Hi,

Am I correct in assuming there's no Linux package available as of now?

For the record, I run:
1xAMD64x2/4400 + 8400 / Ubuntu 8.10 / standard Ubuntu Boinc + AK_V8_linux64_sse3
1xAMD64x2/4850 + 8400 / Ubuntu 8.10 / standard Ubuntu Boinc + AK_V8_linux64_sse3

I'm planning on buying a better-but-cheap 100W-or-so GPU for the processing power for the 4400.

I could do some testing if it's useful, but I might need some hand-holding.

---

Crunch3r did some Linux (and maybe OS X too ? ) build of CUDA MB.
I don't know if it available for free download but its results in discussion on Lunatics site (in general - if you want news about opt app - register on Lunatics site ;) )

By reports of Linux testers (I don't have Linux with CUDA enabled GPU so can't do tests by myself now) this build uses 100% of CPU so it would be very interesting to establish will this additional CPU usage increase GPU throughput and in what degree.
Perfect test environment would be some host with dual boot Windows/Linux, CUDA -enabled GPU and experienced operator who can use KWSN testbench properly.

ADDON: Yes, it available for regular Lunatics users at least:
http://lunatics.kwsn.net/linux/seti-mb-cuda-for-linux.0.html

Crunch3r did some Linux ...

ADDON: Yes, it available for regular Lunatics users at least:
http://lunatics.kwsn.net/linux/seti-mb-cuda-for-linux.0.html

I'm running that here and it works fine.

Check carefully the app_info.xml file for any typos! Also, you may need to add your own app_info mods if you already have other optimised apps available for fallback to CPU processing if the GPU runs out of VRAM.

Happy fast crunchin',
Martin

---

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

SETI@home error -12 Unknown error
cudaAcc_find_triplets erroneously found a triplet twice in find_triplets_kernel
File: d:/BTR/seticuda/seti_boinc/client/cuda/cudaAcc_pulsefind.cu
Line: 236
F.

Thanks for report. It's known bug for 6.08 CUDA MB code, I passed bug report to CUDA dev already. This bug was observed with 6.08 stock too (as should be in most cases because CUDA-related codebase for stock and mods is the same).

Keep posting if you will notice some another CUDA error, please.

Looks like I've had something similar on my Linux GPU system:

SETI@home error -12 Unknown error
cudaAcc_find_triplets erroneously found a triplet twice in find_triplets_kernel
File: ./cudaAcc_pulsefind.cu
Line: 236

Happy crunchin',
Martin

---

See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)

How do you set up the cc_config.xml file for a core i7 chip?
I have one coming next week.

Interesting question indeed. It depends of how good AK v8 and AP will utilize HT.
In HT P4 days pair of opt app could perform worse than single opt app with HT disabled.

If app is highly optimized it tends to use all available hardware resources of core - there is too low amount of idle core parts that can benefit from HyperThreading.

So I consider it as open field for experimentation on this stage.

So do I set it for 4 cores+1 or 8 cors+1

Hi
I'm running 8+1. Have tested 4+1 but then CPU was only running at 50%

---

//Peter

Can someone help me with how to install this app? Admittedly, I am a noob, but I like to experiment and do my part. I crunch with a Q6700 and an nvidia 8400 gs. I can download the app (sse3, right?), but can't figure how to install. Thanks.

I am running your v7 app and running non-BOINC projects on the CPU. The non-Boinc project is starving out the GPU. What can I do to make the SetiCuda client run high Priority?

How exactly it starving out GPU ? What project?
What symptoms ? CUDA MB is running on higher priority than other projects already.

When running Community TSC on both cores of my 4450e and SETI CUDA on my 8800GS, SETI gets starved. I guess TSC runs at a high priority too. I changed all the settings in the app_info file from 0.04 to 1.0 and the GPU still staved with TSC on both cores.

The same thing happens when I run TSC with RC5-72 CUDA too.

TSC is just a core hog I guess.

---

Maybe try for roll back to 180.48?
I use that driver on all cards and the 9500's will only work with that or 178.xx, everyting work fine then.
Not sure if the problem but I know I can't use anything above 180. Good luck

Thanks for suggestion westsail.
Couldn't get it to run with any of the versions + 6.6.3 or 6.6.2 and V7. Tried down to 178.24...
Now running 6.6.0 with 181.22 and V7 on my i7 with 3x 9500GT but it's slow. Maybe it's quicker with 1x videocard and V8 instead.

Raistmer, do you think there will be multi-GPU support in next version?

---

//Peter

Peter you should try boinc 6.2.19 with v8 of raistmer pack. very good performance and stability overall.

How do you set up the cc_config.xml file for a core i7 chip?
I have one coming next week.

Interesting question indeed. It depends of how good AK v8 and AP will utilize HT.
In HT P4 days pair of opt app could perform worse than single opt app with HT disabled.

If app is highly optimized it tends to use all available hardware resources of core - there is too low amount of idle core parts that can benefit from HyperThreading.

So I consider it as open field for experimentation on this stage.

So do I set it for 4 cores+1 or 8 cors+1

Hi
I'm running 8+1. Have tested 4+1 but then CPU was only running at 50%

It's not a indicator at all!
As was said already HT technology creates 2 logical (for OS) CPUs from single core. And your OS shows load of it's logical CPU. If OS runs 4 threads on 8 logical CPUs then 4 CPUs will be idle so total CPU load will be 50%. But it will not prevent all 4 cores to be fully loaded and performing best.
So you should sum wall clock times (or RAC at least) to judge what config is faster, not just look on CPU % load (if your host using HyperThreading).

ADDON: Actually, it's quite possible that best config for i7 will be not 4+1 and not 8+1 but something between 4 and 8. Testing needed.

[quote]Maybe try for roll back to 180.48?
Raistmer, do you think there will be multi-GPU support in next version?

Already is - V7.
Kick BOINC devs for better scheduling support, hate to reinvent the wheel ;)

Can someone help me with how to install this app? Admittedly, I am a noob, but I like to experiment and do my part. I crunch with a Q6700 and an nvidia 8400 gs. I can download the app (sse3, right?), but can't figure how to install. Thanks.

extract archived files into SETI project directory.
And try to search Illustrated guide for opt app installation. It was some link on it somewhere - very nice guide indeed.
(Pity I didn't bookmark it at first sight...)

Raistmer.

Nice job on the modifications. After crunching for the first time in 3 weeks it was a very nice thread to find.

I would take my hat of to you sr, but it's too cold at the minute.

---

Thanks,...but I know all that part. My problem is that when I try to download from lunatics, instead of getting the usual Microsoft "Open or Save", I get "Find or Save". If I save the download, I can't open it.

Thanks,...but I know all that part. My problem is that when I try to download from lunatics, instead of getting the usual Microsoft "Open or Save", I get "Find or Save". If I save the download, I can't open it.

Maybe You don't have the proper file archiver? At Lunatics site files are packed into the RAR-format, which can be unpacked by the shareware WinRAR and by the freeware 7-zip file archiver

@ Raistmer

Your V7 mod thread is closed, so I ask here..

In the app_info.xml is this entry:

<avg_ncpus>0.040000</avg_ncpus>
<max_ncpus>0.040000</max_ncpus>

This is for the CPU-support for the GPU?
That the CPU [one Core] support max. 0.04 % the GPU?

If I would like to crunch only on the GPU/s, I could insert there 1.000000 ?
Or I could delete the 6 [all] entries?


BTW.
Which BOINC Version you recommend for your mod?
And which version for stock CUDA?


Thanks!


EDIT:
<ncpus>NUMBER_OF_CORES+2</ncpus> entry is needed for only GPU-crunching?
[Quad-CPU + 2 GPUs]

---

I would take my hat of to you sr, but it's too cold at the minute.

Let me guess your back in Lakewood ,spend so much time in so.cal forget real weather... It's good for a good overclock

I would take my hat of to you sr, but it's too cold at the minute.

Let me guess your back in Lakewood ,spend so much time in so.cal forget real weather... It's good for a good overclock

Sorry satan, just looked at the dog and thought you were high-amps. Doh!

Raistmer.

Nice job on the modifications. After crunching for the first time in 3 weeks it was a very nice thread to find.

I would take my hat of to you sr, but it's too cold at the minute.

Thanks :)
BTW, here too cold too :D