Saenger, I couldn't get my GT240, 8800GT, & GTX260 to work with Linux either, both Mint Linux 8 & Mint Linux 10 failed, I couldn't get it to work no matter what. So I gave up & decided to use Windows 7 90day Enterprise evaluation instead link

I hope this helps. Nothing else did... :-(

---

I also have a gt240 and i downgradet my driverversion to crunch some wu's in a normal time, because i thought the project is worth it, as a backup. And normally i do not spend time on an beta project but after reading the thread i think the project never be out of beta state.
The project is hunting for an application for the newest and fastet gpus that are available.....and not for as many crunchers as possible....that is no science, just my to cents.
Even if i buy today the fastest gpu i can get, i can not be shure i can use it on gpugrid in 12 month.....I think the project should think about this.

Only my opinion.
I would upgrade my driver that i can crunch Einstein.

People have been using 200 series cards for around 2 years. Think about only supporting old GPUs. What sort of science do you think that would be?
GPU crunching is more difficult than CPU crunching, otherwise every project would be doing it.
Every Boinc project does alpha and beta work. Despite having a small team this project works in several fields of science and in computer programming (for molecular dynamics). GPUGrid produces scientific publications (hardly Beta work).

Perhaps you missed the bit where I described the 260 driver issues; it causes some GPUs (especially GT240's) to slow down to lowest power mode, making tasks take about 7 times as long. This will still happen on Einstein or Folding or any other project because it is a driver problem. In itself this is a good enough reason to not force people to use this driver, but by doing so you can only crunch tasks on the 6.13app, which is slower for GT240's; because it was primarily written with Fermi cards in mind.

PS. I have 6 GT240's.

Perhaps you missed the bit where I described the 260 driver issues; it causes some GPUs (especially GT240's) to slow down to lowest power mode, making tasks take about 7 times as long. This will still happen on Einstein or Folding or any other project because it is a driver problem.

Not here, it stays at "Prefer Maximum Performance"

In itself this is a good enough reason to not force people to use this driver, but by doing so you can only crunch tasks on the 6.13app, which is slower for GT240's; because it was primarily written with Fermi cards in mind.

The project knows, that the card is worse off with 6.13, but still it insists on pushing those apps on my machine. I'd like to say "No", but there's no possibility, they refuse to give us one.

Again:
All choices are made alone by the project team, and they are made active by them, and thus this is what they especially and definitely want, I have no choice here but to quit or to extremely babysit each and every single WU.

---

Gruesse vom Saenger

For questions about Boinc look in the BOINC-Wiki

As you have repeatedly noted the project team has made their choices ... perhaps it is now time for you to make yours to either work with the project as best you can or move on.

---

Thanks - Steve

As you have repeatedly noted the project team has made their choices ... perhaps it is now time for you to make yours to either work with the project as best you can or move on.

It would seem that many people have already made their choices and decided to move on. With the attitude that has been expressed by the project/moderators (and others), I'm not surprised either.

This project is starting to become an exclusive club for those with the latest cards/drivers. I am still able to crunch GPU Grid on one of my part-time hosts with a GTX260 (that I went out and bought, just so I could run GPU Grid). Although I do have to be careful when I download and start a wu and make sure I turn the pc on for a couple of hours before I go off to work and start it up again when I get home for dinner etc, otherwise I have trouble returning them for the bonus. I stopped crunchin here when the longer tasks came out and now crunch projects that I do not need to babysit.

Please remember that Boinc's mantra is to "use the idle time of your computer..." Many of us would love to contribute more but cannot afford to buy the latest cards or let their machines run 24/7 or be willing to set their machines up just so they can run a few tasks here.

I feel that GPU Grid should work with the people that are willing to crunch it, not to dictate to the people what they need to do to crunch it. If you continue like this, that graph will only get worse

Mark

The gradual rise during Nov is probably due to outages at PrimeGrid, and the obvious outage at GPUGrid around the 4th Dec meant the same people that came to GPUGrid left and went to other places, including PrimeGrid which by that stage was up and running again.
Aqua seems to have faded badly recently, not sure why.
Einstein seems to have taken a seasonal hit along with some of the big CPU projects cosmology, Rosetta, qmc, abc.

The structure of this project requires a fast turnover of tasks. The contribution of casual crunchers that have average GPUs and only have their systems on for one or two hours a day is negligible at best. Boinc projects differ and the default Boinc settings are not always ideal. So crunchers need to change their settings to suite the project. This is the case with other GPU projects too. For example some projects require specific drivers, and have card limitations. Only the project leader can make the changes that you and others want. You and I are not aware of their financial or time constraints. When there use to be a choice of which task to crunch, there were even more problems than now.

PS. Remember that graph starts at 26M and only rises to 36M; it's not a zero up graph!

Still, for whatever reason(s) there are ups & downs, looking at that graph is like looking at the stock market. There's little love & little trust. Even if the top 5 crunchers do 10% of the work, The top 20 crunchers do 22% and the top 100 crunchers do 45%, that drop looks like there's little trust & little love, maybe all this; if you don't like it, go somewhere else because "science" requires us to tell you to go somewhere else...

---

The last three weeks or so have been almost all KASHIF wu's. They return the least points, at about 85% of the previous average.

I have given up on GPUGRID for good this time... what a waste of time!!!

sk and SAM bring up excellent reasons for some of the PPD decline but I think most of the downward trend was from the SETI crunchers who joined us when their project went down and then returned to their favorite project when it came back on line.

If there was a stats site that displayed more than 60 days of history you would probably also see that this recent decline really is only a bubble and that, overall, the project has continued to increase PPD over time.

While recognizing that the XtremeSystems team is most comprised mostly of dedicated 24/7 crunchers with hardware adequate for this project, our stats have steadily increased with minor flucuations based on normal joins and departures, not based on project complexities, but rather on personal goals. Has anyone looked at other teams to see if the same is true for them?

Are there any stats on the length of time people who have departed have been joined with the project?

If it is a quick turnaround then maybe more focus can be placed on clarifying just how demanding the science of this project is to help set expectations. That might help avoid the hightened disappointment we are seeing in this (and some others) thread.

---

Thanks - Steve

For people who find it hard to run GPUGrid on older cards, you might like to know that Einstein has improved the efficiency of its CUDA app (Windows only, so far), and it's now roughly comparable with SETI on one of my 9800GT cards (6,320 seconds for a BRP3 task, under Windows XP/32). Edit - forgot to say that it still uses quite a bit of CPU, though not nearly as much as the old ABP2 app.

The latest Beta driver (266.35) is no better for the GT240, but it allows some tasks to run on CC1.3 cards that previously failed.
Might be worth testing if you have a GTX570 or GTX580 - anyone?

A GIANNI_DHFR1000 task on a GT240 @1.4GHz (just over ref) W2003x64, i7-860@3.25GHz, 17.56ms per step:

3550471 2232832 10 Jan 2011 23:40:55 UTC 11 Jan 2011 14:27:07 UTC Completed and validated 35,131.58 35,120.72 7,491.18 11,236.77 ACEMD2: GPU molecular dynamics v6.12 (cuda)

Good to see a few of these again.
More would go down well, could even get a card RAC of over 27K at that rate.

- Ran another one on a different GT240, W7x64, stock Q6600, 23.715ms per step. Would still get over 20K RAC, but I guess the difference is the CPU and supporting architecture. The i7 is only using one thread, but it's way faster (much more than 11%). The GT240 on the Q6600 system is also OC'd to 1.59GHz.