TPU's F@H Team

[Steevo]

If you can run your 7970 @ 1240 GHz that is good. I had to go below 1100 lately also after some 99,99 % hanging WUs. OK my GPUs runs a bit warmer but still you seem to had a good one.

I bought it with those custom clocks on it and the water-block installed. I think I might write a bit more aggressive BIOS and up the core voltage to 1.35 as that is the high safe end for this core and pushing it to 1.3V got me to 1313Mhz stable through some benchmarks.

 

[hat]

So far the GT720 appears to be putting out 7331 PPD. My 660 Ti went down to 20k, though... but I am using the computer for other things right now, too. One thing worth noting... On the machine with the GT720, I let FAH have one core (set WCG to leave one core alone). On my main machine, I let WCG have all the cores, so WCG and FAH are kinda fighting over one.

 

[Steevo]

Finally nailed down the source of my instability, this CPU needs NB/CPU volts, bumped it up to 1.35 from the 1.2 ish it was running and I may be able to push more GHz out!!

 

[hat]

The GT720 is surprising me! Almost 13k ppd... not bad for something like that. That's on a core x18 though. My 660ti is stuck with a core x17 at 21k.

You guys can have all the x17s... the x18s run better on my hardware than the x17s lol.

 

[Nordic]

The GT720 is surprising me! Almost 13k ppd... not bad for something like that. That's on a core x18 though. My 660ti is stuck with a core x17 at 21k.

You guys can have all the x17s... the x18s run better on my hardware than the x17s lol.

If I could take all your core 17's I would. I want them ALL!

I wonder what the ppd/watt is on that 720?


By the way team, I am dropping one 750ti for folding. It will be put for sale with in a week. I will be folding again with my remaining 750ti once I get this new itx system stable. Right now I am having OS issues, which might be related to a bad motherboard. I just need to test it out further.

 

[hat]

Apparantly, it consumes 19w, so at 12777PPD, it produces roughly 672PPD per watt.

 

[Nordic]

Apparantly, it consumes 19w, so at 12777PPD, it produces roughly 672PPD per watt.

That is actually a lot better than I thought it would get. It compares to high end 600 series ppd/watt. I am impressed.

Please add some info from it to the charts. It is small, but I think it is pretty cool.

 

[mx500torid]

Are you using the 327 driver hat? The newer drivers kill the ppd of the Keplers. On a x17 you should be around 60k ppd.

 

[revin]

OhMG!!!!!!!!!!!!!!!!!!!!! I'll be back on here after bit, I got HUGE surprise today and I'm speechless
Anyway here's what the ole 5870 girl been doing, man I really love her, but, but but.....I'll Be Back !!!!!!!!!!!!!!!!

@Norton
7950

 

[stinger608]

Hmm, do I hear a new video card being surprised to us @revin

 

[BUCK NASTY]

It appears that the Assignment Server for Linux GPU's is out of P9201's. I shut the GPU rigs down at 10PM EST lat night(they were idling). Still nothing available, but might be back up today when Stanford realizes the problem.

 

[mstenholm]

It appears that the Assignment Server for Linux GPU's is out of P9201's. I shut the GPU rigs down at 10PM EST lat night(they were idling). Still nothing available, but might be back up today when Stanford realizes the problem.

Mine are not spinning either despite the server status shows that the 9201s should be around from time to time. No way that I want to hear coils when they do core 15 WUs.

 

[revin]

Ok.. Sorry haad a couple things come up last couple days !
Most Notable, @Norton sent this HUGE surprise ! Gigabyte 7950 with Accelero Extreme

Some Kisses Mint truffle's Never seen these in Mo. !

Got it in the 932 after trying for couple hours to re-configure bunch of HD's and BDRW,
No luck, still got to figure a way to re-move bunch of Data, but trimmed from 12 to 7 Hdd's ll SATA

Pretty big ass cooler on here brother

Man o man don't think I'd ever use a blower style card again[at least with HAF case !}, this baby is QUIET !}

Got it installed this AM[ tore my back up lastnite lugging this big[HEAVY} case out to garage to clean
Added a support cable to keep'er steady !

Man I still Love that old Ninja Cooler ! Please don't tell me about wiring, it's a 932 not going make it cool better, I know it's not pretty for now it's best I can do

@Norton This is a Dream come true Thank you VERY much , cant wait till Thursday when the grandson get's here to eat kisses and play games
Time to get F@H going ! EDIT : 7950 44C* cool

 

[Norton]

 

[BUCK NASTY]

Nice addition there @revin .

The assignment server issue has been labeled as a "Maxwell Incompatibility" by Pande group. They do not know how it started approx 33 hrs ago, but no Maxwell cards are getting core 17 work in Windows or Linux. Kepler seems unaffected by this issue. Our team production will be down several million until this issue is corrected.

 

[mstenholm]

Nice addition there @revin .

The assignment server issue has been labeled as a "Maxwell Incompatibility" by Pande group. They do not know how it started approx 33 hrs ago, but no Maxwell cards are getting core 17 work in Windows or Linux. Kepler seems unaffected by this issue. Our team production will be down several million until this issue is corrected.

I noticed that the 9201s was going in and out off the servers and even got a few but on my 680 and my 7970 but around 50 tries on my Maxwells left them cold. I wonder why they can say that they didn't change any setting on the feeding priority when it is clear that must be the reason.

 

[BUCK NASTY]

I noticed that the 9201s was going in and out off the servers and even got a few but on my 680 and my 7970 but around 50 tries on my Maxwells left them cold. I wonder why they can say that they didn't change any setting on the feeding priority when it is clear that must be the reason.

Agreed. I'm trying to pinpoint a time when "something changed" so they can review their logs and hopefully identify the issue. This is the 2nd time in a week that this has happened, although unrelated. Luckily the 1st instance only lasted 4 hrs.

 

[stinger608]

Nice @revin !!! @Norton is just soooooooo awesome!

I have an ole 7870 that has the Arctic Twin Turbo II on it. They are just an amazing cooler and as you pointed out, almost silent. Very very well worth the price.

 

[Norton]

OK Team got a question for you- I'm getting ready to setup a dual 7970 rig (water cooled) for folding and I need to make some hardware choices. What do you think the ppd difference would be between:

1- Socket 1366 system (GB X58 board/Xeon L5639)- Already have this setup but will be planning to sell at some point

2- Socket AM3+ system (GB 970 UD3 board/cpu- TBD)- I have the board and will likely get a 63xx/83xx cpu for it. Will the 2nd PCIE slot hurt my output much since it's only 4x?

3- Socket 1155 system (Asrock Z77/2600k)- I have this system running WCG atm but would rather leave as is if the ppd difference is minimal

4- Socket 1150 (board/cpu- TBD)- buy a new/used setup with a basic i3 or i5 cpu. I'll likely sell off some of the above hardware to finance part of the build if I go this route.

Option #3 or 4 I may setup as a folding/crunching/gaming rig but since I don't game too much it will likely fold and/or crunch most of the time. I have 90% of the water cooling parts (cpu block, rads, pump/res, etc) and just need some fresh tubing and a few fittings so no issues there.

Thoughts?

 

[mstenholm]

OK Team got a question for you- I'm getting ready to setup a dual 7970 rig (water cooled) for folding and I need to make some hardware choices. What do you think the ppd difference would be between:

1- Socket 1366 system (GB X58 board/Xeon L5639)- Already have this setup but will be planning to sell at some point

2- Socket AM3+ system (GB 970 UD3 board/cpu- TBD)- I have the board and will likely get a 63xx/83xx cpu for it. Will the 2nd PCIE slot hurt my output much since it's only 4x?

3- Socket 1155 system (Asrock Z77/2600k)- I have this system running WCG atm but would rather leave as is if the ppd difference is minimal

4- Socket 1150 (board/cpu- TBD)- buy a new/used setup with a basic i3 or i5 cpu. I'll likely sell off some of the above hardware to finance part of the build if I go this route.

Option #3 or 4 I may setup as a folding/crunching/gaming rig but since I don't game too much it will likely fold and/or crunch most of the time. I have 90% of the water cooling parts (cpu block, rads, pump/res, etc) and just need some fresh tubing and a few fittings so no issues there.

Thoughts?

Intel/AMD ? No idea. Newer Intel > older Intel. My experience with GTX 970 clearly (for me and not for many others) is that I gain a 5 % reduction in TPF time with a faster CPU in a newer platform. If it is the CPU speed (Core 15 left out since they hardly uses CPU time) or something else I don't know. As for 7970 goes I have noticed a similar reduction from faster i7 970 to a slower i7 920 on a similar GB X58 UD5. This is on 1300x which you will see most of the time. Here the small dark house is that the i7 920 has a slow HD and the 970 a fast SSD.

You will get nothing constructive from this forum since all are in the believe that it doesn't matter. It is even believed that core 18 WU does not use CPU time with Nvidia other then during check points every 5 % point of the WU.

In short - if you want to optimize folding PPD use your fastest CPU/newest platform. That of course will shave a bit of the WCG PPD

 

[Norton]

Intel/AMD ? No idea. Newer Intel > older Intel. My experience with GTX 970 clearly (for me and not for many others) is that I gain a 5 % reduction in TPF time with a faster CPU in a newer platform. If it is the CPU speed (Core 15 left out since they hardly uses CPU time) or something else I don't know. As for 7970 goes I have noticed a similar reduction from faster i7 970 to a slower i7 920 on a similar GB X58 UD5. This is on 1300x which you will see most of the time. Here the small dark house is that the i7 920 has a slow HD and the 970 a fast SSD.

You will get nothing constructive from this forum since all are in the believe that it doesn't matter. It is even believed that core 18 WU does not use CPU time with Nvidia other then during check points every 5 % point of the WU.

In short - if you want to optimize folding PPD use your fastest CPU/newest platform. That of course will shave a bit of the WCG PPD

What are your thoughts on using a PCIE x4 slot for a 7970? I'm thinking 10% reduction in ppd at the most based on the articles @W1zzard does from time to time.

 

[mstenholm]

What are your thoughts on using a PCIE x4 slot for a 7970? I'm thinking 10% reduction in ppd for at the most based on the articles @W1zzard does from time to time.

It is not that bad but I saw a few percentage reductions going from 16 to 8 on my 970s and Buck noticed a decline as well when he put four in one MB. I would say 1-2 % increase in TPF for the 7970.

 

[mx500torid]

You will get nothing constructive from this forum since all are in the believe that it doesn't matter. It is even believed that core 18 WU does not use CPU time with Nvidia other then during check points every 5 % point of the WU.

Dang MS take it easy I just ask a question in the forum and got an answer. If your experience is that a faster cpu = more ppd then it must be so. No one is saying you are wrong and all i said was from what i seen on my rigs. You are way more knowledgeable in these matters that me. If i pissed you off or said something wrong then i apologize.

 

[revin]

Just bout to finish first work thing on the 7900
EDIT :: Went from 13k to 25k @Norton Thank You So Much !!!!!!!

 

[mstenholm]

Just bout to finish first work thing on the 7900

Welcome to GPU folding. There are different kinds of WUs (Work Units). Some get a ERB (Early Return Bonus), some don't. When folding with a 7900 series you will most likely only see a ERB unit. The basic principle is that the faster you do it the more points you get (because the next WU needs the result of your work and hence the project can finish faster). Normally people let a ERB 13001 WU run uninterrupted so they can help the project finish fast and in the process get some more points. Just saying that don't stop unless you have a fairly good reason and I'm sure there can be plenty. I have two 7970s that runs in PC that I don't use for other then distributed calculations, one gets 71500 per WU and one gets 73500 per WU.

For ERB project look at this (not complete list). If it has a k-factor other than 0 it will be better to let it run until it finish. http://www.techpowerup.com/forums/threads/tpus-f-h-team.13038/page-724#post-3221809

Edit: Not related to the above but I will stop my Nvidia rigs for the next two weeks. PG needs to come clean about their recent shift in priorities. They gain 3000 TFLOPS (5 %) in two days at the same time all 970&980 sits and do nothing. It sticks!