AMD’s Ryzen Threadripper 2950X CPU reviewed

AMD’s Ryzen Threadripper 2990WX may have a legitimate claim to being the world’s fastest desktop CPU, but it’s also $1800. Even for a high-end desktop part, that’s rather pricey for anybody who’s not making money with those CPU cycles, and those 32 cores and 64 threads don’t benefit every application. The 2990WX is outstanding when a program can yoke every one of its cores and threads, but our tests suggest most apps on the desktop hit the diminishing-returns point in the scaling curve around 16 cores and 32 threads.

Enter the Ryzen Threadripper 2950X. Compared to the Threadripper 2990WX’s wild core count and rather unusual NUMA arrangement, the 2950X is downright simple. It uses the same two-die multi-chip-module configuration that the Threadripper 1950X and Threadripper 1920X introduced a little over a year ago. It keeps the 1950X’s 16-core, 32-thread configuration, too.

Those two dies are joined using a pair of unidirectional Infinity Fabric links capable of 51.2 GB/s of unidirectional bandwidth for an aggregate of 102.4 GB/s bi-directional throughput, at least assuming they’re talking to DDR4-3200 RAM. As a consequence of the move to Zen+ Zeppelins underneath its heat spreader, the 2950X gains support for DDR4-2933 RAM out of the box.

As a beneficiary of the Zen+ microarchitectural improvements that first appeared on second-generation Ryzen chips in Socket AM4, the 2950X implements three improvements over first-generation Threadrippers. Precision Boost 2 lets the chip gracefully adjust its clock speeds as the number of loaded cores and threads climbs. The Threadripper 1950X, in contrast, maintained a concept of a four-loaded-core boost clock and an all-core boost clock, meaning applications that ended up somewhere in between weren’t necessarily taking full advantage of the chip’s power and thermal headroom. Precision Boost 2 might increase the 2950X’s performance in some workloads as a result. AMD corporate fellow Joe Macri described one of the company’s goals for second-generation Threadrippers as “eliminating performance cliffs,” and Precision Boost 2 does just that.

Second, Extended Frequency Range 2 (XFR 2) allows the 2950X to take advantage of ambient conditions and beefy cooling hardware to deliver better sustained performance under multi-threaded workloads. Unlike the first generation of XFR, which applied a fixed offset to both single-core and all-core clock speeds when conditions allowed, XFR 2 only affects multithreaded speeds.

Finally, the Threadripper 2950X is built on GlobalFoundries’ 12LP process. 12LP allowed AMD to use better-performing transistors in critical parts of the Ryzen die, resulting in better cache and memory latencies.

AMD bins the top 5% of Ryzen dies for use in Threadrippers, and the 2950X puts those select slices of silicon to use by posting a 4.4-GHz single-core peak clock speed, the highest of any Ryzen CPU so far by a nose. Bit by bit, Ryzen chips are eliminating the peak clock speed deficits that have been part of the reason Intel CPUs remain superior for desktop responsiveness and lightly-threaded performance. The 2950X isn’t going to boost to the same peaks as a Coffee Lake Core i7-8700K, to be certain, but 4.4 GHz is not far off the Turbo Boost Max 3.0 speed of 4.5 GHz that the similarly-priced Core i9-7900X puts up.

  Cores/

threads

Base

clock (GHz)

Peak boost

clock (GHz)

L2

cache (MB)

L3

cache (MB)

TDP Suggested

price

Threadripper 2990WX 32/64 3.0 4.2 16 64 250 W $1799
Threadripper 2970WX 24/48 12 64 $1299
Threadripper 2950X 16/32 3.5 4.4 8 32 180 W $899
Threadripper 1950X 16/32 3.4 4.2 8 32 $999
Threadripper 2920X 12/24 3.5 4.3 6 32 $649
Threadripper 1920X 12/24 3.5 4.2 6 32 $799
Threadripper 1900X 8/16 3.8 4.2 4 16 $549

Perhaps most importantly, those benefits don’t make for a more expensive chip. The Threadripper 1950X already made waves in the high-end desktop market by bringing its particular complement of cores and threads to a lower price point than ever before, and the Threadripper 2950X lops another $100 off that chip’s $999 suggested price. Threadripper 1950X prices have fallen far below that initial $999, to be sure, but once old stock of those parts filters out of retailer warehouses, the 2950X will be in a good position to compete against the Core i9-7900X. The 2950X isn’t for sale yet, but it’ll hit retailers August 31.

 

Our testing methods

As always, we did our best to deliver clean benchmarking numbers. We ran each benchmark at least three times and took the median of those results. Our test systems were configured as follows:

 

Processor Intel Core i7-8086K
CPU cooler Corsair H110i 280-mm closed-loop liquid cooler
Motherboard Gigabyte Z370 Aorus Gaming 7
Chipset Intel Z370
Memory size 16 GB
Memory type G.Skill Flare X 16 GB (2x 8 GB) DDR4 SDRAM
Memory speed 3200 MT/s (actual)
Memory timings 14-14-14-34 2T
System drive Samsung 960 Pro 512 GB NVMe SSD

 

Processor AMD Ryzen Threadripper 2990WX AMD Ryzen Threadripper 2950X
CPU cooler Enermax Liqtech TR4 240-mm closed-loop liquid cooler
Motherboard Gigabyte X399 Aorus Xtreme
Chipset AMD X399
Memory size 32 GB
Memory type G.Skill Flare X 32 GB (4x 8 GB) DDR4 SDRAM
Memory speed 3200 MT/s (actual)
Memory timings 14-14-14-34 1T
System drive Samsung 970 EVO 500 GB NVMe SSD

 

Processor AMD Ryzen Threadripper 1950X AMD Ryzen Threadripper 1920X
CPU cooler AMD Wraith Ripper
Motherboard Gigabyte X399 Aorus Gaming 7
Chipset AMD X399
Memory size 32 GB
Memory type G.Skill Flare X 32 GB (4x 8 GB) DDR4 SDRAM
Memory speed 3200 MT/s (actual)
Memory timings 14-14-14-34 1T
System drive Samsung 960 EVO 500 GB NVMe SSD

 

Processor Core i9-7980XE Core i9-7960X Core i9-7900X Core i7-7820X
CPU cooler Corsair H150i Pro 360-mm closed-loop liquid cooler
Motherboard Gigabyte X299 Designare EX
Chipset Intel X299
Memory size 32 GB
Memory type G.Skill Flare X 32 GB (4x 8 GB) DDR4 SDRAM
Memory speed 3200 MT/s (actual)
Memory timings 14-14-14-34 1T
System drive Intel 750 Series 400 GB NVMe SSD

Our test systems shared the following components:

Graphics card Nvidia GeForce GTX 1080 Ti Founders Edition
Graphics driver GeForce 398.82
Power supply Thermaltake Grand Gold 1200 W (Intel X299)

Seasonic Prime Platinum 1000 W (AMD Threadripper 2950X/2990WX)

Corsair RMx 850 W (AMD Threadripper 1950X/1920X)

Seasonic SS660-XP2 660 W (Core i7-8086K)

Some other notes on our testing methods:

  • All test systems were updated with the latest firmware, graphics drivers, and Windows updates before we began collecting data, including patches for the Spectre and Meltdown vulnerabilities where applicable. As a result, test data from this review should not be compared with results collected in past TR reviews. Similarly, all applications used in the course of data collection were the most current versions available as of press time and cannot be used to cross-compare with older data.
  • Our test systems were all configured using the Windows Balanced power plan, including AMD systems that previously would have used the Ryzen Balanced plan. AMD’s suggested configuration for its CPUs no longer includes the Ryzen Balanced power plan as of Windows’ Fall Creators Update, also known as “RS3” or Redstone 3.
  • Unless otherwise noted, all productivity tests were conducted with a display resolution of 2560×1440 at 60 Hz. Gaming tests were conducted at 1920×1080 and 144 Hz.

Our testing methods are generally publicly available and reproducible. If you have any questions regarding our testing methods, feel free to leave a comment on this article or join us in the forums to discuss them.

 

Memory subsystem performance

The AIDA64 utility includes some basic tests of memory bandwidth and latency that will let us peer into the differences in behavior among the memory subsystems of the processors on the bench today, if there are any.

Aside from our memory read test, where Skylake-X parts rule the roost, the 2950X pulls out wins in our memory write and memory copy results. Perhaps we’re seeing the net effect of Zen+’s improvements throughout the memory hierarchy here, as well as the 2950X’s improved peak per-core performance.

While the 2990WX appears to benefit from Windows’ awareness of its NUMA-ness in AIDA64’s memory latency test, concealing the characteristics of its complex memory topology, the 2950X still presents its memory as a unified pool to the operating system. That might give us a more useful picture of Threadrippers’ average latency.

Some quick synthetic math tests

AIDA64 also includes some useful micro-benchmarks that we can use to flush out broad differences among CPUs on our bench. The PhotoWorxx test uses AVX2 instructions on all of these chips. The CPU Hash integer benchmark uses AVX and Ryzen CPUs’ Intel SHA Extensions support, while the single-precision FPU Julia and double-precision Mandel tests use AVX2 with FMA.

The Threadripper 2950X turns in a small but decent improvement over its 16-core, 32-thread predecessor in this test.

The 2950X turns in another modest improvement over the 1950X in the SHA Extensions-accelerated CPU Hash test.

The synthetic FPU Mandel and FPU Julia tests also demonstrate small improvements in performance from Zen+ for the 2950X, thread-for-thread.

The FP32 and FP64 Ray-Trace tests once again show only minor improvements from 1950X to 2950X. Let’s see if any of our real-world benchmarks can coax greater evidence of generation-to-generation improvements for the 2950X.

 

Javascript

The usefulness of Javascript microbenchmarks for comparing browser performance may be on the wane, but these tests still allow us to tease out some single-threaded performance differences among CPUs. As part of our transition to using the Mechanical TuRk to benchmark our chips, however, we’ve had to switch to Google’s Chrome browser so that we can automate these tests. Chrome does perform differently on these benchmarks than Microsoft Edge, our previous browser of choice, so it’s vitally important not to cross-compare these results with older TR reviews.

The 2950X performs remarkably well in these single-threaded tests, even taking home a win over the 5-GHz-endowed i7-8086K in Kraken. More generally, though, the 2950X matches or beats out the i9-7900X in these tests, and that’s an excellent place to be for a high-end desktop CPU.

WebXPRT 3

The WebXPRT 3 benchmark is meant to simulate some realistic workloads one might encounter in web browsing. It’s here primarily as a counterweight to the more synthetic microbenchmarking tools above.

WebXPRT isn’t entirely single-threaded—it uses web workers to perform asynchronous execution of Javascript in some of its tests. Perhaps that’s part of why this test lets the 2950X run away from the high-end desktop pack, trailing only the Core i7-8086K.

All told, our single-threaded benchmarks show that the Threadripper 2950X is more than able to use its 4.4-GHz peak single core speed to deliver plenty of responsiveness in light day-to-day workloads. Next, let’s take a look at whether the chip can deliver better sustained performance than first-generation Threadrippers.

 

Compiling code with GCC

Our resident code monkey, Bruno Ferreira, helped us put together this code-compiling test. Qtbench records the time needed to compile the Qt SDK using the GCC compiler. The number of jobs dispatched by the Qtbench script is configurable, and we set the number of threads to match the hardware thread count for each CPU.

The Threadripper 2950X bolsters the fact that Qtbench performance is bound by Amdahl’s Law. The combination of Precision Boost 2 and a 4.4-GHz peak clock speed allow the 2950X to shave nearly 20 seconds off the 1950X’s compile time.

File compression with 7-Zip

The free and open-source 7-Zip archiving utility has a built-in benchmark that occupies every core and thread of the host system.

No worries about inconsistent performance here. The 2950X trails only the Core i9-7980XE and i9-7960X in 7-Zip compression, and it’s only overshadowed by the jaw-dropping Threadripper 2990WX in the decompression portion of the test. As usual, though, the 2950X’s gains over the 1950X aren’t massive.

Disk encryption with Veracrypt

In the accelerated AES portion of the Veracrypt benchmark, the many-core chips largely cluster up once you hit 16 cores. The 2950X delivers number-crunching performance similar to the i9-7980XE and i9-7960X in the pure-grit Twofish portion of the benchmark, as well.

 

Cinebench

The evergreen Cinebench benchmark is powered by Maxon’s Cinema 4D rendering engine. It’s multithreaded and comes with a 64-bit executable. The test runs with a single thread and then with as many threads as possible.

If our Javascript tests seemed to favor the 2950X’s 4.4-GHz peak clock speed, Cinebench’s single-threaded test is far friendlier to the Intel parts. Cinebench’s single-threaded portion is only of academic interest, though. Its multithreaded mode is where the real action happens.

Core for core and thread for thread, the 2950X and i9-7960X are neck and neck. The 2950X once again fails to put much light between itself and its predecessor in fully-scalable workloads like this one.

Blender

Blender is a widely-used, open-source 3D modeling and rendering application. The app can take advantage of AVX2 instructions on compatible CPUs. We chose the “bmw27” test file from Blender’s selection of benchmark scenes to put our CPUs through their paces.

Blender lets the 2950X win the 16-core fight by a nose, but it only shaves five seconds off the 1950X’s time.

Corona

Corona, as its developers put it, is a “high-performance (un)biased photorealistic renderer, available for Autodesk 3ds Max and as a standalone CLI application, and in development for Maxon Cinema 4D.”

The company has made a standalone benchmark with its rendering engine inside, so it’s a no-brainer to give it a spin on these CPUs.

Corona only gets a second faster on the 2950X.

Indigo

Indigo Bench is a standalone application based on the Indigo rendering engine, which creates photo-realistic images using what its developers call “unbiased rendering technologies.”

Indigo doesn’t give the 2950X any additional room to stretch its legs versus its predecessor. While the 2950X is the top-performing Threadripper in this test, the i9-7960X chalks up a rare victory in both phases of the benchmark.

 

Handbrake

Handbrake is a popular video-transcoding app that recently hit version 1.1.1. To see how it performs on these chips, we converted a roughly two-minute 4K source file from an iPhone 6S into a 1920×1080, 30 FPS MKV using the HEVC algorithm implemented in the x265 open-source encoder. We otherwise left the preset at its default settings.

Handbrake doesn’t give second-generation Threadripper much room to shine versus its predecessor.

SPECwpc WPCcfd

Computational fluid dynamics is an interesting and CPU-intensive benchmark. For years and years, we’ve used the Euler3D benchmark from Oklahoma State University’s CASElab, but that benchmark has become more and more difficult to continue justifying in today’s newly-competitive CPU landscape thanks to its compilation with Intel tools (and hence a baked-in vendor advantage).

Ahead of this review, we set out to find a more vendor-neutral and up-to-date computational fluid dynamics benchmark than the wizened Euler3D. As it happens, the SPECwpc benchmark includes a CFD test constructed with Microsoft’s HPC Pack, the OpenFOAM toolkit, and the XiFoam solver. More information on XiFoam is available here. SPECwpc allows us to yoke every core and thread of our test systems for this benchmark.

In this mostly memory-bound benchmark, the 2950X opens a small lead over the 1950X. That’s consistent with the improved performance from memory operations we saw in our AIDA64 synthetic testing.

SPECwpc NAMD

The SPECwpc benchmark also includes a Windows-ready implementation of NAMD. As its developers describe it, NAMD “is a parallel molecular dynamics code designed for high-performance simulation of large biomolecular systems. Based on Charm++ parallel objects, NAMD scales to hundreds of cores for typical simulations and beyond 500,000 cores for the largest simulations.” Our ambitions are considerably more modest, but NAMD seems an ideal benchmark for our many-core single-socket CPUs.

Mysteriously, the 2950X trails the 1950X in NAMD, although the delta isn’t massive.

That wraps up today’s productivity testing of the 2950X. As our results show, the chip’s 4.4-GHz peak single-core speed is its greatest gain from the move to Zen+. Let’s see if its refined silicon can distinguish itself in games.

 

Crysis 3

Even as it passes six years of age, Crysis 3 remains one of the most punishing games one can run. With an appetite for CPU performance and graphics power alike, this title remains a great way to put the performance of any gaming system in perspective.


In both average frame rates and 99th-percentile frame times, the 2950X proves the best Ryzen chip of the bunch in Crysis 3, but the deltas from generation to generation aren’t that large.


These “time spent beyond X” graphs are meant to show “badness,” those instances where animation may be less than fluid—or at least less than perfect. The formulas behind these graphs add up the amount of time our graphics card spends beyond certain frame-time thresholds, each with an important implication for gaming smoothness. Recall that our graphics-card tests all consist of one-minute test runs and that 1000 ms equals one second to fully appreciate this data.

The 50-ms threshold is the most notable one, since it corresponds to a 20-FPS average. We figure if you’re not rendering any faster than 20 FPS, even for a moment, then the user is likely to perceive a slowdown. 33 ms correlates to 30 FPS, or a 30-Hz refresh rate. Go lower than that with vsync on, and you’re into the bad voodoo of quantization slowdowns. 16.7 ms correlates to 60 FPS, that golden mark that we’d like to achieve (or surpass) for each and every frame.

To best demonstrate the performance of these systems with a powerful graphics card like the GTX 1080 Ti, it’s useful to look at our three strictest graphs. 8.3 ms corresponds to 120 FPS, the lower end of what we’d consider a high-refresh-rate monitor. We’ve recently begun including an even more demanding 6.94-ms mark that corresponds to the 144-Hz maximum rate typical of today’s high-refresh-rate gaming displays.

At the 8.3-ms mark, the 2950X shaves about two-tenths of a second from the time the 1950X keeps our graphics card waiting. That puts it among the best-performing chps here, but as with our productivity tests, the gaps aren’t wide.

 

Deus Ex: Mankind Divided

Thanks to its richly detailed environments and copious graphics settings, Deus Ex: Mankind Divided can punish graphics cards at high resolutions and make CPUs sweat at high refresh rates.


Deus Ex: Mankind Divided doesn’t scale well with Ryzen core counts, so the 1950X and 2950X end up toward the back of the pack in both average frame rates and 99th-percentile frame times.


At the 8.3-ms mark, the 2950X holds up the GTX 1080 Ti for about a second longer than the 1950X. We’ll call it a dead heat. All of the Intel chips feed our GTX 1080 Ti better than the Threadrippers.

 

Grand Theft Auto V

Grand Theft Auto V‘s lavish simulation of Los Santos and surrounding locales can really put the hurt on a CPU, and we’re putting that characteristic to good use here.


GTA V tends to love low memory latency and single-threaded performance, so the Threadripper 2950X is ideally positioned to demonstrate some improvements from the AMD side of the aisle. The Threadripper 2990WX’s 16-core, 32-thread mode upstages the 2950X, though, perhaps thanks to its always-on NUMA-ness and consequent tendency for Windows to keep threads near their local memory.


At the 8.3-ms mark, the 2950X shaves nearly three seconds off the 1950X’s result. Those are nice improvements, to be sure, but they’re not enough to come out on top against the i9-7900X or i9-7960X.

 

Hitman

After an extended absence from our test suite thanks to a frame rate cap, Hitman is back. This game tends to max out a couple of threads but not every core on a chip, so it’s a good test of the intermediate parts of each processor’s frequency-scaling curve. We cranked the game’s graphics settings at 1920×1080 and got to testing.


The 2950X tops the Ryzen bunch with Hitman, but 2990WX aside, the deltas from the worst-performing to the best-performing chips in this test are not large. 99th-percentile frame times are also pleasingly consistent across most every high-end desktop part here. Although we didn’t test Hitman in our original 2990WX review, the uber-Threadripper doesn’t seem to suffer from operating with all of its threads enabled with this title. Its performance deficit can almost certainly be attributed to lower overall clock speeds.


Our time-spent-beyond-X metrics track with the tight grouping of results in our broad picture of Hitman performance. Each chip only spends a couple seconds holding up our GTX 1080 Ti and extending frame times beyond 8.3 ms. Only the i7-8086K performs better than the 2950X, and that’s a fine result in this company.

 

Far Cry 5


Far Cry 5 is another title that seems keen on high single-core clock speeds and low memory latency, so the 2950X turns in a slightly higher average frame rate than the 1950X and the best 99th-percentile frame time of any Threadripper. Once again, however, the 2990WX’s 16-core mode lets it draw close to the 2950X’s performance.


The 2950X cuts nearly five seconds off the 1950X’s time spent beyond 8.3 ms in this test, possibly the single largest improvement in gaming performance we’ve seen from second-gen versus first-gen Threadripper in any of our tests. We’ll take it.

All told, high-refresh-rate gaming at low resolutions is not the natural place of high-end desktop CPUs, but the 2950X mostly improves on the 1950X in that regard. It also trades blows with Skylake-X parts in CPU-bound titles in most of our tests. That’s an important fact to note for single-system streamers who might play a game in a window at 1920×1080 with a powerful graphics card while streaming using CPU encoding, for example. We’ll be testing that very use case soon.

 

A quick look at power consumption and efficiency

To get a sense of just how the chips we’re testing today balance power consumption and performance, we’re going to run our usual back-of-the-napkin numbers using each part’s performance in the Blender “bmw27” benchmark and some handy observations from our Watts Up power meter.

First, we’ll compare the instantaneous power draw of our test systems under load while they run the Blender “bmw27” benchmark. The Threadripper 2950X shaves a few watts off the 1950X’s wall power draw in a system, but it’s basically a wash.

The Threadripper 2950X completes Take that time-to-completion in seconds and multiply it by the instantaneous power draw of the chip, and we get an estimate of the total energy each chip on the bench needs to complete the “bmw27” test scene.

All told, our estimate suggests the 2950X uses slightly less energy than the 1950X on the way to completing our Blender render.

This scatter plot might help visualize the balance between power consumption and performance for the chips we’re testing. The best results on this chart will tend toward the lower-left corner, where energy consumption is lowest and time-to-completion is fastest. The 2950X ends up in an excellent spot versus the Core i9-7900X in this plot by consuming less energy and delivering much better performance.

 

Conclusions

The story of the Ryzen Threadripper 2950X is a simple one: it offers a nice boost to single-threaded performance over the Threadripper 1950X that users will most likely notice in games and desktop use, and it costs $100 less than the 1950X did on its window sticker. For builders trying to put together balanced high-end PCs, getting both of those boxes ticked for less money than ever from an AMD CPU is nothing but upside.

Even with all of the improvements of the Zen+ architecture in its arsenal, the 2950X doesn’t benefit nearly as much from those changes as its Socket AM4 counterparts do in the multithreaded-performance department. That’s all right, though, since the Threadripper 1950X often delivered Core i9-7900X-beating performance even before Zen+ came along. The Threadripper 2950X just polishes an already-appealing package.

AMD Ryzen Threadripper 2950X

August 2018

The 2950X comes backed up by the virtues of the X399 platform, including 60 PCIe 3.0 lanes from the CPU and potential ECC RAM support from motherboard vendors. It’s still markedly more efficient than the Core i9-7900X by our estimation, too, at least in workloads that scale out to all of its cores and threads.

We’re still testing a couple use cases like digital audio workstation performance and same-PC gaming and streaming that could set the Core i9-7900X apart, but even without those results, I think the Threadripper 2950X remains the chip to beat for now in the under-$1000 high-end CPU market. Its formidable combination of high single-threaded and multi-threaded performance, plus the latency improvements in the Zen+ microarchitecture, make it a compelling value and a repeat TR Editor’s Choice.

Comments closed
    • ermo
    • 1 year ago

    Any rumours on the 7 nm process when it comes to clock speed scaling?

    Even if Zen+ in the 2950X guise looks extremely attractive, I can’t help but wonder what the next Threadripper will look like when Zen2 releases.

    • blastdoor
    • 1 year ago

    Anybody know if the Noctura threadripper air cooler will work with the 32 core version, or do we need to wait for a new cooler?

      • dragontamer5788
      • 1 year ago

      Tomshardware suggests that the 2990wx can be cooled with the Noctua U14s TR4 edition.

      [url<]https://www.tomshardware.co.uk/amd-ryzen-threadripper-2-2990wx-2950x,review-34472-12.html[/url<] I don't fully understand their graph. But it [b<]seems[/b<] to suggest that their Air-cooler tests took the 2990wx to roughly 275W before it capped out at 68C+ (suggesting a cap of ~3200 MHz). So the 2990wx can be effectively cooled at stock settings by an air cooler. But don't expect an overclock. Or if you do plan to overclock, it will be difficult to go above 3200 MHz on air alone on the 32-core beast. 68C is when Threadripper begins to downclock itself to protect itself.

      • Srsly_Bro
      • 1 year ago

      There is a new silver arrow hsf coming to TR. It boasts a 320 tdp.

    • dragontamer5788
    • 1 year ago

    2950x look really, really good. Probably the best of the lineup.

    The 51.2GB/s infinity fabric bandwidth (102GB/s bidirectional) allows for better sharing of RAM, the highest clocks, easier to cool 180W TDP. And none of the “Windows Scheduler” issues at the moment.

    Seems like the 2990wx is purely for 3d renderers, which EPYC already served that niche pretty well. In contrast, the 2950x is a well-rounded machine that’s better at gaming and other latency sensitive tasks.

    Besides, if you’re 3d rendering, you might want GPUs and the 2950x supports quad-GPU setups all the same. So the real winner is the 2950x.

      • blastdoor
      • 1 year ago

      The NAMD benchmark appears to be the closest representative of my particular niche and it definitely gives me pause on the 2990wx. From reading the linked website, it looks like NAMD generally scales very well with CPU cores but that’s not happening here, presumably because of memory latency/bandwidth issues.

      With perfect scaling, the 2990wx ought to be about 75% faster than the 2950x (based on core count times base clock), but with NAMD it’s about half that.

      The 2950x seems to be at a noticeable disadvantage relative to the 1950x on this benchmark, though, which makes me wonder if something odd is going on with this benchmark.

    • Unknown-Error
    • 1 year ago

    Thanks for the review Jeff. Something I forget to mention is how clean TR reviews are. Many other sites put OC values along with many other CPUs in and clutters the whole graph, which is really annoying.

    For $900 the 2950X is impressing compared to the 2990WX. Compared Intel it lags in gaming and niches like AVX-512 optimized software, but this is true for all Ryzen CPUs including the 2700X. I think 2950X is well balanced and for $900 it is actually a bit of steal.

      • thx1138r
      • 1 year ago

      For me, the impressive part is how it occasionally beats and frequently gets very close to intel’s 16 core effort, the $1700 core i9-7860X. For a chip that’s almost half the price that’s seriously good value.

        • Srsly_Bro
        • 1 year ago

        7960x and it’s $1399.99 on Newegg.

        For me, it’s impressive how you did little research before that post.

          • thx1138r
          • 1 year ago

          Humble apologies for that single numeral typo and for comparing a list price with another list price. It’s awesome how anonymous strangers on the internet can be so helpful would appearing to be complete Duchbags (oh no! that’s another typo, nobody will know what I mean).

            • Srsly_Bro
            • 1 year ago

            It’s also $1435 on Amazon. Do you always get defensive and call names when you make mistakes instead of acknowledging them and moving on? I’ve seen small children who act out like you. I wish you the best.

            • jarder
            • 1 year ago

            Sheesh, grow a sense of humor bro, and I might suggest that you shouldn’t dish out insults if you are not prepared to take them too.

            • thx1138r
            • 1 year ago

            Lol, glad somebody gets it.

            • moose17145
            • 1 year ago

            ^ This.

            • Srsly_Bro
            • 1 year ago

            Insults don’t bother me. People who when wrong, rather than accepting mistakes, and instead resort to calling names, are the ones who bother me.

            Perhaps I could have said it nicer to spare his feelings and ego.

            In my defense I didn’t also correct his lack of understanding between list and market price.

            I didn’t insult anyone. If you could find the post in reference to the discussion, please direct me to it. I merely made a judgement based off my observation. It was he who called me a name when made aware of errors in a way that caused his ego irreparable harm.

            I might suggest you 1. Mind your business 2. Understand the discussion if you decide to ignore (1.) And 3. Don’t make accusations without evidence.

            I also wish you the best.

            • jarder
            • 1 year ago

            [quote<]1. Mind your business[/quote<] This is a public forum, not a private discussion, if you wanted a private word with thx you should have sent a private message. Also, If you use insulting language here everybody sees it, they can also see the response. [quote<]2. Understand the discussion[/quote<] I think I understand the discussion better than you. Let me recap, after your sharp and insulting post, thx responded with humor. Humor that seems to have gone completely over your head. It's plain to see from his post that he acknowledged his mistake, called you out for being the typo police and he also avoided insulting you by calling you a typo. It's only your fetish for correcting typos that turned it into an insult. [quote<]3. Don't make accusations without evidence.[/quote<] Your post is plain evidence that you did not get the joke.

            • Srsly_Bro
            • 1 year ago

            You didn’t fully quote #2. You have an agenda, bro? We can’t debate of you aren’t honest.

            • jarder
            • 1 year ago

            No Agenda, just unnecessary. The full quote is:
            [quote<]2. Understand the discussion if you decide to ignore (1.)[/quote<] I thought there was no need to remind everyone that I was ignoring your first point.

          • Goty
          • 1 year ago

          It’s funny how A) you knew exactly what he meant despite the typo and B) his point is still entirely valid after your attempt to dilute it by comparing retail prices to list prices (since, y’know, the MSRP of the 2950X is $500 less than the retail price you dug up.)

          Kind of sounds like sour grapes to me.

    • chubbyhorse
    • 1 year ago

    So… it can, indeed, run Crysis.

    • Shobai
    • 1 year ago

    First page, “sentence fragment, consider revising”:

    [quote<]...meaning applications that ended up somewhere in between weren't necessarily taking full .[/quote<] Crisis 3: in the frame time graphs, the 1920X plot is missing from the appropriate tab

      • ermo
      • 1 year ago

      Found another one (also on page 1):

      [quote<]As a beneficiary of the Zen+ microarchitectural improvements that first appeared on second-generation Ryzen chips in Socket AM4, the 2950X implements three[u<] . [/u<]Precision Boost 2 lets (...)[/quote<]

        • ermo
        • 1 year ago

        And another (still page 1):

        [quote<]Second, Extended Frequency Range 2 (XFR 2) allows the [u<][b<]2990WX[/b<][/u<] to take advantage of (...)[/quote<] I'm guessing that was supposed to be 2950X instead (I get that the introductions are probably relatively similar)?

    • DreadCthulhu
    • 1 year ago

    It looks that the odd 2990WX results may be Nvidia’s fault. Golem.de tested that chip with a Vega 64, and found that it massively outperformed the GTX 1080 Ti in 4 of the 5 games they tried. I am guessing Nvidia has some driver weirdness going on with very-high core count chips.

    [url=https://www.golem.de/news/32-kern-cpu-threadripper-2990wx-laeuft-mit-radeons-besser-1808-136016.html<]Results here[/url<]

      • techguy
      • 1 year ago

      Yeah, “massively outperformed” to the point that you’re getting 38.5 FPS @ 720p in Assassin’s Creed Origins and 42.1 FPS @720p in Kingdom Come Deliverance.

      Even the other results being 50FPS or higher is still a joke for 720p.

        • Goty
        • 1 year ago

        I don’t think you can take the actual framerates at face value there. For games shared with TR’s test suite, their framerates with the 1080Ti are significantly lower at 720p than those obtained at 1080p by TR.

          • techguy
          • 1 year ago

          lol wut?

          Can’t take the actual framerates at face value? So they made them up? The numbers are what they are. Either you accept them, or you don’t. This isn’t complicated. AMD’s multi-die Threadripper design isn’t well-suited for games. That’s all there is to it. Ryzen does fine, Threadripper doesn’t. Whether it’s an O.S. scheduler issue, latency issue, NUMA problem, or all of the above – the numbers are what they are.

            • jihadjoe
            • 1 year ago

            [quote<] Whether it's an O.S. scheduler issue, latency issue, NUMA problem, or all of the above - the numbers are what they are.[/quote<] Don't forget the hardware issue coming from big TR having half the die-to-die bandwidth of Epyc or small TR which is the root cause of these problems.

            • Goty
            • 1 year ago

            So how does that explain the performance difference between the 1080Ti and Vega 64 in the link above? If it’s an inter-die communication issue, we still shouldn’t expect the Vega 64 to [i<]outperform[/i<] a 1080Ti in basically any scenario. At best we should expect them to perform roughly the same as they wait on the CPU.

            • dragontamer5788
            • 1 year ago

            Not necessarily. AMD may have simply used affinity settings to lock threads manually when they detect Threadripper CPUs. NVidia could be defaulting to Window’s defaults, which probably didn’t consider the possibility of 1/2 speed bandwidth yet between NUMA cores.

            Threadripper is a very strange CPU: accessing RAM is [b<]FASTER[/b<] than accessing remote L3 cache. Which means if you schedule a task on a remote NUMA, you will incur a significant delay as the data is transferred between the L3 caches. AMD would obviously know about this, and probably wrote their drivers to avoid the issue. (IE: manually setting affinity settings)

            • derFunkenstein
            • 1 year ago

            edit: nvm.

            • Goty
            • 1 year ago

            I don’t know what they did, and neither do you. Nobody is arguing that it’s a good part for gaming, someone has simply pointed out that it appears a large part of the performance deficit appears like it could be due to an issue with NVIDIA’s drivers.

            I say you can’t take the framerates at face value because obviously there’s a difference in how they did their testing compared to TR, for example. How else do you explain the extreme difference in framerates between their 2990WX + 1080Ti at 720p and those obtained by TR at 1080p? Even given that the lower resolution is more CPU bound, we should not expect framerates to [i<]increase[/i<] as resolution increases. Since we don't know the differences in their test setup or methodology, all we can do is compare their two sets of results to one another, and those results indicate an issue when pairing a 1080Ti with this CPU.

            • K-L-Waster
            • 1 year ago

            A Vega 64 at 720 would be running triple digit frame rates on a 2700X.

            So to suggest that the performance problems are *solely* an Nvidia driver problem is ignoring the fact that the performance of *both* cards in that test are horrendously bad. Sure, Nvidia’s are worse, but there’s more going on here than a bad NVidia driver.

            • Goty
            • 1 year ago

            Of course there is. Nobody has made any claim to the contrary. The ONLY claim being made is that performance suffers [i<]more[/i<] when paired with a 1080Ti. Man, critical reading skills have really taken a hit lately.

            • moose17145
            • 1 year ago

            I REALLY have to agree with you here. Someone simply points out something interesting and gets attacked for it.

      • Questar
      • 1 year ago

      .

      1. Their results don’t seem to agree with others on the web.

      2. Why would it be Nvidia’s “fault” as you say?

        • BurntMyBacon
        • 1 year ago

        1) Curious. I haven’t seen anyone else test GF1080Ti vs Vega64, so I’m not really sure I can say they disagree with anyone in that regard. As far as just the GF1080Ti results go, it certainly defies expectations set by many others on the internet. I’m interested in what in their test setup is holding back their framerates. Also, if they managed to fix it, would the GF1080Ti still under perform the Vega64.

        2) I tend to agree with your sentiments about assigning fault. It is likely that nVidia simply hasn’t yet optimized their driver structure for such an esoteric and niche setup. It is also possible that they won’t need to if Microsoft does a good job adapting the windows scheduler to the processor. When you introduce a significantly different architecture to the industry, it takes time to optimize for it. There is really no fault to be assigned.

        AMD likely both had knowledge of the situation long before nVidia and more incentive to optimize to make their processor look good. That said, I’d imagine the bulk of the optimization efforts went into their workstation graphics drivers as that is the more likely pairing for the TR 2990WX. Gaming driver improvements were likely a windfall.

    • chuckula
    • 1 year ago

    I hope Intel knows a good tailor.

    Because its threads just got ripped!

      • jarder
      • 1 year ago

      Intel will be fine, don’t they have a low-cost desktop 8 core in the near-term pipeline, and after that they will surely have a new ‘some-kinda-lake-X’ range which will no doubt give users more cores for fewer dollars to restore competition.

        • chuckula
        • 1 year ago

        I’m not so sure.

        I think ThreaRipper [url=https://techreport.com/forums/viewtopic.php?f=2&t=121239&view=unread#unread<]Foreshadows[/url<] doom for Intel!

          • jarder
          • 1 year ago

          Nah, if spectre and Meltdown had practically no effect on their brand loyalty, then I can’t see another set of bugs having a significant effect.

          • HTarlek
          • 1 year ago

          [I’mNotDeadYet] You’re not foolin’ anyone, you know… [/I’mNotDeadYet]

            • K-L-Waster
            • 1 year ago

            Quiet you, you’ll be dead in a minute.

            • chuckula
            • 1 year ago

            I want to go for a walk!

Pin It on Pinterest

Share This