Was told back in october that my S9000 boards would not work with the new apps. I failed to notice that binary search #1 on GPU works. Was reading another thread here (the 254+ one on O2MDF) and that the R9-280 was too old and got to looking at what apps were available for older boards.
So far, was able to run two work units each on six s9x00 boards under windows
There are 6 GPUs, units are minutes Dev# WU count Avg and Std of avg GPU0 WUs:2 -Stats- Avg:12.3(0.02)\ GPU1 WUs:2 -Stats- Avg:12.2(0.02) GPU2 WUs:2 -Stats- Avg:14.0(1.39) GPU3 WUs:2 -Stats- Avg:12.2(0.03) GPU4 WUs:2 -Stats- Avg:9.8(0.05) GPU5 WUs:2 -Stats- Avg:12.2(0.02)
GPU4 is s9100 rest are s9000 (same chip as HD7950) and elapsed time is significantly faster than GTX1070Ti under windows for same app
There are 1 GPUs, units are minutes Dev# WU count Avg and Std of avg GPU0 WUs:6 -Stats- Avg:21.6(6.19)
and even beats out the Linux version of the same app on gtx1060 class boards.
There are 8 GPUs, units are minutes Dev# WU count Avg and Std of avg GPU0 WUs:24 -Stats- Avg:15.6(2.94) GPU1 WUs:14 -Stats- Avg:18.4(0.03) GPU2 WUs:13 -Stats- Avg:19.5(0.08) GPU3 WUs:12 -Stats- Avg:20.6(0.04) GPU4 WUs:12 -Stats- Avg:20.4(0.05) GPU5 WUs:12 -Stats- Avg:21.4(0.03) GPU6 WUs:12 -Stats- Avg:20.9(0.04) GPU7 WUs:12 -Stats- Avg:20.8(0.04)
Copyright © 2024 Einstein@Home. All rights reserved.
JStateson wrote:Was told back
)
I have an AMD 5870 crunching here and it's also doing just fine on the Gamma-ray pulsar binary search workunits doing one in about 1000 seconds. I have run my AMD 7970 here in the past on the same workunits and it also worked fine. Both pc's are using Win10 as the OS.
Yea, long ago I had a pair of
)
Yea, long ago I had a pair of HD7850s. A new 7870 cost too much back then for me. The last time I looked, the 7850 outperformed the gtx1080 on FP64 math. I used those on Milkyway for years until the price on used HD7950s and S9000s dropped.
JStateson wrote:Yea, long ago
)
I used to several more but have sent them to Teammates as I got newer Nvidia cards, yes they are slower at the dual precision Projects but at the other Projects they are much faster.