Cant run 7900XTX
 

Hello.
I bought Redeon 7900XTX Powercolor 24G-L/OC(190000yen)

I tried to crunch,but failed.

20221217 20:45:24 GpuOwl VERSION v7.2-129-g2a16a0b
20221217 20:45:24 config: -device 1
20221217 20:45:24 config: -prp 77232917
20221217 20:45:24 config: -maxAlloc 19G
20221217 20:45:24 config:
20221217 20:45:24 config:
20221217 20:45:24 config:
20221217 20:45:24 config:
20221217 20:45:24 device 1, unique id ''
20221217 20:45:24 gfx1100-1 77232917 FFT: 4M 1K:8:256 (18.41 bpw)
20221217 20:45:24 gfx1100-1 77232917 OpenCL args "-DEXP=77232917u -DWIDTH=1024u -DSMALL_HEIGHT=256u -DMIDDLE=8u -DAMDGPU=1 -DMM2_CHAIN=1u -DWEIGHT_STEP=0.50132640504118509 -DIWEIGHT_STEP=-0.33392232585653647 -DIWEIGHTS={0,-0.11268106401526803,-0.21266510584292314,-0.30138283945289102,-0.38010376444266458,-0.44995433184432437,-0.023868125978189877,-0.13385970416218501,} -DFWEIGHTS={0,0.12699048723694423,0.27010755832256494,0.43139913599727303,0.61317320970810774,0.81803086160652561,0.024451743266869885,0.1545473692950666,} -cl-std=CL2.0 -cl-finite-math-only "
20221217 20:45:26 gfx1100-1 77232917 OpenCL compilation in 2.24 s
20221217 20:45:26 gfx1100-1 77232917 PRP starting from beginning
20221217 20:45:27 gfx1100-1 77232917 EE 0 on-load: 0000000000000000 vs. 0000000000000003
20221217 20:45:27 gfx1100-1 77232917 PRP starting from beginning
20221217 20:45:27 gfx1100-1 77232917 EE 0 on-load: 0000000000000000 vs. 0000000000000003
20221217 20:45:27 gfx1100-1 Exiting because "error on load"
20221217 20:45:27 gfx1100-1 Bye

How can I? Its first time to
buy AMD Radeon GPU.

How many GPUs are in the system? Have you verified the 7900XTx is the second?
Run Gpuowl or gpuowl-win -h and check the device list it produces between the - options list and the fft lengths list.
OpenCL devices as recognized by Gpuowl versions start at -device 0.

Did you try v 6.11

[URL="https://mersenneforum.org/showpost.php?p=551594&postcount=8"]gpuOwl 6.11.364 windows binary[/URL]

[URL="https://mersenneforum.org/showpost.php?p=556311&postcount=40"]gpuOwl 6.11.380 linux binary (Ubuntu) [/URL]

try to lower the RAM frequenzy f.e. 200 Mhz

20221217 21:43:04 GpuOwl VERSION v7.2-129-g2a16a0b
20221217 21:43:04 config: -device 0
20221217 21:43:04 config: -prp 77232917
20221217 21:43:04 config: -maxAlloc 19G
20221217 21:43:04 config: -mprimeDir mprime
20221217 21:43:04 config:
20221217 21:43:04 config:
20221217 21:43:04 config:
20221217 21:43:04 config:
20221217 21:43:04 device 0, unique id ''
20221217 21:43:07 gfx1100-0 77232917 FFT: 4M 1K:8:256 (18.41 bpw)
20221217 21:43:07 gfx1100-0 77232917 OpenCL args "-DEXP=77232917u -DWIDTH=1024u -DSMALL_HEIGHT=256u -DMIDDLE=8u -DAMDGPU=1 -DMM2_CHAIN=1u -DWEIGHT_STEP=0.50132640504118509 -DIWEIGHT_STEP=-0.33392232585653647 -DIWEIGHTS={0,-0.11268106401526803,-0.21266510584292314,-0.30138283945289102,-0.38010376444266458,-0.44995433184432437,-0.023868125978189877,-0.13385970416218501,} -DFWEIGHTS={0,0.12699048723694423,0.27010755832256494,0.43139913599727303,0.61317320970810774,0.81803086160652561,0.024451743266869885,0.1545473692950666,} -cl-std=CL2.0 -cl-finite-math-only "
20221217 21:43:10 gfx1100-0 77232917 OpenCL compilation in 3.24 s
20221217 21:43:10 gfx1100-0 77232917 PRP starting from beginning
20221217 21:43:11 gfx1100-0 77232917 EE 0 on-load: 0000000000000000 vs. 0000000000000003
20221217 21:43:11 gfx1100-0 77232917 PRP starting from beginning
20221217 21:43:11 gfx1100-0 77232917 EE 0 on-load: 0000000000000000 vs. 0000000000000003
20221217 21:43:11 gfx1100-0 Exiting because "error on load"
20221217 21:43:11 gfx1100-0 Bye




After your post I chnaged my system. It didnt solve issue.

XTX first. No other GPU connct.
now
XTX device number 0
No GPU conncect


previous
RTX4090 first device number 0
XTX device number 1

2022-12-17 21:55:36 config: -device 0
2022-12-17 21:55:36 config: -prp 77232917
2022-12-17 21:55:36 config: -maxAlloc 19000MB
2022-12-17 21:55:36 config:
2022-12-17 21:55:36 config:
2022-12-17 21:55:36 config:
2022-12-17 21:55:36 config:
2022-12-17 21:55:36 config:
2022-12-17 21:55:36 config:
2022-12-17 21:55:36 device 0, unique id ''
2022-12-17 21:55:36 gfx1100-0 77232917 FFT: 4M 1K:8:256 (18.41 bpw)
2022-12-17 21:55:36 gfx1100-0 Expected maximum carry32: 4E8A0000
2022-12-17 21:55:36 gfx1100-0 OpenCL args "-DEXP=77232917u -DWIDTH=1024u -DSMALL_HEIGHT=256u -DMIDDLE=8u -DPM1=0 -DAMDGPU=1 -DMM2_CHAIN=1u -DMAX_ACCURACY=1 -DWEIGHT_STEP_MINUS_1=0x8.056ed6245ec78p-4 -DIWEIGHT_STEP_MINUS_1=-0xa.af7ddf9ea896p-5 -cl-unsafe-math-optimizations -cl-std=CL2.0 -cl-finite-math-only "
2022-12-17 21:55:39 gfx1100-0 OpenCL compilation in 2.59 s
2022-12-17 21:55:39 gfx1100-0 77232917 EE 0 loaded: blockSize 400, 0000000000000000 (expected 0000000000000003)
2022-12-17 21:55:39 gfx1100-0 Exiting because "error on load"
2022-12-17 21:55:39 gfx1100-0 Bye


Also older version wont work

[QUOTE=yuki0831;620077]2022-12-17 21:55:36 config: -device 0
2022-12-17 21:55:36 config: -prp 77232917
2022-12-17 21:55:36 config: -maxAlloc 19000MB
2022-12-17 21:55:36 config:
2022-12-17 21:55:36 config:
2022-12-17 21:55:36 config:
2022-12-17 21:55:36 config:
2022-12-17 21:55:36 config:
2022-12-17 21:55:36 config:
2022-12-17 21:55:36 device 0, unique id ''
2022-12-17 21:55:36 gfx1100-0 77232917 FFT: 4M 1K:8:256 (18.41 bpw)
2022-12-17 21:55:36 gfx1100-0 Expected maximum carry32: 4E8A0000
2022-12-17 21:55:36 gfx1100-0 OpenCL args "-DEXP=77232917u -DWIDTH=1024u -DSMALL_HEIGHT=256u -DMIDDLE=8u -DPM1=0 -DAMDGPU=1 -DMM2_CHAIN=1u -DMAX_ACCURACY=1 -DWEIGHT_STEP_MINUS_1=0x8.056ed6245ec78p-4 -DIWEIGHT_STEP_MINUS_1=-0xa.af7ddf9ea896p-5 -cl-unsafe-math-optimizations -cl-std=CL2.0 -cl-finite-math-only "
2022-12-17 21:55:39 gfx1100-0 OpenCL compilation in 2.59 s
2022-12-17 21:55:39 gfx1100-0 77232917 EE 0 loaded: blockSize 400, 0000000000000000 (expected 0000000000000003)
2022-12-17 21:55:39 gfx1100-0 Exiting because "error on load"
2022-12-17 21:55:39 gfx1100-0 Bye


Also older version wont work[/QUOTE]

Value for maxalloc must be in MB (Exception gpu_bad_alloc: Reached GPU maxAlloc limit 19 MB)

-maxAlloc 19532

and please post a benchmark of gpuowl -iters 20000 -prp 77936867 in this [URL="https://mersenneforum.org/showthread.php?p=620030#post620030"]thread[/URL] , if you can get it to work.

2022-12-17 22:08:43 config: -device 0
2022-12-17 22:08:43 config: -prp 77232917 -maxAlloc 19532
2022-12-17 22:08:43 config:
2022-12-17 22:08:43 config:
2022-12-17 22:08:43 config:
2022-12-17 22:08:43 config:
2022-12-17 22:08:43 config:
2022-12-17 22:08:43 device 0, unique id ''
2022-12-17 22:08:43 gfx1100-0 77232917 FFT: 4M 1K:8:256 (18.41 bpw)
2022-12-17 22:08:43 gfx1100-0 Expected maximum carry32: 4E8A0000
2022-12-17 22:08:43 gfx1100-0 OpenCL args "-DEXP=77232917u -DWIDTH=1024u -DSMALL_HEIGHT=256u -DMIDDLE=8u -DPM1=0 -DAMDGPU=1 -DMM2_CHAIN=1u -DMAX_ACCURACY=1 -DWEIGHT_STEP_MINUS_1=0x8.056ed6245ec78p-4 -DIWEIGHT_STEP_MINUS_1=-0xa.af7ddf9ea896p-5 -cl-unsafe-math-optimizations -cl-std=CL2.0 -cl-finite-math-only "
2022-12-17 22:08:46 gfx1100-0 OpenCL compilation in 2.56 s
2022-12-17 22:08:46 gfx1100-0 77232917 EE 0 loaded: blockSize 400, 0000000000000000 (expected 0000000000000003)
2022-12-17 22:08:46 gfx1100-0 Exiting because "error on load"
2022-12-17 22:08:46 gfx1100-0 Bye

changes max allocation. also didint run.

Eventually it is the same problem how it was with the RTX4090. Try to lower VRAM and gpuclock a bit.

Try running [URL="https://www.mersenneforum.org/showthread.php?t=23394"]mfakto[/URL] on it. And its OpenCL test option.
Try running clinfo or other third-party utilities to check whether OpenCL is working correctly on that GPU.
Also check the system side, not only the GPU side, for overclocking & reliability.

2022-12-17 22:08:43 config: -device 0
2022-12-17 22:08:43 config: -prp 77232917 -maxAlloc 19532
2022-12-17 22:08:43 config:
2022-12-17 22:08:43 config:
2022-12-17 22:08:43 config:
2022-12-17 22:08:43 config:
2022-12-17 22:08:43 config:
2022-12-17 22:08:43 device 0, unique id ''
2022-12-17 22:08:43 gfx1100-0 77232917 FFT: 4M 1K:8:256 (18.41 bpw)
2022-12-17 22:08:43 gfx1100-0 Expected maximum carry32: 4E8A0000
2022-12-17 22:08:43 gfx1100-0 OpenCL args "-DEXP=77232917u -DWIDTH=1024u -DSMALL_HEIGHT=256u -DMIDDLE=8u -DPM1=0 -DAMDGPU=1 -DMM2_CHAIN=1u -DMAX_ACCURACY=1 -DWEIGHT_STEP_MINUS_1=0x8.056ed6245ec78p-4 -DIWEIGHT_STEP_MINUS_1=-0xa.af7ddf9ea896p-5 -cl-unsafe-math-optimizations -cl-std=CL2.0 -cl-finite-math-only "
2022-12-17 22:08:46 gfx1100-0 OpenCL compilation in 2.56 s
2022-12-17 22:08:46 gfx1100-0 77232917 EE 0 loaded: blockSize 400, 0000000000000000 (expected 0000000000000003)
2022-12-17 22:08:46 gfx1100-0 Exiting because "error on load"
2022-12-17 22:08:46 gfx1100-0 Bye


mfakto works well.


RTX4090 its Powerlimt 75% TF over 1000M works well(24/7). System side is still good.
I dont OC runs 24/7.
run on Windows 10

Linux software dont work. And commnd option? i dont know about software config nor compile.

Powercolor dont launch OC tool. So,I Downloaded Msi afterburner.
lower Core clock and PL didnt solve this issue.
MSI afterbuner dont set lower MEM clock.


RC5-72 (distributed.net) OpenCL software works well.
I wait for GPU-Z update. it didnt load MEM colck, power comsumption.