A TRYP WU has with me on GTX260 (216)
following result.

Run Time 84.824 sec
CPU Time 19.643 sec
Credits 17. 699

Am disappointed what the Credits concern.

Hi,

I got this:

Name: 445-IBUCH_revlo_TRYP_100827-1-10-RND5857
GPU/CPU time: 63,717.75 / 62,623.50
Claimed/Granted: 14,159.83 / 21,239.74
GTX 460 Load: ~67%

29-IBUCH_revhi_TRYP_100827-0-10-RND3357
Received: 27 Aug 2010 10:59:28 UTC
Returned: 28 Aug 2010 21:02:00 UTC
Run time: 119,747.21
CPU time: 34,521.56
Claimed: 14,159.83
Granted: 17,699.78

GT240(215) GDDR5

In the same time I could have crunched 4 TONI_CAPBINDsp3's with ~27,000 credits on this card.

Problem with this monsters is, that they usually won't fit in the 24h limit, unless you have a new high-end card.
On my not that slow GT240 the only one I had 'til now took 126,879 seconds runtime, nearly 36h. Even the CPU-time was 90,000 sec., 1h longer than the limit.

You should extend the bonus limits according to the requirements of the WU, for the small, KASHIF_HIVPR probably even 20h would be sufficient, the long IBUCH_freebind_pYEEI are borderline for my machine that runs 24/7, on normal machines running just if in use it will be impossible not to miss the wished deadline, so ~36h would be the right limit, and for this _TRYP_-monsters you probably need an 4xx-one running 24/7 to meet the projects demands, imho at least 48h, better 72h, would be the right bonus limit.

---

Gruesse vom Saenger

For questions about Boinc look in the BOINC-Wiki

If you care about the bonus, I suggest you should simply abort those monster *_TRYP_* WUs, just right after they were downloaded. I've processed a couple of *_TRYP_* WUs while the same WU was processed by someone else at the same time for more than 48 hours. If there is no result after 48 hours, the project automatically reissues the WU to another client. It's no use assigning these long WUs to slow clients.

We will try to change the BOINC software to issue them to fermi only.

gdf

We will try to change the BOINC software to issue them to fermi only.

gdf

Bravo!

Bring them on Dr. G ... we'll crunch them for you even if you push them to 20ns !!!

---

Thanks - Steve

I've got a little complaint with these monster WU's. They're very elitist. I don't know how many thread I've read now with people being told to find some other project to crunch because their wee GPU's just can't follow. These are hard times & not everybody can afford to get that newest new. Yes it doesn't sound noble that people complain about not being able to get that bonus. But for whatever reason people choose to crunch, is it good enough that they have their PC's running 24/7 & pay for all that electricity? As I see, a huge WU processed on two machines is a waste, but also a long WU that doesn't complete within the 5 day deadline or a long WU that fails after crunching for hours or days.

I'd much prefer many more, much shorter WU's that only get sent to one PC at a time & maybe the ability to have multiple GPU's work on the same WU at once on the same machine.

---

They are elitist in that they are large and therefore require more powerful GPUs, in order to get full credit. However, if they are only crunched on Fermi cards then non-fermi card crunchers would not have to worry about crunching them for several days and then not getting full credit.

I have one system with 4 GT240’s and another with 3 GT240’s. While these cards are not in themselves comparable to a Fermi, together they come close; 4 OC’d GT240’s were within 10% of my OC’d GTX470 (when it was in an i7 system) and presently two GT240’s could do about as much work as a GTX460. So I like the idea of sending one task to multiple GPU’s; but only if it is more productive! I know they can do this in the lab, but I am not sure if the package could be used via Boinc or if the techs want to, it would be a bit of an experiment in itself.

But Fermi only WU's already exist. CC2.0 & higher. So what's the point in making monster CC1.1 WU's? Smaller CC1.1 WU would let even the smallest 55nm Geforce 8 contribute. As long as they're stable, there's less waste & early completion bonus' add up, so I can't see how smaller WU's can be bad. Unless it's a matter of bandwidth as sending & receiving WU's will increase as will the number of people able to participate.

BTW, does anyone know if the RSX was doing anything when this was still PS3GRID?

Sorry I complain a lot about this, but it's as if GPUGRID only want the leanest & meanest crew. But there's an Ocean of N00BS, part-timers, & less fortunate who also are willing to contribute.

---

That’s just what the scientists are trying to tackle; to only send the larger WU's to Fermi's and not to CC1.1 cards.
If they can do it, and it may still take a while before it starts happening, things will be better for CC1.1 crunchers. You might still pick up longer tasks for a while - until they start only compiling them for Fermi's.
The team may have to reduce the number of other types of task available to Fermi's until they get through this lot.

I know where you are coming from when you talk about people with lesser GPU's that would contribute if they could. Unfortunately the very basic cards just don’t do enough to be useful and would hinder the project. Even if a separate project within GPUGrid was setup to support cards with for example 16 shaders, there would be a huge server burden that would hinder the overall project.
Fortunately other projects facilitate some of the lesser GPU's, and are still useful there.

You guys are a little confusing.
In an other thread you're talking about what 'green' GPU computing is, and in this thread you want thousands of older GPUs to join (which would be definitely not green).
When I joined this project I had a GT9400 silent VGA card. It was damn slow, I couldn't crunch with it while I was using my PC. The long crunching time was also disappointing, although I like the rosetta project because I can set the designated crunching time in it (and I set it to 24h). Then I 'upgraded' to a GT9600 silent, but it was also very slow. So I realized I need a top NVidia card to participate considerably. NVidia was working on their new card, so I didn't run to buy a GTX295. I was waiting more than a year to buy a fermi card instead. I know that I can use it longer for this project, than a GTX295. You may call this project (and me) elitist, (I prefer perfectionist) but a handful of us can do more work than a hundred of older GPUs (using much less elecricity). The growth of the GPU's computing capabilities makes older cards obsolete quickly in a project like this. GPUgrid will always face complains about not supporting older cards. The team could communicate this in a better way, but they are scientist, not PR managers. To be a perfectionist can be an advantage, if the project could have a clearer image of this, and could catch the attention of the right few people out there with it. A small team like GPUGrid has no resources to develop their application for every level of GPUs, to support hundred thousands of users. They have to choose what is best for the future of the project (and it's clearly not the older GPUs). CPUs also developed a lot, so some of them can catch up with an older GPU. In my opinion it makes more sense to have a CPU client, beacuse there are a lot more CPUs out there to utilize, than older NVidia GPUs.
Maybe I took it a little personally, but I mean no offense.

No offense taken, but even the blackest CUDA capable GPU is still much greener than the greenest CPU. Also, being green isn't junking that GPU before it actually dies or sending it to Africa where they junk it for you. Just because it goes back in the ground, doesn't mean trees like growing there & even if it becomes ashes & dust, that still doesn't solve it, it's still not green.

So I like to keep the thing alive for as long as it lives. If Nvidia makes it last 10 years, they could've made it last 5 good years instead, so I'm doing my part to OC & keep it running 24/7. But there still has to be a good reason to do it in the first place. That's why I'm miffed with ultra long WU's, sending the same WU to multiple PC's, & WU's that aren't robust (unless you choose to test Beta's).

---

I am so tired of hearing the "elitist" remarks.
This is difficult science that requires serious heavy duty processing. Nothing more and nothing less.

I hope everyone who has an older card sticks around, I'm sure they will upgrade someday.

If my opinion gets interpreted as being elitist please don't confuse that, or a misguided sense of entitlement that GPUGrid should change to suit everyone's abilities, with the project's requirements ... heavy duty processing with fast returns.

---

Thanks - Steve

I don't say what I say to offend. I'm just saying how I feel about a communal/social volunteer project (as I thought it was anyways). That it's an expensive hobby to have, for those who view it as a hobby, I understand that it "might" be a sign of prestige to have the latest & the greatest. But I don't psychically get to meet anyone who takes it even as seriously as I do, & I DON'T have a PC farm,

I only have 4 PCs with relatively old hardware. I did this because sometimes I like to build PCs just for fun & like to try new OS' also just for fun. If I'm the only one who's felt the financial crisis, even though I don't feel it that bad, it would be a surprise to me. But I've cut down on my hobby never the less. I'm hoping to wait at least a year before I build a new rig & I don't have the space or electricity to do this unless I get rid of one of my PCs.

---

You are talking to people that take their hobby every bit as seriously as you. For many crunching is not just a hobby it is a way of life; it’s one of the things we live to do. Just because you cannot be a research scientist does not mean that you cannot push science along.
Hopefully we will meet up some day, ditto for anyone else that contributes here, and the scientists.
We contribute, despite the financial situation, because if we did not we would lose part of ourselves.
There is a route to upgrading that is beneficial to all – sell your older parts to people that don’t crunch. They will appreciate them for longer (a green act) and replace them with parts that are more economical to run, in terms of performance per watt (another green act).

We will try to change the BOINC software to issue them to fermi only.

gdf

You haven't succeeded so far ;)
My last one is another of those monsters, but as I just became aware of it, after 13:30h of crunching already done, I will definitely not abort it.

---

Gruesse vom Saenger

For questions about Boinc look in the BOINC-Wiki

I think one of the reasons that these WUs are so slow is that they run at 20% lower GPU than the other WUs. WinXP-64, v258.95 drivers, v6.11.6 BOINC.

I think one of the reasons that these WUs are so slow is that they run at 20% lower GPU than the other WUs. WinXP-64, v258.95 drivers, v6.11.6 BOINC.

These WUs made me overclock my CPUs. I think these WUs need more double precision calculations than other WUs, and this slows them down about 20%.