some questions of mine, in general

jerico2day · 2005-03-30, 07:18

Hello,

I've been running prime95 the last couple days and I've some questions:)

My first one, and I'm all for advancement of science and whatnot, but what's the purpose of finding these numbers? Is it just for kicks, or will it aid in the discovery of .. stuff? I'm all for it even if it's just for kicks, just curious really.

Second question, if I run the program below the recommended RAM recomendations, what's effected? Is it just crazy slow?

Thirdly, what does this mean exactly (in my status on prime95)

days
run / to go / exp
1.0 15.0 75.0

I assume it means, how long I've been running, when I'm expected to finish the current date... and exp I'm not sure of:/

If exp is my expected finish date, why so high? I'm running at 2.75ghz. Does the test rely on fast fsb, or simply clock speed? I'm only running at about 640fsb right now.. Even at that though, 75 days seems awfully high.

garo · 2005-03-30, 07:27

There are several reasons to look for large primes. Kicks yes, bragging rights yes, a way to expedite algorithmic improvements etc. If you do a search on this forum you will find some more reasons. Do a search for "reasons GIMPS".

RAM does not have any effect on the speed of the program. If you have less RAM, one non-critical stage of the test is skipped. This is called P-1 factoring and is used to find factors before the LL test is run. If a factor is found, we know the number cannot be prime thus saving us a long primality test. If you don't run the P-1 test, you check the number primality the hard way. On the whole it means a drop of 1% or less in efficiency.

The third number is days to expiry. This is the number of days within which your computer must report progress to the Primenet server or else the Primenet server will assume that you are not working on the number any longer and assign it to someone else. exp=expiry!

jerico2day · 2005-03-30, 07:56

*phew* was worried about the 75 days thing:P

Okis... so I suppose my last question unanswered... does the program function better with a higher clock rate, or fsb rate?

For example, I can run my pc stable at about 2733mhz with 201mhz to the ram (and back) which yeilds better results in memory benchmarks, or I can run at 3009mhz with only 177mhz (1:1 mem ratio) which results in slightly worse memory benchmarks (only in the neighborhood of 100mb/s on the read speed, or about 3% decrease in performance)

hmm.. calculating that (only if prime95 runs faster at higher mem bus speeds), my current expected finish time would add 10 hours to the total. However, if prime95 depends on processor speeds alone, I would decrease my time by about.. 32 hours!

So I guess this is a good question:)

[edit] I calculated the time by my current memspeed (~2800 or somewhere like that, subtracting 100, and dividing, coming up with 3%. As for proc speed, did the same, coming up with 9%. Taking 15 days times 24 hours, and multiplying by the percentages gave me 10 hours and 32 hours, respectively.

garo · 2005-03-30, 08:18

Use prime95's inbuilt benchmark function to find out which is faster. You can find it in Options->Benchmark. You must stop the test before being allowed to run the benchmark.

Also, when overclocking run the torture test for at least 24 hours before you start testing the numbers as prime95 is very sensitive to errors and if there is one single error in the computation due to an excessive overclock, the entire 15 days of computation will be wasted.

jerico2day · 2005-03-30, 08:23

ok, Ill check tomorrow and report my findings, should be useful for the next person to come along:)

just to be sure, this benchmark will directly coincide with the time of my ll tests?

dave_0273 · 2005-03-30, 09:19

The benchmark tells you (on average) how long each iteration of the LL test will take, for each different FFT size.

Therefore, whichever one gives the fastest (ie lowest) benchmark (for the FFT size that you are using) will complete the LL test the fastest.

All you have to do to work out how long the test will take is multiply the iteration time by the exponent.

However, if you are planning on overclocking, you really need to run the torture test to make sure that your computer can handle it. As garo said, one error anywhere in the test will mean that the whole LL test is wrong.

2005-03-30, 07:18	#1
jerico2day Mar 2005 2² Posts	some questions of mine, in general Hello, I've been running prime95 the last couple days and I've some questions:) My first one, and I'm all for advancement of science and whatnot, but what's the purpose of finding these numbers? Is it just for kicks, or will it aid in the discovery of .. stuff? I'm all for it even if it's just for kicks, just curious really. Second question, if I run the program below the recommended RAM recomendations, what's effected? Is it just crazy slow? Thirdly, what does this mean exactly (in my status on prime95) days run / to go / exp 1.0 15.0 75.0 I assume it means, how long I've been running, when I'm expected to finish the current date... and exp I'm not sure of:/ If exp is my expected finish date, why so high? I'm running at 2.75ghz. Does the test rely on fast fsb, or simply clock speed? I'm only running at about 640fsb right now.. Even at that though, 75 days seems awfully high.

2005-03-30, 07:56	#3
jerico2day Mar 2005 2² Posts	phew was worried about the 75 days thing:P Okis... so I suppose my last question unanswered... does the program function better with a higher clock rate, or fsb rate? For example, I can run my pc stable at about 2733mhz with 201mhz to the ram (and back) which yeilds better results in memory benchmarks, or I can run at 3009mhz with only 177mhz (1:1 mem ratio) which results in slightly worse memory benchmarks (only in the neighborhood of 100mb/s on the read speed, or about 3% decrease in performance) hmm.. calculating that (only if prime95 runs faster at higher mem bus speeds), my current expected finish time would add 10 hours to the total. However, if prime95 depends on processor speeds alone, I would decrease my time by about.. 32 hours! So I guess this is a good question:) [edit] I calculated the time by my current memspeed (~2800 or somewhere like that, subtracting 100, and dividing, coming up with 3%. As for proc speed, did the same, coming up with 9%. Taking 15 days times 24 hours, and multiplying by the percentages gave me 10 hours and 32 hours, respectively. Last fiddled with by jerico2day on 2005-03-30 at 08:02 Reason: added my maths

2005-03-30, 08:23	#5
jerico2day Mar 2005 2² Posts	ok, Ill check tomorrow and report my findings, should be useful for the next person to come along:) just to be sure, this benchmark will directly coincide with the time of my ll tests? Last fiddled with by jerico2day on 2005-03-30 at 08:25

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Computer isn't mine...	CuriousKit	PrimeNet	25	2016-06-02 05:44
gpu72 site - exp 78227507 credited but not mine?	dh1	GPU to 72	1	2015-11-29 14:03
Oh Brother, What betid to mine Haswell 4770?	petrw1	PrimeNet	70	2014-10-10 22:53
General Questions	pastcow	Factoring	10	2013-02-27 07:01
Your end or mine?	davieddy	Lounge	0	2011-12-11 11:31

2005-03-30, 07:27	#2
garo Aug 2002 Termonfeckin, IE 2²×691 Posts	There are several reasons to look for large primes. Kicks yes, bragging rights yes, a way to expedite algorithmic improvements etc. If you do a search on this forum you will find some more reasons. Do a search for "reasons GIMPS". RAM does not have any effect on the speed of the program. If you have less RAM, one non-critical stage of the test is skipped. This is called P-1 factoring and is used to find factors before the LL test is run. If a factor is found, we know the number cannot be prime thus saving us a long primality test. If you don't run the P-1 test, you check the number primality the hard way. On the whole it means a drop of 1% or less in efficiency. The third number is days to expiry. This is the number of days within which your computer must report progress to the Primenet server or else the Primenet server will assume that you are not working on the number any longer and assign it to someone else. exp=expiry!

2005-03-30, 08:18	#4
garo Aug 2002 Termonfeckin, IE 101011001100₂ Posts	Use prime95's inbuilt benchmark function to find out which is faster. You can find it in Options->Benchmark. You must stop the test before being allowed to run the benchmark. Also, when overclocking run the torture test for at least 24 hours before you start testing the numbers as prime95 is very sensitive to errors and if there is one single error in the computation due to an excessive overclock, the entire 15 days of computation will be wasted.

2005-03-30, 09:19	#6
dave_0273 Oct 2003 Australia, Brisbane 2·5·47 Posts	The benchmark tells you (on average) how long each iteration of the LL test will take, for each different FFT size. Therefore, whichever one gives the fastest (ie lowest) benchmark (for the FFT size that you are using) will complete the LL test the fastest. All you have to do to work out how long the test will take is multiply the iteration time by the exponent. However, if you are planning on overclocking, you really need to run the torture test to make sure that your computer can handle it. As garo said, one error anywhere in the test will mean that the whole LL test is wrong.