Memory Timing + Stress Testing Questions

Bman. · Nov 10, 2008

Good morning all-

My machine is currently chugging along on it's 10th hour of Prime95 blend testing with no problems. I did the pencil mod for the P5K modo on the vDroop and that appears to have really helped the cause.

In trying to determine what was keeping me from unstable with a 380x9 @ v1.45 / ram at 2.1v and NB 1.4 - I have either the pencil mod or RAM timings to hang my hat on. I also loosened my timings from 4 4 4 12 to 5 5 5 15.

Question #1 - if I run timing loose at 5 5 5 15 and the machine is stable what exactly is lost with lose timings? I would like to tighten them up but I am not sure what that gains me if the machine is rocking at along at good overclock.

Question #2 - I have been running Prime95 blend test and all is well. I have also been running OCCT also no problem. I guess my question is will OCCT test the stability of my overclock as well as Prime95? I really like the 1 hour test versus the 24 hour test - however, I am not sure they are testing the same levels of stability. If they were, how come everyone wouldn't just use OCCT.

I read a thread on here about stress testing (can't find it now?) and it noted that OCCT was good but that was really about all it said.

Question #3 - What is the best way to measure vDrop / vDroop? I have been monitoring with CPUz (Core Voltage) but I am not positive that is telling me what I should be looking for with regards to vDroop.

As always - thanks!

Benjamin

ObiWong · Nov 10, 2008

there is no real life difference between 5-5-5-15 and 4-4-4-12
if you can some memory benches you'd probably see a slight increase in bandwidth. when i was playing with timings and frequencies i would run the memory benchmark in Everest.

from my own experience, the best way to observe vdrop is to note what vcore you have set in your BIOS then compare the voltage in CPU-Z.
CPU-Z will also show vdroop when your system is at load

jason4207 · Nov 10, 2008

Bman. said:
Good morning all-

My machine is currently chugging along on it's 10th hour of Prime95 blend testing with no problems. I did the pencil mod for the P5K modo on the vDroop and that appears to have really helped the cause.

In trying to determine what was keeping me from unstable with a 380x9 @ v1.45 / ram at 2.1v and NB 1.4 - I have either the pencil mod or RAM timings to hang my hat on. I also loosened my timings from 4 4 4 12 to 5 5 5 15.

Question #1 - if I run timing loose at 5 5 5 15 and the machine is stable what exactly is lost with lose timings? I would like to tighten them up but I am not sure what that gains me if the machine is rocking at along at good overclock.

Question #2 - I have been running Prime95 blend test and all is well. I have also been running OCCT also no problem. I guess my question is will OCCT test the stability of my overclock as well as Prime95? I really like the 1 hour test versus the 24 hour test - however, I am not sure they are testing the same levels of stability. If they were, how come everyone wouldn't just use OCCT.

I read a thread on here about stress testing (can't find it now?) and it noted that OCCT was good but that was really about all it said.

Question #3 - What is the best way to measure vDrop / vDroop? I have been monitoring with CPUz (Core Voltage) but I am not positive that is telling me what I should be looking for with regards to vDroop.

As always - thanks!

Benjamin

1) You will get better bandwidth w/ tighter timings. Tweaking the RAM can help, and it is fun, but the real-world gains are very small compared to CPU OCing, and GPU OCing.

2) I like to use P95 small fft overnight to stress the CPU, OCCT2.0 custom 2hr RAM test to stress the FSB/RAM, and memtest86+ to also test the RAM. How long you test is up to you, but I like to pass the above, and then drop the FSB a few more MHz to give a little bit of cushion.

3) vdrop is the difference b/n the BIOS setting, and what you see in CPU-Z while idle...should be easy enough for you to figure out. Vdrop really has no bearing on anything. Just adjust your BIOS setting to compensate for it. Vdroop is even easier...just watch CPU-Z at idle, and then start up P95 small fft. The difference in observed vcore is vdroop.

deeppow · Nov 10, 2008

Your questions have been addressed well above.

I would just add that I find IntelBurnTest as the ultimate stability test for me. My system passed hours of Prime testing and then failed the load provided by IBT. And you don't need hours of testing with it, much quicker indication of stability. Anyway, another tool to add to your tool chest. :beer:

Bman. · Nov 10, 2008

Thanks for the great responses guys. I will take it all into account. It appears that I am still getting vDroop but I will have to look at it a little more closely. The vCore in BIOS versus CPU-z is in the range of .075v That seems rather large but probably in line with the mobo.

If I don't change anything on the RAM - say the timing (boxed it's 4-4-4-12)...currently set for 5-5-5-15... is there any reason to test it's stability independently? I guess I am unclear how much the FSB plays a roll in the stability of the RAM as an separate function. This is my ignorance to understanding how the FSB translates beyond the CPU.

Benjamin

jason4207 · Nov 11, 2008

To test RAM independently use a moderate FSB that is very stable, and use memory dividers to get higher RAM speeds. Also, keep your CPU multi low to keep the CPU from interfering.

I've passed P95 small fft over night, but then had the OCCT test mentioned above fail almost instantly. If your RAM is loose w/ adequate voltage then you probably don't need to battery it w/ tests, but the FSB is related to the RAM and CPU, and should be tested...especially w/ quad cores where the FSB limits tend to be pushed more.

Memory Timing + Stress Testing Questions

Bman.

Registered

ObiWong

Member

jason4207

Senior Member

deeppow

Senior Member

Bman.

Registered

jason4207

Senior Member

Similar threads