app_config>

<app>
<name>setiathome_v8</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.4</cpu_usage>
</gpu_versions>
</app>

</app_config>

not working

I'm running 1 video card (at 37%) running SETI@home V8 8.00 from the beta test server. I get 1 WU at a time.

Is the beta system limiting downloades or should I look for a config prob on my computer?

Ed F

You need a newer BOINC for that feature. 7.0.42 or later.
Is there a reason you are still on 6.12.34 ? Else I'd recommend upgrading to the latest BOINC client to take advantage of a whole host of new features (and a complete overhaul of the workfetch system...).

---

A person who won't read has no advantage over one who can't read. (Mark Twain)

Hi folks,

There seems to be a problem with OpenCL on Intel HD Graphics 530.
I updated to the latest Intel driver release but this did not help.
After updating I still see tasks with different numbers of signals reported, and I even had one being marked as invalid.
Further research showed that there are other people with similar problems over on Einstein@home.
If there is anyone out there running GPU apps on Intel HD Graphics 530 without problems, please let me know. I am specially interested in the driver version being used.

Tom

Among cuda apps, it seems that the 32s have slightly shorter run times than 42 and 50s, but considerably higher CPU times. Anyone else seeing that trend?

System and application settung dependant, but can be. The main changes between 3.2 and 4.2 involve a complete compiler replacement that tends to favour performance on Kepler onwards, and Cuda 3.2 older technology tends to be a tad lower optimised in the drivers (CPU bits). Not sure where a 630 fits in the scheme of things.

For the time being, choices between OpenCL and Cudas are likely to come down to exactly your own situation and goals (e.g. single instance, fast as possible versus multiple instances low impact preferred). That's a kindof flexibility by having choices I hope to capture. To be able to make configurable within some single future application. But we are talking next generation, and looking what works in different situations, first.

I'm certain once initial v8 is rolled out and any problems nailed flat, that all the applications begin to change with what we learn.

Thanks to Brent for the app_config. I will install that in both of my boxes when I get home.

I will also check the box from my 630 to see if it tells me what flavor it is.

Thanks also to Brent and Raistmer for pointing out where the details are. Does a higher average processing rate mean that app is working faster? Cuda 32 has a significantly higher APR than 42 or 50, but 42 has a much lower turnaround time than 32 and 50. (All on the 630, ignoring my 440 for the moment.) As for those choices mentioned above, on the 630 my choice will be for minimal impact on the CPU.

Now, on the 440. 8.05 opencl had a much higher APR, 77.61, than all the other versions including 8.06 opencl, and there is quite a spread of average turnaround times, from .27 for cuda 32 to 1.02 for 42 (which I might chalk up to a small sample size and other tasks being in the cache at the same time). On this one, I want maximum production.

I realize that in both cases I will have to look at the numbers again a few days after changing the app_config.

P.S. On the 630, Boinc is nagging me to upgrade the driver. It's still 335.28.

---

David
signature sent back to alpha testing

Could any NVidia users - especially those using older, pre-350 drivers - please check their caches and verify that they're starting to receive v8.01 cudaxx tasks?

If you have any of them, please could you run some through quickly, please? It's a final test to verify that nothing has gone wrong with packaging them up before they're transferred to Main.

Thanks.

Got about a dozen on one PC - started pushing them through now, so keep your eyes open on my PCs....

(in fact all four have a few, so I've suspended all GPU work apart from 8.01 CUDA)

Got about a dozen on one PC - started pushing them through now, s keep your eyes open on my PCs....

It's Eric that needs to see them showing up in his statistical monitoring. Providing they run and report successful completion, we're not expecting any surprises from this quick run - the best you can expect is a slightly lower inconclusive rate, which will be hard to detect over just a few hours.

Could any NVidia users - especially those using older, pre-350 drivers - please check their caches and verify that they're starting to receive v8.01 cudaxx tasks?

Yes. 355.82 here, though

If you have any of them, please could you run some through quickly, please?

All OpenCLs suspended for now ... crunching 37 Cudaxx's. Should take about 90 minutes.

---

If I can help out by testing something, please let me know.
Available hardware and software is listed in my profile here.

Thanks Richard.

Passing comment, the first few CUDA v8.01 tasks have run through OK, but I notice that they are more "CPU friendly" that OpenCL tasks on the same PC. OpenCL were using about 1 core per "GPU" task, compared to something less than 0.5CPU per GPU task. Execution times are "comparable".

Ahh, I was beginning to wonder if anything beyond 41zh was going to show up on Beta. I hadn't bothered to run anything since Friday, but 3 boxes have now been re-fired, all running stock. Drivers are 314.22, 335.23, and 337.88. GPUs include 550ti, 630, 660, 670, and 780. Shouldn't take long for results to start coming back.

Ahh, I was beginning to wonder if anything beyond 41zh was going to show up on Beta. I hadn't bothered to run anything since Friday, but 3 boxes have now been re-fired, all running stock. Drivers are 314.22, 335.23, and 337.88. GPUs include 550ti, 630, 660, 670, and 780. Shouldn't take long for results to start coming back.

Yes, v8.01 is a quick burst of x41zi running in stock mode. Thanks - that will help Eric keep within his target of a transfer 'today' (US time).

Thanks Richard.

Passing comment, the first few CUDA v8.01 tasks have run through OK, but I notice that they are more "CPU friendly" that OpenCL tasks on the same PC. OpenCL were using about 1 core per "GPU" task, compared to something less than 0.5CPU per GPU task. Execution times are "comparable".

If you check with a real-time monitoring tool like Task Manager or Process Explorer, you'll probably find it's below 10%.

BOINC Manager's display isn't real-time.

No problems here with v8.01 cuda42/cuda50 (several tasks completed and validated for both). Running driver 361.43 on GTX-750.

Tom

You need a newer BOINC for that feature. 7.0.42 or later.
Is there a reason you are still on 6.12.34 ? Else I'd recommend upgrading to the latest BOINC client to take advantage of a whole host of new features (and a complete overhaul of the workfetch system...).

yes ... I haven't gotten my startup to work yet with boinc v7...

rem
start /affinity 80 Z:\BOINC_test_programs\boinc.exe --gui_rpc_port 31416 --dir Z:\BOINC_test_data_1 --detach
rem
start /affinity 80 Z:\BOINC_test_programs\boinc.exe --gui_rpc_port 31417 --dir Z:\BOINC_test_data_2 --allow_multiple_clients --detach
rem
start /affinity FF Z:\BOINC_test_programs\boinc.exe --gui_rpc_port 31422 --dir Y:\MC_at_Home --allow_multiple_clients --detach
REM
start Z:\BOINC_test_programs\boincmgr.exe /s
exit

maybe later??

Ed F

p.s. I'll get out of everyones way here ... sorry

---

Could any NVidia users - especially those using older, pre-350 drivers - please check their caches and verify that they're starting to receive v8.01 cudaxx tasks?

If you have any of them, please could you run some through quickly, please? It's a final test to verify that nothing has gone wrong with packaging them up before they're transferred to Main.

Thanks.

My 630 with 335.28 is showing two 42s, a 32 and a 50 in progress. I'll be going home in a few minutes and try to rush them through then.

My 440 with 355.98 has three each of 32 and 42.

---

David
signature sent back to alpha testing

Yeah Cuda and VLARS don't really like to play together. On the other hand, VLARS and OpenCL seem to get along well.

It would be an option to go back to the "no VLAR for Cuda" policy. NVIDIA hosts would still get VLARs for the OpenCL app.

It turns out this is currently not possible under BOINC. The scheduler first chooses the fastest app_version for a GPU, then discards any work that isn't feasible. So for a host that thinks the OpenCL app is fastest, it would choose the OpenCL app and then discard any no-VLAR results which would probably result in it getting no work much of the time. It would never use the CUDA app. There's no obvious way to put the apps on a level playing field without rewriting the scheduler. So we're stuck with CUDA alone or OpenCL alone on any given platform.

---

So we're stuck with CUDA alone or OpenCL alone on any given platform.

I had wondered that question a little while back after your statement about the GreenBank data looking like VLAR work.

Will we get VLAR work for OpenCl on main? If so, I plan on dedicating 2 machines to the OpenCl format and my other machines to Cuda as needed.

Might not be something for everyone but for those with capable computers and desire to do so, they could also weigh the decision.

app_config>

<app>
<name>setiathome_v8</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.4</cpu_usage>
</gpu_versions>
</app>

</app_config>

not working

I'm running 1 video card (at 37%) running SETI@home V8 8.00 from the beta test server. I get 1 WU at a time.

Is the beta system limiting downloades or should I look for a config prob on my computer?

Ed F

You need a newer BOINC for that feature. 7.0.42 or later.
Is there a reason you are still on 6.12.34 ? Else I'd recommend upgrading to the latest BOINC client to take advantage of a whole host of new features (and a complete overhaul of the workfetch system...).

It's not working for me either. I installed the app_config and told Boinc to read config files. It found the file, but continued to do just one Beta GPU task. Then I stopped and restarted Boinc and it's still only doing one.

It's running 7.2.42. This is on my 440.

[edit]
Same thing happening on the 630, same Boinc version.

---

David
signature sent back to alpha testing

I will also check the box from my 630 to see if it tells me what flavor it is.

Very unusually for me, I don't seem to have the box. Oh well.

---

David
signature sent back to alpha testing

I will also check the box from my 630 to see if it tells me what flavor it is.

Very unusually for me, I don't seem to have the box. Oh well.

If you have GPU-Z, it should show you which one it is.

app_config>

<app>
<name>setiathome_v8</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>0.4</cpu_usage>
</gpu_versions>
</app>

</app_config>

not working

I'm running 1 video card (at 37%) running SETI@home V8 8.00 from the beta test server. I get 1 WU at a time.

Is the beta system limiting downloades or should I look for a config prob on my computer?

Ed F

You need a newer BOINC for that feature. 7.0.42 or later.
Is there a reason you are still on 6.12.34 ? Else I'd recommend upgrading to the latest BOINC client to take advantage of a whole host of new features (and a complete overhaul of the workfetch system...).

It's not working for me either. I installed the app_config and told Boinc to read config files. It found the file, but continued to do just one Beta GPU task. Then I stopped and restarted Boinc and it's still only doing one.

It's running 7.2.42. This is on my 440.

[edit]
Same thing happening on the 630, same Boinc version.

The app_config that's quoted has an incomplete opening tag. It should be <app_config>. Perhaps that's your problem?