Message from server: No work sent
Message from Server: There was work available, but none for you as your abacus has the wrong size beads.

Great one banditwolf, though I've ran out of sand in my yard and I've broken over a dozen sticks while scribing hash marks to count the number of times my computers have received the phrase "No work sent" from the server. F-U-T-I-L-E is the word that comes to mind. In the mean time, I am working to bring up my credits in other projects. At least in those who have not retired yet!

It came about from a discussion awhile ago. And someone else used it for awhile first. (forget who)

---

Doesn't expecting the unexpected make the unexpected the expected?
If it makes sense, DON'T do it.

It says on the front page of MW@home that there would be a massive update today about MW GPU.
Where does this update would originate from? California or somewhere in the UK?
If from the UK that means it is already evening there. I am in Central Canada.
That would make UK to be about 8-9 pm right now.

Martin Chartrand

Rensselaer Polytechnic Institute
110 8th St.
Troy, NY 12180-3590

---

CUDA Application Update
May 20, 2009
Looks like I've finally gotten the CUDA application working. I'm in the process of cleaning up the code, and it should be available for you to play around with tonight. It's my first crack at a CUDA application, so it's a bit rough, but currently it's crunching the current stripe 82 WUs in around a minute using a GeForce 9600M GT. I'm sure there's quite a bit more optimization we can to do get it running even faster, as this is the first version and not optimized at all. Part of the reason this has taken so long was that I was doing quite a bit of testing comparing double and single precision math on the GPU. I've come up with some schemes that will allow us to use older graphics cards which don't support double precision math (like the GeForce I've been using); which should let a lot more of our users be able to use the GPU application. More news later when I post the code.
--Travis

---

Doesn't expecting the unexpected make the unexpected the expected?
If it makes sense, DON'T do it.

Are we going to be running the stripe 82s with the CUDA release tonight?

Are we gonna run them @ Milkyway@home or GPU.

If its on the GPU side, are WUs gonna be split between projects??

At the moment I'm using an ATI with app_info for Milkyway and a nVidia for GPUgrid in the same PC. Will it be possible to use an ATI card and a nVidia card at the same time for Milkyway?

Travis wrote:

currently it's crunching the current stripe 82 WUs in around a minute using a GeForce 9600M GT.

Nasicus wrote:

At the moment I'm using an ATI with app_info for Milkyway and a nVidia for GPUgrid in the same PC. Will it be possible to use an ATI card and a nVidia card at the same time for Milkyway?

How exciting. That's the card I have in my laptop - will it share with GPUGRID ?

---

wow...
Your the only project that runs ATI and instead you come out with a CUDA GPU app instead of finishing the ATI WU app...
Sorry if this sounds harsh but it is just the way I feel.

---

wow...
Your the only project that runs ATI and instead you come out with a CUDA GPU app instead of finishing the ATI WU app...
Sorry if this sounds harsh but it is just the way I feel.

I feel the same way! Focus should have been on ATI!
Longer workunits to start with... more workunit availability... enough to get done there first....

So the next thing that comes is that there's two projects where there's no work, as all the ATI stay here, all the CUDA are there... I wonder if there's a 3rd project coming.

Milkyway_CPU
Milkyway_GPU_ATI
Milkyway_GPU_CUDA

;-)

---

Jord.

The BOINC FAQ Service.

you should take a look around the forums and read why before making statements.

There doing it just to piss you guys off...

There not doing because they want too, there are many reasons, compatibility, Dave Anderson, much more.

I dont feel like I should have to re-iterate... there around.

But this way I could be sure that I can use my CPU, the ATI and the nVidia all in one PC at the same time for Milkyway ;-)

very true! Then you could!! LOL

wow...
Your the only project that runs ATI and instead you come out with a CUDA GPU app instead of finishing the ATI WU app...
Sorry if this sounds harsh but it is just the way I feel.

I feel the same way! Focus should have been on ATI!
Longer workunits to start with... more workunit availability... enough to get done there first....

I guess I have to repeat this until I'm blue in the face, but CUDA is supported within BOINC as an application. No projects, none, zip, nada, under the BOINC umbrella are running an ATI application as an official project release. As I'm reading the history of GPU computing with BOINC, which you can do by utilizing Google just like I did, you'll find that AMD ("ATI") did not offer help, while nVidia did offer help. Hence, the back end server code as well as code in the later 6.x BOINC client releases that is designed for handling GPU projects is for CUDA. Getting the ATI project done as an "offical" project with "official" support would've taken longer, particularly for any server code interfacing.

As I have mentioned as well, there is a good possibility that people with CUDA cards in systems here that are processing with CPU will migrate to a CUDA-enabled GPU project, thus alleviating the stress on the server here, so even those of you who want to be making more points with your ATI cards may *STILL* have that opportunity because at least some people will shift to the other project.

Alternatively, the project may have taken another 3-6 months for doing an ATI project in-house with no help from other BOINC resources who they could turn to for CUDA.

I don't know why, other than for the want of more points, this is so difficult to understand... If you're an early-adopter of a technology, there is always a risk that you find out that you made the poorer choice once the benefit of hindsight is available...

Or probably not

---

Its new, its relative fast... my new bicycle

Travis wrote:

currently it's crunching the current stripe 82 WUs in around a minute using a GeForce 9600M GT.

Nasicus wrote:

At the moment I'm using an ATI with app_info for Milkyway and a nVidia for GPUgrid in the same PC. Will it be possible to use an ATI card and a nVidia card at the same time for Milkyway?

How exciting. That's the card I have in my laptop - will it share with GPUGRID ?

Probably this card will work. You should expect durations of 24h to more than 50h for the GPUgrid WUs on this 9600M GT which has 32 shaders (on my 8800GTS 512 with 128 shaders they take about 10 to 16 hours). But this are only estimations as I don't have a 9600M GT.

hey, anyone with some answers able to answer this..?

Are we going to be running the stripe 82s with the CUDA release tonight?

Are we gonna run them @ Milkyway@home or GPU.

If its on the GPU side, are WUs gonna be split between projects??

oh and

Or probably not

Quite the necessary post...

***

GPUGRID IBUCH WU's take about 220000 seconds to run on my 8600mGT

and about 50000 seconds on my 9800GT

I would put a 9600mGT around 150000 or about 42 hours

Dear Brian.

I do understand that ATI doesn't offer much support. I have been following the forum. But my question is, why shift to CUDA (if you want to have more work done from the perspective of the project itself), it would have been better to fully utilize the ATI GPUs out there that are idling for work.
Once this is done, and more computing power is needed, the second project supporting CUDA would have been smart enough.

I assume that both CUDA and the "CPU" project in the end process the same data and are therefore working towards the same goal and the MW@home GPU project is not going to be loaded with completely different stuff.

I hope you didn't feel offended by my post and I'd invite you to a nice cold beer if I were any place near you.

Thanks for your reply anyway.

Cheers,

Sysfried

Dear Brian.

I do understand that ATI doesn't offer much support. I have been following the forum. But my question is, why shift to CUDA (if you want to have more work done from the perspective of the project itself), it would have been better to fully utilize the ATI GPUs out there that are idling for work.

As I understand things, the bottleneck of the feeder -> scheduler shared memory segment interfacing is the root of the problem. The only automatic way I can think of to accomplish what you seem to want, more intensive tasks, without strangling CPUs out of the project completely would be to modify Homogeneous Redundancy and get it to figure out what an ATI system looks like and use HR classes to distribute work. A semi-automatic (don't get ideas here!) way would be something similar to the "Run test applications" setting. If you had "Run GPU applications" enabled, you'd get a GPU task. If not, you'd get a CPU task. Well, due to the fact that people like to claim "CHEATER!" a lot, the project would also have to build in some sort of check to totally disallow a GPU from processing a CPU task, else the people who were doing the GPU tasks with their GPU and a CPU task with their CPU would complain if someone were to, say, use the current GPU anonymous platform application to churn through CPU units very quickly.

Once this is done, and more computing power is needed, the second project supporting CUDA would have been smart enough.

As I pointed out above, there are major design issues to contend with keeping the project integrated. Additionally, technically speaking no more computing power is needed at this point for the project as it was originally defined. What they're doing with the GPU project is expanding the design specification of the original project dramatically, so much so that a CPU is deemed not capable of doing the work within a reasonable amount of time.

I assume that both CUDA and the "CPU" project in the end process the same data and are therefore working towards the same goal and the MW@home GPU project is not going to be loaded with completely different stuff.

No, CUDA is going to be handling what I mentioned above, tasks that CPUs are currently not capable of doing in a reasonable amount of time. What is likely going to happen is they'll release the CUDA source code and someone, such as Cluster Physik (or however you spell it) will work on an anonymous platform version just like what was done with the current project. This whole thing is happening exactly like what happened with the current project. The official application, then the 3rd party application.

I hope you didn't feel offended by my post and I'd invite you to a nice cold beer if I were any place near you.

Not offended, just frustrated that people don't think about things logically.

Oh, and I don't like beer, but appreciate the offer... ;-)

As I understand it, ATI is to Folding@home as NVIDIA’s CUDA is to Boinc, (at the very least). Just my observation.