970s...

  • Thread starter Deleted member 88227
  • Start date
D

Deleted member 88227

Guest
Figured I'd see what they can do. Sorry for anyone I mow in the process. Not sorry.

So far they're doing great.
 
Whoa what happened to the team?? How in the heck am I in the top 5 PPD?
 
well..... in short...... there was a move to cancel bigadv or something...... essentially it was worthless to run 4P rigs as 1 or two maxwell cards made the same PPD..... some people got angry. Stanford still does not care what users say.......

It's essentailly a similar story as always, but enough people got sick of the roller coaster and shut their rigs down.

I know the details are not clear and there might be some differences with reality. But keep in mind the Dumpster of FAH is filled with bees and broken glass. You don't really want to crawl into that dumpster by asking questions like yours....
 
Well that makes sense. My two GTX 970s are crushing PPD. No wonder EVGA's daily production is so high now.
 
been out of the game for a long time but I've got a 970 sitting here looking pretty. Whats the best client for me to use? 300K is about what I got with my old 48 core and 12 core rigs running at the same time :( o the days have changed lol

o shit, im 95th on the team now, so much for top 30 lol. Guess I need to try again
 
Just run the regular client with a passkey. You'll get between 240 and 300 ppd depending on luck.
 
we used to have certain flags and what not to note in command line, do we still need to mess with that? I got the client installed and running, and I disabled the GTX650 client as well as the CPU client. Just the 970 for now

anything I need to do special to get teh best work?
 
damn. 6 frames in it's guessing 230k ppd..... damn the times have changed lol
 
No and you can run the CPU and 650 with it as well. You should be in the 300k+ range with everything going.
 
I've got the CPU doing non stop bluray encodes for the next few foreseeable days, I'm gonna let this 970 cook in and put out some numbers tho, not sure how good the cooling is on that 650, may let it run some too, I mean after all it's just a physx card lol. Thanks for the help tho. Any word on if I should use any flags to get certain units or did they stop doing that for gpus?
 
I'm not sure. I didn't look that far into it. I only ran the GPU client for about a week before I shut them down.
 
Do not run a Maxwell and kepler card together at the moment - they don't play nice and need different driver versions to perform at their best.

Unless you want to run beta WU just install the client and let it fold. If you want max points run the gpu under Linux but that would involve a dedicated F@H machine unless you can live with Linux on a daily basis
 
ah my days of chasing top 10 are long gone, though it would seem it's not as near as hard as it was back when I was in 13th lol, that shit took 19 GPUs to maintain lol. I'll stick with windows
 
Imagine what 19 GPUs will do now. With less power at that.

~5.7M PPD if they're all 970s.
 
Wow, my 780ti is cranking out a little over 200k ppd and that gets me into the top 20. Insane.
 
4 8800GT's , 3 GTX 260's, 4 9800 GX2's (8), 2 295's (4)

Those were the days, back when I was [H]ard lol, had a top of 26 clients at one time with these and a few CPUs in the mix and in other peoples boxen

Cards.jpg


You should have seen my power bill lol. And all those were in a slew of P4 powered rigs, not exactly power efficient to start with lol
 
Dang I bet that power was crazy not just for running them but for running the AC more to compensate the extra heat.

I think this winter I'm gonna heat the house with GPUs.
 
Vaulter98c said:
I've got the CPU doing non stop bluray encodes for the next few foreseeable days, I'm gonna let this 970 cook in and put out some numbers tho, not sure how good the cooling is on that 650, may let it run some too, I mean after all it's just a physx card lol. Thanks for the help tho. Any word on if I should use any flags to get certain units or did they stop doing that for gpus?
Click to expand...


You can force the 650 to get core 15 units by adding: unit-type beta & max-packet-size small to the slot config for the card (assuming FAH v7). I find the kepler cards get better ppd on core 15 but theres no QRB for those anymore. My 650Ti only gets 10-19K ppd most of the time. A core 18 takes over a day on it and some core 17's take over 2 days for even less ppd so I flagged it for core 15. Driver version doesn't seem to make a diff on core 15 either so no need to start doing driver trials (I did). Just remember to set the GPU and slot instances to match the cards so they stay in order. Welcome back to the fold ...
 
I guess I lucked out on mixed cards.....

My 570 runs just fine with my 970. These are the only two clients I am running and I am a solid top 10 folder. Not that I would brag about that. I still remember the days of having to run 6 machines and 10 clients just to be in the top 40. My two clients still make more than I did back then.

/waves to vaulter
 
OK, finally got mine up and running. Took a while due to a stupid Z9PE board and B0 stepping cpu's but it works. Power delta is 130w over idle, total system load is 235w from the wall and PPD on core 18 is 283k:)
 
Nathan_P said:
OK, finally got mine up and running. Took a while due to a stupid Z9PE board and B0 stepping cpu's but it works. Power delta is 130w over idle, total system load is 235w from the wall and PPD on core 18 is 283k:)
Click to expand...

Nice PPD, which core 18 project?

What sort of issues are you having with the Z9PE board. I have one that I'm ready to take to the shooting range and leave it behind. ;)
 
A wide variety of core 18 projects - however I can't get hfm to see any of my folding boxes - they all worked before I shut down a couple of months back so it must be a network/windows issue. PPD from that one came from 9113 I think - others have gone from 260-295k PPD. no core 17's yet

As for the Z9PE, well my B0 stepping sandy xeons are not supported after bios 3506, older bios's work better with the cpu's but have all sorts of other issues. Once you have installed bios 3105 you cannot flash back, 3302 just doesn't work and even with a working bios, 3506, it still refuses to boot on occasion, this has been an issue since the start however so its nothing new - just frustrating. Now its running its stable as a rock - no issues or crashes so far.

For me 3506 has worked pretty well but others have had issues. If you have retail or Cx ES chips its a much better behaved board. There is a 200+ page thread over at OCN with all the info you will ever need on it.

Oh and be careful with a replacement BIOS chip from the net - mine went into an endless bios recovery loop, however putting the old chip back in did allow me to boot long enough to flash to 3506.

Also forgot to spec cpu's etc. B0 2665 @2.4, 1.35v quad channel corsair DDR3 @1333 and an AX750 gold. Going to look at down clocking, disabling turbo and HT to see what that does to power consumption
 
Nathan -

A large variety of more "modern" gpu projects are on a super secret psummary. I will dig for a link if I still have one. I ultimately gave up and just stopped using HFM as I have been only folding on one box and or only folding v7 GPU so I can just have the client measure all the other clients I might be running.

John

Edit:

http://assign.stanford.edu/api/project/summary

http://assign2.stanford.edu/api/project/summary

Article is here https://folding.stanford.edu/home/new-psummary-page/
 
Hi Guys-- Been a long time since I was folding. Just wanted to give it a go a little bit more. I have two GTX970s on a p8z77 pro mobo. I can get one GPU folding but when I try to get the other working it appears to be running both clients on the same GPU. Only one GPU is heating up. FAHControl indicates there are two slots in use one for GPU0 and one for GPU1. PPD is only listed as 250-300k/d so it looks like only 1 GPU is actually folding. Anyone know what might be wrong?
 
When you start the client and look in the log does it actually list 2 gpu's or just the one?

The log should look like:-

02:02:58: GPUs: 4
02:02:58: GPU 0: NVIDIA:3 GK110 [Tesla K40m]
02:02:58: GPU 1: NVIDIA:3 GK110 [Tesla K40m]
02:02:58: GPU 2: NVIDIA:3 GK110 [Tesla K40m]
02:02:58: GPU 3: NVIDIA:3 GK110 [Tesla K40m]
02:02:58: CUDA: 3.5
02:02:58:CUDA Driver: 7000

Yours should read gpus 2 and list 2 GM 204. If it doesn't the client has only detected 1 gpu. In which case try reinstalling the drivers and/or reinstalling F@H
 
I got them working now. Not sure what was happening. Went in and set everything manually instead of the default -1 option.
 
Vaulter98c said:
been out of the game for a long time but I've got a 970 sitting here looking pretty. Whats the best client for me to use? 300K is about what I got with my old 48 core and 12 core rigs running at the same time :( o the days have changed lol

o shit, im 95th on the team now, so much for top 30 lol. Guess I need to try again
Click to expand...

I'm with Vaulter98c on this. Right now it looks like my 2 970s are bringing in ~500k ppd. Back in the day I had 4 cpus and 5 gpus folding at one point and didn't even come close to this output.
Back then I was hard core and OC'd the hell out of everything and kept it running 24x7 until it failed. Now however I'm not so over the top. How are these more modern cards holding up to 24x7 folding when they are in the 65-70C range?
 
Not sure, I haven't seen any reports of cards failing but then I suppose they just get RMA'd

I suppose that with the chips running cooler and consuming less power they may last longer but who knows.
 
You must log in or register to reply here.
Back
Top