Hardware Canucks

Hardware Canucks (http://www.hardwarecanucks.com/forum/)
-   Overclocking, Tweaking and Benchmarking (http://www.hardwarecanucks.com/forum/overclocking-tweaking-benchmarking/)
-   -   My i7 Overclocking Experience - Air Cooled (http://www.hardwarecanucks.com/forum/overclocking-tweaking-benchmarking/20474-my-i7-overclocking-experience-air-cooled.html)

keto July 17, 2009 08:25 PM

My i7 Overclocking Experience - Air Cooled
 
If there's interest, I'll continue detailing my ongoing overclocking trials.

-i7 920 C0 (HS NOT lapped)
-True 120 Black (NOT lapped) c/w 2@1600 cfm in push pull, open case, no other case fans, ambient typically 23C in air conditioned house.
-Gigabyte EX58-UD5
-6GB OCZ DDR3 14400 PC1800, triple channel configuration
-Sapphire 4890 1GB OC Edition
-Seagate 500GB HD
-PCP&C 750 Crossfire PSU

As I sit here, I am at 19x200 = 3.8, vcore 1.3650, CPU PLL 1.880
-Load Line disabled
-HT OFF
-Turbo OFF
-power saving features OFF.
-RealTemp3.0 says I'm idling at 45/42/43/44.
-Full load temps with 4 (Not 8, HT is off remember) iterations blend P95 maxed @ 69/68/65/65
-RAM is 8x200 = 1600, vdimm 1.660 (no bios adjustment for spec 1.650), VTT 1.520
-Uncore 17 (RAM MULTI X2 +1) (8*2+1)
-QPI x36 (lowest in bios)

I'm not gonna flood my posts with screenies. You will find my numbers are not going to blow anyone away, so I feel no need to do the extra work to validate. If you have any questions, I will do my best to answer them.

This is stable, based on blend P95. 3DM06 = 19072, SuperPi 1M 10.892s. I haven't loaded up any other benchmarks yet, though I will be shortly.

My only ongoing concern is VTT....but the Gigabyte article on overclocking the EX58 series references 1.515 being used, I don't get that option but .020 increments. 1.500 proved unstable in some earlier testing, so I've just left it at 1.520 for now. Most user experiences I have read show much lower VTT voltages stable at high overclocks. Unless I misunderstand something, Intel's max spec is 1.350. This particular setting does not seem to impact core temps.

Previously, I was stable at 21x172 = 3.612, vcore 1.350, WITH HT WITH powersaving WITH Turbo, mem @ 10 (so 1720) and Uncore at 21. 3DM06 was 18987, SuperPi 1M 11.482s. Idle temps were 41-42, load temps (8 iterations of P95 blend) 76 max. I have had great difficulty going up from this with all features enabled. That said, I have not bumped voltage up to my (floating in my head still, not carved in stone yet) ceiling of 1.450v.

For now, just for the helluvit, I'm going to keep the HT/Turbo/power saving off and see how high I can take this chip. My expectation is that, in this bare bones/reduced features configuration, I will probably see 4.0 and maybe a touch over, at reasonable temps and voltages. Load temp of 69 at current configuration is relatively low, many users who use air cooling report peak temps in the mid 80's as being 24/7 stable - that is, not mid 80's all the time, but an overclock that itself has occasional peak temps there. Most others are in the mid to high 70's peak on air. Built in chip throttling doesn't seem to kick in until well into the 90s.

I've done many hundreds of pages of reading of posts from various forums, reviews, tech articles, etc. in the past week. Still have a couple of minor concepts and wording differences among manufacturers to solidify in my mind, but I think I have a handle on this.

For anyone wondering about D0 stepping vs C0 stepping, the headroom seems to be not with the air cooled guys (n gals) or even the water setups, but more headroom with phase change systems definitely. D0's can run on lower voltage - what I am seeing is D0's @ 4.0 around 1.20-1.25vcore, wheras the C0's need in the 1.3's and up - but the D0's seem to natively generate much more heat, and temps are generally higher at head to head clock speed vs a C0. Beyond that, there are more desireable batches of both C0's and D0's, and lesser chips of both also. But the big numbers you will see from the phase change guys are almost universally D0's.

At least that's my unscientific observation.

Off to try 20X200 or 19X210 or something, see if I can crack 4.0. Even where I'm at 3.8, I haven't owned a chip with this much headroom over stock speed (43% so far!) in many many years if ever.....I'll think about that one some more and try to remember. I missed out on the great Celerons (366? I forget) that would almost double speed back in the day.

Yes, when I get started I am absolutely capable of producing a wall of text :haha:

keto July 19, 2009 12:13 PM

I didn't get a lot higher with HT OFF. 3.83 stable but no higher - again, I haven't played with all the voltages to max potential. 19x210 and 20x20 were definite no-go as I had it set up.

However...as the intention is to run with Turbo & HT, I've refocussed myself in that direction.

21x179 = 3760, 1.375 vcore. 1.880 CPU PLL, Load Line ENABLED (this is new)
8X mem multiplier (1432), 17 Uncore, 1.660 vdimm, 1.40 VTT (glad to see this coming down & stable), Gigabyte memory booster on Extreme (previously default/standard), went with a more conservative memory setting overall while working up cpu speed.
-HT on
-Turbo on
-power saving features disabled
-P95 blend stable, max 78C load, idle 41-42C.
-SuperPi 10.998, 3DM06 19234 (best yet).

somecanadianguy July 19, 2009 01:59 PM

I'm definately interested to see where you end up.
So far I've had pretty good success with my i7 c0 and ud3r combo.
I run 3.8 usually... Vcore 1.36875 in BIOS... core temps barely break 70c in Prime and F@H.
I've been trying a slightly lower OC for folding when it gets hot in here... 3.486GHz with vcore 1.24875 in BIOS (1.184 under load).
Folded for a while with vcore @ 1.23750 in BIOS but can't call it 'stable' yet.

I can get 4GHz+ stable, but my cooling is really holding me back.
Right now I've only got the fan that came with the TRUE, and the ones that came with my case.
Oh, and I zip-tied the fan from the stock HSF to blow across the mobo heatsinks and RAM.
I hope to have that all fixed soon with more (and better) fans.
Maybe even look into watercooling if I can get the cash together... doubt that'll happen for a while though. :blarg:

I've been looking at the board you have, as well as the ud4p and may upgrade.
There are also some new ones coming that look pretty interesting...
Either way I think I'm sticking with Gigabyte (even though I hate their colour scheme).
Please continue to share your experiences.

keto July 19, 2009 10:15 PM

Lookit that, an audience of 1! :rofl:

Well, I would surmise that you have a somewhat better chip than I do, based on voltages alone. I could run 3.8 without hyperthreading quite easily 1.365v, but now when I am at 21x185 (3.885) with HT on, I am having to up my vcore over 1.4. I haven't completed testing so I can't call it stable just yet, but it looks promising. Temps are creeping up, 44 idle, 82 load. Memory still at 8x/Extreme, as above.

SuperPi mod1.5 XS 1M 10.6s, 3DM06 19393.

somecanadianguy July 20, 2009 09:53 AM

Hey, isn't an audience of 1 better than posting to yourself? :bleh:

I'm using 200x19 for 3.8GHz... Prime95 and F@H stable.
I've noticed that pushing it to 3.9 doesn't give much of a performance boost, and 4GHz is just not worth it for me with the cooling I have.
Plus I only have 1333Mhz RAM so it kinda limits what settings I run also.
So I'm staying where I am for now... max temps of 70c... should be safe.
Don't really want it much hotter though if it's gonna run 24/7.

Good luck to you.
Keep posting...

rjbarker July 20, 2009 10:17 AM

Hey guys some will argue this point, but I would prefer to get my I7 Chip stable "with HT On" and therefore be taking full advantage of what the Multi Processor I7 Chip is all about.....with it "off" may as well be a Q9xxx Chip in your rig.

somecanadianguy July 20, 2009 10:26 AM

I agree, rj...
Try to get stable with HT on whenever possible.

gingerbee July 20, 2009 11:26 AM

i too think that ht should be left on. 21 x 180 =3.780@ 1.2vcore qpi@1.21 LL enabled ram@1.5 HT enabled
temps never go over 66c

mattydies July 23, 2009 05:27 PM

Fan Question - 2 X 1600 CFM = enough (if not more than enough) to feed a 2500HP Gasoline Engine

1600 RPM I assume, how many CFM do they flow?

Arinoth July 23, 2009 05:30 PM

Quote:

Originally Posted by somecanadianguy (Post 226507)
I agree, rj...
Try to get stable with HT on whenever possible.

Yep and lets you fold with 2 more instances


All times are GMT -7. The time now is 07:22 PM.