Core i7/X58 Overclocking Thread

[freshy]

thanks for the update unclewebb


over 13 hours prime still going @ 4.137ghz 197 x 19 + turbo , HT on, bios f4k, im gonna stop it , still need to drop voltages or press on for 4.2ghz

edit...... stopped prime after 14 hours

[zlojack]

Now I see why everyone is loving the new platform!!

Some early results on stock air


Love that bandwidth and latency!!

[revenant]

my 920 is a batch 3837a714 that any good? I am waiting for my psu before I can power the new machine on.

[CookiePuss]

Dejanh, very nice post man.

[webwilli]

my new 920

3837B072

[freshy]

4.2ghz ,1 1/2 hours of prime still going HT on, turbo on, loves the vcore , might try f4l next

[dejanh]

thanks for the update unclewebb


over 13 hours prime still going @ 4.137ghz 197 x 19 + turbo , HT on, bios f4k, im gonna stop it , still need to drop voltages or press on for 4.2ghz

edit...... stopped prime after 14 hours

What the hell? How can you have turbo when you are not running your default multiplier? Some kind of feature (bug) in the Gigabyte BIOS?

If it is, I want this on the R2E...

On a different note, I think I finalized my clocks more or less for now. I am running at 179x23 (with turbo on) for a 4117MHz net core clock, 4 cores, 8 threads. Memory is running at 1794MHz and I will probably tighten the timings a bit from the current 8-8-8-24-2T. The voltage required to do this was 1.448V core, 1.35 QPI/DRAM, 1.65V DRAM. All else is auto, including LLC. All CPU power states are enabled as well (i.e., all settings are default).

This seems to be where my air-cooling caps out. If I bump the volts any higher temps are climbing to 90s so I think I will stay right here max. Right now I cap out at about 88C with all 8 threads loaded 100% over extended periods of time.

Maybe I will get better cooling down the line, but we'll see. Not bad for air though I think

[rge]

What the hell? How can you have turbo when you are not running your default multiplier? Some kind of feature (bug) in the Gigabyte BIOS?

If it is, I want this on the R2E...

On a different note, I think I finalized my clocks more or less for now. I am running at 179x23 (with turbo on) for a 4117MHz net core clock, 4 cores, 8 threads. Memory is running at 1794MHz and I will probably tighten the timings a bit from the current 8-8-8-24-2T. The voltage required to do this was 1.448V core, 1.35 QPI/DRAM, 1.65V DRAM. All else is auto, including LLC. All CPU power states are enabled as well (i.e., all settings are default).

This seems to be where my air-cooling caps out. If I bump the volts any higher temps are climbing to 90s so I think I will stay right here max. Right now I cap out at about 88C with all 8 threads loaded 100% over extended periods of time.

Maybe I will get better cooling down the line, but we'll see. Not bad for air though I think

MSR registers on cpu show max multi/turbo as 23,23,23,24 on i940, which is apparently what governs turbo, unless bios can set other limits. With turbo enabled i920 will always be at 21 multi, and i940 23 multi, regardless of setting multi in bios at 15, 17, 18 or whatever. If you want +2 multi, you have to deactive all cores but one in bios...so dont see purpose of +2 multi.

If Asus bios is doing something differently, would be good to know. It would be nice to be able to set turbo limits ourselves...so I can use 21 etc multis with my i940 with turbo enabled.

And yeah 4.1 on air is pretty good, Im on water and that is likely where I will end up for 24/7.

[Hornet331]

finally found my 24/7 settings, tought im not totally through with testing for QPI/VTT voltage and RAM voltage/timings.

Clocks and CPU volts are final.



right now im on 1.33V QPI voltage, maybe i can lower it a bit.

For the Rams i'll try CL6 DDR3-1200.
DDR3-1600 CL8 is not possible with 12GB, i get errors in memtest within a few minutes... i think its the cooling. With 6 stick that close togther i need aktive cooling for the ram, and right now i have no 6cm fan.

[freshy]

--------------------------------------------------------------------------------

started priming at 11.10 pm @ 4.2ghz , computer restarted at 7.48 am
so 8 hours and 38 minutes of prime @ 4.2ghz , now tell me that doesnt suck !!!!


ok voltages for 4.2ghz 8 hours 38 mins of prime
200 x 19 + turbo, HT on
cpu vcore 1.55v
qpi/vtt 1.535v
cpu pll 1.88v
qpipll 1.18v
ioh core 1.24v
ich i/o 1.58v
ich core 1.22v
vdimm 1.66v
qpi link was x36
and uncore was running at 3400mhz

____________________________________

voltages for 4.137ghz stopped after 14 hours of prime
197 x 19 + turbo, HT on
qpi link x 36
uncore frequency x17
vcore 1.49375v
qpi/vtt 1.495v
cpu/pll 1.88v
qpi/pll 1.18v
ioh core 1.14v
ich i/o 1.58v
ich core 1.16v
vdimm 1.66v
_________________________________________

voltages for 4.1ghz 14 hours prime
195 x 19 + turbo HT on
qpi link x36
uncore frequency x18
cpu vcore 1.45v
qpi/vtt 1.435v
vdimm 1.64v
everything else auto
_________________________________________

voltages for 4ghz 13 hours prime
191 x 19 + turbo , HT on
qpi link x36
uncore frequency x18
vcore 1.375v
qpi/vtt 1.315v
vdimm 1.64v
rest auto

and now that ive spent all that time getting system stable, im gonna flash to new bios f4l and start again

edit... scratch that, day off today i got a head ach ! lol , to be continued tomorrow !!!

[bmilos]

How is Everest Ultimate's stability test as far as i7 24/7 use?

[Splave]

freshy how you lovin the gtz, what rad are you using?

[freshy]

hey bobby, yep its a nice block, and im only using a MCR220 and MCP655 ~~Swiftech MCRES , might have to upgrade to a 3x120 rad tho, summer is here in australia gonna start getting real hot soon, 40 to 46c days, ive also got an apogee gt that i had lapped and was using on my q6600, might have to make a mount for it and throw it on , see what temps i get compaired to GTZ,

lol im trying to wait a while before i lap this i7 cpu, even thinking of lapping gtz , but i know guys here at xs dont like lapping water blocks

[dengyong]

What are the best batch numbers to look for ?

[dengyong]

hey bobby, yep its a nice block, and im only using a MCR220 and MCP655 ~~Swiftech MCRES , might have to upgrade to a 3x120 rad tho, summer is here in australia gonna start getting real hot soon, 40 to 46c days, ive also got an apogee gt that i had lapped and was using on my q6600, might have to make a mount for it and throw it on , see what temps i get compaired to GTZ,

lol im trying to wait a while before i lap this i7 cpu, even thinking of lapping gtz , but i know guys here at xs dont like lapping water blocks

The gtz has a bow machined into it, lapping it would be detrimental to it's performance.

[freshy]

mine is batch 3837A727
REVISION CO/C1
got it from mwave U.S.A

edit... and yep thats what ppl said about my apogee gt before i lapped that, and after i lapped it i was @ 1.6v with max temps 54c @ 100% load occt with q6600

[D3mon_Hunt3r]

Here is Winter kicking in, my room's door is always closed and it is still cold and i'm yet to test my WC capacity as i've had a few issues with my 4870X2 (took the chance to get my case worked on) but now i should be jumping to the i7 and will check the performance of my GTZ.

[Thasp]

I emailed Asus about this issue. Their support team is a JOKE!



How the hell can bugs get fixed then? What the hell is their support good for if it can't be in the loop? Thats the most BS I've ever heard!

thanks man, that's my new sig.

[rge]

On my i940, not only does turbo on multi come at cheaper vcore...it looks like dropping the multi and using higher bclk reduces the vcore requirement as well. At least to reach 4.1 via 23x182 (turbo on perma 23 multi) or 21x196 turbo off I can use 2 notches lower vcore at least, than is required for stock multi 22x187 at 4.1 for same 12+hrs prime stable. Like some i920's that dont like their stock 20 multi, maybe my i940 does not like my stock 22 multi. On mine the uncore mhz to qpi mhz does not seem to play a role, as I can run 16 or 17 or 18 uncore, and only alters qpi volts requirment, but not vcore requirement. Maybe just similar to some E8xxx that ran with lower vcore at same mhz by dropping multi and higher FSB, and turbo mode just a different animal.

F4j bios all with LLC enabled
4.1ghz, (182x22, turbo on 23 multi) vcore 1.38bios, 1.34 load for prime stable 12+hrs

4.1 (187x22 turbo off) vcore 1.4bios, 1.36 load for prime stable, 1 notch below fails, 2 notches below at same bios volts as using 4.1 turbo 182x23 is clearly unstable.

4.1 (196x21 turbo off) vcore back to 1.38 bios, 1.34 load prime stable for 12+hrs, and trying lower currently...see pic.

For 4ghz (182x22) I need 1.35 bios, lower fails. When I ran 21x191 4ghz and 20x200 4ghz I only tried same 1.35, which was stable...apparently at least for 21x191 I might need less.

Which is good news for my i940, I can run 21x196 turbo off 4.1, and finally run near my stock memory using 8x multi, instead of either way underclocking or way overclocking memory, and use same lower vcore that 4.1 via permanent turbo on 23 multi x182 allowed.

[dj883u2]

my new 920

3837B072

Very good batch!

[dejanh]

MSR registers on cpu show max multi/turbo as 23,23,23,24 on i940, which is apparently what governs turbo, unless bios can set other limits. With turbo enabled i920 will always be at 21 multi, and i940 23 multi, regardless of setting multi in bios at 15, 17, 18 or whatever. If you want +2 multi, you have to deactive all cores but one in bios...so dont see purpose of +2 multi.

If Asus bios is doing something differently, would be good to know. It would be nice to be able to set turbo limits ourselves...so I can use 21 etc multis with my i940 with turbo enabled.

And yeah 4.1 on air is pretty good, Im on water and that is likely where I will end up for 24/7.

Asus BIOS is doing something significantly differently. On Asus boards turbo can only be used if the multiplier is set to Auto (i.e., default multiplier, but not explicitly keyed in). Otherwise, turbo is automatically disabled and becomes unavailable. So yeah, this is a big PITA. That is a huge implementation difference.

So to reiterate, if you manually specify the multiplier on an Asus board, even if it is 22x for i7 940 or 20x for i7 920, turbo will be disabled regardless and you have no option to turn it on.

Now, in terms of your post on lower multi x higher BCLK requiring less voltage, yes, that is true. I have noticed that as well. For the same clocks, I can usually lower the volts by about 0.01V - 0.02V if using higher BCLK over higher multiplier. This is completly contrary to what you would expect and detrimental to overclocking performance. On an AMD platform being able to increase the multiplier was always the pefered and cheaper way to overclock than raising the base clock generator value.

Btw, I decided to lower my 24/7 OC on air from 4.14GHz to 3.96GHz due to temps that I was not comfortable with. Indeed, I had TM function enabled which was reducing the multipler when the processor got too hot so I disabled it and then the temps peak out at about 97C on air. Not acceptable in my opinion, even though the system is stable. Reduction shaved off about 12C - 14C overall. Until I get better cooling I am not going higher. At least this way my memory runs at a healthy 1804MHz

[Cooperdale]

Hello, I have a strange problem with the configuration in my sig: I am stable at 4Ghz with temps below 80 if I use a 160x25 multi, but with 167x24 I can't get it stable for the life of me.

My voltages:

vcore 1.34375
dram 1.64
qpi/dram 1.35

I don't think raising the vcore should be necessary since the cpu frequency doesn't change. Which voltage do you suggest I tweak? I think qpi would help a lot, but can I safely go above 1.35 with that one? I seem to remember 1.35 as the limit declared by Intel. Am I wrong?

[rge]

That is interesting with the bios differences in turbo between Asus an GB, thanks for the info. Also good to know you found .01 to .02v lower on higher bclk same mhz clocks...I am 2 notches lower so far (.012), I need to keep going and see how low can get vcore.

No kidding with temps, even my dedicated PA120.3 for just cpu with GTZ is barely keeping my temps in check on prime load...hovers at 77C after 10hrs, 79-80max.

[dejanh]

Hello, I have a strange problem with the configuration in my sig: I am stable at 4Ghz with temps below 80 if I use a 160x25 multi, but with 167x25 I can't get it stable for the life of me.

My voltages:

vcore 1.34375
dram 1.64
qpi/dram 1.35

I don't think raising the vcore should be necessary since the cpu frequency doesn't change. Which voltage do you suggest I tweak? I think qpi would help a lot, but can I safely go above 1.35 with that one? I seem to remember 1.35 as the limit declared by Intel. Am I wrong?

Um, I think you got something mixed up here...

25x160MHz = 4.0GHz
25x167MHz = 4.175GHz

That's 175MHz difference between the two. Up your core voltage to about 1.40V and try from there...no need to increase QPI/DRAM voltage. Chances are however that unless you have a very good chip (i.e., one of the golden ones that run low volts and high clocks) you will most likely need about 1.46V - 1.47V at that clock though.

That is interesting with the bios differences in turbo between Asus an GB, thanks for the info. Also good to know you found .01 to .02v lower on higher bclk same mhz clocks...I am 2 notches lower so far (.012), I need to keep going and see how low can get vcore.

No kidding with temps, even my dedicated PA120.3 for just cpu with GTZ is barely keeping my temps in check on prime load...hovers at 77C after 10hrs, 79-80max.

Yeah, indeed. The temps are nuts on these chips. They seem to run hotter than even the 140W AMD parts One thing I am concerned about is that if I buy a nice WC system, and it only ends up shaving like 10C - 15C off the temps that may just get me about 4.2GHz (4.3GHz if I am extremely lucky) out of the chip for 24/7 use and I am not really sure that is really worth it for having spent like $400-$500 on a nice WC system...

After all, I am not going for any OC records here really...

[Xello]

Is that with 1.38v vcore and HT enabled, RGE?