In 'Lunatics ReadMe':

NVidia GPU (CUDA) apps - MB x41zc in CUDA variants 2.3, 3.2, 4.2 and 5.0

rule of thumb for choosing variant:

2.3: for Pre-Fermis ONLY - do NOT use on Fermi/Kepler
3.2: for mixed systems
4.2: for Fermis
5.0: for Keplers

What does this mean at cuda32 app?
The cuda32 app run on Pre-Fermi, Fermi and Kepler cards?

Currently only at arkayn's site for D/L?
A teammate have probs with D/L.
It looks like arkayn's server don't support 'resume'.
He D/L ~ 50 % of the Installer and then the D/L aborts always (56k).

Thanks.


BTW.
I would like to say 'Thank you!' to all Lunatics crew members, and to all others around which help them, for all their hard work in their free time.
My appreciation, respect and recognition!


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

---

In 'Lunatics ReadMe':

NVidia GPU (CUDA) apps - MB x41zc in CUDA variants 2.3, 3.2, 4.2 and 5.0

rule of thumb for choosing variant:

2.3: for Pre-Fermis ONLY - do NOT use on Fermi/Kepler
3.2: for mixed systems
4.2: for Fermis
5.0: for Keplers

What does this mean at cuda32 app?
The cuda32 app run on Pre-Fermi, Fermi and Kepler cards?

Yes, the cuda32 is the nearest there is to a 'universal' cuda app.

It will run, reasonably well, on all those GPUs. Other apps may be better matched and run more efficiently on your particular card. Use your skill and experience, or wait for the comparative reports that I'm sure will soon start to appear on the boards.

Heed the minimum driver versions for each app, and do not select cuda23 if you have a Fermi or Kepler in your system. Apart from that, the choice is yours.

I will host the installer soon as well.

---

With each crime and every kindness we birth our future.

No probs, I'll install the SSE2 app manually. Have a nice dinner.

Edit: AKv8c_Bb_r1846_winx86_SSE2x.exe runs just fine on the Celeron M 370

Sten, I've had another idea. Could you download and run CPUFeatures-TestApp.zip from http://forums.winamp.com/showthread.php?t=355951 - take the updated one from the last post in the thread.

Does that find the SSE2 support that CPU-Z does, indeed, report? That test app uses the same detection code our installer does now - if the test works, and the installer doesn't, blame my programming. Otherwise, we may have to report it back to LoRd_MuldeR.

Yes, it's the known issue where the x264 feature detection code he used deliberately turns off the SSE2 flag for certain processors. That's because for the integer SIMD used by the x264 codec, MMX code is faster on those CPUs.

Laptops with Pentium-M, Core Solo, or Core Duo (and Celeron M processors with the same Family/Model) will all be affected. There are only a few hundred still active, and the workaround of manual installation will be needed by the owners who want SSE2x. Sten got it right, and though I'm sure he'd have preferred a mention of the issue in the readme I feel that document is already as detailed as practical.

                                                                  Joe

Thanks Joe for explaining the issue. It's no problem really, I still have some grey matter left, and can come up with at least a few solutions of my own.

However, running an old Celeron M with an RAC of max 400 (on AP) may actually lead some people to wonder, if I do have any grey matter left. :-)

It would probably be technically possible (still speaking from the pub, without consulting full documentation) to add an 'override' control, allowing 'at your own risk' unlocked access to the non-detected CPU types. Let's take that back into the development team, and wait to see

a) what the demand is
b) how much else emerges to prompt a 'mark 2' refresh release.

guys and gals - the doctor is... OUT (they've called time).

I'll peek in from home, but that's basically it from me for tonight. Others will be around to help overnight, and we'll do it all again tomorrow, I'm sure. Sleep tight, everyone.

Installed and working great. Once again, hats off to Lunatics!

One question regarding the CUDA versions. I'm running a 9800 series GeForce so I selected 3.2. What would actually happen if I were to select a higher version of CUDA? Would it just not work, or would it not work as well as 3.2? If I stick with 3.2, will that prevent me from getting certain kinds of work from Seti?

Thanks!

---

You will be assimilated...bunghole!

Installed and working great. Once again, hats off to Lunatics!

One question regarding the CUDA versions. I'm running a 9800 series GeForce so I selected 3.2. What would actually happen if I were to select a higher version of CUDA? Would it just not work, or would it not work as well as 3.2? If I stick with 3.2, will that prevent me from getting certain kinds of work from Seti?

Thanks!

Your work would probably error out.

---

With each crime and every kindness we birth our future.

The Lunatics installer is now also available on my website.

http://mikesworldnet.de/produkte.html

---

With each crime and every kindness we birth our future.

Installed and working great. Once again, hats off to Lunatics!

One question regarding the CUDA versions. I'm running a 9800 series GeForce so I selected 3.2. What would actually happen if I were to select a higher version of CUDA? Would it just not work, or would it not work as well as 3.2? If I stick with 3.2, will that prevent me from getting certain kinds of work from Seti?

Thanks!

Your work would probably error out.

Actually, not. I have two 9800GT running under Windows XP, and a 9800GTX+ running under Windows 7.

Although 3.2 would probably be the most efficient, and higher versions would run slower, there is no reason - provided you heed the stated minimum driver version - for work to error out.

Installed and working great. Once again, hats off to Lunatics!

One question regarding the CUDA versions. I'm running a 9800 series GeForce so I selected 3.2. What would actually happen if I were to select a higher version of CUDA? Would it just not work, or would it not work as well as 3.2? If I stick with 3.2, will that prevent me from getting certain kinds of work from Seti?

Thanks!

They would work, it just wouldn't be as fast as the Cuda32 app, My testing on my 9800GTX+ showed a slowdown on Cuda32 and Cuda42 apps on drivers later than 301.xx, the Cuda5 app was even slower,
there was no slowdown on the Cuda23 app, so since you're running 311.06, you may want to try the Cuda23 app instead.

Claggy

I am a bit confused today.

Therefore I guess the following may receive a couple of replies about this.

So we are supposed to still be having the Lunatics Installer v0.41 around.

They were not available the last time I checked. In fact I suspected that all of this really was about getting it into one complete package which could be run (possibly Boinc Manager).

For my computer or computers here, I am now happy to receive Seti@home v7.00 tasks for the CPU as well as Seti@home 7.00 (cuda42) / (cuda50) tasks for the GPU.

For nVidia based cards, these GPU-applications supersede the earlier Seti@home Enhanced 6.10 tasks (cuda_fermi).

Running Windows 7 Ultimate I have upgraded from Boinc Manager 7.0.56 (x64) to the same Boinc Manager 7.0.64 less than an hour ago.

These optimized applications are now supposed to be included in the Boinc Manager as an add-on package which helps with the optimization of running given tasks as they are being computed.

This makes autocorrelation possible, among other things.

So what is the point of now reverting back to the "other" optimized applications once again (given whether these are free or available or not)?

Maybe I am getting this a little wrong and thinking that all types of graphics cards (AMD / ATI) may support or include these optimized applications in the applications which are being downloaded for subsequent running of given tasks by the Boinc Manager?

Or what else may be the explanation for this?

Except for that, I am of course very pleased to have access to these optimized applications by means of what I am now being able to currently be running.

Thank you very much Lunatics great work ;)

---

So what is the point of now reverting back to the "other" optimized applications once again (given or whether these are free or available or not)?

First of all you get an optimized app for CPU as well which is not provided by the project.
On the other hand its also for those running under Boinc 6.
I`m sure there are more reasons to use the installer.

---

With each crime and every kindness we birth our future.

Installed and working great. Once again, hats off to Lunatics!

One question regarding the CUDA versions. I'm running a 9800 series GeForce so I selected 3.2. What would actually happen if I were to select a higher version of CUDA? Would it just not work, or would it not work as well as 3.2? If I stick with 3.2, will that prevent me from getting certain kinds of work from Seti?

Thanks!

Your work would probably error out.

As long as you meet the minimum installed driver requirement for the Cuda version, it should run fine, though Cuda 2.3 or 3.2 would likely perform best on a 9800

[Edit:] lol, missed a bunch of posts devising that, more coffee needed...
JAson

---

"Living by the wisdom of computer science doesn't sound so bad after all. And unlike most advice, it's backed up by proofs." -- Algorithms to live by: The computer science of human decisions.

Installed and working great. Once again, hats off to Lunatics!

One question regarding the CUDA versions. I'm running a 9800 series GeForce so I selected 3.2. What would actually happen if I were to select a higher version of CUDA? Would it just not work, or would it not work as well as 3.2? If I stick with 3.2, will that prevent me from getting certain kinds of work from Seti?

Thanks!

They would work, it just wouldn't be as fast as the Cuda32 app, My testing on my 9800GTX+ showed a slowdown on Cuda32 and Cuda42 apps on drivers later than 301.xx, the Cuda5 app was even slower,
there was no slowdown on the Cuda23 app, so since you're running 311.06, you may want to try the Cuda23 app instead.

Claggy

So 2.3 is actually faster than 3.2 for a 9800 series card?

---

You will be assimilated...bunghole!

Just installed and it went smooth as silk. A big thanks to everyone concerned !!

Now the dumb question...

Is there any way, via the app_info file or similar, that I can auto-select the number of CPU's in use depending on whether the machine is crunching MB or AP ?

i.e. So that when crunching MB 100% CPU's is selected but when crunching AP's a core is freed up for every AP task running on the GPU's ?

I think you'll get what I mean.

T.A.

Just installed and it went smooth as silk. A big thanks to everyone concerned !!

Now the dumb question...

Is there any way, via the app_info file or similar, that I can auto-select the number of CPU's in use depending on whether the machine is crunching MB or AP ?

i.e. So that when crunching MB 100% CPU's is selected but when crunching AP's a core is freed up for every AP task running on the GPU's ?

I think you'll get what I mean.

T.A.

For the app_info, you would want to change the average ncpu to 1 and raise the max ncpu to the maximum amount of CPU's that you want to dedicate.

---

Arkayn beat me to it.

---

With each crime and every kindness we birth our future.

Just installed and it went smooth as silk. A big thanks to everyone concerned !!

Now the dumb question...

Is there any way, via the app_info file or similar, that I can auto-select the number of CPU's in use depending on whether the machine is crunching MB or AP ?

i.e. So that when crunching MB 100% CPU's is selected but when crunching AP's a core is freed up for every AP task running on the GPU's ?

I think you'll get what I mean.

T.A.

Change the following values for the OpenCL AP app from:

<avg_ncpus>0.04</avg_ncpus>
<max_ncpus>0.2</max_ncpus>

to:

<avg_ncpus>1.0</avg_ncpus>
<max_ncpus>1.0</max_ncpus>

Now every time an OpenCL AP app starts, one less CPU Wu will be run.

Claggy

Richard that make sense.

It was started under OPENCL_NV rev 1316, but the error happened when it restarted with the new (installer supplied) r1843. I think Raistmer has warned in the past about changing AP OpenCL application versions mid-task, though I don't think he gave us a warning about this particular upgrade path.

---