If I add extra GPUs to my folding machines (I have a GTX 295 and a GTX 460/768MB I can add, depending how many extra power cables I can find), do they still need dummmy plugs, or can I just stick them in the system(s) and set the clients up? I use V7 under Windows 7 on one system and XP on the other (can upgrade that to 7 if it makes a difference, just don't like being constantly told it's a dodgy copy). I've found the power cables for the XP system, and should find them for the other system in the next day or two (too many boxes everywher and no room to sort thing!). Also, will the 295 fold as two GPUs, needing two extra folding slots, or will v7 see it as a single card and use both GPUs in one slot? Makes little, if any, difference right now, but if they bring in QRBs for GPUs it will make a big difference
im sure ive red that GPU's with different architecture dont fold well anymore when im folding with my GTX480 and GTX560 one of then take a serious ppd hit... but if i fold on them separately there fine V7 will detect the different card automatically
Thanks guys. I found some more info at the f@a forums as well. I'll be trying it out over the next couple of days
I'd echo the comments above; - Dummy plugs haven't been required with the newer clients for quite some time (even the newer v6 client) - Mixing GPU's in the same machine is possible, but, they usually don't work as well as if in separate machines. - The 295 should be picked up as two GPU's, but since I don't have one, I can't confirm that.
You shouldn't mix GPUs of a different family (Fermi, Kepler etc.), it still causes big problems. The 295 will show up as two GPUs but not for long. I've had three of them Folding, all blew up under the stain. In fact all of that generation (280, 285 etc.) where well known for destroying themselves.
I'm not concerned if the card blows up, as long as I get a few days out of it To get the 295 working alongside the 680, I had to do the following: Run the client At the top of the log, it lists the GPUs in the order the clients see them. I took a note of that. I ran GPU-Z, and noted that it sees the cards in the opposite order (It's fah that has the order wrong) I edited the setings for the existing gpu slot and added two more. In each slot, I set the gpu index to the index shown by fah I then set the cuda and opencl indeces to the settings shown by gpuz I restarted the clients and all are folding correctly. I'll know soon if they give me proper points
I was, they were "top of the hill" when I folded with them That is why I only do it on a lot of "cheap" cards now, so I don't cry when one dies. Have you tried stopping the client and restarting it? Does it remember the order of the cards? I have a feeling that it doesn't and that is the problem.
Yes, and yes so far. I rarely turn my working computers off or reboot them. I don't think that will be a problem with v7 though as it uses a config file to save all the settings. I restarted it 4 times just to be sure, and so far I'm confident that it is working. Not sure yet if it's worth the extra work, but every little counts
debs... can you explain how you configured v7 to display the correct GPU names as my GTX680 and GTX480 are opposite lol cheers
Can you guys tell me if I need to play around with the v7 client to get my kelper cards working? Been a long time since I used it. Going to try get them folding soon.
How I did it was: Run the v7 app. Check in the log to see the number given to each card, and log that. Run gpu-z, and check the number against each card in that. Configure v7 for each slot. Give gpu-index the number given by the client for each card Give opencl-index and cuda-index the number given by gpu-z for each card. fah lists the cards in the wrong order (eg, I have a GTX 680 in slot 1 one and a GTX 295 in slot 5. FAH has the 295 as using index 0 and 1 and the 680 as index 2, where gpu-z correctly lists the 680 as index 0) If that's not clear enough, I'll t ry to give a better explanation
If you are using the latest v7 it will detect and use the 680. It will say it has a GF114 GPU, but will know it is a 680 and not Fermi. It's not yet optimised for Kepler, but folds fine without any complicated configuration.