I have a 4-core CPU, AMD A10-6700. When I download some program files, all projects give me a maximum of 4 tasks. Bur, since installing the Lunatics version on my PC, when I allow SETI@home it sends me 10 GPU tasks. But this is no problem, since they run one at the time and don't use much CPU (0.04). So I am now running 2 ATLAS@home tasks, 2 GB each required, a vLHC@home (256 MB) and one SETI@home GPU task. All this on 8 GB RAM. Windows 8.1, BOINC 7.4.12, Virtual Box 4.3.12.
Tullio

---

I have been crunching again for SETI after taking the summer off because of the heat in the room. Now I have been running BOINC on 5 machines and still trying to optimize.
One of my machines is an old E6550 2 core with an NVIDIA GT 430. It is running 2 tasks on the gpu and 2 on the cpu. This gives me almost 100% loading on video card and cpu. After several weeks I decided to try the optimized Lunatics applications.
Now my experience with the Lunatics application was mixed. Before the install, the server sent a variety of cuda32, cuda42 and cuda50 work units. Installing the Lunatics apps, requires me to choose only one of the three. I chose cuda42. The optimized code for the cpu reduced the standard SETI wu to run in almost half the time, but I saw no improvement with the cuda42, in fact based on looking at my credits given I ended up with less per cpu time than before.
The biggest issue I had was how to change from cuda42 to cuda32 optimizer. I could find no instructions for that. Secondly after uninstalling the Lunatics apps, it BOINC errored out all my in progress workunits.
So I am asking three questions:
Is there a way to add all 3 cuda apps, and can this be done by editing the app_info script since you only have an option to check one of the boxes?
Is there a good way to uninstall it without damaging the existing workunits in progress?
Will it only install the cpu optimized app and none of the others, by not checking any of the other checkboxes? I am afraid to try this after my experience so far.

Is there a way to add all 3 cuda apps, and can this be done by editing the app_info script since you only have an option to check one of the boxes?

As far as I understand it, you don't want to do that. The stock app sends the different versions because it doesn't know what hardware you have; once enough units have been crunched it selects the most efficient and only targets that version (the work-units are the same, it's just the targeted version that can vary). By selecting the most appropriate version during Lunatics install you can short-circuit this lengthy and inefficient selection process.

---

Ok one questions answered, two more to go:

Is the uninstall supposed to leave the started work units intact?

Can you install the Lunatics optimizer just for the cpu tasks?

Normally, if you stop processing with stock app and switch to the Lunatics apps or vice versa, you are supposed to set NNT and finish all the tasks up before switching apps. That way you don't lose tasks to errors. However, most of the time the servers will just resend the "lost" tasks and no harm, no foul. Yes, you can run the Lunatics installer and just install the enhanced CPU apps. You don't have to select an enhanced GPU app in the installer and just run the stock GPU app the server would send you. BTW, the stock GPU app is just the Lunatics app renamed.

Keith

---

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)

You dont need to set NNT before using the Installer.
Also it doesn`t make any sense to use optimzed apps for CPU only because all stock GPU apps are optimized apps too.

---

With each crime and every kindness we birth our future.

You dont need to set NNT before using the Installer.
Also it doesn`t make any sense to use optimzed apps for CPU only because all stock GPU apps are optimized apps too.

Quite correct. From the test I did on my I7 Vista machine, Running stock apps VS. Opt apps there was statisticaly no differance at all.
Now that we are on AP v7 might have to try AP stock vs. Opt AP. When we get a reliable source of AP work again of course.

---

[/quote]

Old James

Thank you for the answers. My experiment showed me on me weakest cpu, that the optimized cpu aps were executed twice as fast and on my computer it showed twice the GFLOPs number. But the cuda42 showed a loss in performance. Compare here the anonymous vs stock:

computer:
http://setiathome.berkeley.edu/show_host_detail.php?hostid=7421779

Apps:
http://setiathome.berkeley.edu/host_app_versions.php?hostid=7421779

SETI@home v7 7.00 windows_intelx86 10.04 GFPLOS
SETI@home v7 (anonymous platform, CPU) 19.62 GFLOPS

SETI@home v7 7.00 windows_intelx86 (cuda42) 38.33 GFLOPS
SETI@home v7 (anonymous platform, CPU) 19.62 GFLOPS

Do you think there is anything I did wrong?

How long did you run your trial for - as is explained elsewhere tasks that run quickly compared to "normal" duration tasks) are dominated by one set of calculations that are not (properly) included in the calculation of flops, so they show an unrealistically low score...
To get a representative selection of tasks you either have to dig out one of the benchmark sets, or run each scenario for at least a couple of weeks.

---

Bob Smith
Member of Seti PIPPS (Pluto is a Planet Protest Society)
Somewhere in the (un)known Universe?

Thank you for the answers. My experiment showed me on me weakest cpu, that the optimized cpu aps were executed twice as fast and on my computer it showed twice the GFLOPs number. But the cuda42 showed a loss in performance. Compare here the anonymous vs stock:

computer:
http://setiathome.berkeley.edu/show_host_detail.php?hostid=7421779

Apps:
http://setiathome.berkeley.edu/host_app_versions.php?hostid=7421779

SETI@home v7 7.00 windows_intelx86 10.04 GFPLOS
SETI@home v7 (anonymous platform, CPU) 19.62 GFLOPS

SETI@home v7 7.00 windows_intelx86 (cuda42) 38.33 GFLOPS
SETI@home v7 (anonymous platform, CPU) 19.62 GFLOPS

Do you think there is anything I did wrong?

You posted the SETI@home v7 (anonymous platform, CPU) entry twice, it should be:

SETI@home v7 (anonymous platform, NVIDIA GPU) 27.72 GFLOPS

What the APR finishes up as depends on what Angle Range the work processed was, High AR work is relatively inefficient on the x41zc app, causing the APR to fall.

Claggy

If you have the time (and patience), take a look at a thread I started about a year ago, when I had just switched one of my machines over from stock to Lunatics: What, exactly, does APR measure, and is it meaningful?

I used 6-week intervals in order to get what I felt were fairly stable measurements. Later conversions of 3 more of my machines to Lunatics yielded similar results, with a CPU-only laptop showing significant improvement, while a machine that gets most of its productivity from MB GPU work shows essentially no improvement at all (due to the fact that stock and Lunatics is the same app for MB GPU processing).

Also if you have the patience read my Running stock expierement

I went from luntaics to stock just to see what would happen.

---

[/quote]

Old James

Ok, I ran the Lunatics installer and only picked the enhanced cpu application, thinking that the gpu appd would ran as before. Well after i restarted BOINC, all the gpu apps were not in the task list, and nowhere to be found in the BOINC folder. I let it run for a bit and no standard gpu apps were downloaded..

Question is, did I misunderstand what the optimized installer would do? I am not a programmer, so I don't really understand how the scripting works. With the standard BOINC install there is no app_info.xml. So I have no idea what the default script contains that requests the standard gpu applications.

So what I am asking is help with just running the optimized cpu apps and leave the standard gpu apps running as before. Is this hard to do?

Even though the Lunatics MB GPU apps are actually the same as the stock MB GPU apps, you still need to select one of them in the Lunatics installer. Once an app_info.xml file is in place, the original stock apps go away.

The good news is that you don't need to start all over again. Just re-run the Lunatics installer and select the same CPU app as you did before, along with one of the GPU apps. The installer should simply stop BOINC and restart it again without your intervention, then download the files for the GPU app you selected.

The bad news is that S@h will probably run out of new tasks in the next several hours. Such is life!

Since this is bad timing, I decided to uninstall the lunatics installer right now, then delete the whole BIONC folder and replace it with a backed up version before the install. That seemed to work and I got all my original workunits back including the gpu ones. And no wu were errored out on the server from what I can tell.

Will try this again when more work is available so I can do more lengthy testing.

Can I run the installer on a fresh installation of BOINC before receiving any work?

---

Can I run the installer on a fresh installation of BOINC before receiving any work?

So long at BOINC has been attached to the project everything should be OK.

---

SETI@home classic workunits: 93,865 CPU time: 863,447 hours
Join the [url=http://tinyurl.com/8y46zvu]BP6/VP6 User Group[

Oops, didn't see your answer, thx Hal!

---

Hi,

now that SETI is out of work I thought it's a good idea to finaly install APv7 on my laptop.

I was using here (SEE2 capable CPU) untill now ap_6.01r557_SSE2_331_AVX.exe. As I see, there are no more SSE2 CPU builds?

There is a AVX build called AP7_win_x86_AVX_CPU_r2692.exe, but that's SSE3 if I understand the readme right (there's no SSE info in the filename)... so I should use AP7_win_x86_SSE_CPU_r2703.exe?

This part confuses me a bit:

A 32-bit AVX build is offered for the first time - at the SSE3 level, the installer will choose between AMD and Intel builds automatically (there wasn't space for an extra manual choice).

This "at the SSE3 level", does that refer to the 32-bit AVX build (The first one? I have one installed according to the filename.) or does that refer to the automatic AMD/Intel choise?

---

This "at the SSE3 level", does that refer to the 32-bit AVX build (The first one? I have one installed according to the filename.) or does that refer to the automatic AMD/Intel choise?

No, it will choose between AP7_win_x86_SSE3_CPU_r2692.exe and AP7_win_x86_SSE3_CPU_AMD_r2696.exe

The AP7_win_x86_AVX_CPU_r2692.exe app is ment for AVX capable CPUs.

Claggy