6.99 works well, too well. It's fine I have to suspend GPU accelerated tasks when I want to use the GPU for DXVA, but I can't run SETI@Home Beta even when I want to listen to music or browse the web because the whole system becomes laggy. It's very annoying to keep suspending and resuming all the time.

Check your project directory for MB_cmdline.txt.
Close Boinc
Open with notepad and insert -period_iterations_num 50 and save it again.
Restart Boinc again.

---

With each crime and every kindness we birth our future.

6.99 works well, too well. It's fine I have to suspend GPU accelerated tasks when I want to use the GPU for DXVA, but I can't run SETI@Home Beta even when I want to listen to music or browse the web because the whole system becomes laggy. It's very annoying to keep suspending and resuming all the time.

I run on ATI 5850 and zero percent lag here.

6.99 works well, too well. It's fine I have to suspend GPU accelerated tasks when I want to use the GPU for DXVA, but I can't run SETI@Home Beta even when I want to listen to music or browse the web because the whole system becomes laggy. It's very annoying to keep suspending and resuming all the time.

I run on ATI 5850 and zero percent lag here.

Same here with my 4850. No lag whatsoever, 90-92% GPU load, with one CPU core reserved for feeding the GPU.

I wonder what the default value for -period_iterations_num is? My GPU seems able to take a bit more beating.

6.99 works well, too well. It's fine I have to suspend GPU accelerated tasks when I want to use the GPU for DXVA, but I can't run SETI@Home Beta even when I want to listen to music or browse the web because the whole system becomes laggy. It's very annoying to keep suspending and resuming all the time.

I run on ATI 5850 and zero percent lag here.

Same here with my 4850. No lag whatsoever, 90-92% GPU load, with one CPU core reserved for feeding the GPU.

I wonder what the default value for -period_iterations_num is? My GPU seems able to take a bit more beating.

Default is 15.
I wouldn`t change it.
Mine is set to 30 for example.
Doesn`t change calculation speed.
You can add -sbs switch in MB_cmdline.txt instead to tune the app a little bit.

Default is 64 you can go up to 256 .

So switch would look like this.

-sbs 256.

Mike

---

With each crime and every kindness we birth our future.

6.99 works well, too well. It's fine I have to suspend GPU accelerated tasks when I want to use the GPU for DXVA, but I can't run SETI@Home Beta even when I want to listen to music or browse the web because the whole system becomes laggy. It's very annoying to keep suspending and resuming all the time.

I run on ATI 5850 and zero percent lag here.

Same here with my 4850. No lag whatsoever, 90-92% GPU load, with one CPU core reserved for feeding the GPU.

I wonder what the default value for -period_iterations_num is? My GPU seems able to take a bit more beating.

Default is 15.
I wouldn`t change it.
Mine is set to 30 for example.
Doesn`t change calculation speed.
You can add -sbs switch in MB_cmdline.txt instead to tune the app a little bit.

Default is 64 you can go up to 256 .

So switch would look like this.

-sbs 256.

Mike

If the default is 15, and the guy you told to change it to 50 had a laggy system, wouldn't that make it even more laggy then?

On MB_6.10_win_SSE3_ATI_r177.exe the initial value I had for period_iterations_num was 2, and by changing that to 20 (higher introduces lag on my system) the GPU load became much higher, and the tasks runs much faster.

Added: -sbs stands for what, and what does it do?

No.
Higher values decrease screen lags and reduce calculations speed a little.
Up to 100 you shouldn`t notice.

sbs stands for Single buffer allocation size.
So higher the value so more data is loaded in one kernel call in GPU at the same time.
Max is 256 and causes higher GPU memory usage.
You can ignore that on stock apps but it is important if you run multiple instances like i do.

---

With each crime and every kindness we birth our future.

No.
Higher values decrease screen lags and reduce calculations speed a little.
Up to 100 you shouldn`t notice.

sbs stands for Single buffer allocation size.
So higher the value so more data is loaded in one kernel call at the same time.
Max is 256 and causes higher GPU memory usauge.
You can ignore that on stock apps but it is important if you run multiple instances like i do.

Dammit Mike, you're right. Thanks for waking up my sleeping mind. Now I remember why I upped mine to 20, it was ages ago, and I'm not the youngest of players. :-)

It was because I was running 2 at the time, and when it happened to be VLAR's, the computer became very very laggy. Even with one VLAR only running only 1 WU at the time, there was some lag effects with the initial value I had (-2)

Thanks for the -sbs thing though, I may try it one day even though I run stock. Just for the fun of it, and while I'm at it, it's likely that I can't keep away from lowering period_iterations_num a bit too.

Cautious, i could produce calculation errors setting period_iterations_num lower than 5.

---

With each crime and every kindness we birth our future.

A little correction:

#if !( _GNU_SOURCE || __APPLE__ )
int period_iterations_num=20;//R: default for usage in stock app
#elif _GNU_SOURCE || __APPLE__
int period_iterations_num=31;
extern unsigned int BOINCs_dev;
#endif

https://setisvn.ssl.berkeley.edu/svn/branches/sah_v7_opt/AKv8/client/main.cpp
So, current default is 20 for windows and Urs decided to make it 31 for forecoming Linux build.

Perhaps I rised it from 15 on some stage to make it more friendly as stock app.

EDIT: BTW,don't forget that sources are available online for anonymous browsing too so if someone can read code and have will to do it and maybe even improve it - you are welcomed!

added -period_iterations_num=15 -sbs=256 -hp

It is about 10% faster with this setting.

Hello, by boinc is not downloading gpu based exe files and WU-s for the beta projects. (cpu part works but since I only want to stress my gpu, I have turned it off)

I have an AMD HD 7870 vga, latest drivers present, running einstein@home and collatz just fine.

Could someone help me please?

From what I'm seeing in the server logs, your work requests always ask for 0 seconds of ATI GPU work.

I would guess that means your GPU is fully committed with work for other projects. What happens if you suspend all other projects for a short time?

---

Thank you! That did the trick. Tweaked the priority settings on the projects I run. Too bad I can only run collatz when I'm not at my PC because it just kills my desktop. Even my mouse pointer gets all jerky.

GPU usage for the BETA is 80% CPU usage of the process is 5% so it seems to behave nicely so far.

Cool. Glad to hear its working.

---

Maybe it's because I run SETI@Home Beta on my IGP (A8-3850) which shares memory bandwidth with the CPU cores.

finished my first task.
valid but very long runtime on my gpu

http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4163435

finished my first task.
valid but very long runtime on my gpu

http://setiweb.ssl.berkeley.edu/beta/workunit.php?wuid=4163435

Max compute units: 2
Max work group size: 128

Yes, it's low-level GPU, quite slow, but still consumes much less CPU then CPU build you paired with.

Sounds like the description of my card. I have now 5 valid results.

---

Christoph