• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

new GPU for ATI?

M

magoo

[H]F Junkie
2FA
Joined
Oct 21, 2004
Messages
14,683
When does the new GPU program come out for ATI cards, that will make them as good as nvidia's at folding?

I'm wanting to get two 6950s and flash them before the Rev 1.0 cards disappear.

But my 5850 is not so good at this.....I've heard rumors the new GPU4 is supposed to be pretty good.:D
 
It's supposed to be in Beta, bu tnobody that really knows is supposed to be saying dates yet, from what I have heard. I just got a 6950 a month ago, and it's awesome for everything else...
 
um its already out..

http://hardforum.com/showthread.php?t=1596898

check that thread for all the information on running the new open beta AMD/ATI openCL client..


performance is not at Nvidia level yet but the WU's are small test WU's and not science based. once the science based WU's start there should be a nice increase in PPD since the WU's will use more of the shader cores.

HD5970 does around 10k PPD per GPU. as far as the 6900 series goes we don't know since no ones really tested it nor posted numbers in the beta forum on FF. so if we could get some people with 6970's to run it stock and overclocked it would help everyone out. since we know the 6970 did roughly 7600 PPD on the old client. that should put it in the 14-15k PPD range on the openCL client based on the scaling with the 5800/5900 series between brook code and openCL code. but thats all guessing at this point til we get some real numbers.
 
Last edited:
we are gonna need some really large wu before the ati cards shine...
600 atoms just don't do it... 1200 atoms didn't fill up a 3870 back when 320sp was alot... so until wu get ~5-6000 atoms... ati won't be up to par...

each core on my 5970 in sig gets 9k (going off of fah control ppd )
 
I have a 6850 running at 875 MHz on the core right now and it is getting around 6600 ppd. I think I may be a little CPU limited though because my CPU usage jumps a lot and the GPU usage goes down to less than 100%. Now, I'm still running an Athlon 64 X2 (yes, I know it's time for a CPU upgrade) and I don't think my CPU core is keeping up with the GPU.

(edit) I just looked at AMD System Monitor and noticed that the GPU is only running at 300 MHz. Is there any way to force it to run at the non-2D speed? That might be messing up my results.
 
check what core its using and make sure its core16 and not core11 just in case.

why its still sitting in 2D mode i don't know. it should automatically clock up to 3D clocks. you could try forcing 3D clocks with afterburner but since i've never personally run the client patriot would be the better one to answer that question.

the 6850 should be pretty close to the 5850 which i think is around 7700-7800 PPD.


ok after some testing we found out that anything older then the 11.2 drivers will not run the openCL client. engrchris from the irc channel had the same downclocking issues with the 10.12 drivers and was solved with an update to the 11.3 WHQL drivers.
 
Last edited:
Ok well I went to do something else for a little while and when I came back it was back up to the 3D clocks and looks like it's staying that way. I guess I'll just keep an eye on it for now. And yes it is core16.
 
if you ppd is that high I would say its folding in 3d mode...

I use afterburner to keep an eye on temps and clocks....
folding at 950 gets me the 9k .... the fact that your so close says the wu are waay to small... hopefully that will get remedied better than last time...
 
Code: 
17:00:05:Removing old file 'configs/config-20110329-030457.xml'
17:00:05:Saving configuration to config.xml
17:00:06:<config>
17:00:06:  <!-- FahCore Control -->
17:00:06:  <checkpoint v='3'/>
17:00:06:  <cpu-affinity v='true'/>
17:00:06:
17:00:06:  <!-- Folding Slot Configuration -->
17:00:06:  <client-type v='Advanced'/>
17:00:06:
17:00:06:  <!-- Logging -->
17:00:06:  <verbosity v='5'/>
17:00:06:
17:00:06:  <!-- Remote Command Server -->
17:00:06:  <password v=''/>
17:00:06:
17:00:06:  <!-- User Information -->
17:00:06:  <passkey v='********************************'/>
17:00:06:  <team v='33'/>
17:00:06:  <user v='EngrChris'/>
17:00:06:
17:00:06:  <!-- Folding Slots -->
17:00:06:  <slot id='1' type='GPU'/>
17:00:06:</config>
17:00:06:WARNING: Unit 01: 's slot ID 0 no longer exists and there are no other matching slots, dumping
17:00:06:Sending unit results: id:01 state:SEND project:7144 run:0 clone:21 gen:34 core:0xa3 unit:0x00000025fbcb017c4d77872332e2b53d
17:00:06:Connecting to 128.143.199.96:8080
17:00:06:Server responded WORK_ACK (400)
17:00:06:Cleaning up Unit 01
17:00:50:Slot 01 paused
17:00:50:Slot 01: shutting core down
17:00:50:WARNING: FahCore is known to not shutdown cleanly, killing
17:00:50:WARNING: Killing Unit 00
17:00:50:FahCore running Unit 00 exited
17:01:16:Slot 01 unpaused
17:01:16:Starting Unit 00
17:01:16:Running core: F:/FAHClient/cores/www.stanford.edu/~pande/Win32/x86/ATI/R600/Core_16.fah/FahCore_16.exe -dir 00 -suffix 01 -lifeline 4664 -version 701 -checkpoint 3 -gpu 0
17:01:16:Started core on PID 7916
17:01:16:FahCore 0x16 started
17:01:16:Started thread 19 on PID 4664
17:01:16:Unit 00:
17:01:16:Unit 00:*------------------------------*
17:01:16:Unit 00:Folding@Home GPU Core
17:01:16:Unit 00:Version 2.11 (Thu Dec 9 15:00:14 PST 2010)
17:01:16:Unit 00:
17:01:16:Unit 00:Compiler  : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 15.00.30729.01 for 80x86 
17:01:16:Unit 00:Build host: user-f6d030f24f
17:01:16:Unit 00:Board Type: AMD/OpenCL
17:01:16:Unit 00:Core      : x=16
17:01:16:Unit 00: Window's signal control handler registered.
17:01:16:Unit 00:Preparing to commence simulation
17:01:16:Unit 00:- Ensuring status. Please wait.
17:01:26:Unit 00:- Looking at optimizations...
17:01:26:Unit 00:- Working with standard loops on this execution.
17:01:26:Unit 00:- Previous termination of core was improper.
17:01:26:Unit 00:- Files status OK
17:01:26:Unit 00:sizeof(CORE_PACKET_HDR) = 512 file=<>
17:01:26:Unit 00:- Expanded 42478 -> 171163 (decompressed 402.9 percent)
17:01:26:Unit 00:Called DecompressByteArray: compressed_data_size=42478 data_size=171163, decompressed_data_size=171163 diff=0
17:01:26:Unit 00:- Digital signature verified
17:01:26:Unit 00:
17:01:26:Unit 00:Project: 11293 (Run 22, Clone 30, Gen 0)
17:01:26:Unit 00:
17:01:26:Unit 00:Entering M.D.
17:01:28:Unit 00:Tpr hash 00/wudata_01.tpr:  2539059293 555743409 3624593607 1568244673 2333154561
17:01:28:Unit 00:Working on ALZHEIMER DISEASE AMYLOID
17:01:28:Unit 00:Client config unavailable.
17:01:28:Unit 00:Starting GUI Server
17:01:32:Unit 00:Setting checkpoint frequency: 500000
17:01:32:Unit 00:Completed         3 out of 50000000 steps (0%).
17:06:30:Slot 01 paused
17:06:30:Slot 01: shutting core down
17:06:30:WARNING: FahCore is known to not shutdown cleanly, killing
17:06:30:WARNING: Killing Unit 00
17:06:30:FahCore running Unit 00 exited
17:22:40:Slot 01 unpaused
17:22:40:Starting Unit 00
17:22:40:Running core: F:/FAHClient/cores/www.stanford.edu/~pande/Win32/x86/ATI/R600/Core_16.fah/FahCore_16.exe -dir 00 -suffix 01 -lifeline 4664 -version 701 -checkpoint 3 -gpu 0
17:22:40:Started core on PID 6660
17:22:40:FahCore 0x16 started
17:22:40:Started thread 20 on PID 4664
17:22:41:Unit 00:
17:22:41:Unit 00:*------------------------------*
17:22:41:Unit 00:Folding@Home GPU Core
17:22:41:Unit 00:Version 2.11 (Thu Dec 9 15:00:14 PST 2010)
17:22:41:Unit 00:
17:22:41:Unit 00:Compiler  : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 15.00.30729.01 for 80x86 
17:22:41:Unit 00:Build host: user-f6d030f24f
17:22:41:Unit 00:Board Type: AMD/OpenCL
17:22:41:Unit 00:Core      : x=16
17:22:41:Unit 00: Window's signal control handler registered.
17:22:41:Unit 00:Preparing to commence simulation
17:22:41:Unit 00:- Ensuring status. Please wait.
17:22:50:Unit 00:- Looking at optimizations...
17:22:50:Unit 00:- Working with standard loops on this execution.
17:22:50:Unit 00:- Previous termination of core was improper.
17:22:50:Unit 00:- Going to use standard loops.
17:22:50:Unit 00:- Files status OK
17:22:50:Unit 00:sizeof(CORE_PACKET_HDR) = 512 file=<>
17:22:50:Unit 00:- Expanded 42478 -> 171163 (decompressed 402.9 percent)
17:22:50:Unit 00:Called DecompressByteArray: compressed_data_size=42478 data_size=171163, decompressed_data_size=171163 diff=0
17:22:50:Unit 00:- Digital signature verified
17:22:50:Unit 00:
17:22:50:Unit 00:Project: 11293 (Run 22, Clone 30, Gen 0)
17:22:50:Unit 00:
17:22:50:Unit 00:Entering M.D.
17:22:52:Unit 00:Tpr hash 00/wudata_01.tpr:  2539059293 555743409 3624593607 1568244673 2333154561
17:22:52:Unit 00:Working on ALZHEIMER DISEASE AMYLOID
17:22:52:Unit 00:Client config unavailable.
17:22:52:Unit 00:Starting GUI Server
17:22:55:Unit 00:Setting checkpoint frequency: 500000
17:22:55:Unit 00:Completed         3 out of 50000000 steps (0%).
17:26:04:Unit 00:Completed    500000 out of 50000000 steps (1%).
17:29:34:Unit 00:Completed   1000000 out of 50000000 steps (2%).
17:33:03:Unit 00:Completed   1500000 out of 50000000 steps (3%).
17:36:32:Unit 00:Completed   2000000 out of 50000000 steps (4%).
17:40:02:Unit 00:Completed   2500000 out of 50000000 steps (5%).
17:43:31:Unit 00:Completed   3000000 out of 50000000 steps (6%).
17:47:01:Unit 00:Completed   3500000 out of 50000000 steps (7%).
17:50:31:Unit 00:Completed   4000000 out of 50000000 steps (8%).
17:54:00:Unit 00:Completed   4500000 out of 50000000 steps (9%).
17:57:33:Unit 00:Completed   5000000 out of 50000000 steps (10%).
18:01:02:Unit 00:Completed   5500000 out of 50000000 steps (11%).
18:03:48:Server connection id=4 on 0.0.0.0:36330 from 127.0.0.1:52743
18:03:48:Started thread 21 on PID 4664
18:04:33:Unit 00:Completed   6000000 out of 50000000 steps (12%).
18:08:04:Unit 00:Completed   6500000 out of 50000000 steps (13%).
18:11:34:Unit 00:Completed   7000000 out of 50000000 steps (14%).
18:15:03:Unit 00:Completed   7500000 out of 50000000 steps (15%).
18:18:31:Unit 00:Completed   8000000 out of 50000000 steps (16%).
18:22:01:Unit 00:Completed   8500000 out of 50000000 steps (17%).
18:25:30:Unit 00:Completed   9000000 out of 50000000 steps (18%). 
18:28:59:Unit 00:Completed   9500000 out of 50000000 steps (19%).
18:32:28:Unit 00:Completed  10000000 out of 50000000 steps (20%).
18:35:57:Unit 00:Completed  10500000 out of 50000000 steps (21%).
18:39:26:Unit 00:Completed  11000000 out of 50000000 steps (22%).
18:42:56:Unit 00:Completed  11500000 out of 50000000 steps (23%).
18:46:25:Unit 00:Completed  12000000 out of 50000000 steps (24%).
18:49:56:Unit 00:Completed  12500000 out of 50000000 steps (25%).
18:53:25:Unit 00:Completed  13000000 out of 50000000 steps (26%).
18:56:55:Unit 00:Completed  13500000 out of 50000000 steps (27%).
19:00:26:Unit 00:Completed  14000000 out of 50000000 steps (28%).
19:03:54:Unit 00:Completed  14500000 out of 50000000 steps (29%).
19:07:25:Unit 00:Completed  15000000 out of 50000000 steps (30%).
19:10:57:Unit 00:Completed  15500000 out of 50000000 steps (31%).
19:14:28:Unit 00:Completed  16000000 out of 50000000 steps (32%).
19:17:59:Unit 00:Completed  16500000 out of 50000000 steps (33%).
19:21:30:Unit 00:Completed  17000000 out of 50000000 steps (34%).
19:24:59:Unit 00:Completed  17500000 out of 50000000 steps (35%).
19:28:31:Unit 00:Completed  18000000 out of 50000000 steps (36%).
19:32:03:Unit 00:Completed  18500000 out of 50000000 steps (37%).
19:35:33:Unit 00:Completed  19000000 out of 50000000 steps (38%).
19:39:03:Unit 00:Completed  19500000 out of 50000000 steps (39%).
19:42:34:Unit 00:Completed  20000000 out of 50000000 steps (40%).
19:46:04:Unit 00:Completed  20500000 out of 50000000 steps (41%).
19:49:33:Unit 00:Completed  21000000 out of 50000000 steps (42%).
19:53:05:Unit 00:Completed  21500000 out of 50000000 steps (43%).
19:56:36:Unit 00:Completed  22000000 out of 50000000 steps (44%).
20:00:05:Unit 00:Completed  22500000 out of 50000000 steps (45%).
20:03:37:Unit 00:Completed  23000000 out of 50000000 steps (46%).
20:07:07:Unit 00:Completed  23500000 out of 50000000 steps (47%).
20:10:38:Unit 00:Completed  24000000 out of 50000000 steps (48%).
20:14:08:Unit 00:Completed  24500000 out of 50000000 steps (49%).
20:17:41:Unit 00:Completed  25000000 out of 50000000 steps (50%).
20:21:13:Unit 00:Completed  25500000 out of 50000000 steps (51%).
20:24:43:Unit 00:Completed  26000000 out of 50000000 steps (52%).
20:28:12:Unit 00:Completed  26500000 out of 50000000 steps (53%).
20:31:41:Unit 00:Completed  27000000 out of 50000000 steps (54%).
20:35:11:Unit 00:Completed  27500000 out of 50000000 steps (55%).
I'm running core 16 right now with stock 5850. Update your drivers to 11.3+, as older drivers with the client will force the GPU in 2D mode. I'll provide updates.
TPF is around 3 minutes and 30 seconds. - Project 11293, credit 1835. - I'm going to guess 7,000 ppd
 
Last edited:
Patriot said:
...each core on my 5970 in sig gets 9k (going off of fah control ppd )
Click to expand...

Wow! Is that with the old client or the new one? I had no idea that I had 18K PPD sitting there in my main rig (HD5970). I'm wondering how much of a hit my -bidadv client will take by folding with my 5970 at the same time. Any guesses? I could live with a 4 or 5 minute TPF deduction on the -bigadv client.
 
Wyxe said:
Wow! Is that with the old client or the new one? I had no idea that I had 18K PPD sitting there in my main rig (HD5970). I'm wondering how much of a hit my -bidadv client will take by folding with my 5970 at the same time. Any guesses? I could live with a 4 or 5 minute TPF deduction on the -bigadv client.
Click to expand...

it will murder your bigadv...

15% cpu usage per core 3.5ghz hexacore

remember I am at 950 on my 5970 as well
while we are at low sp usage on small atom count clock rate and ram are overly important...

new client 11.3 drivers though 11.2 gave the same for me... but fore some the 2d clocks dont go 3d while folding...
 
Patriot said:
it will murder your bigadv...

15% cpu usage per core 3.5ghz hexacore

remember I am at 950 on my 5970 as well
while we are at low sp usage on small atom count clock rate and ram are overly important...

new client 11.3 drivers though 11.2 gave the same for me... but fore some the 2d clocks dont go 3d while folding...
Click to expand...


Ouch! Sounds like a big increase in power/heat for little gain when paired with -bigadv. Thanks for the information though Patriot.
 
Yea its great for those people that have an older system with new GPUs but for those of us that run bigadv it still hurts more than it helps your PPD.

So its really not worth it....same goes for NV GPUs, but the hit is not as bad.
 
Oh- my bad, and thank you! I will havea 6950 folding tonight, I don't care what I loosee from my unlocked 720fx. Heck, probably better to go back to three cores and overclock it while running the GPU... I should over double my meager output!
 
I have something to add. I have read with core 16, while playing videos, do degrade the performance. I'm not sure if this includes unstable operation, but I can confirm the performance issues with videos.
 
EngrChris said:
I have something to add. I have read with core 16, while playing videos, do degrade the performance. I'm not sure if this includes unstable operation, but I can confirm the performance issues with videos.
Click to expand...

Even with Catalyst 11.4 preview?

Install those and it shouldnt be an issue.
 
fahv7opencl.jpg


This is fantastic. I'm looking forward to science units.
 
69x0 users just need to watch out for any flash based videos.
It would likely freeze their systems.
 
Hey guys, I've been folding the new core 16 WUs on my 6950 (core @ 840 and mem at stock 1250) and I get about 7k PPD depending on whether or not im using my computer.

As far as flash videos go, just disable hardware acceleration and you can fold while watching flash videos. If you do not disable hardware acceleration, and you're folding the new work units and try to play a flash video the system will completely lock up and require a hard reset.

Hope it helps.
 
mm 7k PPD's not bad but i'd expect to see the numbers go up drastically when larger WU's get released..
 
Indeed, I suspect that there is still a lot of room for optimization as well as my friend who has dual nvidia 460s get's about 9k PPD each iirc, and a 6950 should definitely get more than that once it's optimized and receiving bigger WUs as you said. I'm just glad to even finally have a decent AMD client! :)

Also fyi right now, i believe the client allocates about 400mb of RAM; however, the WUs don't necessarily use all of it, to my understanding.
 
wolfofone said:
Indeed, I suspect that there is still a lot of room for optimization as well as my friend who has dual nvidia 460s get's about 9k PPD each iirc, and a 6950 should definitely get more than that once it's optimized and receiving bigger WUs as you said. I'm just glad to even finally have a decent AMD client! :)

Also fyi right now, i believe the client allocates about 400mb of RAM; however, the WUs don't necessarily use all of it, to my understanding.
Click to expand...


yeah right now the WU's are shader clock dependent which is why for example patriots 5970's are doing ~10k PPD running with the gpu at 950 vs your 6950 at 840.. another good example of showing this difference is the non fermi gpu3 WU's for nvidia cards. my 8800GS at 700/1836/1000 gets roughly 1000 more PPD then a gtx 260 with the shaders at 1512 on the same WU. those WU's are tiny as hell and use less than 120 shader cores.

the ones i really wonder about are these 5k and 6k series cards that can hit 1Ghz+ on the gpu's. i really wonder what kind of numbers those cards will do on these small WU's.
 
Last edited:
o wow, that is a small WU then!

Would overclocking the core make any sort of difference? I can get up to at least 880 stable, i just haven't had the need for that sort of overclock since there hadn't been a good AMD client and games weren't really pushing the card much at stock :p.
 
core overclocks generally have little effect on overall performance.

The usual GPU OCing plan is push the shaders as far as you can, then bump up the core until your PPD stops giving you an increase for each bump.
 
jfb9301 said:
core overclocks generally have little effect on overall performance.

The usual GPU OCing plan is push the shaders as far as you can, then bump up the core until your PPD stops giving you an increase for each bump.
Click to expand...

This isn't nvidia.... shaders are linked to core in ati
 
sirmonkey1985 said:
the ones i really wonder about are these 5k and 6k series cards that can hit 1Ghz+ on the gpu's. i really wonder what kind of numbers those cards will do on these small WU's.
Click to expand...

Might get a 6790 and find out...

 
wow, 1ghz core clocks!? :cool:

I might try overclocking the core once the core 16 WUs are more optimized just to see what kind of PPD I can get out of this thing :)
 
wolfofone said:
o wow, that is a small WU then!

Would overclocking the core make any sort of difference? I can get up to at least 880 stable, i just haven't had the need for that sort of overclock since there hadn't been a good AMD client and games weren't really pushing the card much at stock :p.
Click to expand...


you can try.. typically you can hit higher clocks folding then gaming or at least thats the case on the nvidia side.. my 8800GS gaming 700/1728/1000 folding 700/1836/1000 (difference of 900 PPD). my 8800GT 736/1836/1015 gaming 760/1890/1050 folding(difference of 800 PPD(gpu2 really old WU's)). my gtx 260 630/1350/1161 gaming 650/1512/1115 folding(difference of 1000 PPD(gpu2 old WU's))..

that pretty much shows the difference in game stable overclocking and folding stable overclocking along with the difference in PPD. the downside is the shaders are 1:1 linked with the gpu on the AMD/ATI cards so overclocking will play a much bigger role in performance now and down the road. just play with your clocks. since this isn't bigadv or SMP failure rate doesn't play a part in your completion percentage so if a WU EUE's, big deal. it will just send the WU back to PG and it will be redistributed later. typically if i'm playing around with folding clocks i'll let the WU run. change the overclock as its running. let it continue for 2-3 steps, then try raising it again. if it EUE's then i lower it to the last setting i used before the EUE and let it run the entire WU to see if it fails again. if it fails then i lower it to the next lowest working clock and let it run that WU again. i keep doing that til it can run multiple WU's at that clock setting.
 
sirmonkey1985 said:
you can try.. typically you can hit higher clocks folding then gaming or at least thats the case on the nvidia side.. my 8800GS gaming 700/1728/1000 folding 700/1836/1000 (difference of 900 PPD). my 8800GT 736/1836/1015 gaming 760/1890/1050 folding(difference of 800 PPD(gpu2 really old WU's)). my gtx 260 630/1350/1161 gaming 650/1512/1115 folding(difference of 1000 PPD(gpu2 old WU's))..

that pretty much shows the difference in game stable overclocking and folding stable overclocking along with the difference in PPD. the downside is the shaders are 1:1 linked with the gpu on the AMD/ATI cards so overclocking will play a much bigger role in performance now and down the road. just play with your clocks. since this isn't bigadv or SMP failure rate doesn't play a part in your completion percentage so if a WU EUE's, big deal. it will just send the WU back to PG and it will be redistributed later. typically if i'm playing around with folding clocks i'll let the WU run. change the overclock as its running. let it continue for 2-3 steps, then try raising it again. if it EUE's then i lower it to the last setting i used before the EUE and let it run the entire WU to see if it fails again. if it fails then i lower it to the next lowest working clock and let it run that WU again. i keep doing that til it can run multiple WU's at that clock setting.
Click to expand...

Ah, okay, I'll try it out then :)
 
sirmonkey1985 said:
yeah right now the WU's are shader clock dependent which is why for example patriots 5970's are doing ~10k PPD running with the gpu at 950 vs your 6950 at 840.. another good example of showing this difference is the non fermi gpu3 WU's for nvidia cards. my 8800GS at 700/1836/1000 gets roughly 1000 more PPD then a gtx 260 with the shaders at 1512 on the same WU. those WU's are tiny as hell and use less than 120 shader cores.

the ones i really wonder about are these 5k and 6k series cards that can hit 1Ghz+ on the gpu's. i really wonder what kind of numbers those cards will do on these small WU's.
Click to expand...

sirmonkey1985 said:
you can try.. typically you can hit higher clocks folding then gaming or at least thats the case on the nvidia side.. my 8800GS gaming 700/1728/1000 folding 700/1836/1000 (difference of 900 PPD). my 8800GT 736/1836/1015 gaming 760/1890/1050 folding(difference of 800 PPD(gpu2 really old WU's)). my gtx 260 630/1350/1161 gaming 650/1512/1115 folding(difference of 1000 PPD(gpu2 old WU's))..

that pretty much shows the difference in game stable overclocking and folding stable overclocking along with the difference in PPD. the downside is the shaders are 1:1 linked with the gpu on the AMD/ATI cards so overclocking will play a much bigger role in performance now and down the road. just play with your clocks. since this isn't bigadv or SMP failure rate doesn't play a part in your completion percentage so if a WU EUE's, big deal. it will just send the WU back to PG and it will be redistributed later. typically if i'm playing around with folding clocks i'll let the WU run. change the overclock as its running. let it continue for 2-3 steps, then try raising it again. if it EUE's then i lower it to the last setting i used before the EUE and let it run the entire WU to see if it fails again. if it fails then i lower it to the next lowest working clock and let it run that WU again. i keep doing that til it can run multiple WU's at that clock setting.
Click to expand...


im running 1 GHz on my 5870 , it gets just above 10k PPD.
i can run 1 GHz at lower voltage than i do for a gaming clock , but i cant seem to get much higher than this without getting "unstable machine"
guess i might have to one of these bioses that do well at 1+GHz .

how did you find out that these wu's only use 120 shaders?
i would really like to know.

and wouldn't that mean that a 5770 would see identical performance of a 5870 , since they both have 120+ shaders and run at 850 MHz .

5870 has pretty much double everything else of course , but do those things really matter when folding?
 
Devildoll said:
im running 1 GHz on my 5870 , it gets just above 10k PPD.
i can run 1 GHz at lower voltage than i do for a gaming clock , but i cant seem to get much higher than this without getting "unstable machine"
guess i might have to one of these bioses that do well at 1+GHz .

how did you find out that these wu's only use 120 shaders?
i would really like to know.

and wouldn't that mean that a 5770 would see identical performance of a 5870 , since they both have 120+ shaders and run at 850 MHz .

5870 has pretty much double everything else of course , but do those things really matter when folding?
Click to expand...


i have no clue what these WU's use. i was referring to the non fermi GPU3 WU's which are nvidia 8/9/GTX 200 series only WU's.


as far as running the same between a 5770 and a 5870 at same clocks, thats what me and engrchris are trying to figure out, if we can get close to patriots 5970 single gpu numbers which hes running his gpu's at 950. if we can get close to it then we know for sure clocks are more important then shader cores at this point in time. as far as the 6900 series goes they seem to be way under what they should be which makes me wonder if the 4 way shaders are having an odd impact on performance.
 
Right now I think i am running GPU2 WUs on my 285s.

Would a non-fermi GPU3 give me more PPD production?
If so how do I select those WUs?

Currnetly I am seeing about 494 pts per unit and can crunch them in about 80 minutes on a BFG 285 OC+, I have not tried to OC the GPUs.

Thanks for all the info so far.
 
magoo said:
Right now I think i am running GPU2 WUs on my 285s.

Would a non-fermi GPU3 give me more PPD production?
If so how do I select those WUs?

Currnetly I am seeing about 494 pts per unit and can crunch them in about 80 minutes on a BFG 285 OC+, I have not tried to OC the GPUs.

Thanks for all the info so far.
Click to expand...


hell no, stick with the gpu2 client for the GTX 200 series.. non fermi GPU3 is a friggin joke on the GTX 200 series. only run the non fermi GPU3 with the 8k or 9k series cards. an 8800GS @ 700/1836/1000 does 5600 PPD. the GTX 200 series does between 5k-7k PPD. where as the GTX 200 series on gpu2 does 7-11k PPD depending on the overclock.
 
sirmonkey1985 said:
hell no, stick with the gpu2 client for the GTX 200 series.. non fermi GPU3 is a friggin joke on the GTX 200 series. only run the non fermi GPU3 with the 8k or 9k series cards. an 8800GS @ 700/1836/1000 does 5600 PPD. the GTX 200 series does between 5k-7k PPD. where as the GTX 200 series on gpu2 does 7-11k PPD depending on the overclock.
Click to expand...

OK, thank you.:p
 
You must log in or register to reply here.
Back
Top