Folding Guide for ATI 5k and 6k series

EngrChris

Gawd
Joined
Dec 28, 2008
Messages
855
Beta V7 Client Guide for ATI 5000 and 6000 series and above cards.

With the new client, some ATI cards can now use the new OpenCL core. Only the V7 client can use this new core. The client and core is still beta and experimental, other applications that use Flash, DirectX, and OpenGL can create instability with the core and the drivers. Always upgrade to newest Catalyst version while using the new core. Please be careful, as myself and others have found issues such as Flash stated above. While OCing, I’ve found possible card and driver instability of creating a hard BSOD system crash.

UPDATE: I have found a tiny issue for those who wish to use the older 4000 series cards. The client-type/Advanced option would run the OpenCL core on the 4000 cards. The option should only exist on the "slot" level, not "client" wide. The guide has been fixed.
- EngrChris, 6/2/2011.

UPDATE: The guide has been updated for newer V7 Clients, 7.2.9+.
- EngrChris, 12/23/2012.

https://fah-web.stanford.edu/projects/FAHClient/

1. Install the Beta V7 client.
70404503.jpg




2. Click on the “Configure” button for configuration dialog window.
51909700.jpg




3. Fill out donor information under the “Identity” tab.
25981465.jpg




4. Under the “Slots” tab, click “Add”.
37171792.jpg




5. For older clients, use 5A, or newer clients that have gpu-index, use 5B.

5A. Under the GPU group, click the radio button
, then click “Add” under the “Extra slot options (experts only)” group.
28916131.jpg


5B. Normally the client will auto-detect what slot to use for every gpu. If not, select the matching slot for gpu-index, cuda-index, and opencl-index.
CwbHa.jpg


6. This step is only required for older clients that do not have an updated card detection list.
In the “Edit Options” dialog, fill out “client-type” for Name, and “Advanced” for Value. Click “Ok”


80421644.jpg




7. Click "OK", (Older client screenshot used.)
64685764.jpg



8. Click “Save” for the configuration dialog window.

97657445.jpg



58386099.jpg


The client will start downloading the new core and a workunit. Post if you have any issues or comments.

- EngrChris
 
Last edited:
That's pretty sweet. What kind of PPD are we looking at for the higher end 5 and 6 series cards? Is it worth running them now with this beta core?

Also, you might put a handy link at the top to tell everyone where to get the v7 client cause ppl like me who haven't kept up with some of the news in the last few months don't know where it is :)
 
That's pretty sweet. What kind of PPD are we looking at for the higher end 5 and 6 series cards? Is it worth running them now with this beta core?

Also, you might put a handy link at the top to tell everyone where to get the v7 client cause ppl like me who haven't kept up with some of the news in the last few months don't know where it is :)

There is a thread on here for the new client. We are hitting 6-9k ppd on most cards.
https://fah-web.stanford.edu/projects/FAHClient/
 
Just downloaded today. Anyone have any success using HFM.NET with this. If so could you shed some light on what you did.

Thanks
 
Just downloaded today. Anyone have any success using HFM.NET with this. If so could you shed some light on what you did.

Thanks

V7 client was completely rewritten and can not work with HFM.NET, or any other tracker.
 
I've got an app to convert V7 logs to V6, just need to update it to correct the queue.dat - it's coming ........

H.
 
This client seems to like my 5850 quite a bit more than v6! I was getting about 3.5kppd previously, and now it seems to be more like 8k+ppd
 
I wish I could say Folding on ATi/AMD hardware was worth it at the moment.
The core and client still have a long way to go.
Taking up an entire CPU core and 400MB of RAM?
Random accelerated video freezing on the 69x0 series.
I think I'll wait until it all pans out.

Nice guide BTW, OP.
 
With my i7 920, should I just run the SMP client and wait for a better ATI client?
 
With my i7 920, should I just run the SMP client and wait for a better ATI client?
If you can run bigadv on your 920, do it.
Otherwise you might gain PPD by running both clients simultaneously.
Just make sure you lock the ATi client to one core.
 
Yeah, I can run them, but I need to get the mandatory 10 work units before the bonus points come into play. Just started the CPU last night. Read somewhere it was worth it to do the 10 as regular non bigadv units before running bigadv, sound about right?
 
Paging Sirmonkey and/or Tobit. I have an issue.
Fix found, I'll need to update the guide.
 
Last edited:
engrchrisocl48704.jpg


Interesting issue isn't it? Apparently the 4000 series can run the OpenCL core and workunits. I'm not sure how on the software level, but Stanford didn't disabled (blacklisted) the cards. I posted the screenshot to inform how bad the performance, currently, is.
 
interesting discovery Chris

Great job on the guide though. Thanks for putting this together.
 
I used this guide last night. V7 is really slick, the way it should have been for a while. Using the advanced options you can set up and monitor both a CPU and GPU with one download. I did note that it was estimating my 4850 was going to take about 2 days to complete a WU! I don't remember which one, but that seems excessive compared to my roughly equivalent 8800GT (even accounting for the Nvida speed advantage).
 
With the discovery of the 4000 series folding OpenCL units, if its running OpenCL at all, the performance really is terrible. On the 5000/6000 series, the test units have a TPF of around 3-4 minutes, with overclocking, 2:45-3:00 is realistic. My 5850 can do 3:30 at stockm to 2:50 with extreme overclocking.

I added my 4870 card, and ran with the “client-type/Advanced” flag. The 4000 series can run the same OpenCL workunits as on the 5000 and 6000 series cards. PG and others have said and noted of performance issues and other things with the 4000 series.

With testing and overclocking, I found the TPF times of 19:16 with a 4870 at 770 MHz core clock. With overclocking, ~860MHz could do 17:52. I do not know how far theses cards can go. Maybe, they’re at the limit, but I do not believe the 4000 series can have 15 minute frame times.

You can forget about folding on 4000 series cards.
 
Question about the v7 client. It currently shows I'm running a x11 fah core. Since I have a 5770, it "should" be able to run the x16 core. I looked for a way to check for core updates, but can't find any.

Any suggestions?
 
Nice guide!

Seems like 5870 PPD is around 9000ppd @ 900mhz. Increasing & decreasing more or less linearly depending on the OC.
 
Great, they figure this out NOW? right when bitcoining is in crazy mode.. Good news for all those cheap cards after the craze ends though!
 
You need to be in Expert mode.
On the main screen look in the upper right corner, there is a drop box.
Change Novice to Expert.
 
Ah! The window defaults smaller than needs be I see. Had to make the client larger before that window appeared. Thanx.

Edit: So I followed the guide to a T. My 7970 is just sitting there idle. The client found my GPU fine on it's own. When that failed to work, I deleted the GPU and "added" it again. Still nothing. Is it possible the 12.11 beta drivers are an issue?
 
Last edited:
wait what?

did I miss something? FAH sucks balls on AMD cards last I knew, drags down the CPU so bad you make more points without them
 
wait what?

did I miss something? FAH sucks balls on AMD cards last I knew, drags down the CPU so bad you make more points without them

yep it still sucks balls and always will and i have a good bet we'll never see those new WU's in beta testing for nvidia cards ported over to openCL for AMD cards even though they can easily do them, oh well.
 
Stanford is the 2nd best thing ever happened to bitcoin

after paypal
 
Stanford is the 2nd best thing ever happened to bitcoin

after paypal

lol Pande Group is the best thing to happen to all other DC projects.. AMD cards completely destroy nvidia in every DC project besides F@H.
 
hold up Hexen, the straight dope is that it is not worth doing...due to Stanford's continued poor optimization on their non-favored GPU hardware you will make more points idling the cards and folding SMP. I don't like it either.
 
Or do what everyone else does, fold with CPU and make your power costs back with your GPU mining.


Win/Win?
 
hold up Hexen, the straight dope is that it is not worth doing...due to Stanford's continued poor optimization on their non-favored GPU hardware you will make more points idling the cards and folding SMP. I don't like it either.

Sadly, this is true. I have 2 7970's and don't fold on them because it's not worth it. However, I HAVE been able to properly fold on them though..

What client are you using? 7.1.52 or 7.2.9? If 7.1.52 try getting the newer 7.2.9
 
Hello;

I just downloaded the FAH 7.2.9 client and I'm running it. I honestly don't know much about this folding and DC; I'm a complete and total n00b!

Now, I have a Radeon 7970 and an i5 2500K - running at a conservative 4.0GHz. Now, for some reason I thought the GPU would run faster... but after reading this thread I figure that maybe it's not?

It does seem to be folding on the GPU; but in the same time that the SMP did 6.13% of progress the GPU has only done 1.18%. Does that seem correct?

Heck, as I'm reading, is using the 7970 even worth it?

Also, I've seen some people talk about "trackers" and "setups"... so I'm wondering, where could a complete n00b take a look to understand the more interesting parts of this folding thing? Are there optimizations to be done to increase performance or is this just a setup and go thing?

Thanks in advance!

For the [H]orde!
 
there is a sticky at the top of distributed computing that is pretty good reading.

what you running for OS? Windows?
 
Hello; I was out on vacations. Sorry for the delay in answering.

Yep, Windows 8. I'll look at the sticky ;)
 
Back
Top