Folding@Home: THGC Needs You -Team 40051

[06yfz450ridr]

finally running my gaming rig for this besides it getting horrible ppd. 10000ppd seems really off for a 7970 and 8320@ 4.8 right? ive tried everything to get it higher with no avail.

my old athlon x2 clocked at 3ghz and gt 545 seem to get very similar scores at times

 

[EXT64]

Hmm, well the CPU alone should be doing better than that.

For the GPU, you need to be running Advanced or Beta. Otherwise you will get the horrible Core 16 WU which really don't run on the new AMD cards. How familiar are you with the V7 folding interface? Just so i know where to start on the explanation. Also, with such high OCes you'll have to keep an eye on the log and make sure you aren't getting periodic failures. And finally, do you have a passkey? Both CPU and new GPU WUs use quick return bonus for most of their points - without a passkey you'll be losing most of the points.

Just as a reference - once we get everything fixed the CPU should be in the 20-30k range or so and the GPU well over 100k.

 

[06yfz450ridr]

Hmm, well the CPU alone should be doing better than that.

For the GPU, you need to be running Advanced or Beta. Otherwise you will get the horrible Core 16 WU which really don't run on the new AMD cards. How familiar are you with the V7 folding interface? Just so i know where to start on the explanation. Also, with such high OCes you'll have to keep an eye on the log and make sure you aren't getting periodic failures. And finally, do you have a passkey? Both CPU and new GPU WUs use quick return bonus for most of their points - without a passkey you'll be losing most of the points.

Just as a reference - once we get everything fixed the CPU should be in the 20-30k range or so and the GPU well over 100k.

thanks for your help. i know a decent bit with v7. i was thinking of trying the advanced flag to see if it helps. the logs show no errors at all but just have been going really slow. also i do have a passkey on both my machines.

 

[EXT64]

Perfect - yeah you definitely want to use advanced or beta for the GPU. It doesn't really matter for the CPU. But for the GPU, the advanced flag lets you get Core 17 work units. The WU you are currently running is a Core 16 unit which uses a really outdated core that is barely working with new AMD drivers. You are probably getting about 2.5k PPD right now :lol:

Core 17 is a completely new core developed with the intent of running the same work on both AMD and NVidia GPUs and greatly increasing the AMD performance. It also expands the types of science the GPUs can do, making them even more useful than they were previously.

Your GPU PPD should jump over 100k, though Core 17 is very cautious with errors so keep an eye out for errors once you get the advanced flag on there.

Edit: How is your CPU doing currently? At 4.8 it should be pretty decent.

 

[EXT64]

Ah wait - is the 7970 in the 8320 computer? If so, I know why it is doing so badly. Core 16 uses up a CPU core, and if you are running SMP 8 that Core 16 CPU usage will cause work imbalance on your CPU WU, leading to bad performance. With Core 17 and AMD, CPU usage will be much less, though I still leave a free core for the GPU work.

 

[06yfz450ridr]

Perfect - yeah you definitely want to use advanced or beta for the GPU. It doesn't really matter for the CPU. But for the GPU, the advanced flag lets you get Core 17 work units. The WU you are currently running is a Core 16 unit which uses a really outdated core that is barely working with new AMD drivers. You are probably getting about 2.5k PPD right now :lol:

Core 17 is a completely new core developed with the intent of running the same work on both AMD and NVidia GPUs and greatly increasing the AMD performance. It also expands the types of science the GPUs can do, making them even more useful than they were previously.

Your GPU PPD should jump over 100k, though Core 17 is very cautious with errors so keep an eye out for errors once you get the advanced flag on there.

Edit: How is your CPU doing currently? At 4.8 it should be pretty decent.

cpu is only getting about 3k on 0xa4 wu seems really low. the gpu is now getting an x17 but is taking forever to start doing anything but now my gpu usage is at 96+% vs 85-75%


funny thing is on boinic i go though wus on my gpu and cpu like crazy

 

[EXT64]

Yeah, building the problem can take a little while, particularly on Project 8900 (but it is worth the wait )

 

[EXT64]

Folding Work Units are really long compared to BOINC. I'm running some that can take 1-1.5 days on multiprocessor servers - so they are huge. And even the normal CPU and GPU WUs can take half or even a whole day on good hardware.

 

[06yfz450ridr]

Folding Work Units are really long compared to BOINC. I'm running some that can take 1-1.5 days on multiprocessor servers - so they are huge. And even the normal CPU and GPU WUs can take half or even a whole day on good hardware.

cpu went to 5k gpu log said i lost lifeline? so i removed it and readded it

 

[EXT64]

Hmm, lost lifeline is usually when the client is stopped. Not sure what happened there (might have been caused by some of the reconfigurings). If it seems to be starting now I wouldn't worry about it.

 

[06yfz450ridr]

Hmm, lost lifeline is usually when the client is stopped. Not sure what happened there (might have been caused by some of the reconfigurings). If it seems to be starting now I wouldn't worry about it.

gpu is now at 100k ppd haha cpu is only at 7k but still way better than a total of 10k thanks for your help

 

[06yfz450ridr]

anyways my gt 545 in my other computer keeps getting memerrors when starting a wu. it is at stock clocks and is detecting it fine. im running the newest beta drivers 328.xx o well.

 

[warmon6]

Maybe no one will notice. 😗

http://img109.imageshack.us/img109/5401/goje.jpg

 

[beanoslim]

Hmm, lost lifeline is usually when the client is stopped. Not sure what happened there (might have been caused by some of the reconfigurings). If it seems to be starting now I wouldn't worry about it.

gpu is now at 100k ppd haha cpu is only at 7k but still way better than a total of 10k thanks for your help

I have my 7950 set with beta flag, only get 8900 wu's for 85-90k.

Set cpu to smp6 and gpu to BETA as EXT mentioned and you should be maximising your ppd.

@Warmon, no one will notice, now fire it up :lol:

 

[lilcinw]

The best part of this addiction is unboxing the new toys.

 

[beanoslim]

^ But having the balls to buy them in the first place is the problem :lol:

 

[warmon6]

Well the good news is, the HD 7850 "Seams" to be working.


Ran a couple of games (World of Tanks and modded minecraft) on it. Vast improvement over the Geforce 210. Cranked everything to the max in those 2 games (excluding AA or similar stuff like that. Never really bothered with it as I dont see much of a difference) and still had higher FPS than I would of seen at the lowest setting on the 210.

Of course, I didn't get the 210 at the time for gaming. It was meant to be for an HTPC/folder system (cpu folder) but the family decided they wanted the system for something else and that included gaming.





Only issue i'm having with this card is that it is instant EUE'ing WU's. (I'm not even seeing a load being put on the gpu before it EUE's).

Thought there was some remaining Nvidia drivers that i missed somewhere, cleared everything out related to nvidia. Problem still exsist.

Reinstalled the client just be sure there wasn't something wrong with it, still no solution.






The core error code I'm getting is "114 = 0x72"

I'm getting this error on both the Core_16 and 17. So I dont think it's a Core/WU issue but maybe it is....



Now sadly, I'm not able to provide the logs ATM (beyond remembering the error code) as someone else is using that computer.


This gpu is running on Windows XP sp3 (really need to upgrade to windows 7 but there is to many programs that my family doesn't want to lose.....) and it's running with the 13.4 driver.

Haven't even attempted to overclock this card yet so I at least know that isn't the issue.


Maybe you guys that fold on AMD cards might now what's going on? I'm looking around for answers myself but All i'm seeing is the normal response of "overclocking or bad WU's".

http://foldingforum.org/viewtopic.php?nomobile=1&f=19&t=24655





I'll try getting the log post here when I get the chance.

 

[beanoslim]

Only thing I can suggest is uninstalling everything including the f@h client and data files.

Run something like driver cleaner pro in safe mode to clear all Nvidia and AMD file.

Clean registry.

Reboot

Re-install GPU drivers, then f@h client and try again.

I have had to do this a couple of times when switching cards on the same rig and this seemed to be the only cure.

 

[EXT64]

Next time you are on the computer, could you try running fahbench: http://proteneer.com/blog/?page_id=1671

Assuming that still fails we can use it for testing without killing WUs.

 

[warmon6]

Once i get the chance, i'll take your guy's advices.


@EXT, yeah... forgot that fahbench even existed in all my excitement of this card. :lol:


Well I'll report what I find and comeback in a bit.

 

[anti-painkilla]

^ But having the balls to buy them in the first place is the problem :lol:

If I got rid of my vehicle I might be able to afford to buy them. I am sure tanks get better mileage than me.

 

[06yfz450ridr]

Once i get the chance, i'll take your guy's advices.


@EXT, yeah... forgot that fahbench even existed in all my excitement of this card. :lol:


Well I'll report what I find and comeback in a bit.

i thought the beta wu only work for the beta team and you have to sign up in order to even get them. the advanced flag seems to be working for me. may be able to get even higher than 100k if i do tweaking

 

[EXT64]

Yeah, beta is for the beta team but it is open - anyone who puts the beta flag might get beta WUs. In general though they won't get better PPD than normal and advanced WU and there is even the change you'll get more glitches. However if you aren't on the beta team you can't get help with those glitches (or report them). At this stage though with GPUs, beta is the same as advanced. There are no beta GPU Wus currently, so the assignment server will instead assign advanced or normal WU.

I think the beta assignment server may be a different server though, so you might get a slightly different mix of WUs than advanced/normal even if there are no beta WUs.

 

[warmon6]

Welp..... Tried everything everything out and sadly it still didn't work. Even FAHbench was instantly giving out error's.

Although I've finally found out why.....


http://foldingforum.org/viewtopic.php?f=61&t=24391

Re: GPU process fails hash check
Postby Joe_H » Fri Jun 07, 2013 8:48 pm

AMD-ATI dropped support for OpenCL on Windows XP over a year ago. I forget the number of the release of the drivers that had the last supported OpenCL package, but it was around 12.1 or earlier. So if you have a newer version you will need to completely remove it and revert to an older Catalyst version with OpenCL support from the SDK. Also check through the requirements listed here in the Windows Install Guide.

Re: GPU process fails hash check
Postby 7im » Fri Jun 07, 2013 9:51 pm

The 11.12 AMD device driver is the last version to support OpenCL on Windows XP

Well that sucks..... I wasn't even aware of that. Although it's not surprising that such an event for an old OS would happen.

Well, guess I wont be folding long term on this card for now (until I upgade the OS) and I don't feel like rolling the drivers so far back to just get a program or 2 to work.

Although when I get the chance, with the windows 8.1 preview that running on my laptop HDD, i'll run F@H off that and give you guys the numbers.

Guess we all learned something today..... no windows xp gpu folding for me. :lol:





Also here was the log that helped me find the answer.

Spoiler

03:22:14:WU00:FS00:0x16:Folding@Home GPU Core
03:22:14:WU00:FS00:0x16:Version 2.11 (Thu Dec 9 15:00:14 PST 2010)
03:22:14:WU00:FS00:0x16:
03:22:14:WU00:FS00:0x16:Compiler : Microsoft (R) 32-bit C/C++ Optimizing Compiler Version 15.00.30729.01 for 80x86
03:22:14:WU00:FS00:0x16:Build host: user-f6d030f24f
03:22:14:WU00:FS00:0x16:Board Type: AMD/OpenCL
03:22:14:WU00:FS00:0x16:Core : x=16
03:22:14:WU00:FS00:0x16: Window's signal control handler registered.
03:22:14:WU00:FS00:0x16😛reparing to commence simulation
03:22:14:WU00:FS00:0x16:- Looking at optimizations...
03:22:14:WU00:FS00:0x16😀eleteFrameFiles: successfully deleted file=00/wudata_01.ckp
03:22:14:WU00:FS00:0x16:- Created dyn
03:22:14:WU00:FS00:0x16:- Files status OK
03:22:14:WU00:FS00:0x16:sizeof(CORE_PACKET_HDR) = 512 file=<>
03:22:14:WU00:FS00:0x16:- Expanded 44852 -> 171163 (decompressed 381.6 percent)
03:22:14:WU00:FS00:0x16:Called DecompressByteArray: compressed_data_size=44852 data_size=171163, decompressed_data_size=171163 diff=0
03:22:14:WU00:FS00:0x16:- Digital signature verified
03:22:14:WU00:FS00:0x16:
03:22:14:WU00:FS00:0x16😛roject: 11293 (Run 28, Clone 282, Gen 45)
03:22:14:WU00:FS00:0x16:
03:22:14:WU00:FS00:0x16:Assembly optimizations on if available.
03:22:14:WU00:FS00:0x16:Entering M.D.
03:22:16:WU00:FS00:0x16:Tpr hash 00/wudata_01.tpr: 1429627649 967167326 3540074980 3241118914 319290386
03:22:16:WU00:FS00:0x16:Working on ALZHEIMER DISEASE AMYLOID
03:22:16:WU00:FS00:0x16:Client config unavailable.
03:22:16:WU00:FS00:0x16:Finished fah_main
03:22:16:WU00:FS00:0x16:
03:22:16:WU00:FS00:0x16:Successful run
03:22:16:WU00:FS00:0x16😀ynamicWrapper: Finished Work Unit: sleep=10000
03:22:16:WU00:FS00:0x16:Starting GUI Server
03:22:26:WU00:FS00:0x16:Reserved 0 bytes for xtc file; Cosm status=0
03:22:26:WU00:FS00:0x16:Reserved 0 0 786430464 bytes for arc file=<00/wudata_01.trr> Cosm status=0
03:22:26:WU00:FS00:0x16:Allocated 544 bytes for edr file
03:22:26:WU00:FS00:0x16:Read bedfile
03:22:26:WU00:FS00:0x16:- Checksum of file (00/wudata_01.edr) read from disk doesn't match
03:22:26:WU00:FS00:0x16:edrfile file hash check failed.
03:22:26:WU00:FS00:0x16:
03:22:26:WU00:FS00:0x16:Folding@home Core Shutdown: FILE_IO_ERROR
03:22:26:WARNING:WU00:FS00:FahCore returned: FILE_IO_ERROR (117 = 0x75)
03:22:26:WARNING:WU00:FS00:Fatal error, dumping
03:22:26:WU00:FS00:Sending unit results: id:00 state:SEND error😀UMPED project:11293 run:28 clone:282 gen:45 core:0x16 unit:0x000000606652edbc4d94bab05bef9fc2
03:22:27:WU00:FS00:Connecting to 171.67.108.44:8080
03:22:27:WU01:FS00:Connecting to assign-GPU.stanford.edu:80
03:22:27:WU00:FS00:Server responded WORK_ACK (400)
03:22:27:WU00:FS00:Cleaning up

 

[EXT64]

Ah, bummer. I don't think a 7xxx card would even be supported on that old of drivers (the 7970, the first of the 7xxx to be released, was released 12.1 (January 2012), so 11.12 would be no good.