I just noticed the 8900 wus when I the jump in ppd caught my eye. My 560ti stayed the same at 25k but my 660ti increased from 57k to 66k. Temps look like they are the same as the 7660 wus. They still have the periodic drop in usage but are 5 secs long instead of 1 second. They do seem to be less frequent. I'm running the unicore so there's plenty of cpu cycles to spare at 39% usage.
 

The KEPLERS do great but the FERMIS take a big hit. Temps remain the same and there is a huge lag with the 8900 that is not there with 7663. Not sure what all was done with the core change but I do know I can not run it on FERMI.

My best time on 590 tpf 10 minutes. Never saw anything like you got. Some times tpf went to almost 14 minutes

Hockyman

Haven't had any issues with my 480's. I am getting an average TPF at 7:53

Yeah I have my 480s oc to 810/1614/1906 

TheWolf

What doesn't make since is Femi makes more PPD on WU's with NO QRB or k-factor than on WU's with.
Looks like someone just trowed in a fufu for Femi in core 17 to make the newer teh look like its doing
what its suppose to is all. lol

I would not be comparing core 15 with 17
 
issues
old cuda library - no new cuda in core 15 - no JIT cuda compiler
so enhancements with kepler may not be fully utilized under the old cuda platform
 
core 15 never fully implemented the full power of kepler
as been said "when we get the new cuda core it will even be better than opencl"
 
further core 17 is BETA - and MAY NOT BE FULLY STABLE - hence it's still in beta undergoing beta testing
 

 
 
 

Im wondering if it has to do with the opencl support of the cards. Kelper and fermi would have different driver support coding, i would think. As far as the tpf jumps i spent alot of time watching my FHM monitor all week and never noticed more then a 5second tpf jump on any card frame to frame.

I've had no problems with the 660ti.
The 560Ti though, farthest I got was 68% before it dumped. stopped it after folding for 24 hours, crashing 5 units and not completing any even after underclocking the card.
 
nvidia driver 314.07
 

both my 580 lightning's hate the 8900 w/u even at stock settings but my 570 o/c'ed to 890MHz just keeps folding them for around 38,000ppd both 580's are now on core 15.
 
ok the sweet spot for 580's is around 830MHz or so

Since I'm only running one box I may have to switch out my 470's and put in my 660.

reading up over at FoldingForum I found this
 
Top

Re: New GPU core17 Project 8900 released for beta testing

by Quisarious » Sat Jun 01, 2013 2:35 am
11:17:56:WU02:FS01:0x17:ERROR:exception: Error downloading array energyBuffer: clEnqueueReadBuffer (-36)

For what it's worth, for me, these are definitely OC related. I get these when I am two or greater steps beyond stability. When I'm one step beyond, I get the occasional 'Bad state detected, attempting to resume from last checkpoint'. At stable clocks I get nothing but net.

It is very easy to reproduce clxxxxx errors, just OC beyond stability. Go a ways above stable, and you can get them within 1-2% of progress =) (I haven't done this on purpose...but when 8900 was first released to internal, I had successfully completed >300 7663's without a single error, but the first 8900s, using the same OC settings on all 4 cards, produced clxxxxx (there's a while family, clEnqueueReadBuffer (-36) is most common) errors within 5%.)


I'd be happy if I got the error within 3-5%. I getting them at 40-60 % which means 3.5-6 hours wasted each time.

I've upgraded to the latest driver, and now dropping my clocks 20Mhz below reference to see if I can get something completed.  Going to experiment to day and if nothing happens, will go back to standard units.

Well, I give up. seems no matter what I do, they crash on my rig.
I've completed one and crashed about 25 over the last day.
I have no idea what is wrong.
I was getting overclock errors
fixed that and then I get these instead
 
FS00:0x17:ERROR:exception: Error invoking kernel execFFT: clEnqueueNDRangeKernel (-5)
 
FS00:0x17:ERROR:exception: Error invoking kernel gridInterpolateForce: clEnqueueNDRangeKernel (-5)
 
FS02:0x17:ERROR:exception: Error downloading array energyBuffer: clEnqueueReadBuffer (-36)  - 46% completed
 
 
WU00:FS00:0x17:ERROR:exception: Error invoking kernel calcCenterOfMassMomentum: clEnqueueNDRangeKernel (-5) - 72% completed....
 
WU00:FS00:0x17:ERROR:exception: Error invoking kernel copyDataToBuckets: clEnqueueNDRangeKernel (-5) 31% completed
 
and so it goes on.
 
I have no idea.
If anyone wants to send this over to the beta forum, feel free.
 
Win7 64 bit Service Pack 1 - fully up to date
 
nvidia driver 320.18
 

Simba123
Well, I give up. seems no matter what I do, they crash on my rig.
I've completed one and crashed about 25 over the last day.
I have no idea what is wrong.

I'm sorry for your troubles. It could be that the new programming is using a part of the gpu that has errors. Your only option may be to RMA. That's what I did and the new gpu has no errors and is 15 degrees cooler. Of course, I believe this is my third 560ti but who's counting.
 

Simba no offence bud but do you still think its an O/C problem when you in you other post said your-25mhz of stock and still having errors. I would say its rma time on that card. The only other thing i can say im running 320.00 nvidia drivers and they are working(think you said you on 320.14) but idk how much of a change driver make for 400-500 at this point as most release note mention 680-titan updates/fixes not to say that they couldnt be breaking old parts in the process.

Sorry for the late reply, being in Australia means delayed responses!
 
I'm pretty sure it's something with the software somewhere on my rig. These problems seem unique to me, certainly I have not seen anyone else mention them.
I have 2 cards, a 560ti and 660ti; the errors occur on both.  even when underclocked by 100Mhz.
 
I put them back on normal units Core15 openmmgpu and have not had any errors since.
 
I'm currently waiting for the 2 units running at the moment to finish, then I am going to uninstall V7, nvidia drivers , .NET 4 framework and try to re-install them and see what happens.
 
 

boballee
 
I do notice a lag, however, anywhere from 5 - 15 seconds, where utilization drops to 0%.   

 
If this is happening on even percentages of the WU's progress, then it's just the core stopping to make a checkpoint.