HD 5970 Folding Farm

[Aurum]

I've got a pile of 5970s and I'm trying to build multi-GPU folding rigs. It's a dedicated PC with the Catalyst 15.7.1 drivers installed and F@H 7.4.15. F@H crashes like crazy trying to get this thing to run. Yea I know, 6x5970s is probably a lower PPD than one RX 480, but they're collecting dust and they'll warm my house until spring when they'll go in the trash.
Leaving the GPU-Indexes at -1 does not work so I set them as 0,1,2,3. Each 5970 card can run as a single GPU or two GPUs. How should I do it, one GPU-index per card or two ???
MSI 890FXA-GD70
8GB RAM
Corsair 1200
Win7-64
AMD Athlon II X2 445
TightVNC 2.6.5 (to run headless folding farm)

[Aurum]

I did a couple of things that got it running after another "Client: local Updating Inactive" freeze. I disabled CPU Unlock in the BIOS. My AMD Sempron 140 had been unlocked to an AMD Athlon II X2 445.
I uninstalled F@H 7.4.15 and deleted its Roaming folder. I then installed F@H 7.4.4 and it started folding at its default Medium Power setting with all 4 cards (5970,5970,5830,5830) having GPU-index -1. The CPU & all 4 GPUs DLed work and began reporting folding progress until I the Full Power setting at which point it is now frozen again.
Have HD 5800 GPUs been deprecated ???
Is this futile to try to fold with 5970s ???
I thought I had a 4x5830 rig running and then when I went to convert it to 5970s it just fails repeatedly.

I have a dozen multi-GPU rigs running in my Folding Farm using various HD 7000 cards. I really wish I had a way to monitor them from my PC. I switch one by one using TightVNC but that's time consuming. I just want to know if they're running normally. Anyone know of a program to do that ??? TIA

[Aurum]

On client "local" 127.0.0.1 Option 'gpu-index' has no default and is not set.
Again & again I keep getting this error on various rigs but especially this one. This bug is turning into a nightmare. I sure wish it would get fixed. The only way I know to recover from it is to uninstall F@H, delete its Roaming folder and reinstall.

[foldy]

Beta Client 7.4.15 has some problems with gpu index so going for 7.4.4 is a good idea.
I guess CPU unlock should not be a problem.
What is frozen, your PC or folding client?
Can you post the fah logfile?

[Nathan_P]

use HFM.net to monitor all your clients, its available here:- https://groups.google.com/forum/#!topic ... GSmesbvNN0

[bruce]

Can you post the fah logfile?

I don't know how many GPUs can be supported by Win7

Also post the output of FAHClient --lspci

Can you get FAH to configure two of them as 4 GPUs?

[Joe_H]

The 5970 is a dual GPU on a single card with a PCI bridge, so folding is going to see it as two separate GPU's. With six in a single system, that is 12 altogether. I recall a mention some time back of the folding client not being able to handle a two digit index value, so that might be part of the problem.

As for being supported, the HD 5000 series still is for folding. However the newest projects run better on later models that have GCN.

[Aurum]

Beta Client 7.4.15 has some problems with gpu index so going for 7.4.4 is a good idea.
I guess CPU unlock should not be a problem.
What is frozen, your PC or folding client?
Can you post the fah logfile?

I tried several CPUs locked & unlocked it does not seem to cause any problems. I've now unlocked all CPUs that allow it.

I gave up on the 5970s for now and will try in again soon using a different MB. I got in some RX 470 and RX 480 cards and filled that MB with HD 7000 cards.

[Aurum]

use HFM.net to monitor all your clients, its available here:- https://groups.google.com/forum/#!topic ... GSmesbvNN0

Thanks! That's just what I need. I'm tweaking in my Remote Access settings now.
Silly me, I'd seen mention of it before but when I tried go to http://hfm.net I got an error

[Aurum]

Can you post the fah logfile?

I don't know how many GPUs can be supported by Win7

Also post the output of FAHClient --lspci

Can you get FAH to configure two of them as 4 GPUs?

It seems the problem comes from mixing models on the same board. I've assembled 3 rigs using only 5970s and they work fine; 2x5970 and 3x5970. I can also fold with multiple 5830s if they're the only cards on the board. Each 5970 appears as two GPUs.

As soon as I get a chance to build another rack I'll see how many I can get folding on one MB. Someone suggested that the Slots/GPU-index only goes to single digits. I just checked both 7.4.4. & 7.4.15 and mine seems to go to infinity.

[bruce]

Someone suggested that the Slots/GPU-index only goes to single digits. I just checked both 7.4.4. & 7.4.15 and mine seems to go to infinity.

The official word is that there may be a limit in the vicinity of 100, but for all practical purposes you might as well call that infinity.

[Aurum]

I now have a 3rd rig folding with 4x5970s = 12 GPUs. It was a chore to get the 4th 5970 recognized but it turned out that I have several defective risers. Risers are made very poorly. They use a stiff thin ribbon cable that gets stiffer with time. They double the ribbon cable, I guess for redundancy, but it increases the stress on the itty biddy solder joints when the riser is bent. I'm thinking of removing the useless blue tape and coating the solder joints with super glue to embed them and protect them from flexing stress.

I've got 10x5970s deployed and 14 more to go

I've got several PRGC 9175-9178s running and they perform so badly. The base credit is a mere 3300 and they produce an estimated credit that's as much as a tenth of other projects. I read in the FAQs about GPU options and I'm trying using max-packet-size=small for all my 5970s. On the other end of the spectrum I'm setting my RX 480s to max-packet-size=big. Can I expect any improvement in PPD

[bruce]

Max-packet-size has had some problems and it's currently prettty nearly non-functional. FAH has a plan to reinstate it in the future but I don't believe you can depend on it at the present time.

[Aurum]

Max-packet-size has had some problems and it's currently prettty nearly non-functional. FAH has a plan to reinstate it in the future but I don't believe you can depend on it at the present time.

Hmm, seems to be working with my one day of experience. Maybe not, maybe just the luck of the draw. I set it to small (5MB) but get jobs larger but they seem to be <10MB. At least I haven't gotten one of the monster projects 10491-10499 assigned to these old weaklings. When they get small jobs they seem to work very well and use much electricity than cryptomining did. The FAQ does not explain if there's some logic such as asking for <5MB but taking the smallest available in the queue.

There is a real need have some logic to better match the scale of a WU to the capabilities of the GPU.

[bruce]

There is a real need have some logic to better match the scale of a WU to the capabilities of the GPU.

Absolutely. The V7.4.4 client has a default setting of 10. There's a large percentage of Donors who don't change that setting, and since most of the new GPU projects are bigger than that, there's a large percentage of projects that are rarely assigned.

Current plan: The next released client will have a larger default setting so owners of projects that are are larger than 10 can be processed at a reasonable across-the-board rate.