Core i7/X58 Overclocking Thread

[icecold316]

I am testing 3.8ghz @ 1.27v, 1.255v QPI/VTT HT ON
I had to lower my Uncore to 17x multi ... 18x it BSOD'd with the well known "uncorrectable hardware error detected" in linx
Ram is @ 8x 1.6v 7 - 7 - 7 - 21

Little nervous to try higher, may set it to WCG for tonight see if there were any issues come morning..

Max temps went to 70C at this voltage and clockspeed

If you ram is 8x, try for 16x uncore. It should me more stable.

[JWMc]

**NOTE** interestingly, in prime blend, temps are 10 degrees cooler than LinX. I may be able to get a higher prime blend stable OC than LinX simply due to temps...

That is completely normal. LinX is proven to increase by temps by up to that much over Prime blend. However, I think the gap should be about half that between LinX and Prime small-FFT. I bet you already know this but Prime small-FFT is meant for stressing CPU only by using code that fits entirely in processor cache and hardly touches the main memory. The thing is Prime small-FFT is obsolete AFAIK for proper testing on the new i7 platform because of the new cache design. Of course it would still be a useful tool as it still generates very high loads but I don't think it's enough to max out an i7. I just did a test and I am seeing a 5C increase instantly just from flipping back and forth between Prime small-FFT and LinX with HT on @ 3.990ghz-1.312v.

LinX is simply generating the absolute highest individual CPU core loads that we have seen yet so we are still surprised at just how much hotter the i7 gets running it with 8 threads, because this processor is very good at running 8 threads at once.

[rob[GL]]

Hey guys, after solving my little bios problem, by upgrading to beta bios F5f... I finally got around to playing last night;

3.66GHz @ 1.325v (1.3500 in bios)
loadline on
20x multi
turbo off
183 bclk
HT ON
auto memory (7-7-7-20 @ 1.5v) 1066mhz
6x memory
16x NB Freq
36x QPI
qpi/vtt 1.275
ioh 1.200
pll 1.86

i ran prime95 overnight (10 hours) fully stable temps around 80c
stock cooler

i would push for 4.0ghz... but this cpu 3839A560 seems to CRAVE vcore... can post with 1.4vcore but not load windows... can load windows AND prime at 1.55v - but temps immediately hit 100c and cpu begins to throttle...

if i am going to push for 4.0GHz - i will definately need a better cpu cooler. wait for TRUE 1366 to come to stores here

maybe i would have better luck with 19x211? some people say better clock and lower voltage with 19x multi?

here's a screenshot...

[anzial]

gotta upgrade cooling it seems. In my case, 191x21 does the trick. 10hours of prime and no crashes so far

[Chicken Patty]

Have you run any stability programs?

only for short whiles.

about an hour of prime, couple of benchmarks, such as wPrime, 3dmark, etc.

Been running for 5 days already without any errors during a benchmark or anything.

[anzial]

What's considered to be stability programs now? Linx? Prime95? What are the common rules?

[Chicken Patty]

What's considered to be stability programs now? Linx? Prime95? What are the common rules?

I know a lot of people disagree with me if not all, but to me stable is a computer that doesn't crash. I haven't really stability tested my current settings, but they've been running for almost 6 days and havent had a hiccup.

If by any chance it does crash two weeks after, then whatever, its not stable, I simply make a few changes boot up again and thats it.

The difference is with running prime 95 for example, is that if there is any instability you'll find it within seconds or minutes, or at most a few hours. But you have to torture your computer, which I have done many times before, but I dont think it is necessary.


Good programs for stability testing are Prime 95, and LINx, those are the only ones i've used.

[rob[GL]]

I consider 24 hours of prime95 "stable enough" to use day-to-do.

I've had rigs before that would run prime95 for 7-8-9 hours and then mysteriously reboot.

[Chicken Patty]

I consider 24 hours of prime95 "stable enough" to use day-to-do.

I've had rigs before that would run prime95 for 7-8-9 hours and then mysteriously reboot.

my amd rig wont pass prime to save its life, but will not reboot and ran at 3.33 GHz for as long as I remembered, dont ask me how that happens.

[PytonOrm]

Got x 22 multiplier http://valid.canardpc.com/show_oc.php?id=496980 all cores & threads active

[anzial]

Got x 22 multiplier http://valid.canardpc.com/show_oc.php?id=496980 all cores & threads active

Nice! is that a new bios for your board?

[JWMc]

Got x 22 multiplier http://valid.canardpc.com/show_oc.php?id=496980 all cores & threads active

Holy cow did they add a feature to disable the requirement of only one core active for the +2 turbo multi?

I could really use that +2 multi for HT off testing. I suspect I probably won't be hitting 4.4ghz HT stable on water though.

[PytonOrm]

No mhz Gain with 22x multiplier, when i tested pumped up bclk damn... got unstable system when higher bclk than 190

[NBF]

Holy cow did they add a feature to disable the requirement of only one core active for the +2 turbo multi?

I could really use that +2 multi for HT off testing. I suspect I probably won't be hitting 4.4ghz HT stable on water though.

Enable Speedstep, Turbo and C-state and the 22x multi will show up. But as soon as you load all cores it goes back to 21x. It can be usefull for some single thread benchmarks though.

[Pyr0]

Holy cow did they add a feature to disable the requirement of only one core active for the +2 turbo multi?

I could really use that +2 multi for HT off testing. I suspect I probably won't be hitting 4.4ghz HT stable on water though.

i doubt it, the 22 multi is prolly only active on the first core

[NBF]

No mhz Gain with 22x multiplier, when i tested pumped up bclk damn... got unstable system when higher bclk than 190

Seems very unstable the 22x, got it to 206 but not valid.

http://valid.canardpc.com/show_oc.php?id=497060

[SteveRo]

SteveRo, for testing use super pi 32M, 1M is too small a test to test system memory bandwidth and processor

Super pi 32M added - new table below.
Also revised config #1 to be 8-8-8 due to the need for added stabilty.
Interesting that the super pi 32M score for config #2 beat config #1, I wouldn't of expected that.

[SteveRo]

Holy cow did they add a feature to disable the requirement of only one core active for the +2 turbo multi?

I could really use that +2 multi for HT off testing. I suspect I probably won't be hitting 4.4ghz HT stable on water though.

you can get 4.4 stable on water if you turn ht off - done it, see my sig.
yes, cheating, i know.

[jagt]

So how are the newer batches compared to the older? I'm about to take the plunge on an i7 upgrade and I just hope not to end up with a DUD that stops at 3.6GHz.

[SteveRo]

Ok so after adding the 1x3gb DDR3 2000/C9 memory - I have a new 3DMark06 max. This with the 965 @ 4.433 HT off and 1x9800gx2 @ 745/1103/1862

[JWMc]

So how are the newer batches compared to the older? I'm about to take the plunge on an i7 upgrade and I just hope not to end up with a DUD that stops at 3.6GHz.

We still need a lot more data but it looks like people are finding good chips in just about all batches. Some batches more often than others though so we need more data points.

I just made a spreadsheet that anyone can submit data to without a security check. If enough people start using it and you want to help with cleaning bad data out of it (or re-wording/correcting for clarity) I will add you to the access list to modify the spreadsheet directly if you PM me.

http://www.xtremesystems.org/forums/...&postcount=262

[Classic Satch]

you can get 4.4 stable on water if you turn ht off - done it, see my sig.
yes, cheating, i know.

I don't consider that to be cheating. I have turned off HT because it's utterly useless. I don't use a single program that would utilize it and I can keep my 920 at 4.0Ghz with only 1.31v.

[anzial]

Even if one program can't use HT, it doesn't mean it's not useful. IMHO, HT is a nice feature to have rather than to give up for a higher OC. Of course, I can run 4ghz with HT on with vcore1.31, so I guess I'm a bit biased

How many people oc i7 and keep speedstep/powersaving features on? Is there an extra stress on CPU to keep those on?

[GenTarkin]

I don't consider that to be cheating. I have turned off HT because it's utterly useless. I don't use a single program that would utilize it and I can keep my 920 at 4.0Ghz with only 1.31v.

HT sure is useful in WCG crunching =) lol, I just joined that and yeah....I game while its doing that in background. The i7 is simply a monster. No slowdowns in games while WCG is goin woot! 8 threads of goodies!

[Pisklink]

Today I started OC'ing my new hardware. Besides the first results (which look promising, I gues) I also bumped into some questions.

Questions:
As you can see I'm using Asus P6T mobo but there are some settings in bios which I'm not sure of what to select. These are the following items and settings I applied:

C1E Support: disabled
Hardware prefetcher: disabled
Adjecent cache line prefetch: disabled
Intel Virtualization tech: disabled
CPU TM function: enabled
Execute-disable bit: enabled
Intel HT technology: disabled
Active processor cores: all
A20M: disabled
Intel speedstep tech: disabled
Intel turbomode tech: disabled
Intel C-state tech: disabled

- Did I apply the correct settings on overall?
- Is it possible to enable a power consumption saving setting (like speedstep for instance) while overclocking?
- Is it possible to use turbomode on P6T when using 190+ bclk?

First results
After a lot of reading I applied some reference voltages in the bios and started OC'ing right away. It seems to be stable right away. I'm not used to that after switching from P5E3-Premium to this board. These results are without HT enabled. But is there much more vcore necessary when enabling HT? I think my temps are good and still have some potential left. Nonetheless I would like to know your opinion on these results and if possible give some advice for improvement.

Settings:
MP: 20x
BCLK: 200
DRAM: 1200
Uncore: 3200
QPI: 3600
Vcore: 1.35
CPU PLL: 1.92
QPI/DRAM: 1.35
CPU spread spectrum: disabled
PCIE spread spectrum: disabled

Everything else on AUTO