AMD benchmarks NO LONGER needed for v26

[Prime95]

If you have an AMD K8 or K10 class machine and are running Linux, I need benchmarks (I only have a K8 machine running 64-bit Linux)

Download and install in a new directory either 32-bit mprime from ftp://mersenne.org/gimps/test_v262_mprime.tgz or 64-bit mprime from ftp://mersenne.org/gimps/test_v262_mprime64.tgz.

Create prime.txt with:

StressTester=1
MinBenchFFT=4
MaxBenchFFT=32768
OnlyBench5678=0
BenchAllComplex=1
AllBench=1

and create local.txt:

NumCPUs=1

K8 machines can set MaxBenchFFT to 6144 instead.

Then run Options/Benchmark. Post your results.txt file in this thread or email them to me.

Do not use this executable for production work.

Thanks!

[moebius]

AMD Phenom(tm) II X4 955 Processor 3213.85 MHz

[Kaboom]

AMD Opteron 185 Processor, 2.613 MHz

[Gimarel]

AMD Phenom(tm) II X6 1055T Processor Turbo Core enabled, i.e. 3300MHz for 1-3 threads, 2800Mhz for 4-6 threads.

[Prime95]

Ugh. The K10 results were not at all what I was hoping for. I could still use a 32-bit K8 benchmark.

To get v26 properly optimized for the K10 I'll need either remote access to a K10 machine or I'll need to buy one.

[mdettweiler]

Here's a benchmark for the 32-bit application running on a K8 (AMD Phenom(tm) 9750 Quad-Core Processor, to be exact) system. The operating system is Ubuntu 8.04 64-bit; does this still count as a 32-bit benchmark since I used the 32-bit binary?

Note that I stopped this in the middle of when it was doing the benchmarks for 3 threads. I set the prime.txt file exactly as shown in the first post (including the NumCPUs=1 line), but it still went on and did 2, then 3 threads, and probably would have done 4 if I hadn't stopped it when I did. I had only left two cores free for mprime before starting the benchmark anyway, figuring that would be plenty to ensure an unskewed single-thread benchmark result; the other two cores continued on their usual LLR work. Thus, you can ignore all data in the attached benchmarks past line 3557, at which it switched to 3 threads.

Is this sufficient for your purposes, or do you actually need the benchmarks up to and including 4 threads?

[moebius]

[quote=mdettweiler;228754]Here's a benchmark for the 32-bit application running on a K8 .[\quote]

I think the AMD Phenom 9750 quad-core processor is part of the K10 architecture and not off the K8 generation.
You would have to use MaxBenchFFT = 32768

[Prime95]

Originally Posted by mdettweiler

Here's a benchmark for the 32-bit application running on a K8 (AMD Phenom(tm) 9750 Quad-Core Processor, to be exact) system. The operating system is Ubuntu 8.04 64-bit; does this still count as a 32-bit benchmark since I used the 32-bit binary?

I set the prime.txt file exactly as shown in the first post (including the NumCPUs=1 line), but it still went on and did 2

I worried about that too (running 32-bit executable on 64-bit OS). I'm going to try it on my Opteron and hope it yields an accurate benchmark.

Sorry about the NumCPUs line. It belongs in local.txt.

[Batalov]

Originally Posted by Prime95

Ugh. The K10 results were not at all what I was hoping for. I could still use a 32-bit K8 benchmark.

To get v26 properly optimized for the K10 I'll need either remote access to a K10 machine or I'll need to buy one.

I have a 1055T, can give ssh access.
PM me, if this is still needed. (I may answer late - I am traveling now.)

Will post 64-bit results shortly.

[Kaboom]

AMD Phenom II 965 3411 MHz

[retina]

This is a dual core "fusion" CPU running at 1.6GHz with 4GB 533MHz DDR3 RAM.

Note that P95 (64bit) shows the CPU speed as 3130MHz, almost double the actual speed.

In case it matters, the OS here was Win7 64bit SP1.

AMD E-350 Processor
CPU speed: 3130.67 MHz, 2 cores
CPU features: Prefetch, MMX, SSE, SSE2
L1 cache size: 32 KB
L2 cache size: 512 KB
L1 cache line size: 64 bytes
L2 cache line size: 64 bytes
L1 TLBS: 40
L2 TLBS: 512
Prime95 64-bit version 26.6, RdtscTiming=1
Best time for 768K FFT length: 65.510 ms., avg: 69.135 ms.
Best time for 896K FFT length: 79.546 ms., avg: 82.993 ms.
Best time for 1024K FFT length: 86.486 ms., avg: 89.330 ms.
Best time for 1280K FFT length: 112.762 ms., avg: 116.492 ms.
Best time for 1536K FFT length: 140.734 ms., avg: 144.692 ms.
Best time for 1792K FFT length: 172.126 ms., avg: 178.703 ms.
Best time for 2048K FFT length: 187.579 ms., avg: 193.518 ms.
Best time for 2560K FFT length: 240.149 ms., avg: 247.287 ms.
Best time for 3072K FFT length: 297.749 ms., avg: 310.115 ms.
Best time for 3584K FFT length: 349.431 ms., avg: 364.646 ms.
Best time for 4096K FFT length: 408.261 ms., avg: 413.992 ms.
Best time for 5120K FFT length: 520.236 ms., avg: 527.763 ms.
Best time for 6144K FFT length: 669.112 ms., avg: 688.403 ms.
Best time for 7168K FFT length: 845.634 ms., avg: 863.367 ms.
Best time for 8192K FFT length: 1026.842 ms., avg: 1042.803 ms.
Timing FFTs using 2 threads.
Best time for 768K FFT length: 37.153 ms., avg: 39.416 ms.
Best time for 896K FFT length: 44.115 ms., avg: 48.304 ms.
Best time for 1024K FFT length: 48.801 ms., avg: 52.236 ms.
Best time for 1280K FFT length: 62.740 ms., avg: 68.568 ms.
Best time for 1536K FFT length: 77.495 ms., avg: 81.263 ms.
Best time for 1792K FFT length: 100.772 ms., avg: 104.366 ms.
Best time for 2048K FFT length: 102.841 ms., avg: 106.734 ms.
Best time for 2560K FFT length: 140.353 ms., avg: 148.205 ms.
Best time for 3072K FFT length: 175.362 ms., avg: 178.748 ms.
Best time for 3584K FFT length: 193.126 ms., avg: 199.384 ms.
Best time for 4096K FFT length: 235.366 ms., avg: 248.768 ms.
Best time for 5120K FFT length: 271.405 ms., avg: 280.357 ms.
Best time for 6144K FFT length: 353.088 ms., avg: 360.898 ms.
Best time for 7168K FFT length: 439.518 ms., avg: 452.910 ms.
Best time for 8192K FFT length: 576.547 ms., avg: 587.077 ms.
Best time for 61 bit trial factors: 8.032 ms.
Best time for 62 bit trial factors: 8.211 ms.
Best time for 63 bit trial factors: 9.635 ms.
Best time for 64 bit trial factors: 12.157 ms.
Best time for 65 bit trial factors: 13.492 ms.
Best time for 66 bit trial factors: 16.186 ms.
Best time for 67 bit trial factors: 16.130 ms.
Best time for 75 bit trial factors: 16.638 ms.
Best time for 76 bit trial factors: 16.641 ms.
Best time for 77 bit trial factors: 16.555 ms.