Since there has been a lot of )%&^%ing, gripping, complaining about VLARs running on GPUs, I would like to see if there is anyone besides myself that didn't have a problem running them and what their setup is. I would like to have the ability to at least have an opt in/out to process them.

As for me I ran them exclusively on my i7/950 w/ 2 x GTX660SC @ 2Gb, Lunatics 0.41 & Cuda50 since v7 day one. In my app_info.xml file all CPU settings and GPU counts are (.5). Original est. times were approx. 15 - 23 hrs. and settled down quite rapidly to 1 - 1.5 hrs. with the last 2 running this morning. I said that I ran them exclusively on the 950 as I don't remember seeing any on the 930 machine.

---

I don't buy computers, I build them!!

ATI cards never had issues with VLARs.

---

With each crime and every kindness we birth our future.

ATI cards never had issues with VLARs.

Ok, anyone with NVidia cards that didn't have any problems then. I'm trying to have a honest conversation here.

---

I don't buy computers, I build them!!

ATI cards never had issues with VLARs.

Ok, anyone with NVidia cards that didn't have any problems then. I'm trying to have a honest conversation here.

single x GTX 680, on 45nM core2Duo, aggressive priority and pulsefinding settings. 2 VLAR in 1h28m (~44mins each), no particular display or keyboard lag noted as has been described by others.

I would like opt-in too.(for the science value & big work supply)

---

"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.

You have a relatively fast CPU and a mid-range GPU, probaily that´s the clue why you don´t suffer the same problem of our hosts.

But something else calls my attention, what MB you use? that could be another key to solve the quest.

Another interesting info you could share with us, what happening when you run 2 WU on your card, 1 Vlar and 1 normal, did the normal WU slow down the process time too?

I try to look a patern why you don´t have the problem and why i have, and after that try to compare with the others, that could help us to find why and possible a way to avoid the problems.

I realy belive we all will happy crunch any Vlars in our GPUs if we could fix the problems.

@Jason - When your 680 runs 2 WU (1 Vlar and 1 Normal) the normal Wu slow downs too or not? In our case both WU slow down.

And BTW changes in pulfidings realy makes a diference on the video/keyboard lag, the only one host i can´t realy fix the lag was the 2x690 host with the I5 CPU, in that case i belive is the slow MB/I5 combination can´t realy handle 2 big guns at full speed, even if the Mb teoricaly have suport for 3.0 slots.

---

I didn't get any VLAR's until after I had done the maths on running one or two tasks at a time on my 670's. So I was running two tasks at a time when I got the first VLAR's .

They took between 1.5 hrs and 2 hrs to complete and also slowed the task running alongside. So I went back to running one at a time. The original estimate was 16.xx mins but they took about 48 mins to complete. Tasks at other AR's completed close to the original estimate.

When processing VLAR's there was the most display lag I have seen with this graphics card but not excessive when running only one task. When running two tasks, one at least a VLAR, I had to suspend GPU crunching when running some s/ware.

My views are that VLAR's do cause problems;
they take 3 * the estimate when only running one task at a time.
The card can run two tasks at a time if there are no VLAR's around.
At mid range AR = 0.42 running one task time to process is ~640s, running two tasks time to crunch is ~800s each therefore one every 400s.

If it was an opt in or out choice, I would opt out of doing VLAR work on the GPU as more tasks at other AR's are completed without them around.

Jason´s

Winterknight have 670 like me and sees the same problems (670 is almost the same generation of my 690 too) could be something related to our GPU´s since appears that video lag not happening with the 680/660 hosts as reported by you and Cliff?

---

Some higher end cards can handle the VLAR tasks without issue. However it would seem there is a majority that can not. Having an opt-in/out mechanism would be nice, or some way for BOINC to run few instances when VLAR tasks are present. The later I would like to see implemented for CPU tasks as well. As during a VLAR storm running 24 at once on my server slows it down to the point I have to suspend BOINC to do get anything done.

---

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

You have a relatively fast CPU and a mid-range GPU, probaily that´s the clue why you don´t suffer the same problem of our hosts.

But something else calls my attention, what MB you use? that could be another key to solve the quest.

Another interesting info you could share with us, what happening when you run 2 WU on your card, 1 Vlar and 1 normal, did the normal WU slow down the process time too?

I try to look a patern why you don´t have the problem and why i have, and after that try to compare with the others, that could help us to find why and possible a way to avoid the problems.

I realy belive we all will happy crunch any Vlars in our GPUs if we could fix the problems.

@Jason - When your 680 runs 2 WU (1 Vlar and 1 Normal) the normal Wu slow downs too or not? In our case both WU slow down.

Both machines use an Intel DX58SO board, which was specifically designed to support the 1st gen i7 extreme processors and if I could afford it I would bump them both to 990x. It doesn't have all of the bells & whistles that other boards have. I also run BOINC at 80% of processors (6 cores for CPU processing) when in use & 90% when idle (overnight). I take the time to suspend all GPU processing when playing games or watching videos -- therefore no video lag. I could run higher than 85%, then I run into lag, but why do it when it's not needed. 80% has been the sweet spot for these machines for many years even when I was running a Q6600.

Juan - I see that most of your processors are 2nd/3rd gen. i7s w/ Fermi or newer cards, what MB are you using?

I don't really care that when running 2 WUs (1 VLAR & 1 other) if one is slower than the other. Since I'm not in this for the credits per se, but the science, it will balance itself out in the long run. I've also found out that when running AP on the GPUs (2 tasks ea.) that reserving .5 CPU per task is cleaner than 1 full core.

[Edit] BTW, the only and I repeat only changes that I've made to the app_info.xml file are the cpu & count settings. All others are at default including flops, etc. [/Edit]

---

I don't buy computers, I build them!!

I would like opt-in too.(for the science value & big work supply)

If they implement this I would like one more opt-out ;)
When CPU asks for SETI@home v7 work send VLARs in preference (if available)

(all this opt-in/opt-out have to be probably in some 'advanced' section of SETI@home preferences to not confuse 'casual' users)

---

 


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

Some higher end cards can handle the VLAR tasks without issue. However it would seem there is a majority that can not. Having an opt-in/out mechanism would be nice, or some way for BOINC to run few instances when VLAR tasks are present. The later I would like to see implemented for CPU tasks as well. As during a VLAR storm running 24 at once on my server slows it down to the point I have to suspend BOINC to do get anything done.

Just curious, what is the percentage of processors set at for your server?

---

I don't buy computers, I build them!!

Sorry for the delay i was out.

Most of my MB are ASUS P8B75-MLE (one of che cheap ones avaible, here all hardware is very expensive due the taxes) and runs with i5-2310 (a slow cheap one). I only run one I7 with a 590+2x580. No Vlar´s are sending to it because the GPU´s are Fermis or at least what i spect.

Normaly i leave 50% of the cores avaiable to exactly not interfear with the users so on the I5 there are allways 2 cores avaiable and i run only 2WU on each GPU.

On the hosts with a single GPU (most are 670 or 580) there are relatively no video or keyboard lag specialy after follow Jasons instructions and i made some small adjusts on pulsefinding settings allowing me inclusive see a video while typing a word document, spreadsheets or doing others things normaly a VT100 terminal emulator of a mainframe. No games allowed to run here and no aditional program uses the GPU besides SETI work.

On the 2xGPU´s hosts (with 2x690 or 690+670) the problem realy apears, it´s realy complicated to use the host if more than one Vlar´s is running, that´s why i´m ask to find a way to stop sending the vlars to thats hosts. To make the host useable you need to stop the boinc process as described by WinterKnight.

I realy belive my problem is something related to the type of host i run (MB/CPU/GPU) and totaly diferent for the one that HAL9000 has, his 8500 GT should never receive a Vlar WU, as described in the forums, Vlar´s are expected to send only to the Keplers and UP by the servers.

A user selected choice it you want or no run Vlar in your host is the simplest way to keep us all happy, is my intention to keep doing Vlar´s on the hosts without problems even with the slow performance, i just need to find a way to stop receiving Vlar´s on the problematic hosts.

@Hall9000 from you post...

As during a VLAR storm running 24 at once on my server slows it down to the point I have to suspend BOINC to do get anything done.

Something else must be in place on your problem,totaly diferent from what we are looking (V7 VLARs & GPUs?), at least on the host page this hosts has no GPU on it, so you run a CPU only WU, a Vlar WU normaly takes 3 to 4 hours to compleate ( the rela speed depends of the CPU Type/Usage ratio)

---

a VT100 terminal emulator of a mainframe.

Off topic, but it's been a long, long, long, time since I've used at VT100 emulation. Mostly I used 5250 emulation before I retired.

On the hosts with a single GPU (most are 670 or 580) there are relatively no video or keyboard lag specialy after follow Jasons instructions and i made some small adjusts on pulsefinding settings allowing me inclusive see a video while typing a word document, spreadsheets or doing others things normaly a VT100 terminal emulator of a mainframe. No games allowed to run here and no aditional program uses the GPU besides SETI work.

What adjustments did you make to the pulsefinding settings, might want to try them?

---

I don't buy computers, I build them!!

i7 3820 with a 670. Nothing of any noticble lag on 2 vlar units, choppy video at 3 vlars at once and noticable slowdown of other v7 tasks with atleast 1 vlar running. Not so much noticable slowdown when running an AP task on the gpu with above normal priority set on the v7 wu.

---

a VT100 terminal emulator of a mainframe.

Off topic, but it's been a long, long, long, time since I've used at VT100 emulation. Mostly I used 5250 emulation before I retired.

I agree but VT100 still an option when you don't need special keys, color or printing and of course works perfect on old DOS 486 hosts for example... As you could notice I´m very old too... But i allways think, why make changes on something that works for almost 20 years? If you are old as me I´m sure you remember the old (from the computer jurasic era)... NCSA DOS Telnet Emulator... I still have several 486 & PI-100 hosts running it and works perfect at simply no cost (all are recicled machines)

On the hosts with a single GPU (most are 670 or 580) there are relatively no video or keyboard lag specialy after follow Jasons instructions and i made some small adjusts on pulsefinding settings allowing me inclusive see a video while typing a word document, spreadsheets or doing others things normaly a VT100 terminal emulator of a mainframe. No games allowed to run here and no aditional program uses the GPU besides SETI work.

What adjustments did you make to the pulsefinding settings, might want to try them?

See this post from Jason:

http://setiathome.berkeley.edu/forum_thread.php?id=71809&postid=1378668

Realy help a lot in some cases.

---

As per previous posts, no problems running dual
460s or dual 560s two wu per card.

As per previous posts, no problems running dual
460s or dual 560s two wu per card.

Your GPU are no Keplers, so you don´t receive Vlar´s to crunch on the GPU´s... so that problems does not apears on your hosts. Or at least is expected the servers don´t send the Vlar´s to your hosts.

@Cliff - as a sugestion change the thread name to make it more clear to the rest of the comunity. Something like: V7 VLARs & Kepplers GPUs?

---

As per previous posts, no problems running dual
460s or dual 560s two wu per card.

Your GPU are no Keplers, so you don´t receive Vlar´s to crunch on the GPU´s... so that problems does not apears on your hosts. Or at least is expected the servers don´t send the Vlar´s to your hosts.

The 'problem' started back with the first gtx2xx
to any and all higher cards when vlars first came out.
You don't remember our previous exchanges
on this? A search would turn them up.

I'm no longer allowed to run vlars on my gpus due to the tyranny of the majority.

As per previous posts, no problems running dual
460s or dual 560s two wu per card.

Your GPU are no Keplers, so you don´t receive Vlar´s to crunch on the GPU´s... so that problems does not apears on your hosts. Or at least is expected the servers don´t send the Vlar´s to your hosts.

The 'problem' started back with the first gtx2xx
to any and all higher cards when vlars first came out.
You don't remember our previous exchanges
on this? A search would turn them up.

I'm no longer allowed to run vlars on my gpus due to the tyranny of the majority.

There's a line to be found somewhere in the gap between 'tyranny' and 'democracy'. The answer probably gets called 'freedom of choice', but that implies a democratic tyranny over the number of working hours in Eric's day. We've asked for a preference setting, anyway.

As per previous posts, no problems running dual
460s or dual 560s two wu per card.

Your GPU are no Keplers, so you don´t receive Vlar´s to crunch on the GPU´s... so that problems does not apears on your hosts. Or at least is expected the servers don´t send the Vlar´s to your hosts.

The 'problem' started back with the first gtx2xx
to any and all higher cards when vlars first came out.
You don't remember our previous exchanges
on this? A search would turn them up.

I'm no longer allowed to run vlars on my gpus due to the tyranny of the majority.

The change was made because the majority of people were reporting problems running VLAR on NV GPUs.
Future optimizations of the app may change that. If so, I am sure Eric would be happy to reverse the situation.

---

"Freedom is just Chaos, with better lighting." Alan Dean Foster