Originally posted by: phile
Where overclocking is concerned, each individual component has unique characteristics. From what I have observed, some E6600s can achieve very nice overclocks with little added voltage, while others - like mine - are way too hungry for volts to make extreme overclocking manageable. That being said, I highly doubt your E6600 would be load-stable at 350 with vcore at auto. Have you done any load testing? If not, grab a copy of Orthos Stress Prime (
http://sp2004.fre3.com/beta/beta2.htm) and run the
Blend test for a few hours minimum.
Well, there are some observations of my own: first, the most curious part, which is that with all the voltages except the RAM at 'auto' (CPU-Z shows 1.175V for Vcore, but see below) Orthos will, indeed, run for hours on end at 350MHz. No problem. I even ran 3DMark 06 for a few hours (on 650/900 7900GTs in SLI) with no problem. All well and good... except that as soon as I raise the clock to 360, the benchmarks start complaining and anything over 360 is not even boot-stable.
Also, if you have not run a memtest86 diagnostic (
http://www.memtest.org) on your memory, you might want to do so, just to confirm that your mem is fine.
That I didn't think of doing. I usually do, but I figured that I paid for premium RAM for a reason and the memory seems fine if underspec at 350. I think I'll give it a shot.
The bottom line: there is no guarantee that any individual combination of hardware will be able to achieve any particualar overclock. My rig (see link) should be able to do 400, in theory. However, anything beyond 333 either requires too much vcore, or simply will run stable. I have tried every combination/permutation of BIOS settings, without any success. It has been suggested by many that my CPU is the likely culprit. Since it comes from a known weak batch (week 24), I tend to agree.
I realise that, but if you read *every single* review out there, it seems that the E6600s will run 400FSB on a bad day and be MB-limited at 450 on a good one, with 400FSBs being commonplace even with stock coolers. Now, we all know perfectly well that the chips that get sent to reviewers are very, very carefully picked: there is no doubt about that and there hasn't been for at least seven years (I'm being generous). Still, even on the forums, anything below a 400FSB seems to be the exception, not the rule. Therefore, when my CPU seems to have a solid wall at 350-355, I start to suspect that I'm doing something wrong myself.
A few extra considerations:
- enabling SPD only helped marginally: I could boot at higher speeds, but XP would BSOD within 30 seconds of getting to desktop at anything over 370. 360 was still not benchmark-stable, but I could run Orthos for half an hour or so.
- Enabling Hyperpath3 creates a solid wall at 333.
- Clearly OCZ's RAM is not performing to spec: they advertise 4-4-4-12 @ 400, the RAM won't even do that at 333 AND the SPD is indeed programmed for 5-5-5-15 (the MB hten decides to do 5-5-5-20 for some reason). I am back to my (very) long held belief that premium RAM only makes sense for long-term reliability in a server environment and that there is exactly one brand to buy for that, Crucial. For the rest, the Nanya/Elpida chipped generics often are significantly better than 'enthusiast' RAM.
- There must be something going on with my Vcore: with the auto setting, I get 1.185 idle and 1.288-1.296 under load out of CPU-Z, whereas if I set the Vcore manually to anything I get 1.213V *regardless* of the setting. Now, either CPU-Z 1.38 is lieing through its teeth, or my motherboard is doing something very innovative.
BTW, CoreTemp reports the core temperatures as 58-60C under Orthos load.. indefinitely. I am using a Noctua UH-12F cooler. Curiously again, if I set the Vcore manually to something out and out ridiculous, like 1.55, the load temperature goes way up and like every Conroe I've used, as soon as CoreTemp hits 80 or above, problems surface very rapidly.
I have to assume from the above that CPU-Z is reporting the manual Vcores incorrectly. Even with this said, however, I don't understand how or why overvolting the hell out of everything won't even get me 10 extra FSB MHz. This seems inconsistent with what everyone is saying.
For addtional fun and giggles: I own another P5W, 1.02G revision, one of the early ones, with 150x BIOS. It has an E6300 in it which has been running at 333FSB for half of forever with generic RAM at 1:1 (despite the RAM's ratings, it runs 350FSB at 4-4-4-12 without skipping a beat... which the OCZ premium does not). It works very nicely up to about 340FSB, stock Vcore (which is materially higher than the 6600 for some reason). It will not go higher than that and again, changing voltages does not gain me anything.
I own two P5B Deluxes, 1.02G, very very recent (I believe like the second P5W they're november boxes), with 804 BIOS. They both have E6300s in them and they both run 333 with the generic RAM in them (this generic has ELPIDA chips, the generic on the P5W has Nanya ones). I have not tried any further OC with these, because they're secondary desktop machines.
For further reference, all these machines run in Antec cases, with fanless Noctua or ThermalRight heatsinks and with PC Power&Cooling silencer PSs. All have 7900GTs and Maxtor SATA HDDs, no other peripherals. One of the P5Ws has an X-Fi in it.
Sorry for the long post and thanks for your reply!