mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > PrimeNet

Reply
 
Thread Tools
Old 2021-04-14, 13:53   #1
drkirkby
 
"David Kirkby"
Jan 2021
Althorne, Essex, UK

101111112 Posts
Default How does one interpret the CPU efficiency meters/scales ?

I'm puzzled on how to interpret the attachment.



Two PRP tests should be checked in within the next two hours and 10 minutes (see below). I assume the 24 hour one is likely to go up, but the whole concept seems a little odd to me. Perhaps because I have not been on here long, the scales seem to go up/down a huge amount.



[Worker #2 Apr 14 14:47] Iteration: 109180000 / 110725829 [98.60%], ms/iter: 1.543, ETA: 00:39:45
[Worker #1 Apr 14 14:47] Iteration: 105300000 / 110317973 [95.45%], ms/iter: 1.548, ETA: 02:09:26
Attached Thumbnails
Click image for larger version

Name:	Capture.PNG
Views:	22
Size:	140.6 KB
ID:	24674  
drkirkby is offline   Reply With Quote
Old 2021-04-14, 14:20   #2
Viliam Furik
 
"Viliam Furík"
Jul 2018
Martin, Slovakia

1C816 Posts
Default

It is most probably because your last result is from the 10th of April, so more than 24 hours ago.
Viliam Furik is online now   Reply With Quote
Old 2021-04-14, 15:22   #3
drkirkby
 
"David Kirkby"
Jan 2021
Althorne, Essex, UK

191 Posts
Minus

Quote:
Originally Posted by Viliam Furik View Post
It is most probably because your last result is from the 10th of April, so more than 24 hours ago.
One was checked in a few minutes ago. Now the 24-hour reading is 246.2%, and the 90-day 55.5%. But what do those numbers actually mean? A PRP test of 110317973 will complete in 30 minutes or so. I will check again, but I like to know what the numbers actually mean.
Attached Thumbnails
Click image for larger version

Name:	Capture-3.PNG
Views:	10
Size:	158.4 KB
ID:	24677  

Last fiddled with by drkirkby on 2021-04-14 at 15:26
drkirkby is offline   Reply With Quote
Old 2021-04-14, 16:26   #4
drkirkby
 
"David Kirkby"
Jan 2021
Althorne, Essex, UK

191 Posts
Default

Another PRP test was checked in half an hour or so ago, so the 24 hour figure is now 491.5%, and the 90 day figure 58.2%. Again I have no idea what they mean. I will not be checking any more in for 36 hours or so, in which case I assume the 24 hour scale will drop to zero in just under a day.

I'm seriously thinking of giving up the PRP test of
https://www.mersenne.org/report_expo...2646233&full=1
That's the first, and last attempt I make at a 100 million digit prime. I imagine quite a few people start such tests, then give up.
Attached Thumbnails
Click image for larger version

Name:	cpu4.png
Views:	6
Size:	842.8 KB
ID:	24678  

Last fiddled with by drkirkby on 2021-04-14 at 16:27
drkirkby is offline   Reply With Quote
Old 2021-04-14, 16:46   #5
axn
 
axn's Avatar
 
Jun 2003

25×5×31 Posts
Default

Quote:
Originally Posted by drkirkby View Post
One was checked in a few minutes ago. Now the 24-hour reading is 246.2%, and the 90-day 55.5%. But what do those numbers actually mean? A PRP test of 110317973 will complete in 30 minutes or so. I will check again, but I like to know what the numbers actually mean.
The server is aware of certain machines registered under your account. It has certain expectations of how much these machines can produce. It merely lists what % of expected work was actually reported as complete in past 24 hours / past 90 days.
Evidently, 2 PRPs are 5x the credit of what is expected of your computers in 24 hours. Obviously the "24 hr" figure is useless since it will just fluctuate between 0 and some ridiculous number (because of the granularity of the PRP test completion).
The 90 days figure is more useful; once you've been steadily crunching for 90 days, it should show something closer to 100%.

Honestly, these two are the least useful stats in the whole site. Only useful for someone running a large number of computers, and who can use it to quickly check if there is something abnormal in their production. For normal people running just one or a handful of computers, it is useless.
axn is offline   Reply With Quote
Old 2021-04-14, 17:22   #6
drkirkby
 
"David Kirkby"
Jan 2021
Althorne, Essex, UK

191 Posts
Default

Quote:
Originally Posted by axn View Post
The server is aware of certain machines registered under your account. It has certain expectations of how much these machines can produce. It merely lists what % of expected work was actually reported as complete in past 24 hours / past 90 days.
Evidently, 2 PRPs are 5x the credit of what is expected of your computers in 24 hours. Obviously the "24 hr" figure is useless since it will just fluctuate between 0 and some ridiculous number (because of the granularity of the PRP test completion).
The 90 days figure is more useful; once you've been steadily crunching for 90 days, it should show something closer to 100%.

Honestly, these two are the least useful stats in the whole site. Only useful for someone running a large number of computers, and who can use it to quickly check if there is something abnormal in their production. For normal people running just one or a handful of computers, it is useless.
Okay, thank you for the explanation, and the warning they are of little use to someone with only a handful of computers.

I have probably been giving the Primenet server a bit of a headache trying to work out the capabilities of my machines, as I had some which I had only access for a short period, so moved the jobs to a faster machine. At the minute the server is expecting to receive 110601373 and 110469433 in 8 days, but it should get them in 36 hours. The server indicates it will get 332646233 in 189 days, but that may be very optimistic or very pesimestic. Running that exponent appears to cause a serious slowdown on my computer. Allocating 26 cores to that increases the time per iteration of the 110 million PRP checks by a factor of about 4. This is despite the machine has two 26-core CPUs. I'm guessing there might be a memory bandwidth problem. I know the memory is not configured optimally this machine. I'll have to check if running 110 million exponents slows down the 332 million one. If so, it might be sensible to just let the 332 million exponent run on its own for however long it will take to complete. I reckon about a week or two in order to finish that huge exponent on 26-cores, with the other 26-cores idle.
Attached Thumbnails
Click image for larger version

Name:	x1.png
Views:	9
Size:	896.7 KB
ID:	24679  

Last fiddled with by drkirkby on 2021-04-14 at 17:26
drkirkby is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
8 Meters being Made Uncwilly Lounge 28 2021-05-07 13:40
Time scales davieddy Lounge 8 2011-05-08 03:38
Different Scales mfgoode Puzzles 5 2007-08-04 11:22
How do I interpret Prime95 factoring benchmarks? jasong Hardware 1 2006-10-23 12:08
How to interpret the report nimzo PrimeNet 5 2004-09-03 02:03

All times are UTC. The time now is 08:52.

Mon May 10 08:52:47 UTC 2021 up 32 days, 3:33, 0 users, load averages: 1.64, 1.83, 1.88

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.