Pages:
Author

Topic: AMD Stream SDK 2.6 (Catalyst 11.12/12.1) - Get your performance back! (Phoenix) (Read 14755 times)

sr. member
Activity: 271
Merit: 250
Is there a good app that let's me see the SDK version installed? I installed 11.12 drivers on a new build (using hardware that I know hashes higher - ie my 6870s are only getting 270MH even with DiabloMiner) and then uninstalled everything and reinstalled 11.7 drivers but I don't think it cleaned out the 2.6 SDK properly.

GPU Caps Viewer does a good job of providing plenty of information on wat ATI Stream/APP SDK's are installed on your system and allows you to sellect each GPU/CPU individually and displays a read out of wat the openCL/GL software is doing.

Seen it also allow you to select between either ATI Stream or APP SDK within the app when I had both installed at one time. Have not yet see it offered for linux systems but I suspect something similiar will soon surface.

Here is a link to the latest version.

http://www.geeks3d.com/20120202/gpu-caps-viewer-1-15-0-opengl-opencl-cuda-graphics-card-utility/
sr. member
Activity: 378
Merit: 250
I find it funny how everyone sort of ignored me there.  If you're going to test the kernels, test them to the fullest extent of their capabilities.  I would like to see how they measure up completely.  Which means include VECTORS8 with the GOFFSET=false
Slower, on all worksizes

(ed: well, cpu was faster)
Hmm, I do know that VLIW is faster with VECTORS8.  I would have thought that GCN could handle it.  I wonder why it decreased like that.  Hmm... Huh
zvs
legendary
Activity: 1680
Merit: 1000
https://web.archive.org/web/*/nogleg.com
I find it funny how everyone sort of ignored me there.  If you're going to test the kernels, test them to the fullest extent of their capabilities.  I would like to see how they measure up completely.  Which means include VECTORS8 with the GOFFSET=false
Slower, on all worksizes

(ed: well, cpu was faster)
sr. member
Activity: 378
Merit: 250
I find it funny how everyone sort of ignored me there.  If you're going to test the kernels, test them to the fullest extent of their capabilities.  I would like to see how they measure up completely.  Which means include VECTORS8 with the GOFFSET=false
legendary
Activity: 1162
Merit: 1000
DiabloMiner author
Is there a good app that let's me see the SDK version installed? I installed 11.12 drivers on a new build (using hardware that I know hashes higher - ie my 6870s are only getting 270MH even with DiabloMiner) and then uninstalled everything and reinstalled 11.7 drivers but I don't think it cleaned out the 2.6 SDK properly.

DM says what your SDK is on startup. If it says AMD-APP (851.4), you're on SDK 2.6.
DrG
legendary
Activity: 2086
Merit: 1035
Is there a good app that let's me see the SDK version installed? I installed 11.12 drivers on a new build (using hardware that I know hashes higher - ie my 6870s are only getting 270MH even with DiabloMiner) and then uninstalled everything and reinstalled 11.7 drivers but I don't think it cleaned out the 2.6 SDK properly.
legendary
Activity: 1162
Merit: 1000
DiabloMiner author
I finally got a 5870 to mess around with, and, yeah, the "best" memory setting on that is diff than the 5830.  

At 1000/395, it gets 440mhash, at 1000/310, 448mhash, etc.

1/3rd core speed is considered the rule. So, in your case, 1000/333.

Not a good rule. I did a SI shit ton of benchmarking every 10MHz (and 5MHz around the peak) on SDK 2.5 (worksize 256). On a 5770 at 950MHz, the performance peak was at 295MHz RAM; at 800MHz it was around 260MHz. On a 5830 at 1050MHz peak performance was obtained at 375MHz-390MHz RAM; at 800MHz, it was still around 360MHz. There is probably a better formula that involves number of stream processors and memory bus width vs worksize and vector size, but benchmarking your particular configuration is ideal.

Of course this topic is about SDK 2.6, where any memory clock other than ~1000Mhz will hurt your hashrate.

Memory speed is largely voodoo magic. Every "best" speed I've seen that doesn't fit the 1/3rd rule (such as on 1200mhz mem) 57xx/58xx) seems to still be between 1/3rd /1200*1000 (1000mhz core == 278mhz) and 1/3rd /1000*1200 (== 400 mhz), your lower and higher settings seem to be within that margin.
hero member
Activity: 497
Merit: 500
I finally got a 5870 to mess around with, and, yeah, the "best" memory setting on that is diff than the 5830.  

At 1000/395, it gets 440mhash, at 1000/310, 448mhash, etc.

1/3rd core speed is considered the rule. So, in your case, 1000/333.

Not a good rule. I did a SI shit ton of benchmarking every 10MHz (and 5MHz around the peak) on SDK 2.5 (worksize 256). On a 5770 at 950MHz, the performance peak was at 295MHz RAM; at 800MHz it was around 260MHz. On a 5830 at 1050MHz peak performance was obtained at 375MHz-390MHz RAM; at 800MHz, it was still around 360MHz. There is probably a better formula that involves number of stream processors and memory bus width vs worksize and vector size, but benchmarking your particular configuration is ideal.

Of course this topic is about SDK 2.6, where any memory clock other than ~1000Mhz will hurt your hashrate.

Except for on a 7970. It helped!
legendary
Activity: 1512
Merit: 1036
I finally got a 5870 to mess around with, and, yeah, the "best" memory setting on that is diff than the 5830.  

At 1000/395, it gets 440mhash, at 1000/310, 448mhash, etc.

1/3rd core speed is considered the rule. So, in your case, 1000/333.

Not a good rule. I did a SI shit ton of benchmarking every 10MHz (and 5MHz around the peak) on SDK 2.5 (worksize 256). On a 5770 at 950MHz, the performance peak was at 295MHz RAM; at 800MHz it was around 260MHz. On a 5830 at 1050MHz peak performance was obtained at 375MHz-390MHz RAM; at 800MHz, it was still around 360MHz. There is probably a better formula that involves number of stream processors and memory bus width vs worksize and vector size, but benchmarking your particular configuration is ideal.

Of course this topic is about SDK 2.6, where any memory clock other than ~1000Mhz will hurt your hashrate.
legendary
Activity: 1162
Merit: 1000
DiabloMiner author
Will the CPU bug be gone with reversing to 2.1? How do I do that when I'm on 2.6 already? Heard all over the board going back isn't easy AT ALL.

Thx!

Actually Quite easy.. Let me know if you want help with it.

Except his problem isn't the SDK. There are two known CPU bugs, the first is SDK side and exists in 2.2 and 2.3 only. 2.1, 2.4, 2.5, and 2.6 do not exhibit it, and 2.2 and 2.3 exhibit it on any driver version; the other bug is in driver and exists in 11.7 through 11.11* and exhibit it with any SDK including both 2.1 and 2.6.

* Depends on the user, some had it fixed in 11.9 and 11.10. There are no known instances of this bug in 12.1 and up.
sr. member
Activity: 378
Merit: 250
*Cough*Diapolo VECTORS8 GOFFSET=false*cough*
Play around with the worksize.
hero member
Activity: 497
Merit: 500
Will the CPU bug be gone with reversing to 2.1? How do I do that when I'm on 2.6 already? Heard all over the board going back isn't easy AT ALL.

Thx!

Actually Quite easy.. Let me know if you want help with it.
zvs
legendary
Activity: 1680
Merit: 1000
https://web.archive.org/web/*/nogleg.com
I finally got a 5870 to mess around with, and, yeah, the "best" memory setting on that is diff than the 5830. 

At 1000/395, it gets 440mhash, at 1000/310, 448mhash, etc.

1/3rd core speed is considered the rule. So, in your case, 1000/333.
Small test size, but results were similar to 1000/310.  I know if I set my 5830 to 1000/333 instead of 1000/395, it drops it quite noticeably (agg 9, ~316mhash @ 395 to ~305mhash @ 333)
legendary
Activity: 1162
Merit: 1000
DiabloMiner author
I finally got a 5870 to mess around with, and, yeah, the "best" memory setting on that is diff than the 5830. 

At 1000/395, it gets 440mhash, at 1000/310, 448mhash, etc.

1/3rd core speed is considered the rule. So, in your case, 1000/333.
zvs
legendary
Activity: 1680
Merit: 1000
https://web.archive.org/web/*/nogleg.com
I finally got a 5870 to mess around with, and, yeah, the "best" memory setting on that is diff than the 5830. 

At 1000/395, it gets 440mhash, at 1000/310, 448mhash, etc.
zvs
legendary
Activity: 1680
Merit: 1000
https://web.archive.org/web/*/nogleg.com
I'm having issues keeping my 2nd GPU busy.  Huh

Using the above method with the updated miner script for 2.6, I'm able to keep GPU1 at 99% utilization, but GPU2 bounces between 91-98% on my graph. I've put both on their own CPU to see if that would help but it doesn't seem to have an impact.

GUIMiner Settings on both:
-k phatk AGGRESSION=12 FASTLOOP=false VECTORS2 WORKSIZE=64

Has anyone tried the above with multiple GPU's??

Just switch to DiabloMiner or cgminer already.

i still use phoenix rising  v1.3, hoho
zvs
legendary
Activity: 1680
Merit: 1000
https://web.archive.org/web/*/nogleg.com
My card's highest possible mhash with 2.4-2.5 are at 1060/366Mhz with Phoenix 1.7.2 (exe) phatk2 VECTORS AGGRESSION=13 FASTLOOP=False WORKSIZE=256. The odd 366MHz optimal RAM speed differs on 6xxx, 5850, and 57xx, where 300MHz is where the peak performance is usually found.

Lets see what we find, I was curious what Diapolo's new 2011-12-21 kernel could do, so lots of Phoenix 1.7.2 benchmarks with different kernels, worksize, and RAM speeds (the scaling is similar to standard phatk2), tests were ran for several minutes and shares (unless the setting was obviously poor), with the -a averaging option:

SDK 2.6/11.12 (12.1 is identical)
worksize:25612864
DiapoloVECTORS2366MHz332.99327.86321.16
DiapoloVECTORS21000MHz326.03327.90325.99
DiapoloVECTORS4366MHz219298329.74
DiapoloVECTORS41000MHz255278216
phatk2VECTORS366MHz307.37306.78297.25
phatk2VECTORS1000MHz300.87304.61298.81
phatk2VECTORS4366MHz217.42289.25288.78
phatk2VECTORS41000MHz262.69339.5340.21
phatkVECTORS366MHz326323.5315.3
phatkVECTORS1000MHz317321.2320.5


5830, SDK 2.6 (10.0.851.6)

Diapolo from December 12th, 2011

1060 core, 385 memory (phoenix 1.7.2)

phoenix -k phatk device=1 AGGRESSION=11 fastloop=false VECTORS4 WORKSIZE=64 -a 1000

332.96

395 raises it another point or so.  Never tried 399.

Not going to run my card at 1060 core and 1000 memory setting.  I imagine the ~2% gain you have listed here is more than offset by energy costs associated w/ cooling (and wear and tear on the card, theoretically resulting in faster depreciation of the value of the equipment)

I usually run it at 1000/385.

(ed: for comparison, back in June or so when I put my score on the 'Mining Hardware comparison' wiki page, I got 334, with 1030 core and 385 memory...   but it's worth it for a multi-card setup, anyway...  don't have that core getting smashed 100% anymore)

(ed2:  same thing at 399 memory setting - [335.09 Mhash/sec] [18 Accepted] [0 Rejected] [RPC (+LP)])

legendary
Activity: 1512
Merit: 1036
I'm having issues keeping my 2nd GPU busy.  Huh

Using the above method with the updated miner script for 2.6, I'm able to keep GPU1 at 99% utilization, but GPU2 bounces between 91-98% on my graph. I've put both on their own CPU to see if that would help but it doesn't seem to have an impact.

GUIMiner Settings on both:
-k phatk AGGRESSION=12 FASTLOOP=false VECTORS2 WORKSIZE=64

Has anyone tried the above with multiple GPU's??

On Vista and Win7, try the command line
Start /REALTIME phoenix.exe ...
This sets the miner to the highest possible process priority, ensuring that other OS background stuff don't steal any CPU time from the miner sending work to your GPU.

Additionally, after the realtime option, you can add the command line option setting single-CPU affinity for each miner. /AFFINITY 01 = first CPU core
/AFFINITY 02 = second CPU core
/AFFINITY 04 = third CPU core
/AFFINITY 08 = fourth CPU core

Finally, if you have two console windows open, each with a miner, the selected/highlighted one will get more CPU. Click on the wallpaper so neither are selected. Alternately, under Control Panel -> System -> Advanced -> Performance settings, you can optimize for "background services" instead of "Programs" to avoid a foreground window hogging resources.
newbie
Activity: 7
Merit: 0
Has anyone tried this on a 5850? I'm curious if the same settings would be optimal?
legendary
Activity: 1162
Merit: 1000
DiabloMiner author
Will the CPU bug be gone with reversing to 2.1? How do I do that when I'm on 2.6 already? Heard all over the board going back isn't easy AT ALL.

Thx!

The CPU use bug currently is a flaw of the driver, not the SDK. I currently have no issues on 11.12 combined with any SDK, yet 10.7 through 11.10 cause me problems. SDK 2.2 and 2.3, however, had an identical CPU use bug and were the fault of the SDK (2.1 didn't do it, 2.2 and 2.3 do it with any driver revision); they may be identical bugs, but the source is different than the one people suffer from now. Don't get them confused.

On Windows, it is difficult to revert SDKs, on Linux it is easy. So, ymmv.
Pages:
Jump to: