Prime95 and 4GB ram system

Bourni · 2007-06-27, 07:11

How can I fully test my ram using the new version (25.3) of Prime95? Even in custom test, no matter what amount of memory I set, Prime95 just stresses about 75% of my 4GB ram system and a 100% of each core (Intel E4300 dual core CPU). These are the values seen in a Windows multimeter program and running blend torture test.
In the task manager, prime95 is using about 2,67GB of memory and a 100% of CPU. But in the Processes tab of task monitor, it's like 80~90% of CPU usage and 1.625.268k of memory usage (it doesn't vary too much).
This just happens when I test 4G of ram system. For the same system, with 2GB of ram, everything runs fine.
But I'm receiving errors after 6, 8 and even 10 hours. I wonder if I have all my memory been tested by Prime95 these tests would be shorter since errors would occur faster.
If I run 2 instances of old Prime95 at the same time, I can see my memory fully stressed but I can even move the mouse...

So please tell me what is the right way to run Prime95 in a dual core CPU in a system with 4Gb ram?

Thanks

Edited: I tried to run Orhos custom test with ffts between 1024 and 4096k and memory to use in 1800MB while running prime95's blend test and I can see 99% of my system memory beeing tested. Again, I could hardly open an internet page, and I don't know if one program would interfere with other and errors would be normal. In this case if errors happens in one program or the other, it means that my system is instable or erros runnning Prime95 blend test and Orthos at the same time are expected to happen? I don't know what to do...

Jamiaz · 2007-06-27, 14:17

Unfortunately you will probably not be able to use the memory between 3gb and 4gb due to how the memory system was set up a long time ago. That space is reserved for graphics memory and other system stuff. It is sort of like the space betwwen 640kb and 1 MB. Here is an article about I saw recently.

http://www.dansdata.com/askdan00015.htm

Matt

Cruelty · 2007-06-27, 14:20

Are you using 32bit OS? If yes, then AFAIR there is a memory limit for Prime95 @ 1.6GB. The only way to test larger amount of memory is to use 64bit OS and also 64bit Prime95.
BTW: if you insist on not being able to move your mouse during stress-testing then rise the priority of Prime95 to 10 (realtime)

Bourni · 2007-06-27, 16:53

Thanks for your replies. I know about the 32bit system/3~3,5Gb limitation, but I would like to test the larger amount of memory possible. In my case, can I run two instances of Prime95 (one blend and the other custom, for instance)? Is it worth to do this or do you guys mean that Prime95 and other stressing componentes programs are not trustable for testing 4G of memory in a 32bit OS? I mean, if errors occur or not, it doesn't mean my system is stable... That's the major question...

Because if I just run one instance of prime95 (blend), it shows 75% of memory usage in the task manager. If I open one more custom instance of Prime (the old one) and set 600MB of memory, I can see my memory usage in 92% wich is good since I can work on my PC while the tests run and testing 92~95% of the amount of memory that I can test is better than testing just 75%. Am I wrong?

VBCurtis · 2007-06-27, 19:06

Google/get memtest86. Prime is excellent for testing CPU loads and heat tolerance, but does not directly test memory stability. memtest fits on a floppy, and will test all possible memory usage methods.
-Curtis

Bourni · 2007-06-27, 19:42

Quote:

Originally Posted by VBCurtis

Google/get memtest86. Prime is excellent for testing CPU loads and heat tolerance, but does not directly test memory stability. memtest fits on a floppy, and will test all possible memory usage methods.
-Curtis

I already did it Curtis... But with another memory settings and more overclocked system. I didn't spend a lot of time in memtest... I tested the memory (just test 5 and 8 because I read that these 2 are the ones that most stresses memory) for 2 hours more or less with no errors. And it's weird because it tooks a lot of time to finish the steps. At least in test number 8, it simply doesn't move... I took off two sticks and tested the other two and then it was fast again.
I put back the other sticks and ran two instances of old Prime95 and I received an error fast. I concluded that it's better to run Prime95, but maybe I was wrong. I can give it another try. But it's sloooooooow dude...
How many time should I test in memtest 86 and should I choose an specific test number?
Thanks

VBCurtis · 2007-06-28, 17:52

When I get an overclocking error I fear may be memory-related, I let memtest run all day while I'm at work.

Note most DDR2 intel boards need more relaxed memory settings with 4 sticks populated instead of 2. Try slowing down every memory timing by 1 step (if you're at 4/4/4/12, go to 5/5/5/13 or 5/5/5/14) for 4 sticks, see if your crashes cease.

If you have no errors with 2 sticks, it is almost certainly a case of too-tight timings for 4 sticks. Another path would be to go to the loosest settings, get stable, then see which gets you more speed: more CPU speed or tightening memory timings.
-Curtis

japelprime · 2007-07-02, 11:08

I have one question related to this subject.

If I have 1 GB memory but choose to give prime 95 2GB available memory what will habbend? I am wondering if the program will try to swap the extra 1 GB memory to the disk.

I am asking this because I have 16 Gb Solid state disk that is faster than ordinary hard disk and maybe can be used as extra (swap) memory.

sdbardwick · 2007-07-02, 11:26

If that is a flash based SSD, do not use it for a swap or paging file. Flash can only tolerate a finite number of writes, and heavy swap/page file use will dramatically shorten the life of the SSD.

Given that warning, what you propose is possible, but not practical. The speed of standard storage interfaces is slower than system RAM by at least an order of magnitude; flash is even slower (although that can be mitigated by clever engineering - think of an internal RAID 0 for example).

Cruelty · 2007-07-02, 12:27

Quote:

Originally Posted by sdbardwick

Flash can only tolerate a finite number of writes

I think this is true for any existing mass-storage device regardless of technology being used

hhh · 2007-07-04, 09:25

Quote:

Originally Posted by japelprime

I have one question related to this subject.

If I have 1 GB memory but choose to give prime 95 2GB available memory what will habbend?

I think remember that prime95 can't be fooled and takes no more than q<100% of your memory. H.

2007-06-27, 07:11	#1
Bourni Jan 2005 1011₂ Posts	Prime95 and 4GB ram system How can I fully test my ram using the new version (25.3) of Prime95? Even in custom test, no matter what amount of memory I set, Prime95 just stresses about 75% of my 4GB ram system and a 100% of each core (Intel E4300 dual core CPU). These are the values seen in a Windows multimeter program and running blend torture test. In the task manager, prime95 is using about 2,67GB of memory and a 100% of CPU. But in the Processes tab of task monitor, it's like 80~90% of CPU usage and 1.625.268k of memory usage (it doesn't vary too much). This just happens when I test 4G of ram system. For the same system, with 2GB of ram, everything runs fine. But I'm receiving errors after 6, 8 and even 10 hours. I wonder if I have all my memory been tested by Prime95 these tests would be shorter since errors would occur faster. If I run 2 instances of old Prime95 at the same time, I can see my memory fully stressed but I can even move the mouse... So please tell me what is the right way to run Prime95 in a dual core CPU in a system with 4Gb ram? Thanks Edited: I tried to run Orhos custom test with ffts between 1024 and 4096k and memory to use in 1800MB while running prime95's blend test and I can see 99% of my system memory beeing tested. Again, I could hardly open an internet page, and I don't know if one program would interfere with other and errors would be normal. In this case if errors happens in one program or the other, it means that my system is instable or erros runnning Prime95 blend test and Orthos at the same time are expected to happen? I don't know what to do... Last fiddled with by Bourni on 2007-06-27 at 07:42

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Problems with Prime95 28.9 on my old Phenom II x6 System	Rhanceed	Information & Answers	24	2016-06-05 15:28
Prime95 exhausting system in Vista	Rackers	PrimeNet	1	2008-03-08 20:04
Prime95 system freeze	coolul007	Software	1	2007-06-20 09:03
Running prime95 and NFSNET together on a HT enabled system	TauCeti	NFSNET Discussion	1	2003-07-02 16:26
prime95 on a dual boot system	hmeyer	Software	1	2003-01-14 22:30

2007-06-27, 14:17	#2
Jamiaz Sep 2005 Detroit, MI 2³ Posts	Unfortunately you will probably not be able to use the memory between 3gb and 4gb due to how the memory system was set up a long time ago. That space is reserved for graphics memory and other system stuff. It is sort of like the space betwwen 640kb and 1 MB. Here is an article about I saw recently. http://www.dansdata.com/askdan00015.htm Matt

2007-06-27, 14:20	#3
Cruelty May 2005 65C₁₆ Posts	Are you using 32bit OS? If yes, then AFAIR there is a memory limit for Prime95 @ 1.6GB. The only way to test larger amount of memory is to use 64bit OS and also 64bit Prime95. BTW: if you insist on not being able to move your mouse during stress-testing then rise the priority of Prime95 to 10 (realtime)

2007-06-27, 16:53	#4
Bourni Jan 2005 11 Posts	Thanks for your replies. I know about the 32bit system/3~3,5Gb limitation, but I would like to test the larger amount of memory possible. In my case, can I run two instances of Prime95 (one blend and the other custom, for instance)? Is it worth to do this or do you guys mean that Prime95 and other stressing componentes programs are not trustable for testing 4G of memory in a 32bit OS? I mean, if errors occur or not, it doesn't mean my system is stable... That's the major question... Because if I just run one instance of prime95 (blend), it shows 75% of memory usage in the task manager. If I open one more custom instance of Prime (the old one) and set 600MB of memory, I can see my memory usage in 92% wich is good since I can work on my PC while the tests run and testing 92~95% of the amount of memory that I can test is better than testing just 75%. Am I wrong?

2007-06-27, 19:06	#5
VBCurtis "Curtis" Feb 2005 Riverside, CA 2³×661 Posts	Google/get memtest86. Prime is excellent for testing CPU loads and heat tolerance, but does not directly test memory stability. memtest fits on a floppy, and will test all possible memory usage methods. -Curtis

2007-06-28, 17:52	#7
VBCurtis "Curtis" Feb 2005 Riverside, CA 2³·661 Posts	When I get an overclocking error I fear may be memory-related, I let memtest run all day while I'm at work. Note most DDR2 intel boards need more relaxed memory settings with 4 sticks populated instead of 2. Try slowing down every memory timing by 1 step (if you're at 4/4/4/12, go to 5/5/5/13 or 5/5/5/14) for 4 sticks, see if your crashes cease. If you have no errors with 2 sticks, it is almost certainly a case of too-tight timings for 4 sticks. Another path would be to go to the loosest settings, get stable, then see which gets you more speed: more CPU speed or tightening memory timings. -Curtis

2007-07-02, 11:08	#8
japelprime "Erling B." Dec 2005 91₁₀ Posts	I have one question related to this subject. If I have 1 GB memory but choose to give prime 95 2GB available memory what will habbend? I am wondering if the program will try to swap the extra 1 GB memory to the disk. I am asking this because I have 16 Gb Solid state disk that is faster than ordinary hard disk and maybe can be used as extra (swap) memory.

2007-07-02, 11:26	#9
sdbardwick Aug 2002 North San Diego County 2³·7·13 Posts	If that is a flash based SSD, do not use it for a swap or paging file. Flash can only tolerate a finite number of writes, and heavy swap/page file use will dramatically shorten the life of the SSD. Given that warning, what you propose is possible, but not practical. The speed of standard storage interfaces is slower than system RAM by at least an order of magnitude; flash is even slower (although that can be mitigated by clever engineering - think of an internal RAID 0 for example).