Prime95 always uses nighttime settings
 

I've been having a little problem with prime 95 recently.

I currently have the CPU settings set like this:

Daytime : 1024MB
Nighttime: 2048MB
Daytime begins @ 7:30 AM and ends @ 11:30 PM

If I switch around the night and day values, Prime95 still uses 2048MB of RAM regardless of the time.

How do I fix this?


Also, it took me almost 60 days to do a full LL test and I had all 4 cores of my Core2Quad assigned to it (2.4GHz). This seems a little long or is this about right?


Thanks!

[QUOTE=fox_mccloud_123;339066]Also, it took me almost 60 days to do a full LL test and I had all 4 cores of my Core2Quad assigned to it (2.4GHz). This seems a little long or is this about right?[/QUOTE]LL tests do not use that much RAM. A test on an exponent in the 60,000,000 range is about 133 GHz-days worth of work. What else do you run on the machine? Did it do the P-1 before the L-L started?

Your day- and nighttime memory settings apply only to P-1 stage 2 and ECM stage 2 work, including P-1 done on an LL assignment prior to doing the LL itself. The settings are disregarded for all other types of work.

The client should not use anywhere near 1GB for any work other than these two.

[QUOTE=fox_mccloud_123;339066]Also, it took me almost 60 days to do a full LL test and I had all 4 cores of my Core2Quad assigned to it (2.4GHz). This seems a little long or is this about right?[/QUOTE]
[QUOTE=Uncwilly;339074]A test on an exponent in the 60,000,000 range is about 133 GHz-days worth of work. What else do you run on the machine? Did it do the P-1 before the L-L started?[/QUOTE]
One core of your quad should be able to do 133 GHz-days worth of work in around 60 days, running 24x7 & not losing much of the CPU cycles to other stuff. Roughly what size is your exponent, and how much of the time is your machine on & running prime95?

You will have better throughput with one exponent per core*. For that, the settings in Test / Worker Windows would be: Number of worker windows to run = 4, CPUs to use = 1. (Presumably you currently have those values reversed.)

* Physical core, which is all a core2quad has, since no hyper-threading.

I don't think it did any P1 or ECM testing, at least if it did I never saw any messages about it.

I understand that only P-1 and ECM testing use the higher amounts of RAM, but when I'm actually doing P1 testing, the day/night settings are ignored and it always uses the maximum amount, regardless of time of day.


I had all 4 cores assigned to the single worker, which was 55861867 (finished, now awaiting double check), finished in 61 days, and gave me 118.3961 GHz days of credit.

After that L1 finished, I set everything up differently :

Worker 1 : Trial Factoring , 1 core (estimated 2 days per output)
Worker 2 : P1 factoring, 1 core (estimated 4 days per output)
Worker 3 : World Record L1 tests , 2 core (61837319, estimated 72 days @ ~18 hours a day)


The computer is on usually about 18 hours out of the day, running prime95 all that time.
The computer is mostly used at various times through the day for small-business work (Scanning, email, and web-based work). I'd say the computer is totally idle about 8 of those 18 hours.

Bump.

I'm having this problem on other computers too, when I P-1 factor, Prime95 always uses the biggest amount of RAM I put into either category, regardless of the time.

For example, If I have it set up to be 1024MB day/ 2048MB night, it will use 2048MB all day, no matter what I put for the cut-off times.

[QUOTE=Mr. P-1;339091]Your day- and nighttime memory settings apply only to P-1 stage 2 and ECM stage 2 work, including P-1 done on an LL assignment prior to doing the LL itself. The settings are disregarded for all other types of work.

The client should not use anywhere near 1GB for any work other than these two.[/QUOTE]

Skipped over your post completely on accident.

I see what your talking about. I went back and looked and you're right. I just misinterpreted that the memory limits applied to stage 1 as well. My bad!

Thanks for clearing that up for me!