prime small FFTs = ok. blend = 124 BSOD in 2 minutes

[Deders]

I do have VCCIO - which from what I read "boosts the signal from CPU to the memory" (in very laymans terms), so based on that it seems like it could be relevant, but I also read about it that stock values should be more than sufficient up to 2000++ Mhz memory (and I'm only running 1333Mhz), so I wouldn't think it should be necessary.

That said I have only barely tried fiddling with that value - and only up as much as to 1.1 (from 1.05 stock). The range for stable operation apparently isn't very wide on this, and I read that no more than 1.2 is recommended for extreme memory bandwith overclocks.

I'm a little sceptical that this will help me - but suppose I could give it another whirl and see if blend is stable at a notch or two less Vcore than what I have established it needs to complete an 6-hour overnight (if it has no effect it will probably crash in less than an hour, so not so difficult to test). Do you have any recommendations on how high to test setting it though?

-Stigma

Keep going until you hit 1.2 if you feel that is the highest you can safely go. It's the same as VTT on my lynnfield and was essential for me to get a stable overclock. When I changed ram I had to up it again ever so slightly.

I would definitely do at least 24 hours, I've seen it fail well into the 23rd hour. One notch up on the VTT and it was fine.

 

[flexy]

Memtest, the one from http://www.memtest.org/ which you put on a USB or DVD and then boot onto and test memory is "relatively" good test to check whether your memory is OK but it's not really a good "stability test".

For me, this memtest is only the FIRST stage to make sure essential stuff is ok, so you can boot into Windows safely before you do "real" mem testing. (For this you should use HCI's memtest for Windows or even OCCT large).

It can well be you run memtest from a stick fine but it fails WITHIN WINDOWS stress-testing once there is more strain on the entire system.

TLDR: Bootable memtest is not really a good "stability test"

 

[the_stigma]

Well, short status update:

I tried my hand at OCing via offset(+). It didn't take me long to find a good curve that seems to be stable at both idle, light and max load (prime blend).

I'm blend stable @ 46x100, LLC medium, actual Vcore 1.432-1.440, idle around 1.000-1.050'ish (hard to get exact reading)

I am still performing a long-test for this (about 1hr 30 min so far) but...
I also suspect to be blend stable @ 47x100 LCC medium, actual Vcore 1.472-1.480, idle 1.000-1.0050'ish

The thing is - while I am hesitant to run 1.450+ Vcore all the time, when I am using the offset method my CPU will very rarely be 4x100% loaded in normal use (and even when it is - not as heavily stressed as in prime95) - so I should be operating at significantly lower Vcore 99% of the time. That makes me consider accepting the current 47x100 settings - assuming they can prove stability in an overnight-prime95-blend.

Thoughts on this? It's a bit high but should be ok no? My temps peak at 75C in prime95 (realtemp) and I'm not even running full fan speed yet because I'm waiting for a PWM controller. When I have that I'll set up a curve that won't let it pass 70C (not that I expect any normal application to push that high).

For reference again, this is a 2500K Sandy Bridge. I've read that 1.425 well within 24/7 safe limits given adequate cooling (source: Asus's Juan Jose tech guy). I don't think in actual use the current settings will dip much past that point except perhaps in tiny peaks. (I intend to do a little testing - play a few games and log the Vcore and see if that assumption holds true). Any relevant feedback is welcome even though I know there is no "correct" answer to this question.

EDIT: ok so my hypothesis was a bit off. I played an hour of Planetside2 (very CPU limited) and the Vcore stayed near or at max most of the time, so even though it was very stable in practice at 47x that's too high for me to have peace of mind for daily use It wouldn't be such a problem if i lost this old 2500K. It's not new expensive hardware or anything, but I think I'll settle for 46x and have some peace of mind instead. 100hz less isn't going to kill me. I'll see how low I manage to stabilize it, but it should be barely above the 1.425V that reputable people have recommended as a general guideline.

EDIT: By the way - while graphing the Vcore I noticed that HWinfo only detected something like 1.370'ish Vcore consistently (while at the same time CPU-Z was showing 1.480 under load). Is one or the other more trustworthy, and is there a way to see Vcore more accurately?

-Stigma