• 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.

Preliminary numbers for Nvidia GPU client

MixManSC said:
A paltry 3ghz.... It runs hot too - it's the older core 6600. At least 2 of the cores are always around 70 degrees.... I'm probably going to pull the heatsink and proc and lap them.
Click to expand...

Hey Mix, please keep us posted as to whether the lap job gets gooder' results. The reason I'm asking is I have 2x Q6600's, both WC'ed (bowed Apogee GT), both OC'ed to 3.0 GHz and both G0 steppings. (both fold 24/7 :D)

IMO one of my Q6600's is reading too much delta (7c degrees to sometimes as high as 10c degrees) betwixt the two dies and the cores in the dies. (if that makes any sense :confused:). Although, thank God, I never see any temps as high as 70c probably because of the low OC', the using of the stock vcore (1.35v according to the BIOS) , the G0 steppings, the WC'ing and I try to keep my ambient about 21c to 22c degrees, it's still troubling to a "temp anal" individual like me :p. (of course that may change if it really gets to be a "corker" of a summer, light bill city with the AC on all the time :mad:).

My other quad, identical setup as stated above, has a delta of about 0c degrees (mostly) to 4c degrees (infrequently). The vcore for both processors is 1.35v and both are running the stock vcore of 1.35v according to BIOS. (luck of the draw on the vcore deal too bad I didn't get a chip with a lower lower vcore :( but I only OC' to 3.0 GHz to 3.2 GHz anyway ;))

I have reseated (OBTW in all reseatings I've used AS5 TIM with appropriate "burn in" time) the WB many times trying to cure the temp discrepency, but to no avail. I've thought seriously about lapping the cpu HS, but I don't think I can lap the WB (Apogee GT) because it's already "bowed" :confused:

Thanks :D

 
Has anyones PPD gone down significantly.. for no reason? I was getting about 5200 PPD with just the GPU2 client, but since last night it's been ~1000...
 
Syribo said:
Has anyones PPD gone down significantly.. for no reason? I was getting about 5200 PPD with just the GPU2 client, but since last night it's been ~1000...
Click to expand...
I haven't run the client yet but didn't see anyone post about a major drop. Have you double checked to see if everything is running with no problems?
 
APOLLO said:
I haven't run the client yet but didn't see anyone post about a major drop. Have you double checked to see if everything is running with no problems?
Click to expand...

Yeah :( I haven't changed anything... My temperatures are all the same... The cards clocks are all the same... WTF :(
 
My PPD are consistent. I would guess that maybe your GPU2 client is sitting on a heavily used core and might not be getting the loving it should. Check to see if you SMPs have suddenly jumped.

 
Maybe you're getting some kind of errors in fah or maybe not getting work at all? I assume you checked the log?
 
APOLLO said:
I haven't run the client yet but didn't see anyone post about a major drop. Have you double checked to see if everything is running with no problems?
Click to expand...

Imitation said:
Maybe you're getting some kind of errors in fah or maybe not getting work at all? I assume you checked the log?
Click to expand...

I use FahSpy (For some reason FahMon just will not work for me), and my GPU2 client is running fine.. no errors or anything.

Even with my SMP clients not running, I'm still only getting ~1000 PPD.. I've restarted, shut it down and started again, and nothing is changing it..
 
almost sounds like it is using 2d clocks which generally are much lower than 3d clocks...

 
Syribo said:
Even with my SMP clients not running, I'm still only getting ~1000 PPD.. I've restarted, shut it down and started again, and nothing is changing it..
Click to expand...
Strange. I'm no expert on the GPU2 client but when I was running the GPU1 client and I saw anomalous drops in performance it was almost always due to the clocks on the card slowing down because the temperature monitor would kick in. If that's the case with your card, you will need some app to disable any safety monitoring feature. I don't believe that is the cause but something worth looking into.
 
Sunin said:
almost sounds like it is using 2d clocks which generally are much lower than 3d clocks...
Click to expand...
Could be.

Syribo, see if you can also raise the shader clocks on your card if you haven't already. Apparently it can make a huge boost in performance.
 
GPU client by itself:

GPU: 8800GTS 640MB
CPU: Intel E6400 stock
CPU load while folding: ~55%
Board Connection (PCI-E, AGP, etc.): PCI-E
Memory used by the client: 15MB
Temps of the GPU (NVMonitor): 76C (@100% fan speed)
GPU Clocks: 513/792
PPD (as stated by FAHMON or other program): 3088

 
Dan-515 said:
GPU client by itself:

GPU: 8800GTS 640MB
CPU: Intel E6400 stock
CPU load while folding: ~55%
Board Connection (PCI-E, AGP, etc.): PCI-E
Memory used by the client: 15MB
Temps of the GPU (NVMonitor): 76C (@100% fan speed)
GPU Clocks: 513/792
PPD (as stated by FAHMON or other program): 3088

Click to expand...


Wow, that makes me feel really good about my 8800m gts getting 2500ppd if your 640 is gettting 3088. :D My mobile is running at about 60c with about 50%fan speed.
 
PuppeteerNessus said:
Wow, that makes me feel really good about my 8800m gts getting 2500ppd if your 640 is gettting 3088. :D My mobile is running at about 60c with about 50%fan speed.
Click to expand...

FahMon is now saying 3183 :p

I'm going to OC it a bit and see what I can achieve. I've got up to 680/1020 before...

 
APOLLO said:
Could be.

Syribo, see if you can also raise the shader clocks on your card if you haven't already. Apparently it can make a huge boost in performance.
Click to expand...

My card(s) have both been OCd for quite a while. They're at the highest I can get them to. 765/1911s/1020m
 
Actually, this has always been interesting to me.. The highest I can get my shader clock to is 1910. But it only shows it at 1890. When I upped it from 1910 to 1915, the shader goes up to like 1940 and I run ATI Tool and get artifacts. So I have to leave it at 1910 (Or I guess.. 1890).

I don't know why it stays at 1890... But it's always like that, and I was getting over 5000PPD so... :(
 
Syribo said:
Actually, this has always been interesting to me.. The highest I can get my shader clock to is 1910. But it only shows it at 1890. When I upped it from 1910 to 1915, the shader goes up to like 1940 and I run ATI Tool and get artifacts. So I have to leave it at 1910 (Or I guess.. 1890).

I don't know why it stays at 1890... But it's always like that, and I was getting over 5000PPD so... :(
Click to expand...

From what I remember, shaders only go up by ~50mhz increments. You'll notice the same on the GPU and memory although the steps are much smaller for those two.
 
Syribo said:
Actually, this has always been interesting to me.. The highest I can get my shader clock to is 1910. But it only shows it at 1890. When I upped it from 1910 to 1915, the shader goes up to like 1940 and I run ATI Tool and get artifacts. So I have to leave it at 1910 (Or I guess.. 1890).
Click to expand...
What about trying the client without any tuning app and see what the card does on stock clocks for comparison? I used to do that for troubleshooting.

I don't know why it stays at 1890... But it's always like that, and I was getting over 5000PPD so... :(
Click to expand...
AtiTool is finicky. I always found it difficult to get consistent results from using that app. Maybe you should try something else? What about Riva Tuner, is that any good?
 
Riva Tuner is what I'm using. Before last night I've never once tried overclocking a video card. Took me a bit to figure it out. I need to check the temps. Hope it's not on fire when I get home. The thermal goo that came on it stock looked pretty bad. I might shut it down and redo the stock goo with AS5.
 
MixManSC said:
Riva Tuner is what I'm using. Before last night I've never once tried overclocking a video card. Took me a bit to figure it out. I need to check the temps. Hope it's not on fire when I get home. The thermal goo that came on it stock looked pretty bad. I might shut it down and redo the stock goo with AS5.
Click to expand...
I'm going to give Riva Tuner a try when I finally install the GPU2 client later this summer. BTW, is there now a better thermal compound released by Arctic Silver? Someone mentioned that on another forum.


 
APOLLO said:
I'm going to give Riva Tuner a try when I finally install the GPU2 client later this summer. BTW, is there now a better thermal compound released by Arctic Silver? Someone mentioned that on another forum.


Click to expand...

I use MX-1... I read that it's a small improvement over AS5.


Also, I will try it with my card at stock clocks... But like I said, I have been running the GPU2 client since the day it came out and I haven't overclocked or underclocked my cards at all since then. They've been at the same clocks for months now.

I used to use RivaTuner but then it stopped working with the latest Nvidia drivers.. But I guess it seems to be working for everyone again. I use the EVGA Precision Tool to overclock my cards now. It's actually way more simple to use.
 
APOLLO said:
I'm going to give Riva Tuner a try when I finally install the GPU2 client later this summer. BTW, is there now a better thermal compound released by Arctic Silver? Someone mentioned that on another forum.


Click to expand...

Arctic Cooling MX-2 might be what you're thinking of, and the MX-1 that Syribo mentioned. Noctua also has a paste out, but I don't think it beats AS5.
 
IcedEmotion said:
Arctic Cooling MX-2 might be what you're thinking of, and the MX-1 that Syribo mentioned. Noctua also has a paste out, but I don't think it beats AS5.
Click to expand...

These days you can't really go wrong with many of the aftermarket pastes, I think it comes down to the application of them more than anything IMO. Factory paste is probably decent, or good enough, but the put a ton on or worse, use a thermal pad!
 
Imitation said:
IcedEmotion said:
Arctic Cooling MX-2 might be what you're thinking of, and the MX-1 that Syribo mentioned. Noctua also has a paste out, but I don't think it beats AS5.
Click to expand...
These days you can't really go wrong with many of the aftermarket pastes, I think it comes down to the application of them more than anything IMO. Factory paste is probably decent, or good enough, but the put a ton on or worse, use a thermal pad!
Click to expand...
Yes, the MX-2 compound is what I remember mentioned but haven't seen it for sale. There are good ones from other companies but they're on the expensive side. IC Diamond comes to mind, it's probably the best but at the cost of a HSF.
 
APOLLO said:
Yes, the MX-2 compound is what I remember mentioned but haven't seen it for sale. There are good ones from other companies but they're on the expensive side. IC Diamond comes to mind, it's probably the best but at the cost of a HSF.
Click to expand...

They have the MX-2 on NewEgg, I posted the link for it above :)
 
I'm still unhappy :( Still getting only 1100PPD.. What is going on here!? I can't think of anything that would be wrong... I'm very unhappy :(
 
Syribo said:
I'm still unhappy :( Still getting only 1100PPD.. What is going on here!? I can't think of anything that would be wrong... I'm very unhappy :(
Click to expand...
Syribo, can you post your WU project and frame time average or maybe copy/paste a portion of your recent log.txt from client startup?
 
[17:47:47] - Ask before connecting: No
[17:47:47] - User name: Syribo (Team 33)
[17:47:47] - User ID: 6EF4FD04563E01F8
[17:47:47] - Machine ID: 2
[17:47:47]
[17:47:47] Loaded queue successfully.
[17:47:47] Initialization complete
[17:47:47] + Benchmarking ...
[17:47:48]
[17:47:48] + Processing work unit
[17:47:48] Core required: FahCore_11.exe
[17:47:48] Core found.
[17:47:48] Working on Unit 01 [June 26 17:47:48]
[17:47:48] + Working ...
[17:47:48]
[17:47:48] *------------------------------*
[17:47:48] Folding@Home GPU Core - Beta
[17:47:48] Version 1.05 (Thu Jun 12 15:39:34 PDT 2008)
[17:47:48]
[17:47:48] Compiler :
[17:47:48] Build host: amoeba
[17:47:48] Preparing to commence simulation
[17:47:48] - Ensuring status. Please wait.
[17:47:58] - Looking at optimizations...
[17:47:58] - Working with standard loops on this execution.
[17:47:58] - Previous termination of core was improper.
[17:47:58] - Files status OK
[17:47:58] - Expanded 42527 -> 246249 (decompressed 579.0 percent)
[17:47:58] Called DecompressByteArray: compressed_data_size=42527 data_size=246249, decompressed_data_size=246249 diff=0
[17:47:58] - Digital signature verified
[17:47:58]
[17:47:58] Project: 5004 (Run 8, Clone 9, Gen 4)
[17:47:58]
[17:47:58] Entering M.D.
[17:48:04] Will resume from checkpoint file
[17:48:04] Working on p5002_supervillin_e1
[17:48:05] Client config found, loading data.
[17:48:05] Starting GUI Server
[17:48:05] Resuming from checkpoint
[17:48:05] Verified work/wudata_01.log
[17:48:05] Verified work/wudata_01.edr
[17:48:05] Verified work/wudata_01.trr
[17:48:05] Verified work/wudata_01.xtc
[17:48:05] Completed 38%
[17:49:22] Completed 39%



And then, to be safe, here is after a WU is completed and a new one starts:



[19:31:23] Completed 99%
[19:32:40] Completed 100%
[19:33:40]
[19:33:40] Finished Work Unit:
[19:33:40] - Reading up to 3550680 from "work/wudata_01.trr": Read 3550680
[19:33:40] trr file hash check passed.
[19:33:40] - Reading up to 1131660 from "work/wudata_01.xtc": Read 1131660
[19:33:40] xtc file hash check passed.
[19:33:40] edr file hash check passed.
[19:33:40] logfile size: 110923
[19:33:40] Leaving Run
[19:33:42] - Writing 4794335 bytes of core data to disk...
[19:33:43] Done: 4793823 -> 4113426 (compressed to 85.8 percent)
[19:33:43] ... Done.
[19:33:43] - Shutting down core
[19:33:43]
[19:33:43] Folding@home Core Shutdown: FINISHED_UNIT
[19:33:46] CoreStatus = 64 (100)
[19:33:46] Sending work to server


[19:33:46] + Attempting to send results
[19:34:08] + Results successfully sent
[19:34:08] Thank you for your contribution to Folding@Home.
[19:34:08] + Number of Units Completed: 180

[19:34:13] - Preparing to get new work unit...
[19:34:13] + Attempting to get work packet
[19:34:13] - Connecting to assignment server
[19:34:13] - Successful: assigned to (171.64.65.20).
[19:34:13] + News From Folding@Home: GPU folding beta
[19:34:13] Loaded queue successfully.
[19:34:14] + Closed connections
[19:34:14]
[19:34:14] + Processing work unit
[19:34:14] Core required: FahCore_11.exe
[19:34:14] Core found.
[19:34:14] Working on Unit 02 [June 26 19:34:14]
[19:34:14] + Working ...
[19:34:14]
[19:34:14] *------------------------------*
[19:34:14] Folding@Home GPU Core - Beta
[19:34:14] Version 1.05 (Thu Jun 12 15:39:34 PDT 2008)
[19:34:14]
[19:34:14] Compiler :
[19:34:14] Build host: amoeba
[19:34:14] Preparing to commence simulation
[19:34:14] - Looking at optimizations...
[19:34:14] - Created dyn
[19:34:14] - Files status OK
[19:34:14] - Expanded 42768 -> 246249 (decompressed 575.7 percent)
[19:34:14] Called DecompressByteArray: compressed_data_size=42768 data_size=246249, decompressed_data_size=246249 diff=0
[19:34:14] - Digital signature verified
[19:34:14]
[19:34:14] Project: 5004 (Run 1, Clone 319, Gen 0)
[19:34:14]
[19:34:14] Assembly optimizations on if available.
[19:34:14] Entering M.D.
[19:34:21] Working on p5002_supervillin_e1
[19:34:21] Client config found, loading data.
[19:34:21] Starting GUI Server
[19:35:38] Completed 1%
[19:36:56] Completed 2%
 
Dan-515 said:
GPU client by itself:

GPU: 8800GTS 640MB
CPU: Intel E6400 stock
CPU load while folding: ~55%
Board Connection (PCI-E, AGP, etc.): PCI-E
Memory used by the client: 15MB
Temps of the GPU (NVMonitor): 76C (@100% fan speed)
GPU Clocks: 513/792
PPD (as stated by FAHMON or other program): 3088
Click to expand...

Just tried OC to 680 core/1577 shader/1020 mem

4450 PPD :cool:

CPU usage still at 55%, GPU temp 81C

 
Well now it's saying 4403PPD.. I didn't change anything.. Strange.
 
Syribo said:
Well now it's saying 4403PPD.. I didn't change anything.. Strange.
Click to expand...

I guess there is a struck process sucking a core, deprivating the GPU2 of the needed cycles. After it is gone, this fixed the issue.

 
Yeah - watching pron flicks in media player really drags the ppd down.... :D ;)

 
Right now i am getting 7200-7400 ppd from my gtx280 at 700/1604/1250, this thing kicks my 8800gtx's ass handidly
 
oshia86 said:
Right now i am getting 7200-7400 ppd from my gtx280 at 700/1604/1250, this thing kicks my 8800gtx's ass handidly
Click to expand...

All of a sudden my 8800 GTX doesn't look so shiny and new anymore ;)

This is just what one would need if they were in the middle of a fold down with someone else..... Ok Kendrak, you go buy one and tell me how it works :p

 
HighYield said:
All of a sudden my 8800 GTX doesn't look so shiny and new anymore ;)

This is just what one would need if they were in the middle of a fold down with someone else..... Ok Kendrak, you go buy one and tell me how it works :p

Click to expand...

Well look on the bright side though, the 8800GTX was shiny and new for like 18 flipping months! We probably won't ever see cards that reign supreme for that long ever again.



 
You must log in or register to reply here.
Back
Top