Ugh, that one again. Tweeted Eric & Marius a link to the other thread, just because they happened to be on my twitter feeds.

---

"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.

These settings suck on my 980Ti !!!!


Brent Norman Wrote:

Found this in the distribution info:

High end cards x8x 780TI Titan / Titan Z
-sbs 256 -spike_fft_thresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 64 -oclfft_tune_cw 64 (*requires testing)


I'm gonna give it a try.

A little more details please.

---

With each crime and every kindness we birth our future.

As the title says. Beta test only, try at your own risk.

I think we should dig out the old "Advanced Users only" tag for this one - I'm not sure whether it will fly or not.


I have set up the app_info segment for NVidia cards to handle the following version/plan_class combinations:

8.12 (opencl_nvidia_sah)
8.12 (opencl_nvidia_SoG)
8.00 (opencl_nvidia_sah)
8.00 (opencl_nvidia_SoG)
8.00 (cuda50)
8.00 (cuda42)
8.00 (cuda32)
8.00 (cuda23)

If you have been running any of:
Stock download
Lunatics v0.44
Raistmer's cloud download

this installer is designed to preserve your downloaded tasks and pick up where your previous application left off. That includes when switching between CUDA and SoG, in either direction. But it's possible I may have missed a special case - if your NV MB cache disappears, my apologies, but please tell me about it - with as much detail as you can about what application(s) and settings you were running before.

Any pre-existing application tuning should be preserved if you placed it in the 'mb_cmdline_win_x86_SSE3_OpenCL_NV.txt' file, but if won't be if you placed it directly into app_info.xml - whichever, you'd be wise to make backups.

If you have any observations about the installer, feel free to make them in this thread. If you have any observations about the SoG application, please use the thread which Raistmer is monitoring.

And with no further ado, download from here.

. . First comment, boy does my timing suck. Being a strict believer in Murphy's Law I took the cautious approach. Having just added a new GTX950 to my rig on Saturday I was still in the settling in running stock apps phase, so I let the cache run dry before implementing Lunatics 0.45 Beta. Install went perfectly as usual. D/l'ed about 15 to 20 WUs and began crunching away happily. As soon as I walked away from the machine Seti went down and after a few hours or productive crunching my rig remained in a WU drought for the next 24 hours, hard to evaluate the changes when that happens :(


. . Machine is i5-6400 now with a GTX950, I selected AVX for CPU tasks and SoG (I really like the translation of Same Old Grind for that) for the GPU. All seems AOK so far. Oh yes Windows 10 core 64 for OS.

. . I was going to add some details of the effects of SoG but I will abide by your directions and find the other thread to post those.

. . Stephen


. . BTW which thread is Raistmer monitoring??

These settings suck on my 980Ti !!!!


Brent Norman Wrote:

Found this in the distribution info:

High end cards x8x 780TI Titan / Titan Z
-sbs 256 -spike_fft_thresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 64 -oclfft_tune_cw 64 (*requires testing)


I'm gonna give it a try.

A little more details please.

This seems to work well on my 750Ti, but I added -use_sleep

I have had to add -period_iterations_num 250 to get my 980Ti to reduce lag to manageable amounts.


These settings suck on my 980Ti !!!!

High end cards x8x 780TI Titan / Titan Z
-sbs 256 -spike_fft_thresh 4096 -tune 1 64 1 4 -oclfft_tune_gr 256 -oclfft_tune_lr 16 -oclfft_tune_wg 256 -oclfft_tune_ls 512 -oclfft_tune_bn 64 -oclfft_tune_cw 64 (*requires testing)


A little more details please.[i]

---

I have tinkered with this software for the last few weeks and no matter what you do VLAR always introduces lag somwehere unless you disable SETI whilst doing other things on the PC. Put simply the software we are being issued with sucks at handling VLAR WU's. We need VLAR filtered to cpu's or all new software that handles them smoothly.

I know folks here will not agree with this, but let me put it out there right now they are plain and simply in denial.

---

I have tinkered with this software for the last few weeks and no matter what you do VLAR always introduces lag somwehere unless you disable SETI whilst doing other things on the PC. Put simply the software we are being issued with sucks at handling VLAR WU's. We need VLAR filtered to cpu's or all new software that handles them smoothly.

I know folks here will not agree with this, but let me put it out there right now they are plain and simply in denial.

Can you please post one link with those values in use.

---

With each crime and every kindness we birth our future.

Hey Zoom,

The -use_sleep so only be used if you are seeing significant lag on your computers because the SoG is utilizing too much of the CPU cores.

This of course will also vary depending on your equipment, ie.. speed of processing the data.

So a system that is totally devoted to crunching, screen lag and unresponsiveness isn't a problem so -use_sleep isn't necessary.

But if the computer is going to be used for other things, then the lagging will be a problem for doing other things on the computer.

As such, the -use_sleep decreases the CPU usage and prolongs the time to finish, but frees up the CPU for doing non-seti work.

I'd rather abort any that I find on My 580 gpu.

Better not to crunch at all than to abort.

---

Grant
Darwin NT

I'd rather abort any that I find on My 580 gpu.

Better not to crunch at all than to abort.

Well I aborted all the cuda guppi's/vlar's, it's done...

So sue Me.

The only effect of the Guppies on my systems is they take forever to finish & don't pay much credit.

So it would be better just to block all those that abort from using the project. They have the option of not using the GPU in their settings.


EDIT-
Just in case you didn't get around to reading the initial post in this thread, i'll repost the first couple of lines.

"As the title says. Beta test only, try at your own risk.

I think we should dig out the old Advanced Users only tag for this one"

---

Grant
Darwin NT

I'd rather abort any that I find on My 580 gpu.

Better not to crunch at all than to abort.

Well I aborted all the cuda guppi's/vlar's, it's done...

So sue Me.

The only effect of the Guppies on my systems is they take forever to finish & don't pay much credit.

So it would be better just to block all those that abort from using the project. They have the option of not using the GPU in their settings.

Also, not crunching while the PC is in use is a forgotten option. Interesting fact for another discussion, My Windows, Linux and Mac machines all crunch Guppies just fine while the machine is in general use. For the first two cases those are old Core2duo hosts prone to CPU overcommit.

---

"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.

Interesting fact for another discussion, My Windows, Linux and Mac machines all crunch Guppies just fine while the machine is in general use. For the first two cases those are old Core2duo hosts prone to CPU overcommit.

Likewise.
With 2*GTX750Tis my C2D had no issues crunching Guppies while in general use. It's only when I made it a dedicated cruncher & played with the settings to boost throughput that it became sluggish to input.

---

Grant
Darwin NT

Fine then more people will leave, vlars it was agreed would not be on the gpu again We were told, promised really, sounds like that was an empty(hollow) promise...

That was MB data & that is still the case. There are still no MB VLARs going to GPUs.

---

Grant
Darwin NT

I don't suffer from lag here, I have an i7 3820 and the coolest 580(temperature wise) gpu, a PNY LC 580.

I'm not a programmer, I'm a hardware guy, but it would be nice if running vlars could be optimized.

If you run the stock application then it will (eventually) select the application that gives the best throughput.


If you choose to run something that is only at the Beta stage of release- not yet close to general release- then it's up to you to select the application & then tweak it to give the best results. The information you provide back will allow those developing the applications to make the release version better. Unfortunately the documentation on what each variable is & does & what's best & not best doesn't really exist- this is still a Beta release remember.
But there are people around that can help.

---

Grant
Darwin NT

I run Lunatics 0.44...

Lunatics is an installer, nothing more.
It allows you to select the applications that you want to run.

This thread is about Lunatics 0.45 which allows people to select the r3430 SoG app.
If you are having issues with the installer, this is the thread to post in.

If you are having issues with the SoG application, then the "OpenCL NV MultiBeam v8 SoG edition for Windows" is the thread to post in.

If you are having other issues, check for an existing thread on that topic. If there aren't any around, then I suggest starting one.

---

Grant
Darwin NT

Note that to test the SoG application you have to run the v0.45 Beta installer, and choose that particular application from the versions offered.

When your SoG test is complete, you can - if you wish - re-run the v0.45 installer and choose one of the CUDA applications instead. There's no need to keep two separate installers on your hard disk.

My AMD 4200 does not like this installer for CPU tasks :(

I have tried SSE2 SSE3 but all are Computation Errors. And it was fine before.

<snip>

From Jason

Ugh, that one again. Tweeted Eric & Marius a link to the other thread, just because they happened to be on my twitter feeds.

Is there any news on this front? My AMD is CPU dead now.

My AMD 4200 does not like this installer for CPU tasks :(

I have tried SSE2 SSE3 but all are Computation Errors. And it was fine before.

<snip>

From Jason

Ugh, that one again. Tweeted Eric & Marius a link to the other thread, just because they happened to be on my twitter feeds.

Is there any news on this front? My AMD is CPU dead now.

Did you try SSE version from my website ?

---

With each crime and every kindness we birth our future.

My AMD 4200 does not like this installer for CPU tasks :(

I have tried SSE2 SSE3 but all are Computation Errors. And it was fine before.

<snip>

From Jason

Ugh, that one again. Tweeted Eric & Marius a link to the other thread, just because they happened to be on my twitter feeds.

Is there any news on this front? My AMD is CPU dead now.

Saw activity before the weekend on Boinc Dev, and discussions on the other thread here as well. So I believe wheels are in motion

---

"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.

Saw activity before the weekend on Boinc Dev, and discussions on the other thread here as well. So I believe wheels are in motion

What activity/discussions are you referring to, please? I didn't think I'd made any changes to the CPU path for this test installer, except removing the MB_v7 support.