Cant run 7900XTX

[Magellan3s]

Originally Posted by kriesel

My bad (memory). Retested, on 3 Radeon VII (not pro) GPUs, ~630 (good) - 640 (failing fan) usec/iter, gpuowl-win V6.11-38x, if the GPU VRAM is Hynix not Samsung. Set to -20% power, & VRAM as fast as it will reliably run, which varies from one GPU to the next, with "AMD Radeon Software", launched from the Windows start menu, using its tuning tab. That software can be installed along with the driver during initial installation from the same download package, or during a driver update.
George has reported being able to run VRAM at 1200 MHz; my Radeon VII GPUs mostly range up to 1170 with Hynix, ~870 with Samsung (yes, considerably below nominal clock for Samsung, and it's still too error prone to run P-1 or LLDC).
Typically gpuowl V6.11-38x is several percent faster than v7.x on the same work, hardware and tune, on Radeon VII & Windows. I have no Linux data.
If reliable enough, use a large block size which reduces GEC checking overhead and lifts performance slightly.
Same everything,including fft length, ~74M runs 580 usec/iter in LLDC or PRP.
For comparison, RX6900XT, 705 usec/iter., -10% power, +7% VRAM (allowed min and max respectively).

I ran the benchmark on a new linux install and got the following.

jesus@Magellan:~/gpuowl$ ./gpuowl -maxAlloc 16G  -prp 77936867
20230102 15:34:00  GpuOwl VERSION v7.2-131-gca22dce
20230102 15:34:00  GpuOwl VERSION v7.2-131-gca22dce
20230102 15:34:00  Note: not found 'config.txt'
20230102 15:34:00  config: -maxAlloc 16G -prp 77936867 
20230102 15:34:00  device 0, unique id 'd78a416172e17d3c'
20230102 15:34:00 d78a416172e17d3c 77936867 FFT: 4M 1K:8:256 (18.58 bpw)
20230102 15:34:00 d78a416172e17d3c 77936867 OpenCL args "-DEXP=77936867u -DWIDTH=1024u -DSMALL_HEIGHT=256u -DMIDDLE=8u -DAMDGPU=1 -DMM_CHAIN=1u -DMM2_CHAIN=2u -DWEIGHT_STEP=0.33644726404543274 -DIWEIGHT_STEP=-0.25174750481886216 -DIWEIGHTS={0,-0.44011820345520131,-0.37306474779553728,-0.29798072935699788,-0.21390437908665341,-0.11975874301407295,-0.014337887291734644,-0.44814572555075455,} -DFWEIGHTS={0,0.78609128957452257,0.5950610473469905,0.42446232150303748,0.2721098723818392,0.1360521812214803,0.014546452690911484,0.81207258201996746,}  -cl-std=CL2.0 -cl-finite-math-only "
20230102 15:34:01 d78a416172e17d3c 77936867 OpenCL compilation in 0.87 s
20230102 15:34:01 d78a416172e17d3c 77936867 PRP starting from beginning
20230102 15:34:02 d78a416172e17d3c 77936867 OK         0 on-load: blockSize 400, 0000000000000003
20230102 15:34:02 d78a416172e17d3c 77936867 validating proof residues for power 8
20230102 15:34:02 d78a416172e17d3c 77936867 Proof using power 8
20230102 15:34:07 d78a416172e17d3c 77936867     10000 fc4f135f7cf4ad29  530
20230102 15:34:12 d78a416172e17d3c 77936867     20000 3cd1bd9d5e09cbc5  530
20230102 15:34:18 d78a416172e17d3c 77936867     30000 c4e0ff35e3290d98  531
20230102 15:34:23 d78a416172e17d3c 77936867     40000 dffe1b1b0d748128  531
20230102 15:34:28 d78a416172e17d3c 77936867     50000 52e286945371ed29  532
20230102 15:34:34 d78a416172e17d3c 77936867     60000 0945da4dc08bdd95  532
20230102 15:34:39 d78a416172e17d3c 77936867     70000 7131fa4eb77f4bb2  533
20230102 15:34:44 d78a416172e17d3c 77936867     80000 8d76071d27ee4221  533
20230102 15:34:50 d78a416172e17d3c 77936867     90000 0bacff453b2f470e  533
20230102 15:34:54 d78a416172e17d3c 77936867 Stopping, please wait..
20230102 15:34:55 d78a416172e17d3c 77936867 OK     99200   0.13% f9443cde8dee98f4  533 us/it + check 0.25s + save 0.09s; ETA 11:32
20230102 15:34:55 d78a416172e17d3c  Exiting because "stop requested"
20230102 15:34:55 d78a416172e17d3c  Bye
jesus@Magellan:~/gpuowl$

[kriesel]

Originally Posted by Magellan3s

I ran the benchmark on a new linux install and got the following.

jesus@Magellan:~/gpuowl$ ./gpuowl -maxAlloc 16G  -prp 77936867
20230102 15:34:00  GpuOwl VERSION v7.2-131-gca22dce
20230102 15:34:00  GpuOwl VERSION v7.2-131-gca22dce
20230102 15:34:00  Note: not found 'config.txt'
...
20230102 15:34:55 d78a416172e17d3c 77936867 OK     99200   0.13% f9443cde8dee98f4 533 us/it + check 0.25s + save 0.09s; ETA 11:32
20230102 15:34:55 d78a416172e17d3c  Exiting because "stop requested"
20230102 15:34:55 d78a416172e17d3c  Bye
jesus@Magellan:~/gpuowl$

For which GPU model? (I mentioned at least two) At what sclk settings, GPU power level, clocking details?

[Prime95]

Originally Posted by kriesel

Typically gpuowl V6.11-38x is several percent faster than v7.x on the same work, hardware and tune, on Radeon VII & Windows. I have no Linux data.

I've tuned all my Radeon VIIs for maximum energy efficiency. Here is the report from one:

Temp: 71.0c
Power: 117.0W (probably 130+W at the wall, I think rocm-smi's reported watts are a bit low)
Sclk: 1157Mhz 725mV
Mclk: 1201Mhz

Timings at 4M FFT: 546us (actually 2 linux instances at 1091 or 1092us, gpuowl v6.11-351-ge930f9e)

[yuki0831]

Hello, I sold the XTX in auction. It sold at 174000yen postage incl. 1month cost 15000yen.
I stand in line for long time .2hours to get queue ticket.
Have some trouble in setup,
I finally found that I dont expect AMD advertising blurb.FLOPS FP64etc...
If I will buy AMD GPU,I must watch Gpuowl benchmark.

Thank you for advicing me to install Ubuntu Gpuowl supoort.

[moebius]

174.000 Yen = 1.330,44 US-Dollar = 1266,89 Euro

[M344587487]

Hope you had fun benchmarking, thank you for putting the effort in.

[Xyzzy]

https://old.reddit.com/r/hardware/co...ps_vs_61tflop/

[Magellan3s]

Originally Posted by kriesel

For which GPU model? (I mentioned at least two) At what sclk settings, GPU power level, clocking details?

Radeon PRO VII - I haven't touched any sclk settings, power level settings, or clock settings. I've actually been running GPU OWL on a windows build for the majority of the work I've submitted to GIMPS using this particular card. I like windows a little better than Ubuntu.

[PhilF]

Originally Posted by Magellan3s

Radeon PRO VII - I haven't touched any sclk settings, power level settings, or clock settings. I've actually been running GPU OWL on a windows build for the majority of the work I've submitted to GIMPS using this particular card. I like windows a little better than Ubuntu.

If you've not done any tuning you are almost certainly using considerably more power than necessary (and generating more heat too).

[Magellan3s]

Originally Posted by PhilF

If you've not done any tuning you are almost certainly using considerably more power than necessary (and generating more heat too).

The software I downloaded from the AMD website doesn't allow me to change those values. Is there a program anyone recommends?

[preda]

Originally Posted by moebius

https://docs.google.com/spreadsheets...f=true&sd=true

That table is very nice!

The exponent used for benchmarking (77936867) is a bit dated though, we could start using another exponent, e.g. around 115M, as that's what is relevant at the PRP wavefront now.

115M would be towards the upper limit of 6M FFT; 120M would be towards the upper limit of 6.5M FFT.