Posts by cristipurdel
log in
1) Message boards : Wish list : Mobile GPU support (Message 20806)
Posted 925 days ago by cristipurdel
Hello,

I would like to see support for the GPU cores in any mobile device since the current generation is quite powerful (over 100 GFLOPs for mid to high range)

for Android maybe only the smartphones which support OpenCL could be a possibility
for upcoming Windows 10 Mobile I hope it will be easy enough :)
for UbuntuPhone the Linux driver support (E.g. Mali) maybe it is helpful

Thank you in advance
2) Message boards : Number crunching : Future collatz applications? (Message 6888)
Posted 2853 days ago by cristipurdel
Would it be possible that, not in the distant future, collatz apps would be run on OpenCL compute units, to take advantage of all the power from the shaders?
I think it would be more efficient if the apps could be run, let's say using my own 4650 mobility card, on 8 OpenCL compute units (8 * 8 * 5 = 320 shaders) independently, in the same way that boinc has an option for how many cores to use on multiprocessor systems.

Don't worry, the GPU applications use all available shaders automatically. But a single WU uses all shaders.

It is currently not possible to "split" the GPU and let the compute units (an OpenCL compute unit is a SIMD on ATI GPUs or a SM on nvidia GPUs) work independently on different tasks. The hardware supports that starting with the HD5000 line or nvidias GF100 (GTX480/470), but the APIs don't do it really right now. Furthermore this would increase the memory requirements (each compute unit would need the amount necessary for a WU). Generally, it only provides benefits for the manageability and especially for tasks with a low parallelism (i.e. tasks which can't use the full GPU). Collatz is an embarrassingly parallel problem, it is very easy to use all units for a single WU without compromising the scaling. That's the reason a GPU with twice the shader units (at the same clock and with enough bandwidth) is exactly twice as fast.

That was the answer I was looking for.
I asked this, in order to avoid some problems regarding the lag that may appear when running collatz in the background.
I think this could be avoided if the first ATI SIMD or NVIDIA SM could be "told" not to crunch, but to deal with OS rendering. Since is better than the Intel HD Graphics, it could be a start to "split" the GPU just for these 2 tasks (OS & BOINC)
3) Message boards : Number crunching : Future collatz applications? (Message 6868)
Posted 2854 days ago by cristipurdel
With the current SDK, it has been proven that while you can compute on the 4xxx and down models, the 5xxx models are not crunching correctly.

???
4) Message boards : Number crunching : Future collatz applications? (Message 6863)
Posted 2854 days ago by cristipurdel
Would it be possible that, not in the distant future, collatz apps would be run on OpenCL compute units, to take advantage of all the power from the shaders?
I think it would be more efficient if the apps could be run, let's say using my own 4650 mobility card, on 8 OpenCL compute units (8 * 8 * 5 = 320 shaders) independently, in the same way that boinc has an option for how many cores to use on multiprocessor systems.
Not trying to pressure anyone, just asking for a possible e.t.a. :)
5) Message boards : Number crunching : ATI application Q&A (Message 6200)
Posted 2897 days ago by cristipurdel
I agree the problem is the driver and ATI/AMD versus HP; i.e., I have to get it from the computer manufacturer instead of ATI. But when I tried following Arkayn's link a while back and installing an updated driver, Catalyst Control Centre would crash-something about Data Execution Protection. (See http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=1537&nowrap=true#36560.) I did not see my type of GPU listed at the site Scott provided, but I might have overlooked it and will look again.

BTW, I still don't understand how BOINC Manager can be checking for usable GPUs when what can be used by various projects itself varies.

Or you could try these drivers http://www.station-drivers.com/page/ati%20catalyst.htm if you trust the frenchmen :)
Version 10.x (8.700) WHQL - worked on my mobility radeon 4650 win 7 64 and after I've installed ati sdk 2.0.1, I also have OpenCL ;)
6) Message boards : Number crunching : Are you ready? (Message 3617)
Posted 2993 days ago by cristipurdel
Okay Thanks, that's pretty much how I had it set up anyway, not that I'm getting any work from the Project but if I do I'll get a chance to see what I think of it & if it's worth running them ...

http://lunatics.kwsn.net/12-gpu-crunching/cpu-ati-gpu-hybrid-astropulse-for-windows-released.msg23139.html#msg23139

my app_info is:
<app_info>
<app>
<name>astropulse_v505</name>
</app>
<file_info>
<name>ap_5.05_win_x86_SSE3_BROOK_r280_NO_DOUBLE.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>astropulse_v505</app_name>
<version_num>506</version_num>
<avg_ncpus>0.1</avg_ncpus>
<max_ncpus>1</max_ncpus>
<coproc>
<type>ATI</type>
<count>0.01</count>
</coproc>
<file_ref>
<file_name>ap_5.05_win_x86_SSE3_BROOK_r280_NO_DOUBLE.exe</file_name>
<main_program/>
</file_ref>
</app_version>
</app_info>
I'm not receiving any 'regular CPU' WU
7) Message boards : Number crunching : Are you ready? (Message 3610)
Posted 2993 days ago by cristipurdel
I took a look at the open beta at the lunatics forum, which requires ATI GPU a few days ago but it was way over my head how to get it to work plus you have to compile your own app_info.xml file which is out of my league too ... It made my eyes hurt to even read that Rainbow of colors they use at the Lunitics forum ... :P

It's no longer in open Beta, as the hybrid AstroPulse app has been released, app_info is in the first post:

CPU/ATI GPU hybrid AstroPulse for Windows released

Claggy


From my read on it you would be running regular SETI Wu's along with the Hybrid ones, that's something I'm not willing to do is run the Regular SETI Wu's.

Not quite. I managed to run the AP ATI with 0.01 GPU + 1 CPU and Collatz app with 0.99 GPU + 0.02 CPU . They say to run regular with hybrid simply because the hybrid doesn't utilize too much the GPU.
8) Message boards : Number crunching : Are you ready? (Message 3594)
Posted 2993 days ago by cristipurdel
First of all I'm not trying to say that one project is better than the other. I was just trying to say that there is an open beta at the lunatics forum, which requires ATI GPU, and since a lot of people over at MW were using them, I was trying to point out that they could also be helpful over at SETI as well as they could be over at gpugrid.
Regarding the pornographic part, I real don't know about that part, I've been crunching over 3 years now, and I started with seti and after with rosetta, cpdn, einstein, wcg, folding and collatz.
Don't want to influence anybody, but since the pornographic thing got out of hand let me say just one thing: when members of your family die of cancer, and there isn't any cure, would you choose an astronomy or math project over a medicine one?
9) Message boards : Number crunching : Are you ready? (Message 3571)
Posted 2993 days ago by cristipurdel

Yep, MW will be down for some days and gpugrid doing some ati 4850+ testing.

Don't forget SETI ;)
10) Message boards : Number crunching : Nvidia CUDA-Performance Thread (Message 3189)
Posted 3008 days ago by cristipurdel


28/10/2009 09:18:02 NVIDIA GPU 0: Quadro FX 580 (driver version 19100, CUDA version 2030, compute capability 1.1, 512MB, 72 GFLOPS peak)


One question though, it seems that the single precision float for single precision is calculated like: Peak = Shaders * Shader Clock * 2
where as on wikipedia it is Peak = Shaders * Shader Clock * 3

http://en.wikipedia.org/wiki/Comparison_of_Nvidia_graphics_processing_units

Which is it the total peak?
11) Message boards : Number crunching : OpenCL (Message 2850)
Posted 3021 days ago by cristipurdel
The ATI 4xxx series are already over a year old. If an OpenCL project will appear after 6 months, it will be the same thing what's happening on the folding project right now, old software running on old & new hardware. From my point of view, Pushing the OpenCL client seems more important, so that it would unify the rest of the clients and make it easier for other projects to develop GPU projects.
12) Message boards : Number crunching : OpenCL (Message 2812)
Posted 3022 days ago by cristipurdel
ATI Stream Software Development Kit (SDK) v2.0 Beta Program
With OpenCLâ„¢ 1.0 Support

http://developer.amd.com/gpu/ATIStreamSDKBetaProgram/Pages/default.aspx

So, who needs a beta tester for OpenCL? :)
13) Message boards : Number crunching : ATI application Q&A (Message 2716)
Posted 3025 days ago by cristipurdel
True. I was trying to set up a couple of BOINC clients on Win 7, and I found out that Win 7 also comes with .NET 3.5 support. I 'almost' miss the 'old' days where I had to copy the 3 dlls, and install the .NET packages. Now that's what I call ...technology. ....Deutsch-Technologie
14) Message boards : Number crunching : ATI application Q&A (Message 2714)
Posted 3026 days ago by cristipurdel
I remember that Gipsel posted somewhere a sticky on how to run the ATI client, which framework .NET and so on to install, can somebody put this also here, or at least put a link. I've been searching for over half hour and I couldn't find it.
15) Message boards : Number crunching : ATI 2.04 Applications Released (Message 2217)
Posted 3036 days ago by cristipurdel
For me the 2.04 application is running ok (can really feel the speed bump), but boinc manager says it's running collatz 2.00. Do I have to wait until the servers finish the 2.00 version?
16) Message boards : Number crunching : v2.0 Application Issues (Message 1861)
Posted 3043 days ago by cristipurdel
23/09/2009 14:32:39 Collatz Conjecture [error] collatz_1253542062_28824_0: negative FLOPs left -2.000000
what's this error supposed to mean?
17) Message boards : Number crunching : Collatz v2.0 Status (Message 1613)
Posted 3045 days ago by cristipurdel
I changed the <cmdline>n1</cmdline> with <cmdline>n1 w1.2 f50</cmdline> and the system is more responsive now.
Problem now is that while using the app_info, BOINC is requesting GPU WUs, doesn't receive them, asks for CPU WUs, receives them (it says only 0.05 CPUs ), and starts running them on the GPUs. Another small bug is that in a couple of minutes I received around 15 WUs, and while only one was running, the others were loaded in memory(around 15 MBs each) and did nothing.
Anyway, since it's more responsive, and the application it will improve, I'm not complaining too much :)
18) Message boards : Number crunching : Collatz v2.0 Status (Message 1568)
Posted 3045 days ago by cristipurdel
Haha yes, or you could try experimenting with the w parameter. Just add it to the app_info.xml as previously explained by Gipsel. You could start off with w1.1 or w1.2 and see how you go. If that isn't sufficient then you try increasing it further to w1.3 or more.

I used the app_info.xml and it used 1.4 and the gpu % went down to 84%.
But, my system (video card) has some freezes, and it is laggy, even with the low GPU %.
If in a future release if you could do something about the interference with the gpu, it would be perfect.
Or maybe Cruncher can put the GPU % in a new release ;)

Crunch3r had nothing to do with the ATI app so far ;)
And the option you are demanding already exists. Just add f50 or something in that range to the list of command line parameters. It should look like this:
<cmdline>w1.4 f50</cmdline>

And the short freezes you observe may be caused by the CCC, GPU-Z or similar monitoring tools (often depends on OS and driver version if they are causing problems). So close this tools (or just the CCC monitoring/overclocking tab) and see if it helps.

Sorry, I was thinking of...a new BOINC release from Cruncher.
I'm not using any monitoring tools whatsoever. Don't even have CCC.
Point is, when I'm watching a movie, the video starts to lags, also firefox tabs behave in the same manner.
From what I see, collatz v 2.0 ati uses core priority below normal, while for instance folding uses low. And while I'm folding, I don't have this problems while watching movies (from what I remember folding was limited to 95%). My point is, maybe it would be a better idea to limit the gpu, somehow (I'll try TThrottle) to say 95%, so to let the system 'breathe' a little bit.
19) Message boards : Number crunching : Collatz v2.0 Status (Message 1560)
Posted 3045 days ago by cristipurdel
Haha yes, or you could try experimenting with the w parameter. Just add it to the app_info.xml as previously explained by Gipsel. You could start off with w1.1 or w1.2 and see how you go. If that isn't sufficient then you try increasing it further to w1.3 or more.

I used the app_info.xml and it used 1.4 and the gpu % went down to 84%.
But, my system (video card) has some freezes, and it is laggy, even with the low GPU %.
If in a future release if you could do something about the interference with the gpu, it would be perfect.
Or maybe Cruncher can put the GPU % in a new release ;)
20) Message boards : Number crunching : Collatz v2.0 Status (Message 1386)
Posted 3047 days ago by cristipurdel
Which parameter should I tinker with, in order to decrease a little bit the gpu usage? On my 4650 is running 99% and when I play some movies, there are some problems with audio sync.


Next 20

Main page · Your account · Message boards


Copyright © 2018 Jon Sonntag; All rights reserved.