GPU72 and BOINC a match made in ....

[Uncwilly]

Originally Posted by rebirther

different, 1-3 times

An hour, day, month?

[rebirther]

Originally Posted by Uncwilly

An hour, day, month?

a day

[pinhodecarlos]

Originally Posted by pinhodecarlos

I did not manage to post at Gridcoin forum about the app add on. SETI.USA and the Scottish are on it. Hard OCP with a 13no. GPU guy on it. Was expecting to see XtremeSystems but let’s wait, news will spread out by itself.

Edit and BTW: if each host has 3 wus queued then more than 900 GPU’s are connected to the server. Only Reb can confirm this. Can you count the number of GPU hosts connected to the server?

There’s away to trick the server when you have more GPU’s than the real ones to grab more wus, therefore discard the 900 GPU figures.

[chalsall]

Originally Posted by rebirther

a day

While the firepower is certainly impressive (and I'm repeated told this is just the beginning), "breadth-first" is not really all that immediately useful to GIMPS.

If you *really* want to impress us, why not make your next fetch to 78???

As you and KEP requested, GPU72 is ready to facilitate this. And it would immediately assist the P-1'ers, who have to do (slightly) less work the higher a candidate is TF'ed.

[rebirther]

Originally Posted by chalsall

While the firepower is certainly impressive (and I'm repeated told this is just the beginning), "breadth-first" is not really all that immediately useful to GIMPS.

If you *really* want to impress us, why not make your next fetch to 78???

As you and KEP requested, GPU72 is ready to facilitate this. And it would immediately assist the P-1'ers, who have to do (slightly) less work the higher a candidate is TF'ed.

Soon, I still cannot fix the opencl linux app, I dont know why its not running (have no test PC for this). The 78 is a bit too high at the moment but Iam sure we can handle this in some weeks.

[chalsall]

Originally Posted by rebirther

The 78 is a bit too high at the moment but Iam sure we can handle this in some weeks.

Why? You have demonstrated your system can go from 72 to 73 just fine. What's the difference (other than the longer run-time, of course)?

[rebirther]

Originally Posted by chalsall

Why? You have demonstrated your system can go from 72 to 73 just fine. What's the difference (other than the longer run-time, of course)?

the runtime.

[chalsall]

Originally Posted by rebirther

the runtime.

I fail to understand how this is a problem. This is the nature of the work.

[Prime95]

Originally Posted by chalsall

I fail to understand how this is a problem. This is the nature of the work.

It could be the he BOINC culture prefers work-units that take a short period of time.

If so, one could split the 2^72 - 2^77 work into two work units 2^72 - 2^76 and 2^76 - 2^77. Both will take about the same amount of time.

In the end all work helps, I'm sure reb will get it working smoothly.

[chalsall]

Originally Posted by Prime95

It could be the he BOINC culture prefers work-units that take a short period of time. If so, one could split the 2^72 - 2^77 work into two work units 2^72 - 2^76 and 2^76 - 2^77. Both will take about the same amount of time.

However... Reb asked that GPU72 provide blocks of assignments only going a single bit-level at a time. And then he reserved everything available to go from 72 to 73...

Originally Posted by Prime95

In the end all work helps, I'm sure reb will get it working smoothly.

Indeed. Although, again, there's the question of timeliness... Your FC'ers and P-1'ers are hungry!

[LaurV]

You can even split large jobs in classes. I used to do that when factoring my 666M expo to 86 bits, or some 332M to 85. To start from any class is easy, you just create the right checkpoint file (a text editor is enough). To stop at a class, you have to watch (capture the output) and send CTRL+C. You can split the classes in how many cards you have, create 2 or 3 or 4 checkpoint files to resume for the class you want and watch them to stop when the right class is reached. Or, modify mfaktc (not complicate) to stop at some class that he reads from a file similar to the checkpoint file. When you split the job in 4 cards, you do it 4 times faster.

On the other hands, what times are you talking about? It take like 10 minutes to do a 73 and 20 minutes to do a 74 in an average GPU (??), well, more or less, but I do not believe that BOINC community is so sensitive to "10 vs 20 minutes jobs". They may get sensitive when you talk days versus weeks, but when everything is measured in minutes?