hmmm... perhaps someone needs to ask the f@h guys how they do it :)

I doubt that they use OpenCL, probably CAL.

openMM, aka gpu3, uses opencl but F@H is running into similar problems as gpugrid. two of their kernels wouldnt compile.

Hi thare I'm receiving work's for GPu ATI.
My 5870 is not occ, but it sending errors...
it is ACEMD beta version 6.03 (ati14)

estimated time is 1.20min but few sec ater starting procesing it info "error"
System win 7 64 bit
boinc manager 6.10.18

My fault. I forgot to remove the ATI app, while we are giving out the new beta app for nvidia.
Fixed now.
gdf

OK....

so there is still no work for ati...:(

Take a look here.. you're not going to get any work soon, as GPU-Grid doesn't want to crash your machine.

MrS

---

Scanning for our furry friends since Jan 2002

it will take several months before the SDK becomes stable enough to be of practical use.

GDF

maybe the situation is not so tragic in AMD...
SDK v2.01 is out, with many bugs fixed and some FFT samples (if you still need them). You could retry now and maybe report new bugs found, to help development!

We are testing it out.

gdf

Please excuse my ignorance. But is GPUGRID using OpenCL or porting CUDA to OpenCL? If so, are there plans to go native with OpelCL or is ATI happy to ride around on Nvidia's horse? (BTW Kuda in Indonesian means horse, I know it's spelled with a C!)...

Punn set aside. Nvidia is a sponsor of GPUGRID, but if Einstein, SETI, or Collatz Conjecture can better utilize ATI cards, ATI card owners will be tempted to go there instead.

---

is GPUGRID using OpenCL or porting CUDA to OpenCL?

We use CUDA for our application. We are experimenting with OpenCL with a view to being able to use ATI cards, but results to date have not been encouraging.

MJH

but if Einstein, SETI, or Collatz Conjecture can better utilize ATI cards, ATI card owners will be tempted to go there instead.

Collatz and Milkyway make excellent use of the ATIs, computationally wise. Whether you think their science is worthwhile or not is up to you ;)

MrS

---

Scanning for our furry friends since Jan 2002

We are testing it out.

gdf

any good news? does it still crash when compiling?

could i suggest to release a beta app and work-units for ATi, even if performances are poor and maybe first results not really important for science?
We need an OpenCL app to start the chain: opencl boinc implementation -> optimizations -> newer releases -> drivers getting better -> users more satisfied -> new science -> new apps -> optimizations and so on...

Now you're in the position of being able to benefit another time gpgpu in boinc (such as when you developed the first cuda on boinc app) releasing early (and often, maybe) and breaking this stalled situation, if you have a working app (even if it's not great in performances, it will become better one day)

Have a read of this AMD Developer Forum thread - started by a programmer working on a BOINC project.

Doesn't seem to be much chance of a viable general-purpose OpenCL implementation in the near future.

Have a read of this AMD Developer Forum thread - started by a programmer working on a BOINC project.

Doesn't seem to be much chance of a viable general-purpose OpenCL implementation in the near future.

There're some bugs, I know. And I read really often AMD developers forums, but there's not much activity going on there. I'm sure that a lot of discussions are going on behind the scenes via email, which I do not like in this stage because I think that on a forum everyone would benefit.

Raistmer has many problems porting multibeam to ati because of these problems with Apple FFT opencl implementation. But this is not the whole story: not every app needs FFT, and more important everybody could write a fast fourier transform implementation on OpenCL if this one doesn't fit your needs or is broken. We couldn't ask Raistmer to do this (he is in the lunatics crew), but here there's a whole project behind...
In this thread GdF (or someone of his group) said that they hadn't the possibility to implement one themselves too, but they were watching MANY implementations and that they could do some science at the beginning also without FFT-enabled app.

Do not forget that MANY apps written in OpenCL are working NOW (nothing on boinc, but boinc is just a little program that close-to-no one knows).

Also, I'm sure that we will have another refresh of the Stream SDK before those "few months" cited by genaganna: they said the same after SDK 2.0 but we have just had refresh 2.01 with many bug fixed. ATi is doing many efforts in their software departments (having learnt that hardware is not the only important thing) and maybe (maybe) genaganna doesn't know the whole story.

On the other hand: are gpugrid project developers writing on some developers forum (Apple, nvidia, ATi) about their problems or are they using emails??

Hi,
we are talking to ATI directly by email.
We managed to compile and run the old application. On a 5850 is at present 2.5 times slower than a GTX 275 and furthermore it does not allow to be used scientifically really. If we could work with the newer application that would be much better. Hopefully soon.

We will do a little boinc test to see how reliable is the ATI driver in the next couple of weeks.

gdf

Hi,
we are talking to ATI directly by email.
We managed to compile and run the old application. On a 5850 is at present 2.5 times slower than a GTX 275 and furthermore it does not allow to be used scientifically really. If we could work with the newer application that would be much better. Hopefully soon.

We will do a little boinc test to see how reliable is the ATI driver in the next couple of weeks.

gdf

thanks a lot for sharing the app status. I was sure that also you were talking privately with ati support.

ps: within 30 minutes 10.2 drivers will be out.
Watch on those for your reliability tests!

Hi, have crunched for GPUGrid before, on a 9800GTX+, which is now in use for other purposes. But I've got a rig ASUS P5E; X38; Q6600 and a ATI HD5770.
This is a 'single precision' unit, runs at Collatz C. at the moment.
Can I use this card with GPUGrid, also?

---

Knight Who Says Ni N!

GPUGrid does not currently have work for ATI but they continue to review all new developments on the SDK from ATI and spin up new betas to test things out but afaik there is nothing imminent.

---

Thanks - Steve

Please excuse my ignorance. But is GPUGRID using OpenCL or porting CUDA to OpenCL? If so, are there plans to go native with OpelCL or is ATI happy to ride around on Nvidia's horse? (BTW Kuda in Indonesian means horse, I know it's spelled with a C!)...

Punn set aside. Nvidia is a sponsor of GPUGRID, but if Einstein, SETI, or Collatz Conjecture can better utilize ATI cards, ATI card owners will be tempted to go there instead.

Are you aware of any software for automatically converting CUDA to OpenCL? I'm not.

Are you aware of any ATI SDKs that include a reliable version of the FFT software? I'm not.

Are you aware of any BOINC versions, even still in alpha test, that can use OpenCL directly instead of compiling it to CUDA or ATI's equivalent first? I'm not. How do you expect an OpenCL-only BOINC application to be produced before there is a suitable version of BOINC to test it under, or do you want one with a very high error rate even for pre-alpha testing?