Core i7/X58 Overclocking Thread

[bmilos]

I will look into this today, as I planned on doing some stability searching...

[bfghornet]

Ok looks like I've finally cracked it for a stable (15mins prime) setup & good temps ..

Here are the settings:

Multi: 200x20
SpeedStep/Turbo = OFF
DRAM = 1600Mhz 8-8-8-24
Uncore = 3208Mhz
QPI = 7218MT/s
CPU Volts = 1.4V
LLC = Enabled
C-State = On/Auto

Rest is ASUS Default

4Ghz 1600Mhz Prime 8K HT On


4Ghz 1600Mhz Prime 1024K HT On


4Ghz 1600Mhz Cinebench HT On


4Ghz 1600Mhz Prime HT Off


I refitted the Noctua today with horizontal mount (fans blowing up out top of case) + smooth cover of paste over whole CPU + thin invisible layer rubbed into HSF side ... brought my temps down 10-15C at load and sub-30's at Idle as you can see ..

Finally happy with the build !!

Hmm, I have a HAF932 aswell. I have my True facing to the back. Do you think I should point it up? So far the general consensus Ive seen is to blow to the back.

Thankyou for any help on this matter.

[SteveRo]

rge - what would you recommend for 1st cut at settings to go after high bclk x lower multi - including settings for volts?

I will give it another try, seems like I was not having much luck at higher bclocks - this - when leaving qpi and uncore on auto.

On GGBT boards do we need to manually set qpi and uncore multi's when going for high bclck?
If yes, what would you reco as a first cut?

[SteveRo]

rge - from your signature - I will try this -
4.2Ghz (200x21) with 1.412 bios LLC, 1.376 load, prime 12+hrs.
QPI/UC/Mem 36/17/8
QPI/Vtt 1.335
Mem 1600mhz 1.64v
CPU PLL, QPI PLL +1

correction - I will try x6 for memory - this the first time as I think I have weeker memory than you.

[ReverendMaynard]

rge - from your signature - I will try this -
4.2Ghz (200x21) with 1.412 bios LLC, 1.376 load, prime 12+hrs.
QPI/UC/Mem 36/17/8
QPI/Vtt 1.335
Mem 1600mhz 1.64v
CPU PLL, QPI PLL +1

Those are my exact same settings, but with 1.4v opposed to 1.412 and I'm using 1.66v dimm. I get 1.344 load.

[rge]

Edit: yep what you are trying is exactly my settings, let us know how it goes...is your mem not 1600 rated? (btw by +1, I mean 1 notch above stock, qpi pll 1.14 instead 1.1, cpu pll 1.84 instead 1.8 in case that was not clear)

If that fails, can leave all settings same except change blck to 196x21 for 4.1...if your 4.0 run was at stable vcore load 1.344, then 4.1 on those settings should be close to stable if not 4.2, then you can use ~scaling to find higher stable points quickly. But hard to compare yours to mine, since I have not seen many i965 yet and dont know the effect of dropping multi.

[SteveRo]

rge - got it, running p95 at your multi's and volts, except vcore at 1.36875 (same as what was stable for me for 12hrs p95 at 134x30).
currently priming at 191x21 - 4.01ghz, temps are lower by approx 5C so far! excellent!
max temp so far down from 75 to 70C
I will let this run for an hour then drop another -1 on cpu v.

[dejanh]

Bios voltage settings

Man, you're gonna kill your CPU. Your vcore is past maximum at load (Asus is BIOS + 0.03V at load), and your VTT is way to high too...

But, you are entitled to your opinion...

Hmm, I have a HAF932 aswell. I have my True facing to the back. Do you think I should point it up? So far the general consensus Ive seen is to blow to the back.

Thankyou for any help on this matter.

I would not try this TBH. For starters, I think the TRUE will block some of your DRAM slots. Also, the temps he is reporting are still quite high (if you look at his MAX on the screenshots you will see 89C on the 1024K test). If he says he shaved off 10C-15C his original mount must have not been good as those temps would have put him at or over 100C.

[bmilos]

Man, you're gonna kill your CPU. Your vcore is past maximum at load (Asus is BIOS + 0.03V at load), and your VTT is way to high too...

But, you are entitled to your opinion...

I was thinking the same thing... Those are CRAZY voltages...

[ReverendMaynard]

I was thinking the same thing... Those are CRAZY voltages...

suicidal, but somebody's gotta do it lol.

[Splave]

Man, you're gonna kill your CPU. Your vcore is past maximum at load (Asus is BIOS + 0.03V at load), and your VTT is way to high too...

But, you are entitled to your opinion...

QPI has to be that high to run such memory speed does it not?

[freshy]

im not going over 1.35v for qpi/vtt for now

11 hours into prime still going , 4.1ghz , vcore 1.475v qpi/vtt 1.35v still have to try lower vcore a bit, i had prime get an error after 18 hours other day small ftt

saaya was telling me that temps over 80 are no good for 24/7 use , what do you guys recon ?

my temps under 100% load are sitting around 80c but sneaked up to 83 for a few seconds earlyer

[bmilos]

Im all for going over the 1.35v QPI limit that Intel set, but going THAT far over, just to run memory speeds? I dunno...

Also, my board seems to be running BIOS + 0.05 at load.

[rge]

rge - got it, running p95 at your multi's and volts, except vcore at 1.36875 (same as what was stable for me for 12hrs p95 at 134x30).
currently priming at 191x21 - 4.01ghz, temps are lower by approx 5C so far! excellent!
max temp so far down from 75 to 70C
I will let this run for an hour then drop another -1 on cpu v.

Be interesting to see final vcore/temp difference high bclk to low bclk for same mhz...keep us posted.

saaya was telling me that temps over 80 are no good for 24/7 use , what do you guys recon ?

my temps under 100% load are sitting around 80c but sneaked up to 83 for a few seconds earlyer

I would not worry the least about 80C load temps for 2 reasons. First, if its only prime that gets you there, then prime temps dont really matter, unless you are folding 24/7 with same temps. And 2nd, all cpus per intel are calibrated such that at max load tjmax is reached (~100C on corei7) when tcase max is reach (~67C on IHS on core i7 if memory serves me correctly). At 80C prime load you are definitely going to be within tcase max specs.

Granted for every 10-15C drop in temp you double the life of the component, and I will do everything I can to get them lower (except reduce my stable OC), but given upgrade cycle is rarely longer than 3 year intel warranty period, my main concern with prime temps is stability, and if not hindering that, not even 85C max temps would I let hinder my OC.

[Splave]

Im all for going over the 1.35v QPI limit that Intel set, but going THAT far over, just to run memory speeds? I dunno...

Also, my board seems to be running BIOS + 0.05 at load.

Well if its a requirement for these beasts to do it, cant be that bad for the CPU.

http://www.xtremesystems.org/forums/...7&postcount=34

[freshy]

I would not worry the least about 80C load temps for 2 reasons. First, if its only prime that gets you there, then prime temps dont really matter, unless you are folding 24/7 with same temps. And 2nd, all cpus per intel are calibrated such that at max load tjmax is reached (~100C on corei7) when tcase max is reach (~67C on IHS on core i7 if memory serves me correctly). At 80C prime load you are definitely going to be within tcase max specs.

Granted for every 10-15C drop in temp you double the life of the component, and I will do everything I can to get them lower (except reduce my stable OC), but given upgrade cycle is rarely longer than 3 year intel warranty period, my main concern with prime temps is stability, and if not hindering that, not even 85C max temps would I let hinder my OC.

i agree,

the thing with i7 that is spinning me out is a 13 to 20 hour prime run and it might still not be 100% stable, like i said ive had prime error after 18 hours, and the other day i primed for 21 hours small ftt, then ran 2 smp folding clients over night and by the time i got up the pc had crashed, so 1 prime run isnt enough,

edit... that was @4ghz that i ran the folding clients so load temps were around 70c max

[Xello]

like i said ive had prime error after 18 hours, and the other day i primed for 21 hours small ftt, then ran 2 smp folding clients over night and by the time i got up the pc had crashed,

As has been said small ftt doesn't seem to do much to the uncore (l3 cache, IMC), i'd stick to running blend.

[bmilos]

Well if its a requirement for these beasts to do it, cant be that bad for the CPU.

http://www.xtremesystems.org/forums/...7&postcount=34

I've read that thread. Those are some monster sticks.

Maybe QPI is supposed to be within .05, not .5

I'd much rather run with a high VTT as it def helps with my OC stability. I'd just rather not blow out my CPU.

Also, what does 'good cooling' mean as far as high VTT? would that fall under CPU cooling or is QPI cooling seperate?

[Xello]

Maybe QPI is supposed to be within .05, not .5

If that was the case most of us would have dead chips by now

I think the cpu itself is where volts are fed for QPI, but QPI itself runs through the whole chipset, ie from cpu to nb to SB, between memory and CPU, etc, i think...

[bmilos]

If that was the case most of us would have dead chips by now

I think the cpu itself is where volts are fed for QPI, but QPI itself runs through the whole chipset, ie from cpu to nb to SB, between memory and CPU, etc, i think...

Yeah, that was kinda sarcasm, kinda not. Maybe, for good OC'ing, .05 is a good rule of thumb.

Also, I remember reading some reviews where guys were using 1.55 vtt right out of the gate... Don't know how good it is for 24.7 as only time will tell.

As far as QPI cooling goes, I guess that means one using high QPI should have solid airflow/low ambient. I'm trying to figure out a way I can mount my rig in a glass/acrylic box in my window. Kinda like a window extension of sorts... Think air conditioner, hollow inside. It is winter after all - 12 inches of snow on the ground. Oh, to be young and energetic.

[Astratuner]

In the manual of the Asus Rampage II Extreme stands that QPI Voltage is green from 1.20000 - 1.39375, from 1.4= oranje.

So 1.35V is maybe not the safe min?

[Astratuner]

http://blue.ap.teacup.com/kazukun/img/1229638228.jpg
i7 965　3837B209　WaterCooled
Core Speed：4200Mhz
BCLK：200MHz
Multiplier：21
QPI：3600MHz
QPIV：BIOS1.65V
Vcore：BIOS1.55V
LLC：ON
MEM:KHX16000D3T1K3/3GX
Vdimm：BIOS1.65V
Uncore：4000MHz
DRAMFrequency：2000MHz
DRAMTiming：9-9-9-27 1T
HT：ON
http://blue.ap.teacup.com/kazukun/

i c on your pictures that you run a QPI voltage of 1.45V.

You know this could damage your cpu in time? Or you don't run this 24/7?

How long do you run on 2000MHz?

greetz

[rge]

I would bench 2000 ram if I had it, but not 24/7. I just ran 22x182, 17x uncore, 8x mem for spi, everest, pcmark5. Then ran same except uncore at 20x. Then ran same except both uncore 20x and mem at 10x. Wanted to see what had more effect the 20x uncore (necessary for 10x mem) or 10x mem.

I gained ~0.5% in spi and pcmark 5 scores from just increasing uncore to 20X from 17X, keeping mem at 8x. I did not see any more gain in pc mark5 or in spi at then increasing mem to 10x, but had to loosen timings to 10's from 8's, not to mention use high qpi/vtt and DRAM voltages. Everest increased read and copy speeds 6% and write speeds 13% from uncore increase alone. Additional 3% increase from increasing mem to 10X. Other than benching, I wont be using mem speeds over my 8X divider.

For an extra 0.5% speed for 24/7, raising bclk by 1 point is a lot easier all the way around.

[speckled10]

Hmm, I have a HAF932 aswell. I have my True facing to the back. Do you think I should point it up? So far the general consensus Ive seen is to blow to the back.

Thankyou for any help on this matter.

All depends what sorta temps you currently getting Idle & Load ??

I think my better reapplication of grease probably had more to do with the lower temps .. at Idle and moderate loads I'm down and the Noctua is still cold ..

But I reckon blowing out top still helps as theres more airflow with 240mm fan + surrounding airholes (and hot air rises ainnit?) .. at load I feel a lotta hot air coming out ..

Blowing out the back its a much more enclosed space .. smaller fan no surrounding vents .. and the fans dont quite line up with the P6T mount ..

[fornowagain]

I'm trying to figure out a way I can mount my rig in a glass/acrylic box in my window. Kinda like a window extension of sorts... Think air conditioner, hollow inside. It is winter after all - 12 inches of snow on the ground. Oh, to be young and energetic.

Have a quick look at the radbox link in my sig.