For those of you with big (and I mean BIG) NVidia cards taking a rest from SETI currently, GPUGrid has a big batch of work available today.

The rules of their game are very different from the ones we play by here.

1) they like their tasks to be returned within 24 hours. Since a single task may run for 18 hours or more, you need to start them near enough immediately - so close to zero cache.

2) for the same reason, the project is really only suitable for people who run their machines 24/7.

3) I'd suggest that the starting GPU level is around the GTX 770. 750Ti barely cuts the mustard.

4) Driver version must support CUDA 6.5 - sorry, I can't give you an exact version number for that offhand.

5) They pay an unbelievable number of Zimbabwean Dollars, but you still can't buy a toaster with them.

6) Watch those temperatures.

7) Did I mention the badges? For your personal contribution to a published scientific paper, after publication.

Got my 32 thread machine up and running for testing, and it will be a killer (I think). And the best thing is, with 2 E5-2670, Asrock Rack EP2C602 motherboard with 32GB of desktop RAM (4x8gb), and 2 x GTX 980s (reference) it draws less than 550 watts at the wall running Lunatics, 3 WUs/GPU and 29 threads of CPU. Am going away for a week tomorrow, but will unleash the beast when I get back...

Used 2 x Corsair H50 for CPU cooling, in a Xigmatek Elysium case (which had the correct mounting holes for the SSI-EEB motherboard). Running full out, got to low-to-mid 70s on one CPU, about 70 on the other, and that's with the fans controlled to hold down the noise. If necessary, can up them a hundred or two rpms. (The board has set voltages (10 steps) for non-CPU fans, though Smart Fan for the CPU fans). And no sound and a very few USB ports. Definitely cruncher grade only!

Unfortunately, the fact that its a server MB means lots of quirks and a pain to set up. Had troubles with setting up 2nd vid card. Got it working, somehow.

---

My Team here has just been deleted, I'll think of shutting down here too.

---

That is sad, the targeted GBs data is probably the best we ever had.

---

That is sad, the targeted GBs data is probably the best we ever had.

Maybe it is the best data, but the Credit per WU is not.

I know... it's all just for science, but after 15 jears of crunching for seti@hohme it's time to push the button...

=> seti@home := off
=> einstein@home := on

Maybe i'll be back if this mess is fixed...

---

I've been crunching einstein since about a week and i think i reach about 1 mill rac by end of next week i'm place 25 or so right now over there

i'm fairly disappointed how seti handles this so untill they fix all the things that should have been fixed years ago i be crunching einstein ....at least their stuff works

---

I came down with a bad case of i don't give a crap

I also deleted my smart car team about a week ago. Didn't make sense to keep it online. I am sure other team members where in agreement.

All the best,
Aaron Lephart

---

Well if this keeps going all the good crunchers are gonna be gone exept that bloke chris (i think he's australian) so they persistent

Either way it's time for seti to look into their software and backend and the creditnew

---

I came down with a bad case of i don't give a crap

Not everyone is leaving..

Some are coming back.

why not run Seti on the CPU and another project on the GPU ? so you'll still be contributing valuable time to a very worthwhile project...

---

Not everyone is leaving..

Some are coming back.

I can't imagine why! Please. Keep up enthusiast!

---

I can't imagine why! Please. Keep up enthusiast!

That's easy.

Green Bank was what we been waiting for.

Sustained observation of the sky versus the rapid scanning of Arecibo.

We stand a better chance of finding something with this data than we did with previous data.

If people chose to leave, that is their right.

Others will eventually make their way back.

I can't imagine why! Please. Keep up enthusiast!

That's easy.

Green Bank was what we been waiting for.

Sustained observation of the sky versus the rapid scanning of Arecibo.

We stand a better chance of finding something with this data than we did with previous data.

If people chose to leave, that is their right.

Others will eventually make their way back.

I think change is tough for people to handle too, and can understand that to a point. I once saw projections that GBT reobservation APs may take 6 months per task on CPU. Don't know if the projections have since changed, or alternative parameters examined, but that'd probably be a severe culture shock to many IMO, at least those not familiar with the Climate change Wus or similar.

---

"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.

They might copy the "trickle" mechanism from CPDN, by which your credits go up as long as you keep crunching even before finishing a long task.
Tullio

---

why not run Seti on the CPU and another project on the GPU ? so you'll still be contributing valuable time to a very worthwhile project...

Cause all this thread not about how really deal with problems but about directing own frustration out to others. Definitely cafe thread and nothing common with number crunching.

Astropulse for GBT data? Can you tell us more?

---

Astropulse for GBT data? Can you tell us more?

Haven't heard any more about it than the prediction a year or so ago that they might be large. If and when they do come, from what I understand it may be a fairly big server & application overhaul. (probably depending on which telescope etc.)

---

"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.

Astropulse for GBT data? Can you tell us more?

Haven't heard any more about it than the prediction a year or so ago that they might be large. If and when they do come, from what I understand it may be a fairly big server & application overhaul. (probably depending on which telescope etc.)

I vaguely remember the same remark, but on a quick search I can't find who made the statement - so I can't judge at this distance whether it was made by someone with the right mathematical knowledge.

The more I think about it, the less likely it seems. Even keeping the existing datapak size (8 MB), there's only so much analysis that can be done on it.

Einstein's Gravity Wave search uses much larger datapaks, and searches a much larger parameter space - the first confirmed gravity wave detection was at a distance of 1.3 billion light years, with all the de-dispersion range that implies (although the actual detection was done in-house, not via the BOINC application).

Einstein handle the search envelope problem by re-using the datapaks, and setting each task to study a limited part of the parameter range. That keeps the processing time down to something comparable to the existing Arecibo AP tasks. Wouldn't that be a better approach than the CPDN 'megatask and trickle' needed to simulate the global atmosphere/ocean interactions all in one go?

I vaguely remember the same remark, but on a quick search I can't find who made the statement - so I can't judge at this distance whether it was made by someone with the right mathematical knowledge.

The more I think about it, the less likely it seems. Even keeping the existing datapak size (8 MB), there's only so much analysis that can be done on it.

Einstein's Gravity Wave search uses much larger datapaks, and searches a much larger parameter space - the first confirmed gravity wave detection was at a distance of 1.3 billion light years, with all the de-dispersion range that implies (although the actual detection was done in-house, not via the BOINC application).

Einstein handle the search envelope problem by re-using the datapaks, and setting each task to study a limited part of the parameter range. That keeps the processing time down to something comparable to the existing Arecibo AP tasks. Wouldn't that be a better approach than the CPDN 'megatask and trickle' needed to simulate the global atmosphere/ocean interactions all in one go?

Probably the logistics would have to be decided during prototyping I'd imagine.

The prediction was made by Eric, possibly longer than a year ago, I suspect well before details of the telescopes were known. In that context he might have been referring to some particular telescope we haven't seen the data for yet. The main thing that stood out was the likliehood of needing 2M point FFTs (as opposed to our MB 128k element transforms, and AP (iirc) 32k element chunks. Which application would warrant such large transforms I wasn't clear on then or now).

At least in the case of MBv8, I interpreted from more recent contact that extra precision/search depth wasn't likely necessary at this point. So if similar holds for AP then the logistic & search questions might be moot.

I did congratulate Eric via Twitter when v8 CPU went out relatively smoothly over newyears, and jokingly suggested he could get some sleep. His quip response was along the lines of 'I'll sleep when v9 is out...'

---

"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.

Let's hope that the scientists - that means astronomers, mathematicians, and sociologists with experience of distributed computing - are allowed to get on with the prototyping at their own speed and with their own goals in mind, then.

We don't want more releases like v8 to a timetable set by financiers and marketing, which turned out not to be a timetable after all (because the same finance/marketing machine held back the data until the financier's name-day).