TPU's F@H Team

[AnomalouS]

I just remoted in and installed FAH.

How do you run multiple clients? or should I not be concerned with that?
Current PPD = 58k

***and then my PC crashed...lol. Shut down and I cannot access it, will have the wife reset it after while.
so i gotta re think the multi clients for a while.

[thebluebumblebee]

Originally Posted by AnomalouS

I just remoted in and installed FAH.

How do you run multiple clients? or should I not be concerned with that?
Current PPD = 58k

***and then my PC crashed...lol. Shut down and I cannot access it, will have the wife reset it after while.
so i gotta re think the multi clients for a while.

Remote access and GPU Folding don't mix. 58K indicated that you were Folding on both the CPU and GPU.

[AnomalouS]

so teamviewer is a bad idea huh, live and learn.

[johnspack]

Nah, teamviewer is fine. Just next time you install fahclient, make sure to select gpu only folding, or make sure your cpu is set up properly for folding. Sounds like your cpu wasn't stable enough for folding, hence the crash.

[thebluebumblebee]

It's official. I despise P764x's. They will not run on both of my GTX460's at the same time. I'm now having to "finish" each one at a time. I've set my options to small WU's so I hope I'm able to get something else. My DSL connection only does 300kb on the upload, so it takes a long time to upload on of these 10MB P764x's.

[flmatter]

OK 600+ pages in this thread, since I am new to the team
can anyone PM me and explain the badges and how I can run my F@H the best in between gaming and leaving my machine to fold. I normally game from 7PM until Midnight-1AM then go to bed. I can leave my machine on to fold while sleeping and at work. System specs under my name on right. Thanks in advance and I appreciate your time and patience with me.

[theoneandonlymrk]

what you all saying bout the chimp ,capn's

[Jstn7477]

I'm having some frustration with some of my NVIDIA GPUs. Apparently my GTX 460s which are both on one machine have been endlessly reporting UNKNOWN_ENUM and finally decided to both say "failed" today (both are only OCd to 800/2000). Last week I had the same issue with my GTX 550 Ti and GT 440 both failing, and my dad's Dell would fail both my GT 220 (which works flawlessly overclocked in my AMD E350 shoebox) and the GT 440.

So, why are my seemingly working cards coming up with this "failed" bullshit? Is it the drivers, overclocks, temperatures, busted motherboards, what? I upgraded to the latest clients, the 296.10 or whatever drivers and that seemed to increase the problem. Any ideas?

[thebluebumblebee]

^^ Do you know what projects they failed on?

[Jstn7477]

I'll have put the IDE drive back in and look at the log, but I'm pretty sure it was those new 7xxx ones that don't even seem to save your progress at all either. I'm reformatting the machine and using an old 36GB Raptor, and I swapped the CPU from the PII X4 955BE to a Sempron 130 because this Crosshair III Formula is probably my only board left that can unlock an AM3 with a post-2009 BIOS. I'll have to see how it runs with the fresh OS, latest NV drivers and the Sempron.

Not sure if my board was unstable because the CPU used to be at 3.9/1.5v but it was failing WCG units all the time so I lowered it to 3.8 and it seemed to do fine. ROG Timer Keeper says my board has been on for 512 days (it's been in use since Aug. 2009 and overclocked in some way or another for most of its life). I remember folding 9800 GTs on it until they died lol.

[jsfitz54]

Originally Posted by Jstn7477

I'm having some frustration with some of my NVIDIA GPUs. Apparently my GTX 460s which are both on one machine have been endlessly reporting UNKNOWN_ENUM and finally decided to both say "failed" today (both are only OCd to 800/2000). Last week I had the same issue with my GTX 550 Ti and GT 440 both failing, and my dad's Dell would fail both my GT 220 (which works flawlessly overclocked in my AMD E350 shoebox) and the GT 440.

So, why are my seemingly working cards coming up with this "failed" bullshit? Is it the drivers, overclocks, temperatures, busted motherboards, what? I upgraded to the latest clients, the 296.10 or whatever drivers and that seemed to increase the problem. Any ideas?

It's not you.

My 460 and 450 failed 4/11 so I deleted the slots on separate machines. Didn't bother with looking at unit number, I've had so many fail in last 2-3 months. I had a few work for a couple of days then failure. Same 296.10 drivers. It's not an overheat issues, mine ran at 62c with fan on curve, at 83%. It's not an overclock issue either as I've gone stock on a couple machines to try it.

[thebluebumblebee]

I'm now convinced that the "tweak" that they did to Core_15 is causing problems. (they admitted tweaking it, but I don't have the time to find the post now) I thought I was having problems with one system but then last night, I looked at my rock-solid-steady system with a GTX 460, I found it only getting some 3000 PPD on an 803x, where it normally gets close to 14K. It picked up that WU after a 764x WU. I had to reboot to get it to work properly. I'd like to know how the WU's can engage at different rates? The different TPF rates that I'm seeing have nothing to do with the system, but how the client starts. I have seen a lot with having to babysit almost every WU start since March 8th. Different GPU memory amounts. Different GPU usage %. Different %'s while starting.

WARNING! Rant:
I'm still trying to Fold as hard as ever but look at my points. I cannot use both of my SLI'd GTX460's in the one system (because of the P764x's) so 1/3 of my GPU's sits doing nothing. I only got one P764x WU done yesterday for a grand total of 7718 points for the day. Waho! Then Win7 magically reset updates to auto and rebooted in the middle of the night - 2 GPU's doing nothing overnight.

[jsfitz54]

Originally Posted by thebluebumblebee

WARNING! Rant:

Yea, I'm F'in annoyed as well.

When I said it would be a good idea to limit GPU by percentage like adjusting smp core amounts to limit heat, have less failures, retain people's interest, etc...I was told to stop griping about it.

I'm a little put off by the comment. Especially due to the lack of quality info as to what is causing the failures or work arounds until a more permanent solution is found.


jsfitz54 wrote:
I still feel strongly that one should be able to set GPU slots by percentage. I think this would help retain donors.

bruce:
Everyone agrees with you on that point ... but Development is actively working on making that possible in a future version of the FahCores, and they're not there yet. Griping about it won't make them able to fix the software any faster than they already hope to be able to.

[Jstn7477]

I'm pissed off too. I (re)started late last year with a 6950, 5770, 6670 and a GTX 550 Ti and was getting 25K PPD. Added a GT 440 and got 30K PPD. Added one then another GTX 460, had 60K PPD for about a week. Now I'm scraping along at 30-40K again and furious that the $300 worth of GTX 460 cards and $170 PSU is getting me literally NOTHING. I even resorted to turning my 6950 on because I am literally getting ****** by F@H.

[Jstn7477]

If anyone is interested, I have the entire log for my 2 GTX 460s on client 7.1.52 using NVIDIA 296.10 drivers. The machine was just reformatted last night at 8PM and I got to work this afternoon to find both cards failed at stock frequencies.

My P4 651 machine with a 550 Ti and a GT 440 has not failed yet.

[theoneandonlymrk]

one thing ive noticed with the latest beta and nv folding with my Gt240 on a dual core Pentium G850 i thinks @3.2 is that i now Need smp or uniprocessor off as i said before , but after watching my processor usage a day i can see now why that is , the gpu only client is using about 50-60% of cpu utilisation , on a 5767Wu ox11 but its not level its frequently momentarily peaking to 100% and the spectrogram (cpu usage in Task manager) looks like a decent Metal band is on, usage bar shows a steady 58 ish , it didnt used to do that im sure it was a constant 20-30% before, ie previouse V7 beta

[johnspack]

My gtx2xxx cards are rock solid in folding, must be the new client for gtx4xx and up, too bad, I want to upgrade. I'm more disappointed in the ppd value for my i7.. they should give more love for a formerly $500 cpu! Now I have to find a decent priced 6 core, which I haven't so far.

[Jstn7477]

Originally Posted by theoneandonlymrk

one thing ive noticed with the latest beta and nv folding with my Gt240 on a dual core Pentium G850 i thinks @3.2 is that i now Need smp or uniprocessor off as i said before , but after watching my processor usage a day i can see now why that is , the gpu only client is using about 50-60% of cpu utilisation , on a 5767Wu ox11 but its not level its frequently momentarily peaking to 100% and the spectrogram (cpu usage in Task manager) looks like a decent Metal band is on, usage bar shows a steady 58 ish , it didnt used to do that im sure it was a constant 20-30% before, ie previouse V7 beta

Core 11 (old AMD/NV) and Core 16 (new AMD) need a substantial amount of CPU to fold on them. Core 15 (Fermi) actually needs literally no CPU to run, hence why one of my NV Fermi folders is a Pentium 4 651 that is extremely slow.

[hertz9753]

I still use Nvidia 266.58 drivers on my GTX 275's and 400's. I also don't fold with SLI enabled or use a SLI cable.

[thebluebumblebee]

He's alive!

[Jstn7477]

My situation is weird since I have unmatched 460s (a Gigabyte 768MB and an EVGA 1GB) and they are on an AMD 790FX board. I might try separating them onto two machines and seeing if they will work individually or not. Either way, I'm sick of F@H's dumb issues like being unable to fold core 15 and core 16 on the same Windows 7 machine or running GTX 460s or whatever.

Time to pull out my Socket 754/GeForce 6150SE/Turion 64 system and drop a 460 in with my recently freed up Rosewill Capstone 550w and see if I can get my +30K PPD back.

[hertz9753]

Originally Posted by thebluebumblebee

He's alive!

Yes I am.

Originally Posted by Jstn7477

My situation is weird since I have unmatched 460s (a Gigabyte 768MB and an EVGA 1GB) and they are on an AMD 790FX board. I might try separating them onto two machines and seeing if they will work individually or not. Either way, I'm sick of F@H's dumb issues like being unable to fold core 15 and core 16 on the same Windows 7 machine or running GTX 460s or whatever.

Time to pull out my Socket 754/GeForce 6150SE/Turion 64 system and drop a 460 in with my recently freed up Rosewill Capstone 550w and see if I can get my +30K PPD back.

I have a Foxconn Destroyer with a GTX 275, GTX 275 co-op and a GTX 460 1GB. Your problem is the driver.

[Feanor]

Originally Posted by Jstn7477

Core 15 (Fermi) actually needs literally no CPU to run, hence why one of my NV Fermi folders is a Pentium 4 651 that is extremely slow.

True, but has some exception. Some of the 76xx (not really the 762x, the 764x and 761x for sure) work units needs up to 3% of my i7 970 cpu to fold.

When my three gtx 460 are folding 803x wu, i can fold using the -smp 12 flag no problem (each fah_core 15 use between 0 and 0,5% of my cpu). 35k ppd for the cpu and 16k ppd for each of the gpu.

But if one of the gpu client gets a 764x or 761x, then my cpu ppd drops in the 25k zone, and i have to free a core (i.e. use the -smp 11 flag) to be able to feed the gpu client. Then my cpu nets around 30k ppd and the gpu folding the non-803x work units gets 10k ppd.

I know it is all for science, and i will not stop folding because i get less points, but moving from 80-85k ppd (35k+16k*3) to around 60k (30k+10k*3) is still a huge drop!

[jsfitz54]

Originally Posted by hertz9753

I have a Foxconn Destroyer with a GTX 275, GTX 275 co-op and a GTX 460 1GB. Your problem is the driver.

So what Nvidia driver are you using?

[BUCK NASTY]

Originally Posted by Feanor

True, but has some exception. Some of the 76xx (not really the 762x, the 764x and 761x for sure) work units needs up to 3% of my i7 970 cpu to fold.

When my three gtx 460 are folding 803x wu, i can fold using the -smp 12 flag no problem (each fah_core 15 use between 0 and 0,5% of my cpu). 35k ppd for the cpu and 16k ppd for each of the gpu.

But if one of the gpu client gets a 764x or 761x, then my cpu ppd drops in the 25k zone, and i have to free a core (i.e. use the -smp 11 flag) to be able to feed the gpu client. Then my cpu nets around 30k ppd and the gpu folding the non-803x work units gets 10k ppd.

I know it is all for science, and i will not stop folding because i get less points, but moving from 80-85k ppd (35k+16k*3) to around 60k (30k+10k*3) is still a huge drop!

I see the same, but my drop is in the 11k PPD range with and i7-860 and 2x GTX460's.