Zen4 7950X Benchmarks - Page 2

joblack · 2022-10-28, 12:21

Quote:

Originally Posted by M344587487

Data is good but it might be better to wait for someone to test with faster RAM as that's what'll typically be used, AMD officially supports 5200 and reviewers seem to be going for 6000 or even 6400. I can't even buy 4400 DDR5 from my normal retailer, the speeds on offer range from 4800 to 6400.

As nobody came forward with a faster RAM could you please still test it with your setup?

scan80269 · 2022-11-06, 00:13

Ryzen 9 7950X PBO enabled and 2x16GB DDR5-6000 30-38-38-96 memory

Code:

[Sat Nov  5 15:52:45 2022]
Compare your results to other computers at http://www.mersenne.org/report_benchmarks
AMD Ryzen 9 7950X 16-Core Processor            
CPU speed: 2980.15 MHz, 16 hyperthreaded cores
CPU features: 3DNow! Prefetch, SSE, SSE2, SSE4, AVX, AVX2, FMA, AVX512F
L1 cache size: 16x32 KB, L2 cache size: 16x1 MB, L3 cache size: 2x32 MB
L1 cache line size: 64 bytes, L2 cache line size: 64 bytes
Machine topology as determined by hwloc library:
 Machine#0 (total=30641516KB, Backend=Windows, OSName=Windows, WindowsBuildEnvironment=MinGW, OSRelease=10, OSVersion=10.0.20348, Hostname=PROARTX670E, Architecture=x86_64, hwlocVersion=2.6.0, ProcessName=prime95.exe)
  Package (total=30641516KB, CPUVendor=AuthenticAMD, CPUFamilyNumber=25, CPUModelNumber=97, CPUModel="AMD Ryzen 9 7950X 16-Core Processor            ", CPUStepping=2)
    L3 (size=32768KB, linesize=64, ways=16, Inclusive=0)
      L2 (size=1024KB, linesize=64, ways=8, Inclusive=1)
        L1d (size=32KB, linesize=64, ways=8, Inclusive=0)
          Core (cpuset: 0x00000003)
            PU#0 (cpuset: 0x00000001)
            PU#1 (cpuset: 0x00000002)
      L2 (size=1024KB, linesize=64, ways=8, Inclusive=1)
        L1d (size=32KB, linesize=64, ways=8, Inclusive=0)
          Core (cpuset: 0x0000000c)
            PU#2 (cpuset: 0x00000004)
            PU#3 (cpuset: 0x00000008)
      L2 (size=1024KB, linesize=64, ways=8, Inclusive=1)
        L1d (size=32KB, linesize=64, ways=8, Inclusive=0)
          Core (cpuset: 0x00000030)
            PU#4 (cpuset: 0x00000010)
            PU#5 (cpuset: 0x00000020)
      L2 (size=1024KB, linesize=64, ways=8, Inclusive=1)
        L1d (size=32KB, linesize=64, ways=8, Inclusive=0)
          Core (cpuset: 0x000000c0)
            PU#6 (cpuset: 0x00000040)
            PU#7 (cpuset: 0x00000080)
      L2 (size=1024KB, linesize=64, ways=8, Inclusive=1)
        L1d (size=32KB, linesize=64, ways=8, Inclusive=0)
          Core (cpuset: 0x00000300)
            PU#8 (cpuset: 0x00000100)
            PU#9 (cpuset: 0x00000200)
      L2 (size=1024KB, linesize=64, ways=8, Inclusive=1)
        L1d (size=32KB, linesize=64, ways=8, Inclusive=0)
          Core (cpuset: 0x00000c00)
            PU#10 (cpuset: 0x00000400)
            PU#11 (cpuset: 0x00000800)
      L2 (size=1024KB, linesize=64, ways=8, Inclusive=1)
        L1d (size=32KB, linesize=64, ways=8, Inclusive=0)
          Core (cpuset: 0x00003000)
            PU#12 (cpuset: 0x00001000)
            PU#13 (cpuset: 0x00002000)
      L2 (size=1024KB, linesize=64, ways=8, Inclusive=1)
        L1d (size=32KB, linesize=64, ways=8, Inclusive=0)
          Core (cpuset: 0x0000c000)
            PU#14 (cpuset: 0x00004000)
            PU#15 (cpuset: 0x00008000)
    L3 (size=32768KB, linesize=64, ways=16, Inclusive=0)
      L2 (size=1024KB, linesize=64, ways=8, Inclusive=1)
        L1d (size=32KB, linesize=64, ways=8, Inclusive=0)
          Core (cpuset: 0x00030000)
            PU#16 (cpuset: 0x00010000)
            PU#17 (cpuset: 0x00020000)
      L2 (size=1024KB, linesize=64, ways=8, Inclusive=1)
        L1d (size=32KB, linesize=64, ways=8, Inclusive=0)
          Core (cpuset: 0x000c0000)
            PU#18 (cpuset: 0x00040000)
            PU#19 (cpuset: 0x00080000)
      L2 (size=1024KB, linesize=64, ways=8, Inclusive=1)
        L1d (size=32KB, linesize=64, ways=8, Inclusive=0)
          Core (cpuset: 0x00300000)
            PU#20 (cpuset: 0x00100000)
            PU#21 (cpuset: 0x00200000)
      L2 (size=1024KB, linesize=64, ways=8, Inclusive=1)
        L1d (size=32KB, linesize=64, ways=8, Inclusive=0)
          Core (cpuset: 0x00c00000)
            PU#22 (cpuset: 0x00400000)
            PU#23 (cpuset: 0x00800000)
      L2 (size=1024KB, linesize=64, ways=8, Inclusive=1)
        L1d (size=32KB, linesize=64, ways=8, Inclusive=0)
          Core (cpuset: 0x03000000)
            PU#24 (cpuset: 0x01000000)
            PU#25 (cpuset: 0x02000000)
      L2 (size=1024KB, linesize=64, ways=8, Inclusive=1)
        L1d (size=32KB, linesize=64, ways=8, Inclusive=0)
          Core (cpuset: 0x0c000000)
            PU#26 (cpuset: 0x04000000)
            PU#27 (cpuset: 0x08000000)
      L2 (size=1024KB, linesize=64, ways=8, Inclusive=1)
        L1d (size=32KB, linesize=64, ways=8, Inclusive=0)
          Core (cpuset: 0x30000000)
            PU#28 (cpuset: 0x10000000)
            PU#29 (cpuset: 0x20000000)
      L2 (size=1024KB, linesize=64, ways=8, Inclusive=1)
        L1d (size=32KB, linesize=64, ways=8, Inclusive=0)
          Core (cpuset: 0xc0000000)
            PU#30 (cpuset: 0x40000000)
            PU#31 (cpuset: 0x80000000)
Prime95 64-bit version 30.8, RdtscTiming=1
Timings for 2048K FFT length (16 cores, 1 worker):  0.97 ms.  Throughput: 1035.55 iter/sec.
Timings for 2048K FFT length (16 cores, 2 workers):  0.50,  0.89 ms.  Throughput: 3119.29 iter/sec.
Timings for 2048K FFT length (16 cores, 4 workers):  1.08,  1.05,  1.65,  1.71 ms.  Throughput: 3066.52 iter/sec.
Timings for 2048K FFT length (16 cores hyperthreaded, 1 worker):  1.01 ms.  Throughput: 990.06 iter/sec.
Timings for 2048K FFT length (16 cores hyperthreaded, 2 workers):  0.48,  0.55 ms.  Throughput: 3919.92 iter/sec.
Timings for 2048K FFT length (16 cores hyperthreaded, 4 workers):  1.13,  1.12,  1.35,  1.20 ms.  Throughput: 3354.57 iter/sec.
Timings for 2100K FFT length (16 cores, 1 worker):  0.94 ms.  Throughput: 1065.62 iter/sec.
Timings for 2100K FFT length (16 cores, 2 workers):  0.59,  0.97 ms.  Throughput: 2722.04 iter/sec.
Timings for 2100K FFT length (16 cores, 4 workers):  1.20,  1.12,  1.69,  1.97 ms.  Throughput: 2829.11 iter/sec.
Timings for 2100K FFT length (16 cores hyperthreaded, 1 worker):  1.08 ms.  Throughput: 925.77 iter/sec.
Timings for 2100K FFT length (16 cores hyperthreaded, 2 workers):  0.55,  0.58 ms.  Throughput: 3557.60 iter/sec.
Timings for 2100K FFT length (16 cores hyperthreaded, 4 workers):  1.20,  1.19,  1.20,  1.20 ms.  Throughput: 3344.71 iter/sec.
Timings for 2160K FFT length (16 cores, 1 worker):  0.98 ms.  Throughput: 1024.06 iter/sec.
Timings for 2160K FFT length (16 cores, 2 workers):  0.59,  0.96 ms.  Throughput: 2749.72 iter/sec.
Timings for 2160K FFT length (16 cores, 4 workers):  1.14,  1.15,  1.79,  1.99 ms.  Throughput: 2807.15 iter/sec.
Timings for 2160K FFT length (16 cores hyperthreaded, 1 worker):  1.06 ms.  Throughput: 943.70 iter/sec.
Timings for 2160K FFT length (16 cores hyperthreaded, 2 workers):  0.61,  0.66 ms.  Throughput: 3137.22 iter/sec.
Timings for 2160K FFT length (16 cores hyperthreaded, 4 workers):  1.39,  1.28,  1.43,  1.39 ms.  Throughput: 2921.37 iter/sec.
Timings for 2240K FFT length (16 cores, 1 worker):  1.09 ms.  Throughput: 914.85 iter/sec.
[Sat Nov  5 15:57:48 2022]
Timings for 2240K FFT length (16 cores, 2 workers):  0.60,  1.06 ms.  Throughput: 2613.00 iter/sec.
Timings for 2240K FFT length (16 cores, 4 workers):  1.36,  1.36,  1.80,  2.28 ms.  Throughput: 2465.16 iter/sec.
Timings for 2240K FFT length (16 cores hyperthreaded, 1 worker):  1.08 ms.  Throughput: 924.39 iter/sec.
Timings for 2240K FFT length (16 cores hyperthreaded, 2 workers):  0.57,  0.64 ms.  Throughput: 3307.97 iter/sec.
Timings for 2240K FFT length (16 cores hyperthreaded, 4 workers):  1.74,  1.49,  1.78,  1.78 ms.  Throughput: 2368.79 iter/sec.
Timings for 2304K FFT length (16 cores, 1 worker):  1.11 ms.  Throughput: 899.06 iter/sec.
Timings for 2304K FFT length (16 cores, 2 workers):  0.58,  1.04 ms.  Throughput: 2671.07 iter/sec.
Timings for 2304K FFT length (16 cores, 4 workers):  1.47,  1.47,  2.07,  2.12 ms.  Throughput: 2318.98 iter/sec.
Timings for 2304K FFT length (16 cores hyperthreaded, 1 worker):  1.13 ms.  Throughput: 886.85 iter/sec.
Timings for 2304K FFT length (16 cores hyperthreaded, 2 workers):  0.63,  0.59 ms.  Throughput: 3283.45 iter/sec.
Timings for 2304K FFT length (16 cores hyperthreaded, 4 workers):  1.69,  1.69,  2.05,  1.39 ms.  Throughput: 2387.71 iter/sec.
Timings for 2400K FFT length (16 cores, 1 worker):  1.03 ms.  Throughput: 974.57 iter/sec.
Timings for 2400K FFT length (16 cores, 2 workers):  0.64,  1.07 ms.  Throughput: 2507.64 iter/sec.
Timings for 2400K FFT length (16 cores, 4 workers):  1.37,  1.44,  1.86,  2.28 ms.  Throughput: 2397.75 iter/sec.
Timings for 2400K FFT length (16 cores hyperthreaded, 1 worker):  1.02 ms.  Throughput: 985.20 iter/sec.
Timings for 2400K FFT length (16 cores hyperthreaded, 2 workers):  0.59,  0.64 ms.  Throughput: 3257.36 iter/sec.
Timings for 2400K FFT length (16 cores hyperthreaded, 4 workers):  1.57,  1.57,  1.97,  1.39 ms.  Throughput: 2500.05 iter/sec.
Timings for 2520K FFT length (16 cores, 1 worker):  1.09 ms.  Throughput: 921.11 iter/sec.
Timings for 2520K FFT length (16 cores, 2 workers):  0.68,  1.14 ms.  Throughput: 2341.20 iter/sec.
Timings for 2520K FFT length (16 cores, 4 workers):  1.54,  1.47,  2.05,  2.50 ms.  Throughput: 2214.39 iter/sec.
[Sat Nov  5 16:02:51 2022]
Timings for 2520K FFT length (16 cores hyperthreaded, 1 worker):  1.12 ms.  Throughput: 890.09 iter/sec.
Timings for 2520K FFT length (16 cores hyperthreaded, 2 workers):  0.62,  0.69 ms.  Throughput: 3047.49 iter/sec.
Timings for 2520K FFT length (16 cores hyperthreaded, 4 workers):  1.78,  1.78,  1.97,  1.65 ms.  Throughput: 2235.79 iter/sec.
Timings for 2560K FFT length (16 cores, 1 worker):  1.14 ms.  Throughput: 880.48 iter/sec.
Timings for 2560K FFT length (16 cores, 2 workers):  0.68,  1.10 ms.  Throughput: 2379.96 iter/sec.
Timings for 2560K FFT length (16 cores, 4 workers):  1.71,  1.71,  2.29,  2.54 ms.  Throughput: 1999.28 iter/sec.
Timings for 2560K FFT length (16 cores hyperthreaded, 1 worker):  1.04 ms.  Throughput: 963.14 iter/sec.
Timings for 2560K FFT length (16 cores hyperthreaded, 2 workers):  0.67,  0.67 ms.  Throughput: 2987.08 iter/sec.
Timings for 2560K FFT length (16 cores hyperthreaded, 4 workers):  2.05,  2.53,  2.42,  2.41 ms.  Throughput: 1709.89 iter/sec.
Timings for 2592K FFT length (16 cores, 1 worker):  1.16 ms.  Throughput: 862.78 iter/sec.
Timings for 2592K FFT length (16 cores, 2 workers):  0.68,  1.09 ms.  Throughput: 2381.99 iter/sec.
Timings for 2592K FFT length (16 cores, 4 workers):  1.64,  1.79,  2.41,  2.56 ms.  Throughput: 1972.75 iter/sec.
Timings for 2592K FFT length (16 cores hyperthreaded, 1 worker):  1.13 ms.  Throughput: 886.61 iter/sec.
Timings for 2592K FFT length (16 cores hyperthreaded, 2 workers):  0.63,  0.65 ms.  Throughput: 3129.89 iter/sec.
Timings for 2592K FFT length (16 cores hyperthreaded, 4 workers):  2.27,  2.71,  2.35,  2.77 ms.  Throughput: 1595.41 iter/sec.
Timings for 2688K FFT length (16 cores, 1 worker):  1.15 ms.  Throughput: 866.25 iter/sec.
Timings for 2688K FFT length (16 cores, 2 workers):  0.71,  1.16 ms.  Throughput: 2270.05 iter/sec.
Timings for 2688K FFT length (16 cores, 4 workers):  1.81,  2.16,  2.40,  2.83 ms.  Throughput: 1782.69 iter/sec.
Timings for 2688K FFT length (16 cores hyperthreaded, 1 worker):  1.18 ms.  Throughput: 848.32 iter/sec.
Timings for 2688K FFT length (16 cores hyperthreaded, 2 workers):  0.65,  0.72 ms.  Throughput: 2918.60 iter/sec.
[Sat Nov  5 16:07:54 2022]
Timings for 2688K FFT length (16 cores hyperthreaded, 4 workers):  2.61,  2.65,  2.31,  3.10 ms.  Throughput: 1515.76 iter/sec.
Timings for 2880K FFT length (16 cores, 1 worker):  1.32 ms.  Throughput: 760.29 iter/sec.
Timings for 2880K FFT length (16 cores, 2 workers):  0.75,  1.26 ms.  Throughput: 2137.51 iter/sec.
Timings for 2880K FFT length (16 cores, 4 workers):  2.00,  2.94,  2.80,  3.15 ms.  Throughput: 1513.93 iter/sec.
Timings for 2880K FFT length (16 cores hyperthreaded, 1 worker):  1.33 ms.  Throughput: 752.02 iter/sec.
Timings for 2880K FFT length (16 cores hyperthreaded, 2 workers):  0.72,  0.77 ms.  Throughput: 2679.84 iter/sec.
Timings for 2880K FFT length (16 cores hyperthreaded, 4 workers):  3.08,  3.02,  3.70,  2.39 ms.  Throughput: 1343.06 iter/sec.
Timings for 2940K FFT length (16 cores, 1 worker):  1.29 ms.  Throughput: 776.12 iter/sec.
Timings for 2940K FFT length (16 cores, 2 workers):  0.83,  1.30 ms.  Throughput: 1968.81 iter/sec.
Timings for 2940K FFT length (16 cores, 4 workers):  2.67,  2.63,  3.28,  3.39 ms.  Throughput: 1354.08 iter/sec.
Timings for 2940K FFT length (16 cores hyperthreaded, 1 worker):  1.32 ms.  Throughput: 757.14 iter/sec.
Timings for 2940K FFT length (16 cores hyperthreaded, 2 workers):  0.78,  0.82 ms.  Throughput: 2496.45 iter/sec.
Timings for 2940K FFT length (16 cores hyperthreaded, 4 workers):  3.91,  3.66,  3.90,  4.14 ms.  Throughput: 1026.87 iter/sec.
Timings for 3000K FFT length (16 cores, 1 worker):  1.28 ms.  Throughput: 780.24 iter/sec.
Timings for 3000K FFT length (16 cores, 2 workers):  0.81,  1.32 ms.  Throughput: 1985.41 iter/sec.
Timings for 3000K FFT length (16 cores, 4 workers):  2.64,  2.64,  3.01,  3.17 ms.  Throughput: 1406.95 iter/sec.
Timings for 3000K FFT length (16 cores hyperthreaded, 1 worker):  1.59 ms.  Throughput: 628.60 iter/sec.
Timings for 3000K FFT length (16 cores hyperthreaded, 2 workers):  0.77,  0.83 ms.  Throughput: 2508.79 iter/sec.
Timings for 3000K FFT length (16 cores hyperthreaded, 4 workers):  3.44,  3.11,  2.86,  3.94 ms.  Throughput: 1215.67 iter/sec.
Timings for 3072K FFT length (16 cores, 1 worker):  1.34 ms.  Throughput: 745.30 iter/sec.
[Sat Nov  5 16:12:58 2022]
Timings for 3072K FFT length (16 cores, 2 workers):  0.76,  1.30 ms.  Throughput: 2080.80 iter/sec.
Timings for 3072K FFT length (16 cores, 4 workers):  3.23,  2.98,  2.93,  3.73 ms.  Throughput: 1253.90 iter/sec.
Timings for 3072K FFT length (16 cores hyperthreaded, 1 worker):  1.29 ms.  Throughput: 776.85 iter/sec.
Timings for 3072K FFT length (16 cores hyperthreaded, 2 workers):  0.72,  0.78 ms.  Throughput: 2658.43 iter/sec.
Timings for 3072K FFT length (16 cores hyperthreaded, 4 workers):  2.55,  4.36,  2.35,  4.75 ms.  Throughput: 1257.77 iter/sec.
Timings for 3136K FFT length (16 cores, 1 worker):  1.52 ms.  Throughput: 657.38 iter/sec.
Timings for 3136K FFT length (16 cores, 2 workers):  0.85,  1.49 ms.  Throughput: 1840.72 iter/sec.
Timings for 3136K FFT length (16 cores, 4 workers):  3.56,  3.56,  3.61,  4.46 ms.  Throughput: 1062.95 iter/sec.
Timings for 3136K FFT length (16 cores hyperthreaded, 1 worker):  1.40 ms.  Throughput: 716.21 iter/sec.
Timings for 3136K FFT length (16 cores hyperthreaded, 2 workers):  0.86,  0.92 ms.  Throughput: 2246.46 iter/sec.
Timings for 3136K FFT length (16 cores hyperthreaded, 4 workers):  4.90,  4.96,  5.51,  4.61 ms.  Throughput: 804.06 iter/sec.
Timings for 3200K FFT length (16 cores, 1 worker):  1.37 ms.  Throughput: 731.72 iter/sec.
Timings for 3200K FFT length (16 cores, 2 workers):  0.87,  1.44 ms.  Throughput: 1849.03 iter/sec.
Timings for 3200K FFT length (16 cores, 4 workers):  3.25,  3.26,  3.30,  3.91 ms.  Throughput: 1173.39 iter/sec.
Timings for 3200K FFT length (16 cores hyperthreaded, 1 worker):  1.74 ms.  Throughput: 575.11 iter/sec.
Timings for 3200K FFT length (16 cores hyperthreaded, 2 workers):  0.84,  0.84 ms.  Throughput: 2388.41 iter/sec.
Timings for 3200K FFT length (16 cores hyperthreaded, 4 workers):  4.44,  3.84,  4.68,  4.18 ms.  Throughput: 938.31 iter/sec.
Timings for 3360K FFT length (16 cores, 1 worker):  1.49 ms.  Throughput: 670.70 iter/sec.
Timings for 3360K FFT length (16 cores, 2 workers):  0.90,  1.50 ms.  Throughput: 1773.82 iter/sec.
Timings for 3360K FFT length (16 cores, 4 workers):  3.56,  3.95,  3.54,  4.60 ms.  Throughput: 1033.55 iter/sec.
[Sat Nov  5 16:18:03 2022]
Timings for 3360K FFT length (16 cores hyperthreaded, 1 worker):  1.42 ms.  Throughput: 702.06 iter/sec.
Timings for 3360K FFT length (16 cores hyperthreaded, 2 workers):  0.88,  0.95 ms.  Throughput: 2183.74 iter/sec.
Timings for 3360K FFT length (16 cores hyperthreaded, 4 workers):  4.63,  4.63,  5.09,  5.08 ms.  Throughput: 824.67 iter/sec.
Timings for 3456K FFT length (16 cores, 1 worker):  1.51 ms.  Throughput: 660.37 iter/sec.
Timings for 3456K FFT length (16 cores, 2 workers):  0.90,  1.50 ms.  Throughput: 1775.85 iter/sec.
Timings for 3456K FFT length (16 cores, 4 workers):  4.03,  4.02,  3.83,  4.90 ms.  Throughput: 962.25 iter/sec.
Timings for 3456K FFT length (16 cores hyperthreaded, 1 worker):  1.39 ms.  Throughput: 719.22 iter/sec.
Timings for 3456K FFT length (16 cores hyperthreaded, 2 workers):  0.89,  0.95 ms.  Throughput: 2178.03 iter/sec.
Timings for 3456K FFT length (16 cores hyperthreaded, 4 workers):  5.09,  5.09,  5.43,  5.32 ms.  Throughput: 765.32 iter/sec.
Timings for 3600K FFT length (16 cores, 1 worker):  1.54 ms.  Throughput: 647.94 iter/sec.
Timings for 3600K FFT length (16 cores, 2 workers):  0.99,  1.53 ms.  Throughput: 1665.93 iter/sec.
Timings for 3600K FFT length (16 cores, 4 workers):  4.11,  4.19,  4.54,  5.06 ms.  Throughput: 899.59 iter/sec.
Timings for 3600K FFT length (16 cores hyperthreaded, 1 worker):  2.07 ms.  Throughput: 482.67 iter/sec.
Timings for 3600K FFT length (16 cores hyperthreaded, 2 workers):  0.99,  1.00 ms.  Throughput: 2006.25 iter/sec.
Timings for 3600K FFT length (16 cores hyperthreaded, 4 workers):  5.87,  5.45,  6.13,  5.75 ms.  Throughput: 691.16 iter/sec.
Timings for 3840K FFT length (16 cores, 1 worker):  1.61 ms.  Throughput: 620.90 iter/sec.
Timings for 3840K FFT length (16 cores, 2 workers):  1.05,  1.74 ms.  Throughput: 1529.15 iter/sec.
Timings for 3840K FFT length (16 cores, 4 workers):  4.98,  4.63,  5.42,  5.38 ms.  Throughput: 786.98 iter/sec.
Timings for 3840K FFT length (16 cores hyperthreaded, 1 worker):  1.52 ms.  Throughput: 659.99 iter/sec.
Timings for 3840K FFT length (16 cores hyperthreaded, 2 workers):  1.10,  1.10 ms.  Throughput: 1825.15 iter/sec.
[Sat Nov  5 16:23:07 2022]
Timings for 3840K FFT length (16 cores hyperthreaded, 4 workers):  5.89,  6.13,  5.90,  6.76 ms.  Throughput: 649.93 iter/sec.
Timings for 3920K FFT length (16 cores, 1 worker):  1.68 ms.  Throughput: 593.53 iter/sec.
Timings for 3920K FFT length (16 cores, 2 workers):  1.14,  1.75 ms.  Throughput: 1450.41 iter/sec.
Timings for 3920K FFT length (16 cores, 4 workers):  4.68,  5.49,  6.29,  5.92 ms.  Throughput: 723.81 iter/sec.
Timings for 3920K FFT length (16 cores hyperthreaded, 1 worker):  1.57 ms.  Throughput: 636.32 iter/sec.
Timings for 3920K FFT length (16 cores hyperthreaded, 2 workers):  1.28,  1.34 ms.  Throughput: 1529.56 iter/sec.
Timings for 3920K FFT length (16 cores hyperthreaded, 4 workers):  6.94,  6.93,  7.30,  7.55 ms.  Throughput: 557.85 iter/sec.
Timings for 4032K FFT length (16 cores, 1 worker):  1.70 ms.  Throughput: 587.71 iter/sec.
Timings for 4032K FFT length (16 cores, 2 workers):  1.08,  1.79 ms.  Throughput: 1480.94 iter/sec.
Timings for 4032K FFT length (16 cores, 4 workers):  5.22,  5.22,  5.69,  5.68 ms.  Throughput: 735.29 iter/sec.
Timings for 4032K FFT length (16 cores hyperthreaded, 1 worker):  1.57 ms.  Throughput: 635.54 iter/sec.
Timings for 4032K FFT length (16 cores hyperthreaded, 2 workers):  1.18,  1.18 ms.  Throughput: 1698.24 iter/sec.
Timings for 4032K FFT length (16 cores hyperthreaded, 4 workers):  6.80,  5.33,  6.03,  7.28 ms.  Throughput: 637.70 iter/sec.
Timings for 4200K FFT length (16 cores, 1 worker):  1.71 ms.  Throughput: 584.68 iter/sec.
Timings for 4200K FFT length (16 cores, 2 workers):  1.20,  1.95 ms.  Throughput: 1348.96 iter/sec.
Timings for 4200K FFT length (16 cores, 4 workers):  5.17,  5.84,  5.81,  6.40 ms.  Throughput: 692.98 iter/sec.
Timings for 4200K FFT length (16 cores hyperthreaded, 1 worker):  1.64 ms.  Throughput: 609.45 iter/sec.
Timings for 4200K FFT length (16 cores hyperthreaded, 2 workers):  1.24,  1.34 ms.  Throughput: 1554.74 iter/sec.
Timings for 4200K FFT length (16 cores hyperthreaded, 4 workers):  6.88,  6.77,  7.99,  7.08 ms.  Throughput: 559.50 iter/sec.
Timings for 4320K FFT length (16 cores, 1 worker):  1.78 ms.  Throughput: 562.52 iter/sec.
[Sat Nov  5 16:28:12 2022]
Timings for 4320K FFT length (16 cores, 2 workers):  1.24,  1.89 ms.  Throughput: 1336.18 iter/sec.
Timings for 4320K FFT length (16 cores, 4 workers):  6.25,  5.54,  6.46,  7.05 ms.  Throughput: 636.84 iter/sec.
Timings for 4320K FFT length (16 cores hyperthreaded, 1 worker):  2.07 ms.  Throughput: 483.47 iter/sec.
Timings for 4320K FFT length (16 cores hyperthreaded, 2 workers):  1.63,  1.53 ms.  Throughput: 1266.72 iter/sec.
Timings for 4320K FFT length (16 cores hyperthreaded, 4 workers):  7.17,  8.37,  8.53,  8.36 ms.  Throughput: 495.76 iter/sec.
Timings for 4480K FFT length (16 cores, 1 worker):  1.82 ms.  Throughput: 549.43 iter/sec.
Timings for 4480K FFT length (16 cores, 2 workers):  1.34,  2.07 ms.  Throughput: 1228.55 iter/sec.
Timings for 4480K FFT length (16 cores, 4 workers):  6.16,  6.49,  7.42,  6.97 ms.  Throughput: 594.85 iter/sec.
Timings for 4480K FFT length (16 cores hyperthreaded, 1 worker):  1.77 ms.  Throughput: 563.76 iter/sec.
Timings for 4480K FFT length (16 cores hyperthreaded, 2 workers):  1.60,  1.68 ms.  Throughput: 1219.73 iter/sec.
Timings for 4480K FFT length (16 cores hyperthreaded, 4 workers):  8.03,  8.03,  8.90,  8.01 ms.  Throughput: 486.12 iter/sec.
Timings for 4608K FFT length (16 cores, 1 worker):  1.96 ms.  Throughput: 510.46 iter/sec.
Timings for 4608K FFT length (16 cores, 2 workers):  1.47,  2.15 ms.  Throughput: 1143.96 iter/sec.
Timings for 4608K FFT length (16 cores, 4 workers):  7.29,  7.28,  8.34,  8.00 ms.  Throughput: 519.64 iter/sec.
Timings for 4608K FFT length (16 cores hyperthreaded, 1 worker):  1.81 ms.  Throughput: 551.20 iter/sec.
Timings for 4608K FFT length (16 cores hyperthreaded, 2 workers):  1.86,  2.07 ms.  Throughput: 1020.56 iter/sec.
Timings for 4608K FFT length (16 cores hyperthreaded, 4 workers):  8.84,  8.84,  9.51,  9.33 ms.  Throughput: 438.52 iter/sec.
Timings for 4704K FFT length (16 cores, 1 worker):  1.96 ms.  Throughput: 511.05 iter/sec.
Timings for 4704K FFT length (16 cores, 2 workers):  1.60,  2.30 ms.  Throughput: 1059.00 iter/sec.
Timings for 4704K FFT length (16 cores, 4 workers):  7.50,  7.29,  7.33,  8.82 ms.  Throughput: 520.39 iter/sec.
[Sat Nov  5 16:33:17 2022]
Timings for 4704K FFT length (16 cores hyperthreaded, 1 worker):  1.87 ms.  Throughput: 535.41 iter/sec.
Timings for 4704K FFT length (16 cores hyperthreaded, 2 workers):  2.08,  2.14 ms.  Throughput: 948.74 iter/sec.
Timings for 4704K FFT length (16 cores hyperthreaded, 4 workers):  9.11,  9.11,  9.69,  9.70 ms.  Throughput: 425.68 iter/sec.
Timings for 4800K FFT length (16 cores, 1 worker):  2.00 ms.  Throughput: 500.08 iter/sec.
Timings for 4800K FFT length (16 cores, 2 workers):  1.57,  2.33 ms.  Throughput: 1064.40 iter/sec.
Timings for 4800K FFT length (16 cores, 4 workers):  7.27,  7.34,  8.84,  9.34 ms.  Throughput: 494.04 iter/sec.
Timings for 4800K FFT length (16 cores hyperthreaded, 1 worker):  2.22 ms.  Throughput: 451.43 iter/sec.
Timings for 4800K FFT length (16 cores hyperthreaded, 2 workers):  2.50,  2.64 ms.  Throughput: 778.56 iter/sec.
Timings for 4800K FFT length (16 cores hyperthreaded, 4 workers):  8.67,  9.35,  9.79,  9.78 ms.  Throughput: 426.58 iter/sec.
Timings for 5040K FFT length (16 cores, 1 worker):  2.07 ms.  Throughput: 482.72 iter/sec.
Timings for 5040K FFT length (16 cores, 2 workers):  1.64,  2.33 ms.  Throughput: 1040.32 iter/sec.
Timings for 5040K FFT length (16 cores, 4 workers):  7.29,  7.44,  8.08,  8.46 ms.  Throughput: 513.73 iter/sec.
Timings for 5040K FFT length (16 cores hyperthreaded, 1 worker):  1.88 ms.  Throughput: 531.50 iter/sec.
Timings for 5040K FFT length (16 cores hyperthreaded, 2 workers):  2.19,  2.19 ms.  Throughput: 912.90 iter/sec.
Timings for 5040K FFT length (16 cores hyperthreaded, 4 workers):  8.36,  8.74,  9.32,  9.53 ms.  Throughput: 446.41 iter/sec.
Timings for 5120K FFT length (16 cores, 1 worker):  2.08 ms.  Throughput: 480.83 iter/sec.
Timings for 5120K FFT length (16 cores, 2 workers):  1.77,  2.47 ms.  Throughput: 972.01 iter/sec.
Timings for 5120K FFT length (16 cores, 4 workers):  7.74,  7.75,  8.70,  8.41 ms.  Throughput: 491.88 iter/sec.
Timings for 5120K FFT length (16 cores hyperthreaded, 1 worker):  1.93 ms.  Throughput: 518.52 iter/sec.
Timings for 5120K FFT length (16 cores hyperthreaded, 2 workers):  2.34,  2.54 ms.  Throughput: 821.51 iter/sec.
[Sat Nov  5 16:38:23 2022]
Timings for 5120K FFT length (16 cores hyperthreaded, 4 workers):  9.48,  9.55,  9.91,  9.98 ms.  Throughput: 411.31 iter/sec.
Timings for 5184K FFT length (16 cores, 1 worker):  2.08 ms.  Throughput: 479.84 iter/sec.
Timings for 5184K FFT length (16 cores, 2 workers):  1.79,  2.50 ms.  Throughput: 960.09 iter/sec.
Timings for 5184K FFT length (16 cores, 4 workers):  8.17,  7.41,  9.13,  8.52 ms.  Throughput: 484.27 iter/sec.
Timings for 5184K FFT length (16 cores hyperthreaded, 1 worker):  1.99 ms.  Throughput: 501.42 iter/sec.
Timings for 5184K FFT length (16 cores hyperthreaded, 2 workers):  2.54,  2.61 ms.  Throughput: 776.89 iter/sec.
Timings for 5184K FFT length (16 cores hyperthreaded, 4 workers):  9.93,  9.94, 10.11, 10.77 ms.  Throughput: 393.10 iter/sec.
Timings for 5376K FFT length (16 cores, 1 worker):  2.19 ms.  Throughput: 457.19 iter/sec.
Timings for 5376K FFT length (16 cores, 2 workers):  2.05,  2.64 ms.  Throughput: 866.45 iter/sec.
Timings for 5376K FFT length (16 cores, 4 workers):  8.50,  8.20,  9.73,  9.38 ms.  Throughput: 449.03 iter/sec.
Timings for 5376K FFT length (16 cores hyperthreaded, 1 worker):  2.03 ms.  Throughput: 493.55 iter/sec.
Timings for 5376K FFT length (16 cores hyperthreaded, 2 workers):  2.68,  2.93 ms.  Throughput: 714.87 iter/sec.
Timings for 5376K FFT length (16 cores hyperthreaded, 4 workers): 10.44, 10.31, 10.89, 10.82 ms.  Throughput: 376.99 iter/sec.
Timings for 5760K FFT length (16 cores, 1 worker):  2.32 ms.  Throughput: 430.38 iter/sec.
Timings for 5760K FFT length (16 cores, 2 workers):  2.67,  3.43 ms.  Throughput: 666.37 iter/sec.
Timings for 5760K FFT length (16 cores, 4 workers):  9.55,  9.62, 11.04, 11.20 ms.  Throughput: 388.50 iter/sec.
Timings for 5760K FFT length (16 cores hyperthreaded, 1 worker):  2.45 ms.  Throughput: 408.20 iter/sec.
Timings for 5760K FFT length (16 cores hyperthreaded, 2 workers):  4.12,  4.27 ms.  Throughput: 476.65 iter/sec.
Timings for 5760K FFT length (16 cores hyperthreaded, 4 workers): 11.40, 11.40, 12.23, 12.23 ms.  Throughput: 339.01 iter/sec.
Timings for 6048K FFT length (16 cores, 1 worker):  2.43 ms.  Throughput: 412.02 iter/sec.
[Sat Nov  5 16:43:28 2022]
Timings for 6048K FFT length (16 cores, 2 workers):  2.83,  3.52 ms.  Throughput: 638.26 iter/sec.
Timings for 6048K FFT length (16 cores, 4 workers): 11.13,  8.64, 11.24, 11.77 ms.  Throughput: 379.54 iter/sec.
Timings for 6048K FFT length (16 cores hyperthreaded, 1 worker):  2.26 ms.  Throughput: 442.74 iter/sec.
Timings for 6048K FFT length (16 cores hyperthreaded, 2 workers):  3.76,  4.12 ms.  Throughput: 508.68 iter/sec.
Timings for 6048K FFT length (16 cores hyperthreaded, 4 workers): 12.89, 11.67, 13.50, 12.88 ms.  Throughput: 314.96 iter/sec.
Timings for 6144K FFT length (16 cores, 1 worker):  2.40 ms.  Throughput: 416.85 iter/sec.
Timings for 6144K FFT length (16 cores, 2 workers):  3.03,  3.55 ms.  Throughput: 611.45 iter/sec.
Timings for 6144K FFT length (16 cores, 4 workers):  8.44, 11.80, 11.57, 10.91 ms.  Throughput: 381.29 iter/sec.
Timings for 6144K FFT length (16 cores hyperthreaded, 1 worker):  2.26 ms.  Throughput: 443.23 iter/sec.
Timings for 6144K FFT length (16 cores hyperthreaded, 2 workers):  3.83,  4.02 ms.  Throughput: 509.68 iter/sec.
Timings for 6144K FFT length (16 cores hyperthreaded, 4 workers): 12.19, 12.25, 13.13, 13.13 ms.  Throughput: 316.02 iter/sec.
Timings for 6272K FFT length (16 cores, 1 worker):  2.46 ms.  Throughput: 406.06 iter/sec.
Timings for 6272K FFT length (16 cores, 2 workers):  3.16,  3.53 ms.  Throughput: 599.82 iter/sec.
Timings for 6272K FFT length (16 cores, 4 workers): 10.57,  9.85, 11.21, 11.29 ms.  Throughput: 373.92 iter/sec.
Timings for 6272K FFT length (16 cores hyperthreaded, 1 worker):  2.30 ms.  Throughput: 434.02 iter/sec.
Timings for 6272K FFT length (16 cores hyperthreaded, 2 workers):  3.78,  4.03 ms.  Throughput: 512.48 iter/sec.
Timings for 6272K FFT length (16 cores hyperthreaded, 4 workers): 12.36, 11.85, 13.32, 13.30 ms.  Throughput: 315.51 iter/sec.
Timings for 6400K FFT length (16 cores, 1 worker):  2.53 ms.  Throughput: 395.42 iter/sec.
Timings for 6400K FFT length (16 cores, 2 workers):  3.53,  4.29 ms.  Throughput: 516.50 iter/sec.
Timings for 6400K FFT length (16 cores, 4 workers): 11.31, 11.30, 13.42, 12.72 ms.  Throughput: 330.04 iter/sec.
[Sat Nov  5 16:48:36 2022]
Timings for 6400K FFT length (16 cores hyperthreaded, 1 worker):  2.39 ms.  Throughput: 418.77 iter/sec.
Timings for 6400K FFT length (16 cores hyperthreaded, 2 workers):  4.94,  5.43 ms.  Throughput: 386.59 iter/sec.
Timings for 6400K FFT length (16 cores hyperthreaded, 4 workers): 12.95, 12.97, 13.89, 14.11 ms.  Throughput: 297.14 iter/sec.
Timings for 6720K FFT length (16 cores, 1 worker):  2.61 ms.  Throughput: 383.13 iter/sec.
Timings for 6720K FFT length (16 cores, 2 workers):  3.85,  4.83 ms.  Throughput: 466.34 iter/sec.
Timings for 6720K FFT length (16 cores, 4 workers): 11.57, 11.79, 13.90, 14.12 ms.  Throughput: 314.01 iter/sec.
Timings for 6720K FFT length (16 cores hyperthreaded, 1 worker):  2.48 ms.  Throughput: 402.55 iter/sec.
Timings for 6720K FFT length (16 cores hyperthreaded, 2 workers):  5.08,  5.57 ms.  Throughput: 376.40 iter/sec.
Timings for 6720K FFT length (16 cores hyperthreaded, 4 workers): 13.56, 13.57, 14.90, 14.33 ms.  Throughput: 284.37 iter/sec.
Timings for 7056K FFT length (16 cores, 1 worker):  2.77 ms.  Throughput: 360.57 iter/sec.
Timings for 7056K FFT length (16 cores, 2 workers):  4.16,  5.31 ms.  Throughput: 428.79 iter/sec.
Timings for 7056K FFT length (16 cores, 4 workers): 12.74, 12.08, 14.75, 14.31 ms.  Throughput: 299.00 iter/sec.
Timings for 7056K FFT length (16 cores hyperthreaded, 1 worker):  2.59 ms.  Throughput: 385.58 iter/sec.
Timings for 7056K FFT length (16 cores hyperthreaded, 2 workers):  5.87,  6.01 ms.  Throughput: 336.67 iter/sec.
Timings for 7056K FFT length (16 cores hyperthreaded, 4 workers): 14.14, 14.11, 15.19, 15.19 ms.  Throughput: 273.22 iter/sec.
Timings for 7168K FFT length (16 cores, 1 worker):  2.65 ms.  Throughput: 378.04 iter/sec.
Timings for 7168K FFT length (16 cores, 2 workers):  4.25,  4.89 ms.  Throughput: 440.02 iter/sec.
Timings for 7168K FFT length (16 cores, 4 workers): 12.50, 12.21, 13.92, 14.17 ms.  Throughput: 304.37 iter/sec.
Timings for 7168K FFT length (16 cores hyperthreaded, 1 worker):  2.53 ms.  Throughput: 395.67 iter/sec.
Timings for 7168K FFT length (16 cores hyperthreaded, 2 workers):  5.26,  5.52 ms.  Throughput: 371.04 iter/sec.
[Sat Nov  5 16:53:44 2022]
Timings for 7168K FFT length (16 cores hyperthreaded, 4 workers): 14.59, 14.43, 15.78, 15.53 ms.  Throughput: 265.60 iter/sec.
Timings for 7200K FFT length (16 cores, 1 worker):  2.68 ms.  Throughput: 373.06 iter/sec.
Timings for 7200K FFT length (16 cores, 2 workers):  4.09,  4.66 ms.  Throughput: 458.94 iter/sec.
Timings for 7200K FFT length (16 cores, 4 workers): 12.14, 12.14, 13.34, 13.37 ms.  Throughput: 314.49 iter/sec.
Timings for 7200K FFT length (16 cores hyperthreaded, 1 worker):  2.55 ms.  Throughput: 391.66 iter/sec.
Timings for 7200K FFT length (16 cores hyperthreaded, 2 workers):  4.96,  5.27 ms.  Throughput: 391.33 iter/sec.
Timings for 7200K FFT length (16 cores hyperthreaded, 4 workers): 13.82, 14.16, 15.06, 15.29 ms.  Throughput: 274.74 iter/sec.
Timings for 7680K FFT length (16 cores, 1 worker):  2.79 ms.  Throughput: 358.54 iter/sec.
Timings for 7680K FFT length (16 cores, 2 workers):  4.83,  5.31 ms.  Throughput: 395.24 iter/sec.
Timings for 7680K FFT length (16 cores, 4 workers): 13.39, 13.40, 14.73, 14.47 ms.  Throughput: 286.27 iter/sec.
Timings for 7680K FFT length (16 cores hyperthreaded, 1 worker):  2.67 ms.  Throughput: 374.96 iter/sec.
Timings for 7680K FFT length (16 cores hyperthreaded, 2 workers):  5.49,  5.86 ms.  Throughput: 352.99 iter/sec.
Timings for 7680K FFT length (16 cores hyperthreaded, 4 workers): 14.56, 15.05, 15.90, 15.80 ms.  Throughput: 261.30 iter/sec.
Timings for 8064K FFT length (16 cores, 1 worker):  3.05 ms.  Throughput: 327.56 iter/sec.
Timings for 8064K FFT length (16 cores, 2 workers):  5.70,  7.33 ms.  Throughput: 311.76 iter/sec.
Timings for 8064K FFT length (16 cores, 4 workers): 15.40, 15.00, 17.60, 17.06 ms.  Throughput: 247.03 iter/sec.
Timings for 8064K FFT length (16 cores hyperthreaded, 1 worker):  3.09 ms.  Throughput: 323.68 iter/sec.
Timings for 8064K FFT length (16 cores hyperthreaded, 2 workers):  7.51,  8.04 ms.  Throughput: 257.59 iter/sec.
Timings for 8064K FFT length (16 cores hyperthreaded, 4 workers): 16.36, 16.95, 17.83, 17.92 ms.  Throughput: 232.03 iter/sec.

preda · 2022-11-06, 07:48

Quote:

Originally Posted by scan80269

Ryzen 9 7950X PBO enabled and 2x16GB DDR5-6000 30-38-38-96 memory

Code:

Timings for 7200K FFT length (16 cores, 1 worker):  2.68 ms.  Throughput: 373.06 iter/sec.
Timings for 7200K FFT length (16 cores, 2 workers):  4.09,  4.66 ms.  Throughput: 458.94 iter/sec.
Timings for 7680K FFT length (16 cores, 1 worker):  2.79 ms.  Throughput: 358.54 iter/sec.
Timings for 7680K FFT length (16 cores, 2 workers):  4.83,  5.31 ms.  Throughput: 395.24 iter/sec.
Timings for 8064K FFT length (16 cores, 1 worker):  3.05 ms.  Throughput: 327.56 iter/sec.
Timings for 8064K FFT length (16 cores, 2 workers):  5.70,  7.33 ms.  Throughput: 311.76 iter/sec.

So it seems this is becoming competitive with (i.e. catching up with) Skylake-X/CascadeLake-X (x299, LGA2066) from like 7years ago (although Skylake-X supports 4-channel DDR4 RAM, while this is 2-channel DDR5).

Anyway probably this is the way forward (given that Intel seems to have given up on AVX-512, just when they had the leading implementation 7years ahead of AMD). Need to research motherboards now, to see how many PCIe slots, and how many DIMMs they support, and at what price.

Prime95 · 2022-11-06, 18:37

Quote:

Originally Posted by scan80269

Ryzen 9 7950X PBO enabled and 2x16GB DDR5-6000 30-38-38-96 memory

A big ask. Can you run an all FFT implementations benchmark on 16 cores, 2 workers, no hyperthreading for all FFT sizes from 1024K to the maximum 65536K? Email the results to me and I can make your fastest FFT implementations the default for Ryzen CPUs. For extra credit, repeat for all-complex FFTs.

If that goes well, we can do the same for small FFT sizes - likely with a different number of workers.

BTW what is PBO? Running at stock memory speeds and other settings would probably be best for setting defaults.

R. Gerbicz · 2022-11-06, 18:47

Quote:

Originally Posted by Prime95

BTW what is PBO? Running at stock memory speeds and other settings would probably be best for setting defaults.

Precision Boost Overdrive, see:
https://www.pugetsystems.com/labs/ar...formance-2373/

scan80269 · 2022-11-12, 18:10

Quote:

Originally Posted by Prime95

A big ask. Can you run an all FFT implementations benchmark on 16 cores, 2 workers, no hyperthreading for all FFT sizes from 1024K to the maximum 65536K? Email the results to me and I can make your fastest FFT implementations the default for Ryzen CPUs. For extra credit, repeat for all-complex FFTs.

If that goes well, we can do the same for small FFT sizes - likely with a different number of workers.

BTW what is PBO? Running at stock memory speeds and other settings would probably be best for setting defaults.

I need to finish building up a new tower with this 7950X CPU. It should be ready to start FFT benchmarks in a day or two.

One question. Should the FFT benchmark be run with DDR5 memory using AMD EXPO profile (6000, 30-38-38-96, 1.35V) or standard JEDEC profile (4800, 40-40-40-77, 1.1V)?

Prime95 · 2022-11-12, 19:58

Quote:

Originally Posted by scan80269

Should the FFT benchmark be run with DDR5 memory using AMD EXPO profile (6000, 30-38-38-96, 1.35V) or standard JEDEC profile (4800, 40-40-40-77, 1.1V)?

Since this will be used to select default FFTs for many years to come, let's go with the faster settings.

axn · 2022-11-13, 02:20

Quote:

Originally Posted by scan80269

One question. Should the FFT benchmark be run with DDR5 memory using AMD EXPO profile (6000, 30-38-38-96, 1.35V) or standard JEDEC profile (4800, 40-40-40-77, 1.1V)?

Quote:

Originally Posted by Prime95

Since this will be used to select default FFTs for many years to come, let's go with the faster settings.

I'd say, run the full bench with max setting, but also run a selected subset (say 2k-8k) using the slower setting as well, for "scientific purposes"

scan80269 · 2022-11-15, 06:35

FFT benchmark results for Ryzen 9 7950X have been sent to George.

5 completed runs of 1000K to 65536K FFTs:

With 2x16GB DDR5 6000 30-38-38-96 memory (AMD EXPO profile):
1. 7950X standard (no PBO no CPB) - FFT
2. 7950X standard (no PBO no CPB) - all-complex FFT
3. 7950X PBO & CPB enabled - FFT
4. 7950X PBO & CPB enabled - all-complex FFT

With 2x16GB DDR5 4800 40-40-40-77 memory (JEDEC profile)
5. 7950X PBO & CPB enabled - FFT

PBO = Precision Boost Overdrive, CPB = Core Performance Boost

RadosHordosCZ · 2022-11-27, 13:51

CPU speed: 4750 MHz, 16 cores, No SMT, 1,05V PBO
RAM 5600 MHZ CL34-tRCD40-tRP40-tRAS80-tCR125 1,25V
Computer: 160W

FFTlen=6000K, Type=3, Arch=8, Pass1=960, Pass2=6400, clm=4 (16 cores, 2 workers): 3.56, 3.77 ms. Throughput: 546.25 iter/sec.
FFTlen=6000K, Type=3, Arch=8, Pass1=960, Pass2=6400, clm=2 (16 cores, 2 workers): 2.86, 3.13 ms. Throughput: 669.08 iter/sec.
FFTlen=6000K, Type=3, Arch=8, Pass1=960, Pass2=6400, clm=1 (16 cores, 2 workers): 2.97, 3.26 ms. Throughput: 644.02 iter/sec.
FFTlen=6000K, Type=3, Arch=8, Pass1=1920, Pass2=3200, clm=2 (16 cores, 2 workers): 3.47, 3.88 ms. Throughput: 546.13 iter/sec.
FFTlen=6000K, Type=3, Arch=8, Pass1=1920, Pass2=3200, clm=1 (16 cores, 2 workers): 2.86, 3.04 ms. Throughput: 678.76 iter/sec.
FFTlen=6048K, Type=3, Arch=8, Pass1=1152, Pass2=5376, clm=2 (16 cores, 2 workers): 3.04, 3.45 ms. Throughput: 618.69 iter/sec.
FFTlen=6048K, Type=3, Arch=8, Pass1=1152, Pass2=5376, clm=1 (16 cores, 2 workers): 2.89, 3.09 ms. Throughput: 669.70 iter/sec.
FFTlen=6048K, Type=3, Arch=8, Pass1=1344, Pass2=4608, clm=2 (16 cores, 2 workers): 3.10, 3.39 ms. Throughput: 617.80 iter/sec.
FFTlen=6048K, Type=3, Arch=8, Pass1=1344, Pass2=4608, clm=1 (16 cores, 2 workers): 2.80, 3.12 ms. Throughput: 676.99 iter/sec.
FFTlen=6048K, Type=3, Arch=8, Pass1=2304, Pass2=2688, clm=1 (16 cores, 2 workers): 3.01, 3.15 ms. Throughput: 649.93 iter/sec.
FFTlen=6144K, Type=3, Arch=8, Pass1=192, Pass2=32768, clm=4 (16 cores, 2 workers): 4.02, 4.64 ms. Throughput: 464.68 iter/sec.
FFTlen=6144K, Type=3, Arch=8, Pass1=192, Pass2=32768, clm=2 (16 cores, 2 workers): 4.17, 4.25 ms. Throughput: 475.22 iter/sec.
FFTlen=6144K, Type=3, Arch=8, Pass1=192, Pass2=32768, clm=1 (16 cores, 2 workers): 4.05, 4.74 ms. Throughput: 458.02 iter/sec.
FFTlen=6144K, Type=3, Arch=8, Pass1=768, Pass2=8192, clm=4 (16 cores, 2 workers): 3.58, 3.88 ms. Throughput: 536.91 iter/sec.
FFTlen=6144K, Type=3, Arch=8, Pass1=768, Pass2=8192, clm=2 (16 cores, 2 workers): 3.20, 3.41 ms. Throughput: 606.60 iter/sec.
FFTlen=6144K, Type=3, Arch=8, Pass1=768, Pass2=8192, clm=1 (16 cores, 2 workers): 3.20, 3.59 ms. Throughput: 590.62 iter/sec.
FFTlen=6144K, Type=3, Arch=8, Pass1=1024, Pass2=6144, clm=2 (16 cores, 2 workers): 3.16, 3.58 ms. Throughput: 595.59 iter/sec.
FFTlen=6144K, Type=3, Arch=8, Pass1=1024, Pass2=6144, clm=1 (16 cores, 2 workers): 3.10, 3.47 ms. Throughput: 610.75 iter/sec.
FFTlen=6144K, Type=3, Arch=8, Pass1=1536, Pass2=4096, clm=2 (16 cores, 2 workers): 3.42, 3.92 ms. Throughput: 547.91 iter/sec.
FFTlen=6144K, Type=3, Arch=8, Pass1=1536, Pass2=4096, clm=1 (16 cores, 2 workers): 3.09, 3.49 ms. Throughput: 610.75 iter/sec.
FFTlen=6144K, Type=3, Arch=8, Pass1=2048, Pass2=3072, clm=1 (16 cores, 2 workers): 3.25, 3.67 ms. Throughput: 580.02 iter/sec.

https://www.mersenne.org/report_prp/...osHordosCZ&B1=

James Heinrich · 2022-12-17, 00:17

After playing with my 7950X for a bit I came to two conclusions:

1) Getting workers aligned to chiplets is vital. Running 16-thread PRP across 2 chiplets is actually 40% slower than just running 8 threads on one chiplet.

2) You can save a fair amount of power and heat and not lose much performance. I actually dropped the thermal limit from 90°C to 70°C in the BIOS. My PRP iteration times are perhaps 1% slower, but CPU power consumption is down from 235W to 195W, temperature down from 92C to 72C, and (important to me) the fan noise is down from very-noticeable to barely-there. For me at least it's a hugely worthwhile tradeoff.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Zen4's AVX512 Teardown	Mysticial	Hardware	21	2022-11-07 01:10
AVX512 and Zen4 pre-release speculations	Xyzzy	y-cruncher	12	2022-09-15 17:15
RPS benchmarks	pinhodecarlos	Riesel Prime Search	29	2014-12-07 07:13
Benchmarks for 24.12	Prime95	Software	60	2005-06-11 07:35
Benchmarks	Vandy	Hardware	6	2002-10-28 13:45

2022-11-15, 06:35	#20
scan80269 "Sam" Jun 2019 California, USA 2×19 Posts	FFT benchmark results for Ryzen 9 7950X have been sent to George. 5 completed runs of 1000K to 65536K FFTs: With 2x16GB DDR5 6000 30-38-38-96 memory (AMD EXPO profile): 1. 7950X standard (no PBO no CPB) - FFT 2. 7950X standard (no PBO no CPB) - all-complex FFT 3. 7950X PBO & CPB enabled - FFT 4. 7950X PBO & CPB enabled - all-complex FFT With 2x16GB DDR5 4800 40-40-40-77 memory (JEDEC profile) 5. 7950X PBO & CPB enabled - FFT PBO = Precision Boost Overdrive, CPB = Core Performance Boost

2022-11-27, 13:51	#21
RadosHordosCZ Nov 2022 1 Posts	My CPU speed: 4750 MHz, 16 cores, No SMT, 1,05V PBO RAM 5600 MHZ CL34-tRCD40-tRP40-tRAS80-tCR125 1,25V Computer: 160W FFTlen=6000K, Type=3, Arch=8, Pass1=960, Pass2=6400, clm=4 (16 cores, 2 workers): 3.56, 3.77 ms. Throughput: 546.25 iter/sec. FFTlen=6000K, Type=3, Arch=8, Pass1=960, Pass2=6400, clm=2 (16 cores, 2 workers): 2.86, 3.13 ms. Throughput: 669.08 iter/sec. FFTlen=6000K, Type=3, Arch=8, Pass1=960, Pass2=6400, clm=1 (16 cores, 2 workers): 2.97, 3.26 ms. Throughput: 644.02 iter/sec. FFTlen=6000K, Type=3, Arch=8, Pass1=1920, Pass2=3200, clm=2 (16 cores, 2 workers): 3.47, 3.88 ms. Throughput: 546.13 iter/sec. FFTlen=6000K, Type=3, Arch=8, Pass1=1920, Pass2=3200, clm=1 (16 cores, 2 workers): 2.86, 3.04 ms. Throughput: 678.76 iter/sec. FFTlen=6048K, Type=3, Arch=8, Pass1=1152, Pass2=5376, clm=2 (16 cores, 2 workers): 3.04, 3.45 ms. Throughput: 618.69 iter/sec. FFTlen=6048K, Type=3, Arch=8, Pass1=1152, Pass2=5376, clm=1 (16 cores, 2 workers): 2.89, 3.09 ms. Throughput: 669.70 iter/sec. FFTlen=6048K, Type=3, Arch=8, Pass1=1344, Pass2=4608, clm=2 (16 cores, 2 workers): 3.10, 3.39 ms. Throughput: 617.80 iter/sec. FFTlen=6048K, Type=3, Arch=8, Pass1=1344, Pass2=4608, clm=1 (16 cores, 2 workers): 2.80, 3.12 ms. Throughput: 676.99 iter/sec. FFTlen=6048K, Type=3, Arch=8, Pass1=2304, Pass2=2688, clm=1 (16 cores, 2 workers): 3.01, 3.15 ms. Throughput: 649.93 iter/sec. FFTlen=6144K, Type=3, Arch=8, Pass1=192, Pass2=32768, clm=4 (16 cores, 2 workers): 4.02, 4.64 ms. Throughput: 464.68 iter/sec. FFTlen=6144K, Type=3, Arch=8, Pass1=192, Pass2=32768, clm=2 (16 cores, 2 workers): 4.17, 4.25 ms. Throughput: 475.22 iter/sec. FFTlen=6144K, Type=3, Arch=8, Pass1=192, Pass2=32768, clm=1 (16 cores, 2 workers): 4.05, 4.74 ms. Throughput: 458.02 iter/sec. FFTlen=6144K, Type=3, Arch=8, Pass1=768, Pass2=8192, clm=4 (16 cores, 2 workers): 3.58, 3.88 ms. Throughput: 536.91 iter/sec. FFTlen=6144K, Type=3, Arch=8, Pass1=768, Pass2=8192, clm=2 (16 cores, 2 workers): 3.20, 3.41 ms. Throughput: 606.60 iter/sec. FFTlen=6144K, Type=3, Arch=8, Pass1=768, Pass2=8192, clm=1 (16 cores, 2 workers): 3.20, 3.59 ms. Throughput: 590.62 iter/sec. FFTlen=6144K, Type=3, Arch=8, Pass1=1024, Pass2=6144, clm=2 (16 cores, 2 workers): 3.16, 3.58 ms. Throughput: 595.59 iter/sec. FFTlen=6144K, Type=3, Arch=8, Pass1=1024, Pass2=6144, clm=1 (16 cores, 2 workers): 3.10, 3.47 ms. Throughput: 610.75 iter/sec. FFTlen=6144K, Type=3, Arch=8, Pass1=1536, Pass2=4096, clm=2 (16 cores, 2 workers): 3.42, 3.92 ms. Throughput: 547.91 iter/sec. FFTlen=6144K, Type=3, Arch=8, Pass1=1536, Pass2=4096, clm=1 (16 cores, 2 workers): 3.09, 3.49 ms. Throughput: 610.75 iter/sec. FFTlen=6144K, Type=3, Arch=8, Pass1=2048, Pass2=3072, clm=1 (16 cores, 2 workers): 3.25, 3.67 ms. Throughput: 580.02 iter/sec. https://www.mersenne.org/report_prp/...osHordosCZ&B1=

2022-12-17, 00:17	#22
James Heinrich "James Heinrich" May 2004 ex-Northern Ontario 3²·5·7·13 Posts	After playing with my 7950X for a bit I came to two conclusions: 1) Getting workers aligned to chiplets is vital. Running 16-thread PRP across 2 chiplets is actually 40% slower than just running 8 threads on one chiplet. 2) You can save a fair amount of power and heat and not lose much performance. I actually dropped the thermal limit from 90°C to 70°C in the BIOS. My PRP iteration times are perhaps 1% slower, but CPU power consumption is down from 235W to 195W, temperature down from 92C to 72C, and (important to me) the fan noise is down from very-noticeable to barely-there. For me at least it's a hugely worthwhile tradeoff.