GPU72 plans post-announcement

[Xyzzy]

We think this is the first press release where the forum has been linked.

[garo]

So how is GPU72 keeping up with the post-announcement demand? I see that Primenet is assigning stuff in the 62M range but most of what is being assigned is TF to 74 bits but only a few exponents have P-1 done.

[chalsall]

Originally Posted by garo

So how is GPU72 keeping up with the post-announcement demand?

Comfortably.

Originally Posted by garo

I see that Primenet is assigning stuff in the 62M range...

Only occasionally.

Originally Posted by garo

...but most of what is being assigned is TF to 74 bits but only a few exponents have P-1 done.

GPU72 is currently releasing at 73 "bits", holding back 500 candidates for P-1'ing or TFing to 74.

We could use some more P-1 workers. But we're not holding back the show....

[kracker]

Originally Posted by chalsall

We could use some more P-1 workers. But we're not holding back the show....

Rawr, my minions are doing the best they can, about ~300 GHz a week which will go up by another ~100 soon I think, after....
At the moment I believe they are spitting out 10-15 P-1 a day. (haven't checked lately :p)

[garo]

Thanks for the update. Looks like there were 400 additional exponents assigned in the 62M range yesterday. There are 2200 ready to go in Primenet in that range right now - in addition to everything above that range - 400+ right now so we seem to be covered.

I have all my minions with sufficient memory on P-1 too and spitting out about 5-6 a day. Now the question is whether to 73 is the default again.

[chalsall]

Originally Posted by garo

Now the question is whether to 73 is the default again.

313 candidates were assigned in the 62M range yesterday. We're currently completing ~500 candidates a day to 73, so I'm comfortable we can keep ahead.

I'm continuing to monitor the situation, but I think we'll be fine even when Jerry starts taking his monsters off-line. We have also had several new participants join the TFing effort since the "big announcement".

[kracker]

Originally Posted by chalsall

313 candidates were assigned in the 62M range yesterday. We're currently completing ~500 candidates a day to 73, so I'm comfortable we can keep ahead.

I'm continuing to monitor the situation, but I think we'll be fine even when Jerry starts taking his monsters off-line. We have also had several new participants join the TFing effort since the "big announcement".

Also remember Jerry is the second largest P-1 user in gpu72

[chalsall]

Originally Posted by kracker

Also remember Jerry is the second largest P-1 user in gpu72

Yeah, I know...

[ET_]

Originally Posted by chalsall

Yeah, I know...

Would P-1 with just B1=1,500,000 (and no B2) be useful for exponents with no P-1 at all?

Luigi

[chalsall]

Originally Posted by ET_

Would P-1 with just B1=1,500,000 (and no B2) be useful for exponents with no P-1 at all?

I don't feel comfortable enough with the stats and math to speak to this. George or James would be far more authoritative.

[garo]

Originally Posted by ET_

Would P-1 with just B1=1,500,000 (and no B2) be useful for exponents with no P-1 at all?

Luigi

No. Most of the exponents with no P-1 that are assigned for LL testing get a P-1 stage 1 by the same computer since it does not take up much RAM. The bound chosen is optimal for Prime95. An arbitrary bound like 1.5e6 is more likely to be suboptimal than not.