UnOfficial EVGA X58 Classified Owners Thread

[dejanh]

Not for 24/7 without subzero... believe me, for full stability there's no way to get more than 4.2-4.25GHz UC primestable with watercooling. Chiller may be an option without high humidity. But for benching, in case you do not fear high vdimm (you'd have to raise to get the right vtt-vdimm ratio) you can go some higher.

Can you please explain what you mean by VTT/VDIMM ratio? I keep hearing about this from you and cstkl1 but there is no clear explanation to what this means. I managed to get Prime stable Uncore of about 4.3GHz on water but I do not remember identifying any VTT/VDIMM ratios along the way...ATM I am running ~4.1GHz Uncore as it seems to be a good balance between speed and voltages required.

On a different note, did anyone test the new S22A BIOS for the Classified E759? I tried it and it completely breaks my S3 which has been working perfectly for many many months now. Had to go back to S227 for ability to S3.

[SAE]

There's a silent consent about vtt-vdimm ratio IMHO... i hear it alot in German forums. Truth is, there's an optimal distance between vdimm an vtt voltages, and it's to be kept for maintaining stability, speed and reliability of the memory system. Getting out of range too much either + or - results in bsods of 124 type, spontaneous reboots or plain instability.

My experiences have shown, that a difference between 200 and 240mV is optimal for my set of hypers and this 759 plus xeon cpu.


224mV^^


234mV^^

Vdimm has to be raised, even if there's no need to according to the great elpidas. That's just because vtt has to be increased exponentially - it seems - to maintain stability at high uncore clocks (1.324Vtt 4000 UC -> 1.46Vtt 4200 UC). You have to test it for yourself as hardware differs. Each board and set of ram is different, but you can take those values as a hint for what to try first. RTLs are directly coupled to this as well and have to be set up properly not to get hiccoughs.

Cheers,
Stefan. :cheers:

[davidm71]

Since everyone else has shared their photos thought I would share pictures of my baby. This was my first water cooled build and if I could do it over again I would have the rad facing the other way so that the nozzles are on the other side and installed a better waterblock.

[zalbard]

Nice looking rig!

[davidm71]

Nice looking rig!

Thanks.

The only thing is my center northbridge cooling fan here is dying after not much use! Ahh!

[Jabberwock]

Hi folks,

I'm still in the process of finalizing a new core i7 build but have decided on going with the EVGA classified 760 board.

I've heard that the NB on these boards runs hot. It would be great if people who own this board could post their NB idle and load temps so I can get a better idea of what to expect.

I've also heard that there is no way to directly monitor NB temps in Windows. Has this been fixed?

[nugzo]

Hi folks,

I'm still in the process of finalizing a new core i7 build but have decided on going with the EVGA classified 760 board.

I've heard that the NB on these boards runs hot. It would be great if people who own this board could post their NB idle and load temps so I can get a better idea of what to expect.

I've also heard that there is no way to directly monitor NB temps in Windows. Has this been fixed?

Everest shows NB temps for me. Its on my screensshot page 40 here, 3rd post I idle at 28-31 Load at 60-65.

[nugzo]

Can you please explain what you mean by VTT/VDIMM ratio? I keep hearing about this from you and cstkl1 but there is no clear explanation to what this means. I managed to get Prime stable Uncore of about 4.3GHz on water but I do not remember identifying any VTT/VDIMM ratios along the way...ATM I am running ~4.1GHz Uncore as it seems to be a good balance between speed and voltages required.

On a different note, did anyone test the new S22A BIOS for the Classified E759? I tried it and it completely breaks my S3 which has been working perfectly for many many months now. Had to go back to S227 for ability to S3.

There's a silent consent about vtt-vdimm ratio IMHO... i hear it alot in German forums. Truth is, there's an optimal distance between vdimm an vtt voltages, and it's to be kept for maintaining stability, speed and reliability of the memory system. Getting out of range too much either + or - results in bsods of 124 type, spontaneous reboots or plain instability.

My experiences have shown, that a difference between 200 and 240mV is optimal for my set of hypers and this 759 plus xeon cpu.


224mV^^


234mV^^

Vdimm has to be raised, even if there's no need to according to the great elpidas. That's just because vtt has to be increased exponentially - it seems - to maintain stability at high uncore clocks (1.324Vtt 4000 UC -> 1.46Vtt 4200 UC). You have to test it for yourself as hardware differs. Each board and set of ram is different, but you can take those values as a hint for what to try first. RTLs are directly coupled to this as well and have to be set up properly not to get hiccoughs.

Cheers,
Stefan. :cheers:

I'm real real new to i7. I'm currently reading and reading every chance i get to learn the technology. This is a noob question but how does uncore effect performance? Where would i see an advantage, real life or benchmarks, between say 3600 uc and 4200 uc? Thanks for answering a stupid question. And if somebody has a link to a real good informative, lamens i7 read, please post it for me. Thank You!

[Jabberwock]

Everest shows NB temps for me. Its on my screensshot page 40 here, 3rd post I idle at 28-31 Load at 60-65.

Ah, thanks for that, but pardon my curiosity. Could you check Everest to see if your NB temps reads the same as your cpu temp? (not the core readings but the single main reading) Do both of the temp sensors rise together equally under load?

I'm still having doubts that Windows will only report the cpu temp so the NB temp actually reports only the cpu temp.

Edit: I should of searched harder It seems people who have this motherboard report identical cpu and NB temps in Everest. Apparently, it seems there is still no way of monitoring NB temps in Windows as the NB sensor will just report what the cpu sensor finds. It seems it has something to do with the way the mobo can poll only one sensor at a time.

post 18 on techpowerup
http://forums.techpowerup.com/showth...76464&page=249

[dejanh]

Ah, thanks for that, but pardon my curiosity. Could you check Everest to see if your NB temps reads the same as your cpu temp? (not the core readings but the single main reading) Do both of the temp sensors rise together equally under load?

I'm still having doubts that Windows will only report the cpu temp so the NB temp actually reports only the cpu temp.

The Everest report for NB temp is incorrect. It is reporting CPU temp (albeit incorrect one).

[nugzo]

Ah, thanks for that, but pardon my curiosity. Could you check Everest to see if your NB temps reads the same as your cpu temp? (not the core readings but the single main reading) Do both of the temp sensors rise together equally under load?

I'm still having doubts that Windows will only report the cpu temp so the NB temp actually reports only the cpu temp.

Good Call. They do appear to be the same. 99% of the time they are identical. every now and again they'll be 1 or 2 degrees different. Whats crazy though is when i added 3 fans to the nb my load temps dropped 15 degrees.

[Jabberwock]

Thank you, I just re-edited my previous post right before you replied.

[Jabberwock]

Good Call. They do appear to be the same. 99% of the time they are identical. every now and again they'll be 1 or 2 degrees different. Whats crazy though is when i added 3 fans to the nb my load temps dropped 15 degrees.

I read your previous post on page 40 and you reported NB temps spiking to 90c. I would carefully monitor your cpu temps!

[nugzo]

I read your previous post on page 40 and you reported NB temps spiking to 90c. I would carefully monitor your cpu temps!

My cpu temps were/are fine. Using a TRUE, under load prime small fft or intel burn test never got over 80c. Most of the time under load was 74-77. that 90 spike went away after i added fans to the nb area. Here are some pics of what i added. I just added an 80mm and a 40mmm fan. they are sitting on the video card aiming at the NB HS, either side of the TRUE. For sure this wasn't enough to change the CPU temps much. Since the cpu temp doesnt correspond to the core temps in everest, maybe the cpu/nb temps are actually nb? or is it fact that there is no nb sensor? are there any x58 boards that have a nb sensor?




Oh and by the way, the aluminum 40mm fan has a leather pad veclrod to it so it isnt sitting directly in the vid card..

[Jabberwock]

Looks like you've got everything under control. Very nice setup

I'm planning on an i7 build myself with the Classified and a TRUE120 on a Torture Rack. I just wish the mobo would poll the NB instead of the cpu as i would still be able to monitor the individual core temps and that's what really counts. So far I still cannot get accurate info on what the NB load temps are like

[SAE]

I'm real real new to i7. I'm currently reading and reading every chance i get to learn the technology. This is a noob question but how does uncore effect performance? Where would i see an advantage, real life or benchmarks, between say 3600 uc and 4200 uc? Thanks for answering a stupid question. And if somebody has a link to a real good informative, lamens i7 read, please post it for me. Thank You!

Mate, the uncore is infact the L3 cache and the on-die memory (+ qpi-) controller. That's why its clocks influence L3 cache speeds and memory access times and bandwidth.

You would not notice a higher uncore clock during day-2-day usage as it's more like a benching advantage or for the folks who must know they're squeezing every tiny oz of performance out of the poor hardware parts.

The high uncore speeds get you higher scores in superpi, aquamark 3, 3dmark05 and so on... if you don't do any benching with your rig, you don't have to bother.

[nugzo]

Mate, the uncore is infact the L3 cache and the on-die memory (+ qpi-) controller. That's why its clocks influence L3 cache speeds and memory access times and bandwidth.

You would not notice a higher uncore clock during day-2-day usage as it's more like a benching advantage or for the folks who must know they're squeezing every tiny oz of performance out of the poor hardware parts.

The high uncore speeds get you higher scores in superpi, aquamark 3, 3dmark05 and so on... if you don't do any benching with your rig, you don't have to bother.

Thanks so much. I did up uncore to 4200 and had a definite increase on memory scores.

[SAE]

Never mind. 4200 is kind of a huge jump though... be careful with the vdimm-vtt ratio. It's to be kept smaller than 0.5V.

[Sam_oslo]

My cpu temps were/are fine. Using a TRUE, under load prime small fft or intel burn test never got over 80c. Most of the time under load was 74-77. that 90 spike went away after i added fans to the nb area. Here are some pics of what i added. I just added an 80mm and a 40mmm fan. they are sitting on the video card aiming at the NB HS, either side of the TRUE. For sure this wasn't enough to change the CPU temps much. Since the cpu temp doesnt correspond to the core temps in everest, maybe the cpu/nb temps are actually nb? or is it fact that there is no nb sensor? are there any x58 boards that have a nb sensor?


[IMG]...[/IMG]

Oh and by the way, the aluminum 40mm fan has a leather pad veclrod to it so it isnt sitting directly in the vid card..

There is no NB-sensor on DFI either, but the PWM/system gets hot rally hot under load (specially with hight OC, and high VVT) and should be monitored/cooled carefully.
Putting a good fan on the NB would cool the PWM/system, NB and SB (all under the same heat-pipe) effectively on DFI.
Here is my Intel stock CPU-fan mod (on reduced speed), which has been hanging on the NB for almost 1 year now:

[nugzo]

Never mind. 4200 is kind of a huge jump though... be careful with the vdimm-vtt ratio. It's to be kept smaller than 0.5V.

so keep these 2 within .5v of each other? Is that what your saying or did i mis understand you?

[SAE]

Yep, exactly^^.

[Splave]

good lord that is alot of vtt man

[dejanh]

Yeah, 4.2GHz Uncore definitely does not require that much VTT. Maybe ~+300mV.

[SAE]

Depends on cpu imc and ram. We can be sure it's not the classified's pwm

[dejanh]

Depends on cpu imc and ram. We can be sure it's not the classified's pwm

True