View Full Version : Mem testing
negev
10-09-2008, 03:13 PM
Other than superpi, what is a reliable (and fast) test for memory stability? I know my cpu is stable at the current settings so I don't really want to go through another 24-hour prime session.
How reliable is superpi at detecting memory errors? Should I run it several times to be sure? Or is there something better?
negev
10-09-2008, 03:49 PM
Is this checksum calculator accurate?
http://www.xtremesystems.com/pi/
My checksums fail even on 5-second pi runs, but the ram is blend stable (unless Aegis killed it :/ )
negev
10-09-2008, 04:05 PM
Nope its definitely the site thats wrong, tested my laptop which is all at stock
HDCHOPPER
10-09-2008, 04:23 PM
over at Major Geeks thiers a lot of ram checkers to download most are free too
saaya
10-12-2008, 08:53 PM
the checksum is to make sue you didnt cheat or manipulate the application, but its been broken for a long time already :D
heres my experience from binning hundreds of kits per day at cellshock where i worked before :D
orthos and memtest 86+ test5 will both give you about the same result.
orthos stresses chipset a bit more than memtest
spi 32m is pretty useless, dual pi 32m makes sense and is about as usefull and reliable as running orthos or memtest 86 test5 for the same amount of time it takes to get dual 32m calculated.
test 3d! it stresses the chipset and memory in a different way (vgas use the systems main memory to store quite some stuff and access it agressively)
i would highly recommend you and others to use an approach for testing the max mem speed.
dont try to find the exact max stable speed, you can spend years on that... :D
if x mhz is stable for x minutes/x runs in x benchmark/tool, then you can be pretty sure that going down 25mhz or 50mhz for the memory will mean its 100% stable. at least in that test. 3d usually needs you to drop 25mhz compared to 2d to get the same stability.
10 minutes testing is usually enough already to get a quick idea of what the mem can do.
if 10mins x is stable drop the speed by 25-50mhz and you should be rockstable...
hope this helps :toast:
HDCHOPPER
10-12-2008, 09:39 PM
thanks good info from a memory man
raptor1
10-15-2008, 04:26 AM
spi 32m is pretty useless, dual pi 32m makes sense and is about as usefull and reliable as running orthos or memtest 86 test5 for the same amount of time it takes to get dual 32m calculated.
hey saaya
when you say dual pi 32m, do you mean that two copies of super pi 32m be run at the same time
raptor1
BenchZowner
10-15-2008, 04:41 AM
hey saaya
when you say dual pi 32m, do you mean that two copies of super pi 32m be run at the same time
raptor1
Yes.
That's what Saaya means.
Open two or four ( depending on your CPU core count ) instances of SuperPi, set the CPU affinity of each instance to 1 core ( core 0 for the first instance, core 1 for the second instance, core 2 for the third instance, core 3 for the fourth instance, etc ) and start them simultaneously.
Remember, you have to "setup" SuperPi in 4 different folders and launch each exe separately to work.
saaya
10-16-2008, 10:13 PM
yepp, :D
quite annoying to do for 4, i always just used 2 ^^
is there no script or tool to do this for 4 cores?
i LOVE orthos launching as many prime95 threads as you want and your core can support :D
BenchZowner
10-17-2008, 12:56 AM
Is there no script or tool to do this for 4 cores?
Actually there is.
It's called HyperPi.
Download it here http://home.pages.at/virgilioborges/hyper_pi_0.98b.zip
For sure and you can even find all the tools that he may need in the QF key :D
HyperPI, SuperPI, memtest86+, memtest for Windows, orthos ... ;)