GPUers, please TF my exponents!

[Christenson]

Dubslow, P95 has special database access...and remember that Mr P-1 got quite a few with just his one machine.

[Christenson]

Originally Posted by Prime95

Another small batch:
Test=65000063,70,1
Test=65000081,70,1
Test=65000137,70,1

Umm, somwhat out of scope, not being close to the LL wavefront....but it's short, so I'll do it anyway...

[Dubslow]

Originally Posted by Christenson

Dubslow, P95 has special database access...and remember that Mr P-1 got quite a few with just his one machine.

I'm pretty sure he doesn't get 15 assignments less than 45M. But you're right, P95 is rather privileged as far as this whole GIMPS thing goes.

[davieddy]

Originally Posted by Christenson

Dubslow, P95 has special database access...and remember that Mr P-1 got quite a few with just his one machine.

P95 (aka George) is basically God.
I wonder whether he has ever considered delegating?

David

[Prime95]

Originally Posted by Dubslow

How do you get such low exponents? Just expiries late at (UTC) night like how everyone else does it?
Also, how many computers do you have? 5? 10? That's a lot of assignments.

No special access required. I've set days of work to 90. Once a month at the expiration hour I unreserve all the big exponents and get nice small ones. The downside is that sometimes you get exponents that had errors on the first run but were correct -- so you end up doing a double-check with no chance of finding a prime.

The first batch were from my dual-core laptop, the second batch was from my quad-core Sandy Bridge.

[Prime95]

Originally Posted by Christenson

Umm, somwhat out of scope, not being close to the LL wavefront....but it's short, so I'll do it anyway...

CudaLucas work. I asked the server for manual assignments above 65M that had P-1 already done. Since cudaLucas only supports power-of-two run lengths, it makes sense to give cudaLucas the largest possible assignments. I should probably change to 70 or 75 million....

[kladner]

I am pretty much committed to Mr. P-1's endeavors at the moment. Since he first took up my offer of GPU time, I will stick with him. My hope is that these various needs can be coordinated so that there is no conflict of loyalty involved.

I have great respect for Prime95 and all those who have been long-time contributors to the overall project. I would only like to make the best contribution I can without complications.

[aaronhaviland]

Originally Posted by aaronhaviland

I will gladly take all of these up to 72 for you within the next 12 hours :)

(I'll actually just set the upper bitlevel higher, so it'll just keep going until I wake up in the morning)

44519471, 44917867, 44278103, 44452879, 45260129:

No factors on any of these up to 2^73

[kjaget]

Is it OK for the people doing TF to report the results via the manual results page? I've done this before for my own exponents and it automatically unreserves the LL test at the next checkin if a factor is found. But that's with me knowing the assignment key, so I'm not sure if doing TF for a another user changes the equation at all.

[Prime95]

Originally Posted by kjaget

Is it OK for the people doing TF to report the results via the manual results page? I've done this before for my own exponents and it automatically unreserves the LL test at the next checkin if a factor is found. But that's with me knowing the assignment key, so I'm not sure if doing TF for a another user changes the equation at all.

I think it is OK to report all your results. If you run into any problems, let me know and I'll fix them.

[Mr. P-1]

Originally Posted by davieddy

This is music to my ears!
72 bits should be achievable across the board provided that
the effort is focussed at the wavefront and not two years hence.

After last night's exponent grab, and today's disbursements I am tracking 2054 assignments allocated to 6 GPUers. That figure does not include the several hundred that have been reported back to me as completed, and which I have unreserved. I have 930 assignments in reserve.

My reserves are reasonable, given that I need to be able to meet requests, but bear this in mind: I've only been doing this a week. I grabbed fewer exponents in the first few days, not realising how big this "mini-project" was going to get. Had I grabbed all the exponents I'm now targetting (all those below 53M and TFed below 70) from the very start, my reserves would be more than double what they are.

Over the next week another 4000+ such exponents will expire. I could grab them all, but then what? My current volunteers already have plenty of work queued up. Unless I get a lot more GPUers volunteering, I will have to let many of them go TFed only to 69.

What you said about 72 bit being achievable may be true in theory. But in practice, we can't even get them all up to 70.