OK, I will continue to let run the APv6 r1797 CPU app.

At which file size of the related .wisdom file I could see the app found the best/fastest settings? -> r1797 CPU app reached max performance.

Currently 3 kb.

[EDIT#2: To now, AFAIK, two half (finished) and two complete WUs calculated with r1797 CPU app.]

Thanks.

[EDIT: BTW, no problems so far on my system. All is running smoothly.]

* Best regards! :-) * Philip J. Fry, team seti.international founder. * Optimize your PC for higher RAC. * SETI@home needs your help. *

---

No, we still leave the default to crunch 1 WU at a time, download this file, unzip it and place it into the C:\ProgramData\BOINC\projects\setiathome.berkeley.edu folder, then go to the advanced menu in boinc and select read config files.

Sweet. It's running two at the same time now. I ran three at a time before, is there a way to achieve this still?

With the new files being processed, it is recommended to only do 2 at a time as they take more processing power.

Juat wanted to confirm, but previously I was running 4 units at a time on my i3 2100 (2 cores with hyperthreading). With v7, which I have noticed takes an extra 45~ min to complete, should I limit to processing 2 at a time instead? I didn't catch whether your discussion was about GPU or CPU

---

I have question not shore if you have already been asked but anyway

Is it ok to pass this version on to others ?

Has the copy write issue been solved ?

You can freely redistribute as long as you keeping original content intact.

Thanks Raistmer and thanks again to the team for the work you have done

---

No, we still leave the default to crunch 1 WU at a time, download this file, unzip it and place it into the C:\ProgramData\BOINC\projects\setiathome.berkeley.edu folder, then go to the advanced menu in boinc and select read config files.

Sweet. It's running two at the same time now. I ran three at a time before, is there a way to achieve this still?

With the new files being processed, it is recommended to only do 2 at a time as they take more processing power.

Juat wanted to confirm, but previously I was running 4 units at a time on my i3 2100 (2 cores with hyperthreading). With v7, which I have noticed takes an extra 45~ min to complete, should I limit to processing 2 at a time instead? I didn't catch whether your discussion was about GPU or CPU

They're talking about GPU's. As long as your temps are under control, you can run 4 CPU units at a time (2XHT)

---

AKV8B_SSE3_AMD

Is there a new app released yet.
I installed the installer 41 and the sse2 cashed this machine while running x41zc for multibeam.
It was running fine but the nvidia driver 320.18 failed this BSOD my rig

i am running now x41zc cuda 42 in which I have identified as cuda_fermi and not cuda42 ver 7
running well.
I am still using the old AKV8B_SSE3_AMD app to run 603 and I am still getting them but I do not know how much longer they will send.

need an sse3 app for cpu multibeam

Amd Phenom 2 1100T @ 4.00 ghz win 7 sp1 x64
Nvidia 460 running @ 850 core

You don't actually need an SSE3 app - the SSE2 app runs at the same speed on AMD hardware (we tested one, and decided to take it out to save space - the download has doubled in size since last year as it is).

Joe Segur has posted some technical details of the difference between SSE2 and SSE3 somewhere, and an explanation of why they don't help SETI - I'll try and find a link.

Edit - it wasn't Joe, it was Jason, a few pages down this very thread. Message 1376357

Thanks to all at Lunatics involved in one way or another in THE installer, plus for throwing in the link to the app_config.xml file.

G4-PC is now chewing through wus like they were going out of fashion (compared to standard apps *grin*) and with the 2 GPU wus at a time, winter will vanish... (hopefully)

From Australia, thanks folks.

A big THANX to all the Seti gurus!:)

I've been running the Lunatics apps for a while and, when I ran out of WUs over the weekend, I just waited it out until v0.41 arrived on Monday. The new installation went fine and it's been crunching well since then. A big Thank You to Lunatics and everyone involved.

One quick question: Can the NV 4200M on this laptop handle 2 units at a time?

---

While trying to install v7 on this computer one question came to my mind... no more SSE optimized apps?

---

While trying to install v7 on this computer one question came to my mind... no more SSE optimized apps?

No, i`m afraid.
SSE only has no benefit vs stock.

---

With each crime and every kindness we birth our future.

While trying to install v7 on this computer one question came to my mind... no more SSE optimized apps?

No, i`m afraid.
SSE only has no benefit vs stock.

Largely because the stock app itself has received some extra optimisation this time round.

OK, than I guess it's time to upgrade this machine with the SSE2 capable hardware I have laying around since a year or two...

---

Quick question:

Installed the Lunatics 0.41 on all my computers.

On the Arkayn site; are they optimized app's or the same as the 0.41 installer but manual installation.?

bj

---

...
OK, I will continue to let run the APv6 r1797 CPU app.

At which file size of the related .wisdom file I could see the app found the best/fastest settings? -> r1797 CPU app reached max performance.

Currently 3 kb.

[EDIT#2: To now, AFAIK, two half (finished) and two complete WUs calculated with r1797 CPU app.]

Thanks.

[EDIT: BTW, no problems so far on my system. All is running smoothly.]
...

Actually, within 2 minutes after starting the first r1797 Astropulse task the contents of the .wisdom file were written and haven't really changed since. Each subsequent run will have loaded the file, seen that it contained info suitable for all the plans needed, and rewritten the file. That rewrite may change the order the codelets are listed, which makes a naive comparison think the files are different.

The time saved by using the persistent wisdom rather than doing full planning for each new task is a constant on the order of a minute or so for modern CPUs.

There's always some uncertainty whether the best/fastest plans have been found and the associated wisdom saved, partly because the best plan at any moment depends on whatever else is active in the multitasking environment. FFTW does quick tests of possible plans, starting with those most likely to perform well and then others which could be good too. But if the testing of the plan which would be best long-term happens to be slowed by a system background task or another application then the saved wisdom isn't fully optimal. Fortunately the second-best plan is likely to be so close in performance that there's no significant loss in speed.

Nevertheless, for someone willing to put in extra effort to achieve best Astropulse CPU performance I'll suggest this procedure:

1. Select all tasks in BOINC Manager (or other Manager) and suspend them.
2. Shut BOINC down.
3. Rename the r1797_[stripped CPU name].wisdom file to preserve it but force creation of a new one.
4. Restart BOINC.
5. Unsuspend one task for AP6_win_x86_SSE_CPU_r1797.exe
6. Wait until it has created the new wisdom file, then unsuspend all the other tasks.

If your preferences are set to NOT "Leave tasks in memory while suspended" steps 2 and 4 are not needed. That's a global computing preference where the web setting may be overridden by the local preferences, so using those steps is recommended unless you're very sure they aren't needed.

The purpose is to do the planning with least interference from other crunching, and it would make sense to shut down all other applications too during the procedure. Even this doesn't guarantee the best wisdom, but in testing I've done it gives the best chance of getting optimal wisdom.

Testing whether that has improved the wisdom is a matter of running enough tasks to get a decent comparison, paying attention to the amount of blanking, etc.

                                                                  Joe

"

I've a question about the available CUDA Versions.
On one of my machine I could not change the nvidia driver version.
There ist only the 257.59 available with CUDA 3.1 on a Quadro 4000 (Fermi).
The new v7 support CUDA > 3.2 or 2.3 but with the warning not to try this on Kepler or Fermi cards.

Is there any way or chance to use this card for v7 with cuda 3.1 and Fermi?

The CPU (Xeon L5640) works most on einstein and the gpu should do the seti stuff, this worked fine till v7 came up.

greetz from germany

---

Regards from nothern Germany
Roman

SETI@home classic workunits 207,059
SETI@home classic CPU time 1,251,095 hours

I've a question about the available CUDA Versions.
On one of my machine I could not change the nvidia driver version.
There ist only the 257.59 available with CUDA 3.1 on a Quadro 4000 (Fermi).
The new v7 support CUDA > 3.2 or 2.3 but with the warning not to try this on Kepler or Fermi cards.

Is there any way or chance to use this card for v7 with cuda 3.1 and Fermi?

The CPU (Xeon L5640) works most on einstein and the gpu should do the seti stuff, this worked fine till v7 came up.

greetz from germany

I'm surprised you say 'only 257.59' available for Quadro 4000.

I found every recognised WHQL driver version listed, up to and including v320.00:

http://www.nvidia.de/object/quadro-tesla-grid-win8-win7-winvista-64bit-320.00-whql-driver-de.html

Your only listed NVidia host (6428818) describes the card as 'NVIDIA D12P1-30', and the driver as v267.24 already.

Nevertheless, for someone willing to put in extra effort to achieve best Astropulse CPU performance I'll suggest this procedure:

1. Select all tasks in BOINC Manager (or other Manager) and suspend them.
2. Shut BOINC down.
3. Rename the r1797_[stripped CPU name].wisdom file to preserve it but force creation of a new one.
4. Restart BOINC.
5. Unsuspend one task for AP6_win_x86_SSE_CPU_r1797.exe
6. Wait until it has created the new wisdom file, then unsuspend all the other tasks.

1. For some apps this have to be done twice i.e.:
- after a few minutes suspend the first running task
- wait a few seconds and unsuspend another task of the same kind
- wait a few minutes before "unsuspend all the other tasks"

I tested with AKv8c_Bb_r1846_winx86_SSE2x.exe - the #1 task created the new wisdom file with size 14 KB, the #2 task increased the size to 29 KB
(AP6_win_x86_SSE2_OpenCL_ATI_r1843.exe seems to (maybe) need only one test but it doesn't hurt to run a second task for a while before unsuspend all tasks)

Each subsequent run will have loaded the file, seen that it contained info suitable for all the plans needed, and rewritten the file. That rewrite may change the order the codelets are listed, which makes a naive comparison think the files are different.

I can confirm that #3 task do not change the values in the wisdom file (I made copy of 'wisdom' after each run/task).

The file seems changed but if you sort it alphabetically you will see it is the same for #2 and #3 run tasks.
(I used in CMD prompt: sort <r1846_AMDAthlontmIIX3455Processor.wisdom >>wisdomSorted)


2. I think that (Intel) SpeedStep and (AMD) Cool'n'Quiet / PowerNow! technologies can lead to unpredictable results (skew the tests) during this 'wisdom' files generation.

So disable them (in the BIOS) before the procedure.

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

Attention those failing to getting the Brook app to run. I've posted a revised aistub here. Please give it a try.
We can't update the installer until we've consulted with Eric on this matter (there's a build in beta).

---

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

Nevertheless, for someone willing to put in extra effort to achieve best Astropulse CPU performance I'll suggest this procedure:

1. Select all tasks in BOINC Manager (or other Manager) and suspend them.
2. Shut BOINC down.
3. Rename the r1797_[stripped CPU name].wisdom file to preserve it but force creation of a new one.
4. Restart BOINC.
5. Unsuspend one task for AP6_win_x86_SSE_CPU_r1797.exe
6. Wait until it has created the new wisdom file, then unsuspend all the other tasks.

1. For some apps this have to be done twice i.e.:
- after a few minutes suspend the first running task
- wait a few seconds and unsuspend another task of the same kind
- wait a few minutes before "unsuspend all the other tasks"

I tested with AKv8c_Bb_r1846_winx86_SSE2x.exe - the #1 task created the new wisdom file with size 14 KB, the #2 task increased the size to 29 KB
(AP6_win_x86_SSE2_OpenCL_ATI_r1843.exe seems to (maybe) need only one test but it doesn't hurt to run a second task for a while before unsuspend all tasks)

Each subsequent run will have loaded the file, seen that it contained info suitable for all the plans needed, and rewritten the file. That rewrite may change the order the codelets are listed, which makes a naive comparison think the files are different.

I can confirm that #3 task do not change the values in the wisdom file (I made copy of 'wisdom' after each run/task).

The file seems changed but if you sort it alphabetically you will see it is the same for #2 and #3 run tasks.
(I used in CMD prompt: sort <r1846_AMDAthlontmIIX3455Processor.wisdom >>wisdomSorted)


2. I think that (Intel) SpeedStep and (AMD) Cool'n'Quiet / PowerNow! technologies can lead to unpredictable results (skew the tests) during this 'wisdom' files generation.

So disable them (in the BIOS) before the procedure.

It's Josef's CPU applications which do the double-wisdom jig. Raistmer's ATI applications do all they're going to do the first time.

(Reasoning: the full wisdom generation for the CPU app takes some time, on what may be inherently a slow processor by modern/GPU standards. And until it completes, the task will display no progress in BOINC Manager.

It is likely that the user will be around and anxious to check that it's working, immediately after installation. So we don't want a big startup delay the first time. But by the time the second task starts, it's likely that the user will have lost interest and wandered off to make a cup of tea. So the app can roll up its sleeves and really get to work on that wisdom, without disturbing anybody.

At least that's how I read it.)

:) actually there is some tech reasoning too, not only sociological one :)
first start will be most probably for all tasks. Later finish times will be scattered so delay in one task allows to avoid situation where all 4 or more CPu tasks on multicore host do same FFTW planning, i.e. quite artifical CPU load that will never meet after. Planning on even loaded CPU but loaded with usual work could be more adequate.

---

SETI apps news
We're not gonna fight them. We're gonna transcend them.