** OFFICIAL ** Core i5 Overclocking Results/Tips Thread

[LazyDogNet]

From everything I've read and seen, the ability to overclock particular CPUs can vary based on when and where they were made (for example, certain batch numbers will overclock better than others with less voltage needed while others might not o/c much at all). The i5 is still pretty new, but I would expect the same to hold true for it too. I've read many different posts where someone will trade out their current i5 CPU for a different one and achieve much better overclocking with the exact same supporting components. Even over the lifetime of the manufacturing of a particular chip, adjustments are made during the process that will affect it's o/c ability (for example, stepping).

Nice o/c d4rk ch1ld....what voltage did you use and let us know if it is stable.

[loften]

i left prime blend running all night long. when i woke up i found my son playing cod modern/2.. i stopped the test 0 errors 0 warnings.. i think it is stable.. i will leave it at 4.2 for a while..

[GrossmeisterB]

I still think, that every CPU is different. It can be mainboard, but it doesnt mean every "asus" mboards are bad. Maybe he just got good piece of gigabyte mobo, or you got a bad piece of asus. Gigabytes are good too...

Yeah, every CPU is different, and every motherboard is different...so that's why I was asking, if anyone else noticed similar differences, e.g. "Asus needs more voltage than Gigabyte"...

I'm getting an MSI P55-GD65 now, I hope it works as good or better than my Asus...

[loften]

i ran prime blend at 220x20 with 1.22 vtt and 1.38 cpu.. under load at 1.21-1.22 vtt and 1.40-1.41 cpu, without vdroop, it runs for hours no errors but temps went up to 72 degrees..

[LazyDogNet]

Hey, I'm curious what people are seeing in memory benchmarks. I have mid-range RAM in my system. It's a mix of Corsair and G.Skill DDR3 1600 (4GB each) that run well together.

I'm actually under clocking these a bit @1528MHZ with the CPU at 4GHz (21x191MHz), but I run a little tighter on the memory settings @ 8-8-8-21 vs the rated 9-9-9-24 timings.

The below screen is from Everest

[loften]

i only have the trial version. it says that in most of the boxes.. my read is almost 2k higher then yours.. i have not ventured into the memory timings yet on this board.. they are all on auto except the first 4 settings. 9,9,9,24.. after i play with this thing some more i will venture into the memory timings and see if i can tweak them a little.. i did try to run super pi once at 8,9,8,24 and it wouldn't run it.. i have the trident 2000 they are running at 2100 at 1.65 volts.. i would like to work on them a little but it is hard to get my kid off the games so i can use it..

[loften]

i don't know if it matters but if i remember right my maximus se with double sided crucials ran closer to 11k read and right but this one feels much faster with everything i do on it even at lower clocks.. i know on this at 4.2 with 1 hd4870 3do6 run is faster then my 1 card in my maximus se by over 1k.. on my maximus i was only able to get just over 25 k with both cards over clocked a little and the cpu at 4.5.. i am waiting for the longer crossfire bridge to come so i can run both in this one and run some benches with it at 4.2 and 4.4.
i think maybe 8x on both cards might slow it down compared to 16x on both on the se??

[LazyDogNet]

Yeah, I'm not surprised that your 2000MHz memory (running at 2100MHz) is showing a lot better benchmarks scores than my 1600MHz (running at 1528MHz)....Plus your FSB is about 20MHz faster than mine which will make a big difference too.

I didn't see a lot of people listing their memory read/write/copy scores, so I was just curious what people's scores were. I got some good deals on these 1600MHz RAM chips, but I would like to have purchased something faster so I could have used a higher multiplier...Oh well, that's how it goes when you're really watching the bottom line (you usually got to be willing to sacrifice a little performance to save a little money).

Hopefully your son will give you a little time on your computer (that's why my son has an XBox)

LDN

[loften]

trying to prime it right now at 20x200 at 1.32 cpu 1.22 vtt in bios without vdroop. at load 1.35 cpu and 1.21/2 vtt..

[Biznatch]

So far mine does 20x200 just fine with 1.31 and 1.21vtt. I was able to get up to 210x21 so far with 1.375cpu and 1.25 vtt but it was unstable. Haven't pushed it much farther yet, but it has yet to even see 50c under full load.

[dcl]

Just got my Noctua installed. Straight to 160fsb and 22x multi (3.52ghz) on my i7 860 with 1.3vcore, 1.19vtt. Testing for the last few hours, perfectly stable. Time for 3.8ghz.

[dcl]

Just got my Noctua installed. Straight to 160fsb and 22x multi (3.52ghz) on my i7 860 with 1.3vcore, 1.19vtt. Testing for the last few hours, perfectly stable. Time for 3.8ghz.

Now @ 3960mhz with the same volts. Linx going okay...

One thing I've noticed is that the system seems far more stressed while running Linx than it was before. All other apps are more unresponsive.. This couldn't be because I'm running the ram a little slower no?

[powergiga08]

I am testing running 18x220 @ 3960mhz for now.






pg

[LazyDogNet]

Nice o/c dcl....too bad the i7 860 wasn't in my budget range....looks like one nice chip.

So far mine does 20x200 just fine with 1.31 and 1.21vtt. I was able to get up to 210x21 so far with 1.375cpu and 1.25 vtt but it was unstable. Haven't pushed it much farther yet, but it has yet to even see 50c under full load.

That's one nice cooling system you must have there!

The whole what's stable is always a good question too. I've always heard/read that if it passes P95 for 6-8 hours (or longer), it should be good to go, but I'm not sure I completely agree with that either (even though that's how I generally test my systems too). Running something like P95 or LinX on your system is stressing it out more than most of us would ever do in the real world (running our normal apps or games). Under our normal computer usage, I doubt most of us are seeing 80% CPU load on all the cores at the same time for any length of time. When I'm encoding a video, I might be seeing a 40% overall load with one core getting hit at about 75% load (I love that this CPU can do it in about 10 minutes).

I had a o/c system a couple of years back that would fail P95 after about an 2 hours, but I never saw one single problem when I was using day to day (playing games, encoding videos, photo editing, etc). As far as I was concerned, that system was stable.

I think running P95 or LinX are great indicators, but I'm not sure we should all hold it like the holy grail either. If multiple cores fail in P95 within 2 minutes, then you "do" have an issue that you need to correct. But, if one core "errors" out in P95 after 2 hours of usage (remember, it's a calculation error....not like it BSOD or anything), than my guess would be that the system is pretty darn stable for most of our needs. The real world test should be when you are using your real world applications/games in your day to day activities.

Anyone have any thoughts on the subject?....Not that I'm trying to start a big debate here

[Mav451]

The stability debate is very personal. If you are only benching for WRs, obviously your goal is not a 24/7 machine. So really stability is subjective to your objective. If you need a 24/7 system, yeah you will have lower #'s (and hence less e-peen haha).

[Biznatch]

There is no question it's unstable. Crashes in windows sometimes etc, but I haven't played with it much. Motherboard is the gigabyte ud4. Cooling is HC 3 with a 3x120 and a 2x120 cooling the same loop. I have a waterblock ready to attach to my 5850 that's been ordered for like 6 weeks now. That will increase the temp of the loop some, but I doubt too much since that card is supposed to run cool.

[LazyDogNet]

There is no question it's unstable. Crashes in windows sometimes etc, but I haven't played with it much. Motherboard is the gigabyte ud4. Cooling is HC 3 with a 3x120 and a 2x120 cooling the same loop. I have a waterblock ready to attach to my 5850 that's been ordered for like 6 weeks now. That will increase the temp of the loop some, but I doubt too much since that card is supposed to run cool.

Yep, that sounds unstable alright It sounds like you have more than enough cooling when you decide to push it pass the 4GHz mark.

I hope you don't need to wait too much longer for your 5850 card (that will be sweet).

Nice system there!

[Biznatch]

I'm sure it will do the 210x21 stable with more voltage, but it's not really worth it for the extra 400mhz for daily use. So I'm just staying at 200x20 at 1.31/1.21vtt and can't even tell the difference. Once that card comes though I'm going to push everything way up to run some tests.

I also took a bunch of high res shots of the underside of the CPU before installing it. When the card does come I'm going to pull the CPU out and see how they look after after constant 4ghz.

[OC Nub]

I hear all the time that LinX stresses the cpu harder than p95 and that's why you can test stability with 20 passes of LinX vs hours of blend. Im not sure about the i5 but on the i7-860 the temps do not get any higher in LinX than in blend, with my q9650 after 4.3ghz it was impossible to test with LinX due to temps so I used only blend. Do you guys get the same temps in blend as in LinX with the i5-750? I am a bit ocd when stress testing the cpu and try to run at least 12 hours blend and some LinX to be sure. Some say it is unneeded stress on the hardware but I have yet to kill or degrade a cpu doing this.

[LazyDogNet]

I found my i5 chip running about 3c-4c higher when using LinX over P95, so for me there was a difference. I'm currently able to run 4GHz using 1.24v or I could go down a little and use 1.21v for 3.9GHz, but I also run a lower FSB (191MHz or 186MHz for those two marks plus a 21x multiplier). In the past, I've usually used a higher FSB and lower multiplier because I found it to perform a little better, but I'm so pleased with able to reach 4GHz using less than 1.25v that I'll probably stick where I am for now....plus I just have an average cooler.

I normally just used P95 for all my testing and only started using LinX with this build, so I really don't know much about it. It seemed to me that it was testing the CPU a lot harder than the RAM even if I choose the highest MiB available, but that's only a guess really. I think if you can pass 20 passes of LinX then your CPU is pretty rock solid.

Let us know Biznatch if you notice anything on the CPU or socket (I assume you have a foxconn socket)

[OC Nub]

That is excellent vcore for 4ghz, you have a very good chip.

Messing around with some stability testing. First ss is Blend at 4ghz. Second is linx same settings. If you look at max temps in Real Temp, they are about the same. The reason I was curious is I was thinking of trying an i5-750. Looking around there are a lot of very good clocking 750's out there that require less vcore. I have tried only using 4 threads on this chip but still requires more vcore than a lot of 750's I see.

[mikecdm]

I'm not even sure which puts more heat on my chip. When i first built the rig I used P95 small fft's. After someone suggested I use LinX, my chip errored out on the 18th pass of LinX with the same settings. Ever since, i have onlused LinX. So I really can't say anything on the heat comparison. All I know is that it loads mid 70's in LinX, but that really depends on the ambient temps.

From what I've seen, the i5 generally need less vcore for 4.0ghz than the i7 860. But have also seen a great deal of poor clocking i5. Mine does 4.3ghz at the same vcore that my friends i5 needs for 4.0ghz.

[Dahaka]

Here my config:

Core i5 750
Evga P55 SLI
OCZ Platinum 3x2gb but only ussing two sticks.
Noctua U12P SE powered by hardmodding.net and Noctua

[loften]

this is still running now.. 21x191 is nice and cool with much lower volts than 20x200..

[dcl]

My rig is completely stable at 180x22 (3960) but won't even get in to windows at 200x20. Probably needs more VTT. :o
(i7 860, P55UD4P)