SPEC2006 and SPEC2017 (Single Thread)

Due to some limitations with our systems, we were only able to run SPEC in single thread mode in time for the review. Given that these 7F processors are meant to be the highest frequency EPYC hardware available, in single thread and multi-thread, this is still a very relevant test for the use case. Unfortunately we introduced this test late last year, after testing the bulk of our Intel CPUs. We’re currently re-running on a few and will update this post over the next few days.

*If you are seeing this as the review goes live, we are still waiting for the 6226R results to finish.

SPEC2006 1T Estimated Results
AnandTech AMD
7F52
AMD
7601
AMD
3990X
AMD
3950X
  Intel
6226R
Intel
9900KS
Intel
10980XE
uArch Rome Naples Rome Rome   CLX-R Coffee CLX
Turbo 3900 3200 4300 4700   3900 5000 4800
 
400.perlbench 45.9 29.8 50.8 54.6   40.2 60.1 55.2
401.bzip2 30.9 23.3 34.5 36.6   25.4 37.5 33.5
403.gcc 37.7 28.0 53.4 57.7   30.0 56.1 46.6
429.mcf 35.6 22.6 48.6 52.9   28.5 64.7 45.3
445.gobmk 36.7 23.4 41.8 44.9   32.0 43.3 39.6
456.hmmr 36.8 26.8 41.0 43.3   39.2 51.7 48.2
458.sjeng 32.5 21.9 38.1 41.1   34.7 47.0 43.6
462.libquantum 78.7 50.3 100.4 102.8   38.5 113.2 106.8
464.h264ref 67.7 49.6 75.9 80.4   64.7 83.9 79.1
471.omnetpp 21.1 14.0 27.5 31.9   25.5 31.3 30.0
473.astar 26.9 17.8 30.9 32.8   22.9 30.2 29.5
483.xalancbmk 46.0 29.2 53.8 58.0   37.5 60.4 54.6
 
433.milc 35.0 22.6 46.9 49.3   15.7 31.9 27.9
444.namd 39.0 29.6 43.3 45.9   38.3 52.5 43.9
450.soplex 58.9 39.7 73.7 74.8   21.5 73.0 67.1
453.povray 59.7 37.0 66.3 70.9   58.5 76.2 70.5
470.lbm 101.4 72.4 121.8 126.2   20.2 77.7 102.9
482.sphinx3 94.7 56.2 107.4 113.0   45.3 105.0 72.6
 
Geomean 44.8 30.2 53.6 57.1   32.3 56.6 51.1

The performance jump from the Naples 7601 to the Rome 7F52 is bordering on about 50%. It is worth pointing out that AMD’s consumer Ryzen 9 3950X wins out here due to IPC and single core frequency, closely followed by Intel’s i9-9900KS, the AMD Threadripper 3000s, and the Intel i9-10980XE. This comes down to consumer platforms affording much larger turbos and not being stricter on RAS requirements and such.

SPEC2017 1T Estimated Results
AnandTech AMD
7F52
AMD
7601
AMD
3990X
AMD
3950X
  Intel
6226R
Intel
9900KS
Intel
10980XE
uArch Rome Naples Rome Rome   CLX-R Coffee CLX
Turbo 3900 3200 4300 4700   3900 5000 4800
 
500.perlbench_r 4.3 2.7 5.0 5.3   5.1 6.9 6.3
502.gcc_r 6.1 4.4 8.0 8.6   3.8 9.3 7.4
505.mcf_r 5.0 3.5 6.1 6.6   3.2 6.5 5.4
520.omnetpp_r 2.4 2.0 3.4 3.7   3.1 4.1 3.8
523.xalancbmk_r 4.7 2.5 5.0 5.3   4.0 4.4 5.3
525.x264_r 7.8 5.7 9.0 9.5   6.8 9.7 9.0
531.deepsjeng_r 3.7 3.0 4.4 4.7   4.0 5.5 5.0
541.leela_r 4.1 2.9 4.6 4.9   3.7 5.0 4.6
548.exchange2_r 7.3 4.5 8.2 8.7   6.2 8.3 7.5
557.xz_r 3.0 2.1 3.8 4.1   2.9 4.1 3.8
 
503.bwaves_r 39.7 27.4 46.5 48.5   7.4 38.2 30.6
507.cactuBSSN_r 5.6 4.2 6.4 6.7   4.3 8.3 6.1
508.namd_r 6.0 4.6 6.7 7.0   4.1 7.4 6.3
510.parest_r 7.5 5.5 8.4 9.0   4.4 9.7 7.4
511.povray_r 6.7 4.2 7.5 7.9   6.6 8.7 8.0
519.lbm_r 6.9 5.0 8.0 8.4   1.0 7.7 6.3
521.wrf_r * - - - -   - - -
526.blender_r 6.6 4.7 7.5 8.0   5.2 7.9 7.2
527.cam4_r 6.8 4.8 7.7 8.2   4.8 8.3 6.4
538.imagick_r 7.9 5.8 8.8 9.4   6.4 8.5 7.8
544.nab_r 4.0 3.0 4.4 4.7   3.0 5.2 4.7
549.fotonik_r 14.2 8.1 17.2 16.4   3.5 14.8 11.4
554.roms_r 9.0 5.3 10.9 11.4   3.8 10.0 7.3
 
Geomean 6.3 4.3 7.3 7.7   4.1 7.8 6.8
*512.wrf_r unfortunately doesn't run properly in our SPEC harness at this time

We see a similar result in the newer version of SPEC, again with ~50% jump from the Naples 7601 to the Rome 7F52. The 9900KS has the overall better Geomean here, followed closely behind by the 3950X, then the Threadrippers.

Frequency Ramp, Latency and Power CPU Performance: Rendering and Synthetics
POST A COMMENT

100 Comments

View All Comments

  • anonomouse - Tuesday, April 14, 2020 - link

    Those are all of heavily memory (latency, bandwidth) bound workloads in SPEC, too, which points to something very strange. CLX-R should not be >5x slower than CLX on lbm. Reply
  • eastcoast_pete - Tuesday, April 14, 2020 - link

    Thanks Ian! Two questions: 1. Could you and some of your readers give specific examples of applications for which these high frequency CPUs are of great interest?
    2. Any recent moves by Intel to make software developers use AVX512 even more, basically whenever it would make any sense?
    The reason I am asking the second question is that this seems to be the last bastion Intel holds, almost regardless of CPU class. Except for AVX512, AMD is beating them in price/performance quite badly, now from servers to workstations to desktop to mobile.
    Reply
  • schujj07 - Tuesday, April 14, 2020 - link

    DB servers are one place where you want a fast CPU. SAP HANA for example loves frequency and RAM. I've seen PRD systems with all of 16CPUs but 1.5TB RAM. Reply
  • DanNeely - Tuesday, April 14, 2020 - link

    AVX is a compute feature. Rendering and math heavy scientific/engineering workloads are where it'd shine. Databases, typical webservers, and most other 'conventional' business related software don't care. Reply
  • Shorty_ - Thursday, April 16, 2020 - link

    Web serving is another place where frequency really helps. I run threadrippers with ECC UDIMM for php hosting for this express reason Reply
  • Mikewind Dale - Tuesday, April 14, 2020 - link

    Unfortunately, this breaks AMD's trend of being cheaper than Intel. A 20 core Xeon Xeon Gold 5218R boosts up to 4.0 GHz and costs $1273. This new EPYC is only 16 core, boosts only up to 3.9 GHz, and costs $3100.

    Usually, AMD is cheaper than Intel, but this seems to be an exception. A pity.
    Reply
  • Fataliity - Tuesday, April 14, 2020 - link

    That's because its a specialized processor. If you are buying one of these, you won't be worried about the price.

    To get that much cache, they are using 6-8chiplets. So as many as their top of the line products. So yeah, its going to cost more because theres more silicon.
    Reply
  • schujj07 - Tuesday, April 14, 2020 - link

    The 5218R that you referenced isn't what the 7F52 is competing against. With a base clock of 2.1GHz the 5218R isn't a frequency optimized part. Most of Intel's CPUs have high boost clocks and middle of the road base clocks. The actual competition is the 6246R which has a 3.4GHz base and 4.1GHz boost clock. These high base clocks are for sustained performance in a given scenario. Reply
  • MFinn3333 - Tuesday, April 14, 2020 - link

    That’s is also because it has about 12x as much L3 cache per CPU core. A combined 256MB vs 30MB cache size speaks for itself. Reply
  • edzieba - Tuesday, April 14, 2020 - link

    It's down to two design choices: process choice, and core choice.

    AMDs hands are somewhat tied when it comes to process choice. They get what TSMC has on offer, and what TSMC has on offer is geared towards mobile devices because that's where the volume market is. The high-performance variants are variants, rather than the baseline.
    But even in general, as you shrink your process from 21nm on down, it gets harder and harder to clock up. Gate oxide thickness hit its limit generations ago, which is why gate voltage has remained near constant (~1.1v) for so long. This is only going to get harder as processes shrink further while being stuck with the constant gate oxide thickness but trying to cram closer together without interfering.

    In AMD's hands is the design goal of cramming as many cores as possible in. Great for multi-core workloads, but not so great for single core speed. Getting CPUs to clock higher means using multiple transistors per gate (2-3 or even more as processes shrink), and AMD figured they may as well use these transistors for more cores instead of faster cores. The obvious downside is the difficulty in getting Zen cores to even approach 5GHz (with Zen 2 being notable for getting above 4GHz without overkill cooling), and that any workloads that do not span beyond one thread leave those transistors sitting idle.
    Reply

Log in

Don't have an account? Sign up now