Topic: DiabloMiner GPU Miner - page 22. (Read 866596 times)
Yea I am on the fence. I would like to buy a couple but just not sure.
I'll take that as a No...
Nope, and the whole FPGA mining experience is still rather iffy. Not sure if I want to support that yet.
I'll take that as a No...
Just a question. Can you run Daiblo on the Butterfly Labs "single"?
Sure, when FPGAs can build OpenCL kernels.
Just a question. Can you run Daiblo on the Butterfly Labs "single"?
Thanks again DiabloD3 that -na solved everything. I can get about 2Ghah/s now.
Not on a 6750m you don't. Thats closer to 75-100 mhash.
For 35? 2 ghash is in the right ballpark =P
BTW, I've been told by someone -v 2 -w 64 -na is the most optimum for OSX on that kind of hardware.
Thanks again DiabloD3 that -na solved everything. I can get about 2Ghah/s now.
Not on a 6750m you don't. Thats closer to 75-100 mhash.
Thanks again DiabloD3 that -na solved everything. I can get about 2Ghah/s now.
Not on a 6750m you don't. Thats closer to 75-100 mhash.
Thanks again DiabloD3 that -na solved everything. I can get about 2Ghah/s now.
Update: Remove the previous "fixes" for 12.2 performance.
I think either my testing went wrong, or 12-2-2 on Debian fixed something. -v 2,1 is still the fastest if it works right on your card, -v 4 is back to not fitting, and performance is back to where it was.
I think either my testing went wrong, or 12-2-2 on Debian fixed something. -v 2,1 is still the fastest if it works right on your card, -v 4 is back to not fitting, and performance is back to where it was.
Great work D, thanks for the update!
Update: Remove the previous "fixes" for 12.2 performance.
I think either my testing went wrong, or 12-2-2 on Debian fixed something. -v 2,1 is still the fastest if it works right on your card, -v 4 is back to not fitting, and performance is back to where it was.
I think either my testing went wrong, or 12-2-2 on Debian fixed something. -v 2,1 is still the fastest if it works right on your card, -v 4 is back to not fitting, and performance is back to where it was.
Question: I have an iMac with a 6750M. This is what I am getting in diablominer and btcguild.
-w 128 -v 2 or -w 64
After 7-10 minutes it has 1-3 shares, and 5-10 hardware errors. It was running at 14-20 mhash/s. Then it will crash the system. Is there anything I am doing wrong or is it a mac problem? It is also OSX 10.7 (Lion).
-w 128 -v 2 or -w 64
After 7-10 minutes it has 1-3 shares, and 5-10 hardware errors. It was running at 14-20 mhash/s. Then it will crash the system. Is there anything I am doing wrong or is it a mac problem? It is also OSX 10.7 (Lion).
See the op post. You need to also use -na.
I added -na now I get the error:
[3/9/12 3:05:01 PM] ERROR: OpenCL Build Warning : Compiler build log:
x XA870; x XA871; x XA872; x XA873; x XA874;
But now it actually show what seems to be the correct mhash/s and very little hardware errors.
Yeah, ignore the warnings, its a side effect of -na
Question: I have an iMac with a 6750M. This is what I am getting in diablominer and btcguild.
-w 128 -v 2 or -w 64
After 7-10 minutes it has 1-3 shares, and 5-10 hardware errors. It was running at 14-20 mhash/s. Then it will crash the system. Is there anything I am doing wrong or is it a mac problem? It is also OSX 10.7 (Lion).
-w 128 -v 2 or -w 64
After 7-10 minutes it has 1-3 shares, and 5-10 hardware errors. It was running at 14-20 mhash/s. Then it will crash the system. Is there anything I am doing wrong or is it a mac problem? It is also OSX 10.7 (Lion).
See the op post. You need to also use -na.
I added -na now I get the error:
[3/9/12 3:05:01 PM] ERROR: OpenCL Build Warning : Compiler build log:
x XA870; x XA871; x XA872; x XA873; x XA874;
But now it actually show what seems to be the correct mhash/s and very little hardware errors.
OP post handles it. But shorthand is 58xx and 57xx/67xx are first gen VLIW5, 68xx are second gen VLIW5, 69xx is VLIW4, and 79xx are GCN.
That helps, thanks!
Quick nooish Q: What are the VLIWx?
I keep seeing them mentioned, but I haven't seen an explanation of what they are.
I keep seeing them mentioned, but I haven't seen an explanation of what they are.
OP post handles it. But shorthand is 58xx and 57xx/67xx are first gen VLIW5, 68xx are second gen VLIW5, 69xx is VLIW4, and 79xx are GCN.
Quick nooish Q: What are the VLIWx?
I keep seeing them mentioned, but I haven't seen an explanation of what they are.
I keep seeing them mentioned, but I haven't seen an explanation of what they are.
Update: Improve speed on Catalyst 12.2
So I have surveyed the damage that 12.2 caused. On my 5850, 393 went down to 387, I've been able to bring it up to 390 by reordering a few instructions and switching to -v 2,1 as my recommendation on VLIW5. Another interesting thing is, -v 4 now fits on VLIW5, although it is still slower than -v 2,1.
As for VLIW4? Someone is going to have to run -v 2, -v 2,1 and -v 4 on there and tell me which one wins, I assume -v 2 will be the only one that works right.
So I have surveyed the damage that 12.2 caused. On my 5850, 393 went down to 387, I've been able to bring it up to 390 by reordering a few instructions and switching to -v 2,1 as my recommendation on VLIW5. Another interesting thing is, -v 4 now fits on VLIW5, although it is still slower than -v 2,1.
As for VLIW4? Someone is going to have to run -v 2, -v 2,1 and -v 4 on there and tell me which one wins, I assume -v 2 will be the only one that works right.
Goddamnit, 12.2 just came out for FGLRX and I've confirmed: ITS SLOWER, WHY IS THIS, GODDAMNIT AMD
Haha, that seems to be a common theme among those who program for AMD drivers.
Unwinder (author of MSI afterburner et. al.) expressed similar thoughts a couple days ago.
Please be assured, that while most of us can't do a thing to help you in the code battle against the AMD driver gremlins, we are sympathetic (or is that just plain pathetic?) to your woes and wish you the best!
I, for one, am REALLY appreciative of the work you're doing.
Once I mine enough to start actually making a little cash, I'll be tossing some BTC your way to support further work.
Keep up the fight, you'll beat those critters, I have confidence!
Now if I could just get voltage control back on my 7970....
Goddamnit, 12.2 just came out for FGLRX and I've confirmed: ITS SLOWER, WHY IS THIS, GODDAMNIT AMD
I didnt overclock them alot because they are in standard closed case.
This is just my test box to test stuff on it. I was in cuba for 14 days (no internet there) so im testing all this new drivers and miners now i thought performance would improve but now its like this..
Other rigs are in custom built cases with 4x7970 clocked higher and get around 660 as i do not want temps to exceed 65°C.
Any1 knows whats wrong? Or how do i roll back to old drivers?
This is just my test box to test stuff on it. I was in cuba for 14 days (no internet there) so im testing all this new drivers and miners now i thought performance would improve but now its like this..
Other rigs are in custom built cases with 4x7970 clocked higher and get around 660 as i do not want temps to exceed 65°C.
Any1 knows whats wrong? Or how do i roll back to old drivers?
Woah dude, 65c is too low. 85c is the standard max safe temp for Radeons that aren't mobile. People have almost hit 800 mhash at full overclock potential.
Still, not sure why you should be having problems. You didn't accidentally install SDK 2.6 on top of 12.2 did you? Also, which 12.2, there are several. Try the new 12.3RC release.
Well I didnt dare to leave them alone running at 85°C for 14 days..
I just downloaded 12.2 from ati.com and installed whole package.
edit: installing 12.3 rc doesnt help
Weird. You're only using -v 1 -w 256 as args, right?
Jump to: