[QUOTE=penlu;493251]I assume gpu72 went down briefly and so primetools fell through to primenet...?[/QUOTE]

That's a safe assumption. GPU72 had a bit of a hick-up this morning (sorry about that; SPE with a bad query while a table was locked...).

These assignments didn't come through GPU72 in any way, so I'm presuming your primetools fell back to fetching from Primenet when it got a 500 response from GPU72.

[QUOTE=penlu;493251]Why are these exponents so large?[/QUOTE]

I can't speak to that. Perhaps Aaron, James or George could.

[QUOTE=penlu;493251]I have just received a couple of exponent assignments I found kind of strange since they're so large: [URL="https://www.mersenne.org/report_exponent/?exp_lo=166802869&full=1"]166802869[/URL], [URL="https://www.mersenne.org/report_exponent/?exp_lo=223488289&full=1"]223488289[/URL][/QUOTE]
[QUOTE=chalsall;493313]I can't speak to that. Perhaps Aaron, James or George could.[/QUOTE]I have a machine doing the Primenet based "Trial factoring to low limits" and it just turned in some in the 166,xxx,xxx area and some in the 223,xxx,xxx range. So those fall in the current normal ranges for that type of assignment. HTH

Thanks for your answers; it's interesting that primenet goes so high...

I am not sure if this is the right thread to post in about the following: I have a question about B1 and B2 selection for p-1 factoring. Often the p-1 factoring program will determine these values automatically. I plugged these values into mersenne.ca's [URL="http://www.mersenne.ca/prob.php"]probability/bounds calculators[/URL] and got results that confuse me.

I'm running CUDAPm1 on exponent 88061693 (TF'd to 76 bits). It has selected B1=755000, B2=18120000 and estimates 3.33% factor-finding success probability. The probability calculator at mersenne.ca reports a success probability of ~3.28% and an effort of 8.315589 GHz-days. But when this effort value is entered into the bounds calculator, it reports a selection of B1=858839, B2=16317937 to generate a slightly higher success probability ~3.29%.

I have a few questions. First, why do the success probability estimates differ (3.33% vs. 3.28%) between CUDAPm1 and mersenne.ca? Who should I trust (I assume CUDAPm1's method is more approximate)? Second, should I rely on mersenne.ca's optimal bounds selection for a given amount of work, or do local programs (also wondering about mprime) pick better bounds because they know more about the performance of the machine they're running on? Finally, the [URL="http://www.mersenne.ca/exponent/88061693"]exponent lookup page[/URL] lists B1 and B2 values---in this case, B1=960000, B2=24000000. What should I do with these values---interpret them as upper bounds, or the appropriate amount of work to do?

I understand the difference is slight---possibly just asking this question already exceeds its importance---but I want to understand better the considerations for B1 and B2 selection, and also maximize my machines' overall effectiveness.

[QUOTE=penlu;493641]Who should I trust; is CUDAPm1's estimation method out of date/resource-limited? Second, should I rely on mersenne.ca's optimal bounds selection for a given amount of work, or do local programs (also wondering about mprime) pick better bounds because they know more about the performance of the machine they're running on?[/QUOTE]I personally would trust mersenne.ca's calculations [i]less[/i] than whatever your P-1 program of choice reports, and it's my site. :smile:
Most of the P-1 related calculations on mersenne.ca re approximate at best, guesses at worst, and all written many years ago.

[QUOTE=James Heinrich;493642]I personally would trust mersenne.ca's calculations [i]less[/i] than whatever your P-1 program of choice reports.[/QUOTE]


i was going to say: trust mersenne.ca :)

Seriously, I'd be surprised if prime95 was accurate to more than +/- 5%. This is OK since optimal P-1 bounds is a fairly flat curve. That is, choosing somewhat less than perfect P-1 bounds does not change throughput significantly.

Proxy error GPU72
 

Hey Chris,
I have been seeing the attached for at least a couple of hours.

EDIT:04:13 UTC
This is not happening now. :smile:

[QUOTE=kladner;494625]Hey Chris,
I have been seeing the attached for at least a couple of hours.[/QUOTE]

Ditto

[QUOTE=kladner;494625]I have been seeing the attached for at least a couple of hours.[/QUOTE]

That's interesting.

The fact that you're seeing messages from GPU72 means the networking between your client and the GPU72 server is OK.

There might have been a connectivity issue between the GPU72 server and the Primenet server.

I also have several machines running through the proxy, and none reported any issues, so perhaps this was a brief hiccup.

OK, something weird is going on...
 

I'm going to do an unscheduled upgrade and reboot shortly.

GPU72 should be back within a couple of minutes (longer if things don't go so well)...

[QUOTE=chalsall;494637]GPU72 should be back within a couple of minutes (longer if things don't go so well)...[/QUOTE]

Man, I gotta tell you: I hate to do that kinda thing. Has to happen from time-to-time to ensure sanity, but it is rather nerve racking during the silence from the machine.

GPU72 is currently down...
 

Hey all. FSCK!!!

The GPU72 server just crashed; looks like another bad Seagate drive.

1and1 Tech Support and I are working it. Please stand by....