Core i7/X58 Overclocking Thread

[D3mon_Hunt3r]

eva2000 do you mind sharing voltages for that run?

[eva2000]

cpu = 1.25v bios
mem = 1.68v bios
cpu vtt = 1.44v bios

[Xello]

Seems Uncore clock is the secret to high bandwidth, feed it with vdimm and watch your scores go up!

I got this with my cheap mid range corsair kit @ 1.78v (4ghz uncore)

[D3mon_Hunt3r]

cpu = 1.25v bios
mem = 1.68v bios
cpu vtt = 1.44v bios

WoW
That low VTT and mem. For those mem speeds. Are you using that new breed of chips that we've been seeing popping around in XS?

[eva2000]

WoW
That low VTT and mem. For those mem speeds. Are you using that new breed of chips that we've been seeing popping around in XS?

nah old DDR3-2133Mhz 9-9-9-24 at 2.0 rated stuff - samsung hcf0 http://i4memory.com/f83/corsair-tw3x...x48-t3rs-9904/

[D3mon_Hunt3r]

It's pretty good doing 2000 at 1.65v... IMO.
Thinking if i should be getting this one or the Bloodrage...
How are the memory settings? A lot of them like in X38/X48?

[eva2000]

Bios template for DFI X58-T3EH8 http://i4memory.com/f80/dfi-lp-ut-x5...emplate-12919/ not many memory settings

[rge]

Seems Uncore clock is the secret to high bandwidth, feed it with vdimm and watch your scores go up!

pretty sweet...I hadnt played with mem/uncore too much yet ..even got 4200 uncore by increasing dram volts along with qpi/vtt. I also need to get a paid version of everest

[D3mon_Hunt3r]

Bios template for DFI X58-T3EH8 http://i4memory.com/f80/dfi-lp-ut-x5...emplate-12919/ not many memory settings

Tks a lot. A lot of voltage settings i see...
If i get sick waiting for the Bloodrage i'll go for DFI again... :P
Been having quite the issues with X48 T2R and HD4870X2.

[Xello]

pretty sweet...I hadnt played with mem/uncore too much yet ..even got 4200 uncore by increasing dram volts along with qpi/vtt. I also need to get a paid version of everest

Yeah... It seems uclk is fed by vdimm rather than VTT/QPI, so you can keep VTT low as long as you're not pushing bclk higher. This also dispels that stuff about needing 8:9 / uncore higher than QPI being unstable, i'm totally stable and much higher right now

[Lukee]

This also dispels that stuff about needing 8:9 / uncore higher than QPI being unstable, i'm totally stable and much higher right now

You know, you may be on to something here.. If anyone has read my issues, I crash eveery time i try to Run Everquest II. I raised Vdimm and qpi/vtt a bit. I actually was able to start loading the zone for Eq II before it locked up. i had never gotten that far, so I went back and raised them even higher and I was able to log in and run around. So.... I am running Prime 95 atm and will see if it is stable, but so far so good..

Edit - I pushed it and I'm running 1:1 ratio Qpi:uncore

36x 18x 8x
Qpi/uncore/memory

1 hour and 26 minutes Prime stable and counting, but able to load and play EQ2, so it's working so far.



CHEERS!!

[safan80]

Here's what I have so far

same clock ( 3.7Ghz 20x189) as what I had with the evga board but with less vcore.

[eva2000]

Tks a lot. A lot of voltage settings i see...
If i get sick waiting for the Bloodrage i'll go for DFI again... :P
Been having quite the issues with X48 T2R and HD4870X2.

no probs with hd4870x2 on DFI X58

More DFI LP UT X58-T3EH8 results

Vcore = 1.3375v
VDIMM = 1.680v
CPU VTT = 1.440v





Vcore = 1.25v
VDIMM = 1.815v
CPU VTT = 1.480v





Vcore = 1.25v
VDIMM = 1.830v
CPU VTT = 1.480v

[SteveRo]

When folks calibrate real temp, what are are you coming up with, mine looks like -5C to get to room temp +7F for 77F - see -

http://www.techpowerup.com/realtemp/docs.php

Also - looks like a CANNOT get 4.2 with ht on - on air (thinking about a boreas!) -

[Chicken Patty]

i'm having issues running 3dmark so ill see if I can get it working, but for now its been stable for a few hours of benching, Generic CPU bench, WPRIME, etc

[rge]

When folks calibrate real temp, what are are you coming up with, mine looks like -5C to get to room temp +7F for 77F - see -

http://www.techpowerup.com/realtemp/docs.php

Also - looks like a CANNOT get 4.2 with ht on - on air (thinking about a boreas!) -

I would not calibrate real temp with core i7's, tjmax is read from cpu, and most temps seems to be pretty reasonable from idle to load. But if you do calibrate idle, it is 7C plus ambient in C at idle, undervolted, underclocked state, and that is for duo/quad, no one has tried calibrating for core i7 yet.

[Lukee]

Quick shot of 200x21 turbo on
36x Qpi
16x uncore
8x memory

Right now, I'm working on

36x Qpi
20x Uncore
8x Memory

Will post those once I get back.

[SteveRo]

rge - yep, under volted, underclock, based on the directions (granted for pre i7) real temp was showing 5C over what it should have been.
This should only matter pre 60C based on the directions - http://www.techpowerup.com/realtemp/docs.php

"Calibration

Using this knowledge and some more measurements, I came up with some simple correction factors to help compensate for errors in the DTS at idle. Using an Idle calibration setting in Real Temp of (-2) for my E8400 and using a setting of (+2) for my E6400 now gives me very accurate temperatures for both processors from 10ºC to 100ºC. Typically within 1ºC or 2ºC of what an IR thermometer shows near idle when pointed at the bare core and exactly equal higher up. The formulas were designed to bring the above two processors a lot more in line with each other which they should be, given the identical core voltage. Power or heat generation is proportional to the square of voltage so that is the most important factor to equalize when making comparisons like this.

If you want accurate idle temperatures then you need to calibrate Real Temp to your processor. Avoid changing the default TjMax that RealTemp uses.

The first thing to determine is if your processor reads too high or too low at idle. My method involves reducing the heat output of your processor to a minimum. You do this by lowering the clock frequency (MHz) your processor runs at as well as the core voltage. It's also important to run your CPU fan at its maximum speed. I like to go into the BIOS and set the front side bus speed to 266 MHz and then I set the multiplier to 6.0 which results in an overall speed of approximately 1600 MHz. I also lower the core voltage as low as it will go. My ASUS motherboard goes as low as 1.08 volts which is fine for this test. You can also accomplish the same thing by enabling SpeedStep and C1E in the bios. Start up your computer and use a program like CPU-Z to verify that it is running at approximately 1600 MHz and 1.10 volts when sitting idle at the Desktop.

By reducing your processor's heat output down to a common fixed value, I've found that a well air cooled Dual Core processor will idle at approximately 4ºC to 5ºC above room temperature when the computer case is open. Quad Core processors will idle at 6ºC to 7ºC higher than your room temperature. If your reported temperatures during this test are less than your room temperature then you need to open up the RealTemp Settings window and enter a positive Idle Calibration factor to boost them up into a more believable range. If your CPU is like my E8400 and you notice reported temperatures that are higher than 5ºC above your room temperature at idle, then you'll need to use a negative Idle Calibration factor to bring your temperatures down. In a Dual or Quad Core processor, the temperatures reported during this test should be equal across all of the individual cores.

That's all there is to it. As long as TjMax is chosen correctly and you do a simple calibration, your reported core temperatures will be very accurate and comparable to other users who are using the same software and have also individually calibrated their processor. This is not an exact science. It's just a way to compensate for a DTS sensor that was never designed or calibrated by Intel to report accurate idle temperatures.

Though this test is done when your processor is idle, the calibration settings entered will improve the accuracy of your reported tempertures up to approximately 60C. At this point and higher, most digital thermal sensors are very accurate and no further correction is necessary"

[rge]

Yep the calibration for realtemp was based partly on some of my testing with drilling hole through IHS and then into die and measuring with thermocouple vs software reading. And yep the calibration cuts out around 60C, it is just for idle. But given the core i7s seem to idle at reasonable temps, so far I dont think any calibration is necessary, or at least not like other 65nm or 45nm core duos.

On water I am idling at 8C above ambient, which is probably pretty close to the truth for 6.75W per everest. I would expect on air for you to be a few C higher at that higher watts. But load default settings in bios, then enable C3/C6/C7 deeper sleep states and disable HT, I went down to 2W according to everest, and my temps were then 7C above ambient. I just put pic in realtemp thread of mine at both temps. The spread between water and air is going to increase as watts increase and should decrease as watts decrease, so if you are 12C above ambient, you might come down to 10C above or so. Also make sure your fan is on high when you do the test, and you turn your computer off for about 15 mins or use a fan to blow the internal case temps down to ambient, b/c calibration is with open case.

Had not thought about checking idle temps out with i7, but it is interesting that given similar watts output, it is showing same as I saw with diff cpu, otherwise using similar cooling/setup.
http://www.xtremesystems.org/forums/...35#post3530235,
last post

[SteveRo]

rge - if I understand you right - @ idle - real temp is showing approx 7C higher than it should for your set up?
By the way, thanks for all the help - unclewebb and yourself - re - real temp - a great app!

[rge]

rge - if I understand you right - @ idle - real temp is showing approx 7C higher than it should for your set up?
By the way, thanks for all the help - unclewebb and yourself - re - real temp - a great app!

No, real temp is reading exactly as it should on mine without calibration. It is Unclewebbs app, I am one of his guinea pigs

Core 2 duos/quads idle at about 7C above ambient with high end water cooling (at UC/UV state), via thermocouple to core. So temps on realtemp should be about 7C above ambient for high end cooling.

Core i7 with HT off is also reading about 7C above ambient with basically same cooling system and same watts.

So what I am saying is core sensors seem to be accurate on my i7, and thus realtemp seems to be accurate, and is reading exactly what it should at idle, without calibration.

You probably should be idling at 9-10C above ambient with your air, and my guess is if you had case open so internal case ambients = ambients, and you turn HT off and enable c5/etc, and turn cpu fan on high, your temps would be accurate, and again no need for calibation.

[SteveRo]

vge - very good I will test again - last time I had c1 enabled and not the others, this time I will enable all the c's, turn off HT, turn blockxmulti to lowest and let it sit for a bit and see what I get. My computer closet is probably a bit higher than the rest of my basement - even with the door open - probably - 68f or so.

[Chri$ch]

max on water





i will try later more

[eva2000]

max on water

http://www.abload.de/img/4.45ghz4feh.png

http://www.abload.de/img/3guvh.png

i will try later more

nice

my progress so far with 30C room ambient TRUE120 air cooling

i7 920 and Turbo Mode Enabled
With Turbo Mode enabled you'll have access to 21x multiplier. I tested 21x205 and 21x210 for Super Pi 1M, 8M, 32M and 3dmark06.

i7 920 @4300Mhz

21x 205Bclk = 4300Mhz at 1.45v bios (1.488v idle windows)
DRAM Frequency set to 1333Mhz (1333Mhz/133Mhz = 10x mem multiplier) = DDR3-2050Mhz at 1.725v
Uncore Frequency set to 2666Mhz = 4096Mhz Uncore frequency

• Super Pi 1M = 9.468s
• Super Pi 1M = 1m 47.484s
• Super Pi 1M = 8m 32.516s
• 3Dmark06 = 27,087

i7 920 @4410Mhz

21x 210Bclk = 4410Mhz at 1.4875v bios (1.52v idle windows) for 32M and 1.525v bios (1.55v idle windows) for 3dmark06
DRAM Frequency set to 1333Mhz (1333Mhz/133Mhz = 10x mem multiplier) = DDR3-2100Mhz at 1.875v
Uncore Frequency set to 2666Mhz = 4200Mhz Uncore frequency

• Super Pi 1M = 9.234s
• Super Pi 1M = 1m 44.734s
• Super Pi 1M = 8m 19.344s
• 3Dmark06 = 27,360

[safan80]

Quick shot of 200x21 turbo on
36x Qpi
16x uncore
8x memory

Right now, I'm working on

36x Qpi
20x Uncore
8x Memory

Will post those once I get back.

what voltages are you using?