Conclusion: History is Written By The Victors

I have never used the word ‘bloodbath’ in a review before. It seems messy, violent, and a little bit gruesome. But when we look at the results from the new AMD Threadripper processors, it seems more than appropriate.

When collating the data together from our testing, I found it amusing that when we start comparing the high-end desktop processors, any part that was mightily impressive in the consumer space suddenly sits somewhere in the middle or back, holding its lunch money tightly. While the 16-core Ryzen 9 3950X and the 8-core Intel i9-9900KS enjoy a lot fun in the consumer space, when Threadripper rolls up, they are decidedly outclassed in performance.

AMD has scored wins across almost all of our benchmark suite. In anything embarrassingly parallel it rules the roost by a large margin (except for our one AVX-512 benchmark). Single threaded performance trails the high-frequency mainstream parts, but it is still very close. Even in memory sensitive workloads, an issue for the previous generation Threadripper parts, the new chiplet design has pushed performance to the next level. These new Threadripper processors win on core count, on high IPC, on high frequency, and on fast memory.

Is the HEDT Market Price Sensitive?

There are two areas where AMD will be questioned upon. First is the power, and why 280 W for the TDP? Truth be told, these are some of the most efficient desktop cores we have seen; it's just that AMD has piled a lot of them into a single processor. The other question is price.

Where Intel has retreated from the $2000 market, pushing its 18-core CPU back to $979, AMD has leapfrogged into that $1999 space with the 32-core and $1399 with the 24-core. This is the sort of price competition we have desperately needed in this space, although I have seen some commentary that AMD’s pricing is too high. The same criticism was leveled at Intel for the past couple of generations as well.

Now the HEDT market is a tricky one to judge. As one might expect, overall sales numbers aren’t on the level of the standard consumer volumes. Still, Intel has reported that the workstation market has a potential $10B a year addressable market, so it is still worth pursuing. While I have no direct quotes or data, I remember being told for several generations that Intel’s best-selling HEDT processors were always the highest core count, highest performance parts that money could buy. These users wanted off-the-shelf hardware, and were willing to pay for it – they just weren’t willing to pay for enterprise features. I was told that this didn’t necessarily follow when Intel pushed for 10 cores to $1979, when 8 cores were $999, but when $1979 became 18 cores, a segment of the market pushed for it. Now that we can get better performance at $1999 with 32 cores, assuming AMD can keep stock of the hardware, it stands to reason that this market will pick up interest again.

There is the issue of the new chipset, and TRX40 motherboards. Ultimately it is a slight negative that AMD has had to change chipsets and there’s no backwards compatibility. For that restriction though, we see an effective quadrupling of CPU-to-chipset bandwidth, and we’re going to see a wide range of motherboards with different controllers and support. There seems to be a good variation, even in the initial 12 motherboards coming to the market, with the potential for some of these companies to offer something off-the-wall and different. Motherboard pricing is likely to be high, with the most expensive initial motherboard, the GIGABYTE TRX40 Aorus Extreme, to be $849. Filling it up with memory afterwards won’t be cheap, either. But this does give a wide range of variation.

One of the key messages I’ve been saying this year is that AMD wants to attack the workstation market en mass. These new Threadripper processors do just that.

The Final Word

If you had told me three years ago that AMD were going to be ruling the roost in the HEDT market with high-performance 32-core processors on a leading-edge manufacturing node, I would have told you to lay off the heavy stuff. But here we are, and AMD isn’t done yet, teasing a 64-core version for next year. This is a crazy time we live in, and I’m glad to be a part of it.

AMD Third Generation Ryzen Threadripper

Price no object, the new Threadripper processors are breathing new life into the high-end desktop market. AMD is going to have to work hard to top this one. Intel is going to have to have a shift its design strategy to compete.

Many thanks to Gavin Bonshor for running the benchmarks, and Andrei Frumusanu for the memory analysis.

Gaming: F1 2018
Comments Locked

245 Comments

View All Comments

  • RSAUser - Tuesday, November 26, 2019 - link

    I've only seen the Mozilla benchmarks on LTT, very strange that they're the only ones showing such a workload. I'd be very interested on how these chips handle e.g. large SQL Server DB's and requests, especially with those huge caches.

    The Mozilla benchmark had near 2x the performance for the 3970X vs the 10980X and serve the home has the ryzen chip at near 30 compiles an hour for the Linux Kernel vs around 16 for Intel.

    I'd actually be really interested in the financial market for this TR due to the floating point performance increase. We'll probably be upgrading our servers next year based on current projections, so this has been a really nice development.
  • Dolda2000 - Monday, November 25, 2019 - link

    Why is it that Intel gains so incredibly much more from AVX512 than AMD gains from AVX2?

    In the 3DPM2 test, the AMD CPUs gain roughly a factor of two in performance, which is exactly what I'd expect given that AVX2 is twice as wide as standard SSE. The Intel CPUs, on the other hand, gain almost a factor of 9, which is more than twice what I'd expect given that AVX512 as four times as wide as SSE.

    What causes this? Does AVX512 have some other kind of tricks up its sleeves? Does opmasking benefit 3DPM2?
  • AnGe85 - Monday, November 25, 2019 - link

    The Intel parts are derived from Xeon dies (LCC 10 cores, and HCC up to 18 cores). As such they have two AVX-512-FMA-Units.
    Zen/+ shows a +70 % increase in performance, Zen2 and the 9900K(S) about +90 % with AVX2 in 3DPM2.1 and the Xeon-based parts reach up to +700 %. Ian has obviously done a good job or at least used a good lib ;-)
  • Dolda2000 - Monday, November 25, 2019 - link

    But Zen 1/2 also has two 256-bit FMAs per core. And Intel also has two SSE units per core as well, so I don't see how that would explain the ratios.
  • yeeeeman - Monday, November 25, 2019 - link

    Intel has 512bit units
  • Dolda2000 - Monday, November 25, 2019 - link

    Exactly, which should make it 2× as fast, not 4.5×.
  • abufrejoval - Tuesday, November 26, 2019 - link

    The other element of magic is typically halved operand size=twice the data element throughput.
    Could be FP16 vs FP32 in that code, which means 32 vector elements per 512 bit register and then again of these registers there could be mulitples under SIMD per instruction and clock.
  • Xyler94 - Tuesday, November 26, 2019 - link

    Servethehome also mentioned in their reviews of Epyc Rome Processors, the same basic Zen2 platform that the new TR CPUs are made on, that most programs aren't optimized for AMD's new AVX2 pipes, so the results are lower than they should be. I don't know if that's still the case, but it may be a reason why it's showing such a disparity between the two.
  • Slash3 - Monday, November 25, 2019 - link

    Wow.
    Just wow.
  • shaolin95 - Monday, November 25, 2019 - link

    Why wont yuo enable IGPU for the 9900k on the Premiere test? It will change the performance dramatically.

Log in

Don't have an account? Sign up now