Join PrimeGrid
Returning Participants
Community
Leader Boards
Results
Other
drummers-lowrise
|
Message boards :
Generalized Fermat Prime Search :
RX 6900 XT fails larger GFN ATI searches
| Author |
Message |
|---|
|
|
|
Maybe I'm in the wrong place but I'm a newbe
When I look at std error output for open clati reports I see :
'AMD Accelerated Parallel Processing', device 'gfx1030', vendor 'Advanced Micro Devices, Inc.', version 'OpenCL 1.2 AMD-APP (3188.4)' and driver '3188.4 (PAL,LC)'.
40 computeUnits @ 2135MHz, memSize=3072MB, cacheSize=16kB, cacheLineSize=64B, localMemSize=64kB, maxWorkGroupSize=256.[/i]
The problem is that these parameters dont appear to match the RX6900 XT which 80 ComputeUnits, 128MB Cache, 16GB memory.
has the Open Cl for ATI been upgraded to see later later Radeon GPUs
Could this account CPU crashes when running larger GFN (ie 18 +)
____________
Second:8983 · 21639386 + 1
Second:9613*2^1638624+1 | |
|
Yves Gallot Volunteer developer
Project scientist
Send message
Joined: 19 Aug 12
Posts: 820
ID: 164101
Credit: 305,989,513
RAC: 1,728
|
|
Is it the latest driver?
https://www.amd.com/en/support/kb/release-notes/rn-rad-win-21-2-2 | |
|
|
|
|
Yes that is the driver I am using and the problem existed in earlier drivers. Its been a problem for a month or so with the then current drivers.
I thought it might be temp but have wound back MHz to max at about 2000MHz and allowed a little over voltage and it appears to run cool. Tried two different 1000w power supplies, - no difference, and MB Bios (Aorus X570) is latest. Also changed to run PCIe Gen3 instead of gen4 but again no difference.
Have run Furmark and other GPU stress tests but all pass Ok. I'm at a bit of a loss to know where to go next
____________
Second:8983 · 21639386 + 1
Second:9613*2^1638624+1 | |
|
Yves GallotVolunteer developer
Project scientist
Send message
Joined: 19 Aug 12
Posts: 820
ID: 164101
Credit: 305,989,513
RAC: 1,728
|
40 computeUnits @ 2135MHz, memSize=3072MB, cacheSize=16kB, cacheLineSize=64B, localMemSize=64kB, maxWorkGroupSize=256.[/i][/size]
The problem is that these parameters dont appear to match the RX6900 XT which 80 ComputeUnits, 128MB Cache, 16GB memory.
If the data returned by clGetDeviceInfo are not correct, the RX6900 XT is fast (very fast!) and I see no error reported.
It is not unusual that these static data are not correct with a new driver.
GFN-18 is tested in about 570 sec and GFN-19 in 1900 sec. That's faster than the 2080 Ti but slower than the 3080.
For GFN-21, the RX6900 XT is the fastest GPU: as fast as the TitanV and faster than the 3090!
AMD has made a giant step forward. | |
|
|
|
|
Thank you for your insight.
Yes its a big step forward but only if I can get it stable enough to run GFN.
Last run it managed 3 x GFN18 before crashing the CPU on the 4th.
Each time I tweak another parameter and run again. Throttling the GPU back to about 2000Mhz seems to have made it a little more stable but seems to me that this is not an ideal solution(?)
This time its enabling the "Smart Access Memory" feature of newer AMD CPU/GPU combinations which I had thought was enabled by default on my MB but apparently not.
If all else fails I will just use it for the image processing it was bought for and run GFN 16 and 17 which it appears to handle OK
____________
Second:8983 · 21639386 + 1
Second:9613*2^1638624+1 | |
|
Message boards :
Generalized Fermat Prime Search :
RX 6900 XT fails larger GFN ATI searches |
