storm ahead. We were really satisfied then, to see that both of these Threadrippers scale great in Adobe Premiere Pro, MAGIX Vegas Pro, HandBrake, and even Blackmagic RAW.

We were further amused by the fact that the rendering tests exhibited even greater gains with Threadripper third-gen than we expected. With the last-gen being so good for the task, we just didn’t know there was so much more room to push ahead in a single generation leap.

Ahead of launch, some questioned why AMD raised the price of its new 24-core 3960X ($1,399 USD) and 32-core 3970X ($1,999 USD), and after seeing the test results, the reason seems obvious. This is a very different chip design that scales significantly better than the previous gen, with a lot more cache, redesigned memory architecture, and a new dedicated I/O chip. It’s not just with our Windows benchmarking that can attest to the performance of these new chips; we tackled Linux performance in-depth earlier this week, and found more of the same. Where AVX-512 is involved, Intel has some really strong performance, but if you have more cores with Threadripper, you are likely to leap ahead.

AMD Ryzen Threadripper Third-gen Packaging and Intel Core i9-10980XE Engineering Sample

There is something to be said about how expensive the new Threadripper platform is in comparison to Intel’s Core X-series, which recently saw its flagship 18-core model drop to the $1,000 price point. In the span of three generations, we’ve seen AMD go from being the underdog, with hopes of a truly deserving top-end chip seeming unlikely, to it striking back hard with Zen, to finally twisting the knife with the full Zen 2 rollout.

As amazing as Threadripper is, it is indeed expensive, with the 24-core priced at $1,399, and the 32-core priced at $1,999. The motherboards are also expensive, with $500 being an expected starting point. It wasn’t actually until we sat down to begin writing that we realized the motherboard we did our testing with (ASUS ROG Zenith II Extreme) costs $849. That’s as much as a high-end graphics card! But, a top-end platform doesn’t come cheap.

One thing worth highlighting, since we haven’t yet, is the fact that AMD’s TRX40 platform supports PCIe 4.0, which Intel doesn’t, even with its just-launched Core X-series refresh. The faster 4.0 bus will largely impact storage devices more than other add-in cards, but it would not be unexpected if the next wave of top-end GPUs might benefit from it.

So, you might have to take a deep breath as you hit that order button, but our results have proven that the new Threadripper is seriously fast, and unmatched in performance on the desktop. We also did a bunch of gaming testing, but due to a memory error with the original suite run of the 3970X (the correct settings were not applied like we thought – doh) , we effectively lost our chance to include it. We may explore gaming more later, but clearly, the newest Threadrippers are squarely targeted at creators who demand the most powerful desktops possible.

And to think a 64-core is coming in the new year. We hope we’ve saved enough superlatives for that review.

AMD Ryzen Threadripper 3970X & 3960X - Techgage Editor's Choice
AMD Ryzen Threadripper 3970X & 3960X

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