• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

I've lost hope in memtest.

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.
If ever my thoughts were crystalized in a statement this would be it. There is no quick fix to getting a computer overclock stable in your OS. It takes time and lots of testing

Amen. I just went through about an hour trying to explain to a new overclocker that I can't guarrantee which timings will be fastest and stablest at the maximum clock, that he will have to expirement, as all RAM is different. I gave him some pointers, and I think he'll be off to a good start :)

But, as for stability testing, I really like HL2. It's just such an INTENSE game(gameplay and system stress) it makes everything work, RAM, HDD, VGA, CPU etc. as well as its not as boring as watching 2k5 for all time ;)

The reason I believe this, is because of the physics engine it incorporates, as well as the video power it requires. The massive levels and Model sizes, as well as variables, work your RAM over. All the stress on your other comnents stresses your Mobo and your PSU.

Of course, this isn't the be all end all of all tests. It's merely a very very good stability test, as more often than not, the instability is within 20-30 minutes of playing.
 
Testing my new ram,i found out a few things I did not expect.I had reached the max stability point on my nf7s and decided to try a few of the modded bios's that so many are using to get higher fsb.They seem to change some register settings based on the DFI bios.They did get higher fsb settings but at a cost,3d stability.I also noted that it seemed much slower loading the bios and still scored about the same as the stock bios running Memtes
 
Testing my new ram,i found out a few things I did not expect.I had reached the max stability point on my nf7s and decided to try a few of the modded bios's that so many are using to get higher fsb.They seem to change some register settings based on the DFI bios.They did get higher fsb settings but at a cost,3d stability.I also noted that it seemed much slower loading the bios and still scored about the same as the stock bios running at lower fsb in memtest.The MB/s score became unstable at the same score on both bioses reguardless of fsb.
My conclusion was that the modded bios introduced more wait cycles to the system,slowing everything down including the exchange of the agp to the cpu.It is probably not as noticeable on an NF3 or nf4 platform,but on the nf2 it made 3d Mark jerkey and slow.I suppose that the memory cotroller on the nf2 is less capable then the newer platforms and slows down the agp bus to a point that 3d stability is lost.
The ability to run higher fsb at tighter ram settings seemed a waste as the lower fsb on the stock bios still got the same memory score without the extra wait cycles and is solid at all 3d tests.
 
Most likely you tried a "CPC off" bios, which is similar to running a 2T command rate on the NF3/NF4 plaforms. In a nutshell, CPC off usually allows for higher overclocks, but at the expensve of reduced bandwidth. Often times, it is recommended to stick with 1T/CPC On and see how high you can get with that :)

deception``
 
It's already been said, but i'll throw my .02 in on it as well. First off test #5 is not the only test i run, i always run #5 for at least 15 passes and then #8 for at least another 10 before i bother in windows. It has been my findings in the past that when you can pass both those tests, you're memory is going to be pretty darn stable in windows, but not always the case. As i said in my DDR600 guide testing in Memtest does not equal Windows stable. There are to many other factors that influence that for Memtest to be the final word. Also as stated PI can be a good prelim tester, however that to is heavily affected by other variables. I can't tell you how many times i've been able to loop 01' or run any other bench in the arsenal and still gotten the "not sqr" "not exact" errors in PI. I don't even bother with burn in/Prime, ect because that isn't my thing, but if you want your computer to be stable for what it is you do, i suggest you use that method to test it. If you want to calculate math for 24 hour time periods than i would suggest a nice long run of prime, if gaming is your thing than let 01' loop for a few hours. I myself wouldn't count out Memtest, but like the others have said, i too take the results with a grain of salt. That's my take anyways :)
 
Test #5 only is no good anymore, you need to add 6, 7, and 8 (at least with BH-5) you do. I can run 2x512 278mhz stable test #5, but get tons of error's in test 6, 7, 8

Ultimately 274mhz was Memtest error free with complete passes, yet only 272mhz stable in benching.
 
I spent a few dollars on Memtest Pro and it is more sensitive than memtest86 and picks up more errors.
Hey Collin, who's that guy with my girlfriend?....LOL
 
Back