Hello there..

New on this project. Joined up with a 9600GT and got a WU instantly..

So far the card has been chewing on this wu for 4-5 hours and still only 17% completed...When I look at the top hosts, I see even a 8400GS process faster than mine..

Anything I am missing here??

Hi!

Based on 46 WU crunched on my GTX260 SP216

Min: 18668s
Max: 27513s
Avg: 24071s

if you have 64 sp it schould be around 24h per WU.

---

POLISH NATIONAL TEAM - Join! Crunch! Win!

Hi Beerdrinker

It would help greatly if you unhid your computers, we can then look at your setup and the WU you are crunching.

As mentioned above, 24 hrs to crunch a WU, for a 9600GT, is about par for the course - you get credit immediately on successful crunching - no waiting. There are longer WUs that would take you around 30-36hrs (5600+ credits for those), you might have one of those.

Hang in there for a while, you'll get used to the length of the WU, the server is very reliable and they jump fast to resolve WU/Crunching issues.

Compared to many other Projects its a sea of calm ....

Regards
Zy

No way a 8400GS is faster than yours. The machine you're looking at might have one in, but this doesn't mean it's doing all the crunching.. there are likely some more powerful cards in as well.

MrS

---

Scanning for our furry friends since Jan 2002

Hello, I've been wondering this same question. I'm using a GF8600GT, 256MB DDR3. The job's been running for 14 hours now total and progress is 25%. At this rate I will have to leave it running the whole day everyday to make the tight 1-week deadline.

Are jobs supposed to be as heavy, or did something break on my end?

I have got a little long task 16-KASHIF_HIVPR_n1_for_1hhp_open_ba5-7-100-RND0950_1. Standard tasks at my GTX260top are 5 - 7 hours long, that task looks like 18 hours run (after 46 minutes reached 4.18% progress). Run it out or suspend or abort?

Mine is http://www.gpugrid.net/workunit.php?wuid=560543, I noticed that a peer already computed his copy with a GF9800. I'm at 28%. Does that mean GF8600 is useless, or is there some weird glitch that's slowing me down?

PS: his machine solved it in 1+1/4 hour. I'll probably be done in 60 hours. So, wth?

theultramage, your 8600GT has 32 shaders, whereas 50+ are recommended. A 9600GT with 64 shaders and fairly high clock speeds will need about 1 day for a WU, so you'll need between 2 and 3 days. If you want to try it, maybe with 24/7 crunching, go for it.. otherwise nobody's angry if you take a look at seti, folding@home (non-BOINC) or aqua (commercial). GPU-Grid really needs the fast cards.

nenym, something's going wrong, but before you abort the WU you should first try restarting BOINC (not only the manager, choose "advanced / shutdown connected client"). You may also try rebooting, but if that doesn't work it's probably best to abort the WU.

MrS

---

Scanning for our furry friends since Jan 2002

nenym, something's going wrong, but before you abort the WU you should first try restarting BOINC (not only the manager, choose "advanced / shutdown connected client"). You may also try rebooting, but if that doesn't work it's probably best to abort the WU.

MrS

Thanks for hint, after reboot the task runs OK (1% per ~5').

Thank you for the info. Perhaps BOINC should have a feature that warns the user when he's near/below the minimal requirements :)

You are probably looking at CPU time reported, since it is crunching on the GPU the CPU time looks very fast. My WU's take about 9 hours to run but report about 40 min in CPU time.

I believe you should make it more clear to new users to check the list at http://www.gpugrid.net/forum_thread.php?id=867 before attaching to the project. From what I've observed, GPUGRID's work units are way too large, and the deadlines are way too short for any 'normal' user to participate (unless they dedicate their hardware to full-time computation).

For reference:
peer: < 22 hours on GeForce 9800 GT, cpu time 1.25 hrs
me : > 60 hours on GeForce 8600 GT, cpu time 5 hrs
Deadline was set to 5 days only, which is 120 hours (50%) or 80 awake hours (75%). I made the deadline with only 1 hour to spare, by leaving the machine running overnight. I will not be participating in this project anymore. My suggestion is, either reduce the WU size by a factor of ten to make it more accessible, or make it clear that the project only needs elite dedicated hardware (if you don't want mass participation).

I'm not sure if it helps to write it join) larger, maybe in red or to make it flashing.

And it's not exactly about

the project only needs elite dedicated hardware (if you don't want mass participation)

I've explained this e.g. over here, mainly in the first 3 paragraphs. So, yes, GPU-Grid needs fast GPUs. But this is not because they don't want mass participation - there's a solid scientific reason for it. And a 8600GT is quite slow by todays standards and well below the minimum recommended speed.
And while the 9800GT is much more powerful than any non-gaming PC would ever need (except for CUDA and OpenCL), it's not exactly elite. It's been available since 2.5 years (as 8800GT) and has reached a price point of about 100€ over a year ago. It should actually need about 14h for standard WUs instead of 22h. And the current GTX 260 Core 216 (~150€) need about 7h for one WU. So for these cards the WU size is quite decent.

If I were you I'd also switch that card to a different project - no offense for anyone. But I hope I could give you some insight into why GPU-Grid made these choices and that it's not just to keep users away.

MrS

---

Scanning for our furry friends since Jan 2002

From what I've observed, GPUGRID's work units are way too large, and the deadlines are way too short for any 'normal' user to participate (unless they dedicate their hardware to full-time computation).

I agree. Having one video card and one PC being used for your normal activities AND processing GPUGRID WU's is not ideal unless you have higher-end video card. I have two 9600 GSO's (768MB / 96) on my faster PC and only use that PC when I really need to. This makes for a better operating environment for the larger WUs.

---