Author

Topic: Something weird after installing Catalyst 11-6 (Read 1022 times)

member
Activity: 85
Merit: 10
i had a problem witg the new catalyst too, it "stole" 60Mh/s from my 6950 - simply went back to the previous driver version
newbie
Activity: 19
Merit: 0
Was having the same problem with a 4870 but not with a 5970 on the same machine, removed worksize from all the miners (including the 5970) and actually gained 10MH/s.
member
Activity: 111
Merit: 10
REINSTALL DRIVER. that is all

Tried a normal reinstall but not a full reinstall with clean uninstall. But, worksize 256 >= than worksize 64 on 11-5 catalyst so it's not a big deal yet.
member
Activity: 163
Merit: 10
REINSTALL DRIVER. that is all
member
Activity: 111
Merit: 10
I was on 11-5 with my 4850 and everything was running fine. I upgraded to 11-6 and the following weird things happened:

1. I used to run work size of 64. Now, any work size < 256 crashes the miner in MSVCR90.dll. Any gpu miner. This can be phoenix, diablo, poclbm, whatever. CPU mining is not affected.

Code:
Faulting application name: phoenix.exe, version: 0.0.0.0, time stamp: 0x4918019b
Faulting module name: MSVCR90.dll, version: 9.0.30729.6161, time stamp: 0x4dace5b9
Exception code: 0x40000015

2. However, the miner works ONCE without vectors and once with vectors regardless of work size. Even after rebooting it doesn't work. I ran it with poclbm at work size 64 with and without vectors and now I can't run them with work size < 256. Here's a phoenix example:

Code:
[18/06/2011 18:31:08] Phoenix 1.50 starting...
[18/06/2011 18:31:09] Connected to server
[18/06/2011 18:31:09] Server gave new work; passing to WorkQueue
[18/06/2011 18:31:09] New block (WorkQueue)
[18/06/2011 18:31:09] Server gave new work; passing to WorkQueue
[65.84 Mhash/sec] [0 Accepted] [0 Rejected] [RPC (+LP)]
C:\Users\Michael Hsu\Desktop\guiminer>phoenix.exe -v -u http://MxNbKXPBRqyw31ecG
2V5BNwqqrRLTWwWM3:[email protected]:9098 DEVICE=0 WORKSIZE=1024 -k phatk
[18/06/2011 18:31:40] Warning: Worksize exceeds the maximum of 256, using default.
[18/06/2011 18:31:40] Phoenix 1.50 starting...
[18/06/2011 18:31:40] Connected to server
[18/06/2011 18:31:40] Server gave new work; passing to WorkQueue
[18/06/2011 18:31:40] New block (WorkQueue)
[18/06/2011 18:31:40] Server gave new work; passing to WorkQueue
[64.25 Mhash/sec] [0 Accepted] [0 Rejected] [RPC (+LP)]
C:\Users\Michael Hsu\Desktop\guiminer>phoenix.exe -v -u http://user:[email protected]:9098 DEVICE=0 WORKSIZE=64 VECTORS
[18/06/2011 18:37:29] Phoenix 1.50 starting...
[18/06/2011 18:37:29] Connected to server
[18/06/2011 18:37:29] Server gave new work; passing to WorkQueue
[18/06/2011 18:37:29] New block (WorkQueue)
[18/06/2011 18:37:30] Server gave new work; passing to WorkQueue
[60.94 Mhash/sec] [0 Accepted] [0 Rejected] [RPC (+LP)]
C:\Users\Michael Hsu\Desktop\guiminer>phoenix.exe -v -u http://user:[email protected]:9098 DEVICE=0 WORKSIZE=64 VECTORS
[18/06/2011 18:37:53] Phoenix 1.50 starting...
[18/06/2011 18:37:53] Connected to server
[18/06/2011 18:37:53] Server gave new work; passing to WorkQueue
[18/06/2011 18:37:53] New block (WorkQueue)
[0 Khash/sec] [0 Accepted] [0 Rejected] [RPC]
This application has requested the Runtime to terminate it in an unusual way.
Please contact the application's support team for more information.

C:\Users\Michael Hsu\Desktop\guiminer>phoenix.exe -v -u http://MxNbKXPBRqyw31ecG2V5BNwqqrRLTWwWM3:[email protected]:9098 DEVICE=0 WORKSIZE=256
[18/06/2011 18:38:02] Phoenix 1.50 starting...
[18/06/2011 18:38:02] Connected to server
[18/06/2011 18:38:02] Server gave new work; passing to WorkQueue
[18/06/2011 18:38:02] New block (WorkQueue)
[18/06/2011 18:38:02] Server gave new work; passing to WorkQueue
[66.08 Mhash/sec] [0 Accepted] [0 Rejected] [RPC (+LP)]

Above, I run with worksize 1024 (too large?) and get ~66MH/s. Then I run 64 with vectors. With vectors because I already ran it without vectors and that doesn't work anymore. I then abort it and run it again: crash. Work size of 256 still works.

3. There used to be a performance hit at work size 256 (and 128) but I'm not seeing that now. Don't have current numbers for work size 64 to show it though.
Jump to: