So I haven't folded super actively in a few days because well... the AC in the house is broken OK?

Try to get a WU finished, POW, unstable
 
Starts a new one, folds for over an hour, temps not pushing past 75 I believe, UNSTABLE
 
I'm running the Precision X GPU test and not seeing any artifacts, but I know my NV drivers crashed earlier today playing Need For Speed World... a bit concerned.
 
What do I do folding brethren?

if you bothered checking his mod rigs Panther-X...he's running a stock 570 (no OC)
 
what we need to see is the actual log file portions to make any decisions on whats going on (ie the start of the WU and where the error occurred for that particular unit....like WU info, core info)
 
Im more inclined to believe that core 17 isn't quite that stable yet, or the driver needs improving...which is why it's failing on previous stable cards...core 17 is beta so are any WU's for it
 

 

if you're running beta NV drivers go back to stable whql ones
 
http://forums.evga.com/tm.aspx?m=1937207
new drivers got released in last week - which could be the issue
 

Xavier Zepherious

if you bothered checking his mod rigs Panther-X...he's running a stock 570 (no OC)
...

Please note that while I saw the signature, I wasn't aware that "Mods Rigs" was an actual URL link to the system specification. Never clicked on it before nor knew about it until now.
 
Given general description, the suggestions too were general. However, we now know that the issue is related to FahCore_15 since Drazhar has posted the Projects that crashed. It would be help if the log was posted as Xavier Zepherious suggested. Have you tried a clean installation of the latest WHQL Driver? Moreover, is there any dust build up in your PC?

Its a known fact that some Nvidia drivers stop responding when doing so the video card card can become down clocked among other things when this happens, these things cause FAH to crash.
 
Do a video driver clean and fresh install of the latest known good driver. I'm using 314.22 without problems.

From what I know, TJ Lane handles the points while Proteneer does the coding. Considering that this is a real project, it does open a huge potential for the future.
 
REMOVED
 
FahCore_17 does change the dynamics of points. Hopefully,  it will have a positive impact on the largest possible donor base with very little negative impact.
 
BTW, 100 ns/day Explicit has been achieved in lab on GTX Titan with CUDA  All we have to do is to wait (im)patiently for Nvidia to release JIT and see our GPUs (and PPD) pushed to the limits
 
EDIT -> Sorry for the miscommunication and any inconvenience caused. That wasn't my intention.

Panther, are you having a day off in here or are you just joking?
Those numbers you posted about insane PPD are from internal testing with micro units. Units were decreased in size to catch certain bugs, but point estimate value of that unit was left from unit leaps and bounds bigger than micro unit. You do remember 7970s pulling 5-7 sec in TPF with estimated PPD of millions and millions. If it was April, I would laugh at it, but now it is not really funny, especially in this forum :)

Panther-X


FahCore_17 does change the dynamics of points. Hopefully,  it will have a positive impact on the largest possible donor base with very little negative impact.

BTW, 100 ns/day Explicit has been achieved in lab on GTX Titan with CUDA  All we have to do is to wait (im)patiently for Nvidia to release JIT and see our GPUs (and PPD) pushed to the limits

And nope, failed again.

I'll call EVGA.

muziqaz

Panther, are you having a day off in here or are you just joking?
...

I wasn't having an off day, nor joking. The point that I was making (which didn't come across as I expected) is that when and if bigadv comes to GPUs and they are using the bigadv benchmarking methods, high-end GPUs can make a significant amount of PPD. Now, I am not sure when or if it will happen, but I would say that the future is looking bright for GPU bigadv:

21:27 < proteneer_osx> we might be able to do 500,000 to 1 million atoms
...
21:27 < proteneer_osx> on the GPU
...
23:25 <@proteneer> we've done 1 million attoms on a kepler before successfully

Source -> http://sodaant.com/fah.log
 
Sorry if my post caused any confusion as that wasn't my intention.

Panther-X

muziqaz

Panther, are you having a day off in here or are you just joking?
...

I wasn't having an off day, nor joking. The point that I was making (which didn't come across as I expected) is that when and if bigadv comes to GPUs and they are using the bigadv benchmarking methods, high-end GPUs can make a significant amount of PPD. Now, I am not sure when or if it will happen, but I would say that the future is looking bright for GPU bigadv:

21:27 < proteneer_osx> we might be able to do 500,000 to 1 million atoms
...
21:27 < proteneer_osx> on the GPU
...
23:25 <@proteneer> we've done 1 million attoms on a kepler before successfully

Source -> http://sodaant.com/fah.log

Sorry if my post caused any confusion as that wasn't my intention.

Those quotes were also well before 8900 saw the light of day.  8900 is already taking a long time, and is considerably smaller than a lot of recent non-bigadv SMP units (i.e. 71xx lines from Kasson's lab), and 1/25 the size of current bigadv units.
 
You can't really compare size directly, as difficulty depends at least as much on the types of calculations being performed (just look at the 810x bigadv line, exact same atom count with 8104 having tpfs 1/2 that of 8101), but by any overall measure, 8900 is not big, and yet it already takes a relatively long time to complete.  Extrapolate 8900 to 500k atoms and you're looking at a titan needing 4+ days to finish.
 
But regardless, the points you listed are meaningless.  Not only was the WU completely fake, but just assigning the values from one project to another has no meaning.  You could apply bigadv settings to a uniprocessor WU and get similar ridiculous numbers.  Does that mean uniprocessors will soon be making 1mppd+?  Proteneer did it as a joke, not as any indication of what the core is capable of, or as an indication of the future of GPU folding.  
 
Too many people are simply accepting as fact that gpus are always significantly faster than CPUs.  There is no doubt that for certain simulations that is true, but extrapolating to all cases is simply wrong.
 
I've heard no indication from proteneer, or anyone else official, that the current beta ppd is off in any way.  Any argument against that (that the ppd should be higher), that is based on calculations from fahbench, 7663, previous GPU cores, or worst of all 8052, is faulty logic.  PPD seems right in line with what they want for kepler/AMD.  Fermi's don't do well on core_17, but that's not an indication that the WU settings are off, just that fermi's do not do well on this core.

The reason why people want bigadv gpu is that it appears the only way to get fair points for the work performed.
 
All of this debate could quickly by resolved if benchmark machine TPFs were made public.