Core i7/X58 Overclocking Thread

[Xello]

i have a 965xe. Am i right in saying _ALL_ i need to do to overclock this is just increase the multiplier (along with Vdim) and not worry about the QPI? I am currently running at 4ghz stable with:
BLK: 133Mhz
Multiplier: 30x
Core voltage: 1.275

You answered your own question

Perhaps try a different version of Everest.

[SteveRo]

HT on doesn't look like it is worth it for 3dMark06
Anyone done this compare with 3dvantage yet?

[ReverendMaynard]

HT on doesn't look like it is worth it for 3dMark06
Anyone done this compare with 3dvantage yet?

Yep and I got an 1100 overall score better with HT on.

[Chicken Patty]

My 920 absolutely HATES the 20x multiplier. I can't get it to pass a 3DMark Vantage test at 200x20, Turbo On(4.2Ghz actual) to save my life, even at 1.43v. However, it runs at 211 Bclk x 19 all day long at 1.36vcore.

I think that I'm missing something simple but I've done extensive research and I can't figure out what the hell is holding me back.

Cooling is not an issue because my max temps @4.0 range from 69-73C across all four cores with HT enabled.

Yes, I was lazy and put a Koolance Exos-2 LX with an Apogee GTZ water block on it but my CPU is the only thing in the loop and it seems to be working very well. Plus, I must admit, if you heat this stuff in hot water and then apply it correctly, I think it's paid off royally:

http://www.acousticpc.com/innovation..._compound.html

My temps are consistent with the most "hardcore" H20 rigs and I think it's because I used the best water block and TIM, not to mention that my CPU is the only thing being cooled.

thats my daily settings actually 200x20 turbo mode enabled 4.2 GHz @ 1.400v all day long stable as heck. Mine on the other hand, hates lower multi. These chips are pretty weird, but it takes time and research. Luckily I just hit it right the first couple of times with mine and didn't struggle much

I'm using a apogee gtz with a Koolance 120.1 rad with only one fan pushing, temps max out at 73-75ºc idle mid 40's. Thermal compound is AS5

[bmilos]

I use everest, but for some reason it does not show my real cpu speed or multiplier, for that i have to use cpu-z, any dieas ? is everest not compatible with i7 ?

There is an everest thread where you can find the most reason version -

http://www.xtremesystems.org/forums/...6&goto=newpost

It is compatible with i7 overclocking and the benchmarking apps seem to be more accurate than the older ones.. Expecting a new final release soon.

[Chicken Patty]

@ SteveRo

I score considerably less with HT Disabled on 06, and vantage. My best scores come with HT enabled on either one, at least with my setup.

[Chri$ch]

played today a little bit with my good old Micron D9JNL

all done with 1.90v vDimm







http://valid.canardpc.com/show_oc.php?id=469880

[Chicken Patty]

@ Chri$ch

Thats a great overclock on the CPU, but the clock on the RAM is awesome, great job

[webwilli]

@ Chri$ch

nice

QPI/VTT Voltage please

[Chri$ch]

hehe, thx

vtt is ~1.55v for DDR3-2000 CL8 and ~1.65v for DDR3-2244 CL9

i think the memory can do more, but the limit is the UCLK freq.

[SteveRo]

My best 3Dmark06 to date, this with HT off and with oc'ed 9800GX2 -

[Xello]

i think the memory can do more, but the limit is the UCLK freq.

Good job so far, here's something i seen from Fugger on Saaya's 101 thread that might be of assistance:

Yes, Vdimm and uclk are directly related and raising vqpi doesnt scale uclk, vdimm does. That is the secret to free performance, that is how I hit 25K and 18.9ns latency.

Ill disclose the voltages I am running at on each board.

Voltages

vcore ~1.5v, this is a "leaky" high IDV chip, scaling up with higher voltage is possible at a cost of runaway amp phenomena.
vmem ~2.2v for benching 2v on Vapo, compensate voltage with cooling to be safe.
vqpi ~1.35v depending on the board/bclock, upwards of 1.55v if I am pushing bclock, I run a low bclock so I usually leave it on stock.
vmem is the secret to uclk, 2.2v gets you 4.7Ghz ulck. uclk is also known as uncore.

[freshy]

Looks like you have a really nice OC going there. How far do you intend to drop those voltages, do you think your current ones would risk you eventually burning out your CPU early? I've seen people all over the map with voltages... its come to the point where I'm not even sure what considered OK anymore except Intel's offical quotes (1.65v).

i would love to be able to drop my vcore back to at least 1.375v as i had it stable with that vcore when i was on f4k bios, then i flashed to f4n and i needed more vcore and more qpi/vtt to get stable, im now using 1.35v qpi/vtt as before with f4k i only needed 1.315v,but even if i go back to f4k now i still need more volts, but no use crying over spilt milk , i will prolly push my vcore to 1.45v and qpi/vtt to just under 1.45v to get stable with ht on and turbo for 4.1ghz and stop there and just make sure im not killing my chip by going any higher before going for 4.2 ghz again

you can use LinX which is a gui for linpack

http://www.xtremesystems.org/forums/...highlight=linx

thanks for link guys

Do you mean 10 hr small fft + 10 hour blend, or 20 hour small fft + 20 hour blend? You shouldn't waste your time with small fft, it's for testing max heat output - good for testing your cooling. It puts less stress on the cpu than large fft, which is why Blend is the best option

I always use 24hr blend as the pass mark for my final OC's - has never failed me.

man ive had this rig for 2 weeks and all ive done is prime, prime, prime, ive done so many runs its not funny, and yea 20 hours of small ftt and 20 hours of blend runs, when i first started with blend runs ppl said no do small ftt, then ppl said do blend , lol ppl need to make up there minds , anyway im gonna start hitting it with linpack now

and if ya's think that you are stable try running 2 smp clients for 24 hours with that affinity changer with HT on and turbo on , and see how stable ya's are , good luck

and merry christmas guys all the best to you and your family's in the new year

edit... if i called 2 to 3 hours of prime stable then i could say that i got 4.3ghz stable, but to me even 13 hours of prime isnt stable as ive gotten errors even after 19 hours of prime, small ftt and blend

[ReverendMaynard]

If I can game for 10 hours straight, and I can run 60 different benchmarks without a single hiccup(#4 air cooled Vantage/GTX260 on the planet ) Then I call it stable. I fold with other, cheaper machines that I don't give a rats ass about.

[Xello]

edit... if i called 2 to 3 hours of prime stable then i could say that i got 4.3ghz stable, but to me even 13 hours of prime isnt stable as ive gotten errors even after 19 hours of prime, small ftt and blend

Then that's your experience and your opinion and you're entitled to it

I'll stick with mine based on my experience, until it fails me

About prime, Blend really is the option to go for, small FFT loads data of a size that fits in the cache on the cpu and not so much your ram, which is why it traditionally produced hotter cpu (penryn and previous), this was mistaken by alot of people as being more stressfull but it's not, it's just a hotter cpu because the l1/l2 cache is being used more. Large FFT will utilize the memory and therefore memory controller, so for i7 this stresses the cpu more because the memory controller is located on the Uncore with the L3 cache. That's why large FFT is making cpu's hotter compared to small FFT, relative to how it used to be with penryn.

Blend just covers all the bases, good luck with Linpack though. I might run intel burn test once i have my final OC down, out of curiosity. I done this on my QX9650 and it failed where i was 30hr+ blend stable. I ignored it and never once had a crash/BSOD after. Up to each person themselves i guess.

[freshy]

maybe i should just drop my voltages down then so that im just game stable to then and not bother folding with this rig

but i really dont want to push qpi/vtt v more than 1.45v till we know what is safe for 24/7 use, i might just put it on 4.1ghz prime stable* for now, and wait till we know more

[ReverendMaynard]

maybe i should just drop my voltages down then so that im just game stable to then and not bother folding with this rig

but i really dont want to push qpi/vtt v more than 1.45v till we know what is safe for 24/7 use, i might just put it on 4.1ghz prime stable* for now, and wait till we know more

The great thing about the i7 is that less is more. Settle for a lower clock for folding and it'll still pump out better than average ppd.

[eva2000]

DFI Lanparty UT X58-T3EH8







First 32M Pi result on Core i7/X58 platform! By far, this board is easier Intel/DFI board I have ever overclocked with. Testing the waters with 2x1GB Corsair DDR3-2133C9DF Samsung HCF0 based memory rated for DDR3-2133Mhz 9-9-9-24 at 2.0v.



System specs

• Intel Core i7 920 3836A756
• Thermalright Ultra 120 Extreme + LGA1366 Bolt Thru kit
• DFI LP UT X58-T3EH8 R.A51
• 128MB Gainward FX5200 PCI
• 2x1GB Corsair TW3X2G2133C9DF Dominator 9-9-9-24 at 2.0v rated (Samsung ICs)
• 750GB Samsung HD753LJ
• Pioneer DVD-RW
• 1KW Corsair HX1000 psu
• WinXP Pro SP3 nLite fully updated

Untweaked 32M Pi result

[Solarfall]

great work eva i have the same dfi board comming from germany later on next week(hopefully?) with a tri dominator set.
im curious what is the highest Base Clock you've been able to achieve one this board ??

[Astratuner]

wow you got 2000MHz ram withing the safe spec of intel: 1.35V QPI!

How come?

[ReverendMaynard]

wow you got 2000MHz ram withing the safe spec of intel: 1.35V QPI!

How come?

dual channel maybe?

[dejanh]

Likely dual channel.

[Chicken Patty]

man that board looks awesome!!!!

Hows the BIOS on it?

[bmilos]

Says dual channel on the left side CPU-Z.

[eva2000]

Yeah it's dual channel but for single threaded apps you can lower CPU VTT volts as opposed to multi-threaded loads like prime95/hyper pi..

for Prime95 with TRUE120 air - room temp 29-30C