Cant run 7900XTX

[moebius]

How about 116085643 ?

Radeon RX Vega 64-Vega 10 XT (215-0894200) 8 GB at Win 10 Pro

2023-01-10 02:55:42 gpuowl v6.11-364-g36f4e2a
2023-01-10 02:55:42 config:  -proof 6
2023-01-10 02:55:42 config:  -carry short
2023-01-10 02:55:42 config: -prp 116085643 -iters 10000
2023-01-10 02:55:42 device 0, unique id ''
2023-01-10 02:55:42 gfx900-0 116085643 FFT: 6M 1K:12:256 (18.45 bpw)
2023-01-10 02:55:42 gfx900-0 Expected maximum carry32: 65140000
2023-01-10 02:55:43 gfx900-0 OpenCL args "-DEXP=116085643u -DWIDTH=1024u -DSMALL_HEIGHT=256u -DMIDDLE=12u -DPM1=0 -DAMDGPU=1 -DMM_CHAIN=2u -DMM2_CHAIN=3u -DMAX_ACCURACY=1 -DWEIGHT_STEP_MINUS_1=0xe.cedaab025e09p-5 -DIWEIGHT_STEP_MINUS_1=-0xa.1f97bb12ad1bp-5  -cl-unsafe-math-optimizations -cl-std=CL2.0 -cl-finite-math-only "
2023-01-10 02:55:43 gfx900-0 ASM compilation failed, retrying compilation using NO_ASM
2023-01-10 02:55:48 gfx900-0 OpenCL compilation in 4.87 s
2023-01-10 02:55:49 gfx900-0 116085643 OK    26400 loaded: blockSize 400, 64a8012ccb3c2b85
2023-01-10 02:55:49 gfx900-0 validating proof residues for power 6
2023-01-10 02:55:49 gfx900-0 Proof using power 6
2023-01-10 02:55:51 gfx900-0 116085643 OK    27200   0.02%; 2018 us/it; ETA 2d 17:04; 4095f86bebe9acf3 (check 0.95s)
2023-01-10 02:56:09 gfx900-0 Stopping, please wait..
2023-01-10 02:56:11 gfx900-0 116085643 OK    36400   0.03%; 2025 us/it; ETA 2d 17:16; e328bd20b20c3715 (check 0.93s)
2023-01-10 02:56:11 gfx900-0 Exiting because "stop requested"
2023-01-10 02:56:11 gfx900-0 Bye

20230110 02:23:46  GpuOwl VERSION v7.2-129-g2a16a0b
20230110 02:23:46  config:  -proof 6
20230110 02:23:46  config:  -carry short
20230110 02:23:46  config:  -prp 116085643
20230110 02:23:46  config:  -iters 10000
20230110 02:23:46  device 0, unique id ''
20230110 02:23:46 gfx900-0 116085643 FFT: 6M 1K:12:256 (18.45 bpw)
20230110 02:23:46 gfx900-0 116085643 OpenCL args "-DEXP=116085643u -DWIDTH=1024u -DSMALL_HEIGHT=256u -DMIDDLE=12u -DAMDGPU=1 -DMM_CHAIN=2u -DMM2_CHAIN=3u -DWEIGHT_STEP=0.46275075340503335 -DIWEIGHT_STEP=-0.31635653054891877 -DIWEIGHTS={0,-0.06526321335377705,-0.12626713969029349,-0.18328975376689174,-0.23659088881501888,-0.28641342051450164,-0.33298437369785566,-0.37651595682751576,} -DFWEIGHTS={0,0.069819883293496324,0.14451458269011008,0.22442445728123822,0.30991362999031685,0.40137164676080089,0.49921525158845476,0.60389028548619039,}  -cl-std=CL2.0 -cl-finite-math-only "
20230110 02:23:46 gfx900-0 116085643 ASM compilation failed, retrying compilation using NO_ASM
20230110 02:23:48 gfx900-0 116085643 OpenCL compilation in 1.99 s
20230110 02:23:49 gfx900-0 116085643 OK     20000 on-load: blockSize 400, 4150234a796ca594
20230110 02:23:49 gfx900-0 116085643 validating proof residues for power 6
20230110 02:23:49 gfx900-0 116085643 Proof using power 6
20230110 02:24:10 gfx900-0 116085643 Stopping, please wait..
20230110 02:24:11 gfx900-0 116085643 OK     30000   0.03% c6d23cd4b6e53a5e 2114 us/it + check 0.90s + save 0.15s; ETA 2d 20:10
20230110 02:24:11 gfx900-0  Exiting because "stop requested"
20230110 02:24:11 gfx900-0  Bye

[moebius]

Originally Posted by preda

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.

I will add this exponent to the list and introduce a runtime column in dd:hh:ss.
But it will take a long time to complete the list.