E6600 overclocking & temperture problem

[Bleeble]

LOL Let me see if I can clarify things. The Delta is what is measured and reported by the Tjunction sensors in the CPU. It is the difference between the MAX Tjunction (determined by Intel and stored on the chip) and the actual temp. So [Core temperature = Tjunction - Delta]. There are two sensors for Tjunction, one in each core. These are the temperatures reported by CoreTemp and TAT (and Everest and Speedfan as "Core 0" and "Core 1"). This is the actual temperature of your CPU cores.

These are the dual Core temperatures measured within the hot spot of each Core. Thermal Junction temperatures of 75c is hot, 70c is warm, and 65c is safe. The dual Thermal Junction sensors are how Tjunction is measured, and are the dual Core temperatures displayed in TAT, and SpeedFan: Core 0 / Core 1.

While using CoreTemp, remember this: "(I) Core Temp 0.9X Tjunction 85c or 100c is Intel

[Systemlord]

There are two sensors for Tjunction, one in each core. These are the temperatures reported by CoreTemp and TAT (and Everest and Speedfan as "Core 0" and "Core 1"). This is the actual temperature of your CPU cores.

Tcase is a sensor that sits underneath the IHS between the two cores. This is what is reported by older versions of Speedfan and most other monitoring software as "CPU" temperature.

Basicly I need to keep the Tjunction (two sensors one in each core) temperatures below 75c hot, 70c warm & 65c safe, right?

Also I need to keep Tcase (CPU die temperature measured between the cores) under 60c hot, 55c warm & 50 safe. So I need to make sure that I monitor two aspects of the CPU temperature, right?

[Bleeble]

Basicly I need to keep the Tjunction (two sensors one in each core) temperatures below 75c hot, 70c warm & 65c safe, right?

Also I need to keep Tcase (CPU die temperature measured between the cores) under 60c hot, 55c warm & 50 safe. So I need to make sure that I monitor two aspects of the CPU temperature, right?

Honestly, I'd just keep an eye on your core (Tjunction) temps. Stay out of the 70s and you should be fine. Depending on what software you use, the Tcase measurement can be completely off. CoreTemp and Speedfan show my cores at 41C each right now, Speedfan reports CPU temps around 45... *shrug*

[Systemlord]

Honestly, I'd just keep an eye on your core (Tjunction) temps. Stay out of the 70s and you should be fine. Depending on what software you use, the Tcase measurement can be completely off. CoreTemp and Speedfan show my cores at 41C each right now, Speedfan reports CPU temps around 45... *shrug*

 

 

I fully understand the temps now, :thumbs-up: Speed Fan is showing 37c-39c @3.6GHz (vcore @1.56) with FSB @1600MHz & memory @1200MHz. I do have a few questions about FSB:DRAM ratio's. At stock speeds (2.4GHz) my FSB:DRAM ratio was 1:2, now at 3.6GHz with FSB @1600 the ratio is 2:3. How do I change the FSB:DRAM ratio? I saw a setting in bios called "modify ratio support" in the CPU configuration, not knowing what it does I did not touch it. I would like to run my FSB:DRAM at 1:1, but I don't know what setting or how to do it. I do know that the Asus mobo's don't have an option to run unlinked. Sorry if I'm asking to many questions.

Edited June 9, 2007 by Systemlord

[Bleeble]

Bah! Questions are good! That's how you learn. When you look in your BIOS, there should be an option called DRAM Frequency. That's your memory multiplier. Take a look at post #4 for more information: Asus P5B Overclocking and Tweaking Guide. No doubt you'll find that guide very helpful as you go along.

[Systemlord]

Bah! Questions are good! That's how you learn. When you look in your BIOS, there should be an option called DRAM Frequency. That's your memory multiplier. Take a look at post #4 for more information: Asus P5B Overclocking and Tweaking Guide. No doubt you'll find that guide very helpful as you go along.

 

Hay I just noticed that your at 1.44 Vcore @3.6ghz in your sig, I'm at 1.56 Vcore @3.6GHz. Is there a guide to Vcore setting per GHz range? Maybe I can lower my temps by going to a lower Vcore setting? I ran Orthos 11 hours 14 minutes no errors with TAT showing no more than 66c both cores. You think I might try 1.44 Vcore to?

Edited June 10, 2007 by Systemlord

[road-runner]

Hay I just noticed that your is at 1.44 Vcore @3.6ghz in your sig, I'm at 1.56 Vcore @3.6GHz. Is there a guide to Vcore setting per GHz range? Maybe I can lower my temps by going to a lower Vcore setting? I ran Orthos 11 hours 14 minutes no errors with TAT showing no more than 66c both cores. You think I might try 1.44 Vcore to?

That is something you have to do on your own, all chips are different. Mine will run @ 3.2 on 1.350 volt setting in the bios, but cpuz shows 1.3250 volts. It will run 3.6 on 1.51875 setting. Most people start at stock volts and go as far as they can on that and then up the vcore one incriment at a time until its stable, then increase the clock until it becomes unstable then give it a little more and so on. This can take some time to get you processor to run on the least amount of volts possible for the highest clock instead of running unneeded voltage therefore making extra heat. The cooler the higher you can usually go to a certain point. All chips are different....

Edited June 10, 2007 by road-runner

[Systemlord]

That is something you have to do on your own, all chips are different. Mine will run @ 3.2 on 1.350 volt setting in the bios, but cpuz shows 1.3250 volts. It will run 3.6 on 1.51875 setting. Most people start at stock volts and go as far as they can on that and then up the vcore one incriment at a time until its stable, then increase the clock until it becomes unstable then give it a little more and so on. This can take some time to get you processor to run on the least amount of volts possible for the highest clock instead of running unneeded voltage therefore making extra heat. The cooler the higher you can usually go to a certain point. All chips are different....

 

That info helped as I have kept my Vcore on auto, I 'm thinking that the mobo bios is giving my CPU way more voltage than it really needs just to remain stable. Tonight I think I'll lower the Vcore a little until it becomes unstable then raise it one or two settings up. Of corse this will take a few nights to get it at the point where I find an unstable Vcore. Who said it would be easy, lol.

[Bleeble]

Hay I just noticed that your at 1.44 Vcore @3.6ghz in your sig, I'm at 1.56 Vcore @3.6GHz. Is there a guide to Vcore setting per GHz range? Maybe I can lower my temps by going to a lower Vcore setting? I ran Orthos 11 hours 14 minutes no errors with TAT showing no more than 66c both cores. You think I might try 1.44 Vcore to?

There's no set volts-per-GHz guide. It's different for every CPU. I've seen people get 3.6GHz with only 1.4 Vcore while others need more than 1.5 for a stable OC. You can use someone else's OC (ideally many OCs) to get a ballpark figure for where you should be and go from there.

 

I'll generally set my OC and run some kind of stability test. If it fails, I bump the voltage up a bit and test again. If it passes, I increase the FSB and start the process again. I have to ask, did you just pick 1.56V or do you need that for stability? Whatever you do, make sure you don't put your Vcore too high or you'll greatly shorten the life of your CPU. That said, one thing a lot of people do is set a high(er) Vcore and keep increasing the speed until it becomes unstable. After that they take smaller steps.

 

On a side note, a single stability test is simply not enough if you really want a stable system. When I was OCing this CPU, I was able to get Orthos to run both Blend and Small FFTs at a given speed for over 24 hours each. But when I went to play Quake 4, the game would freeze after only a few minutes of play. After a bunch of messing around, it turns out my Vdimm was too low. I had another similar situation where I got really busy and left Orthos running for 73 hours. That has to be stable, right? Apparently not. Rome: Total War froze after a couple hours of play several times. Same issue. I'm not saying you should run three days of testing every time you increase your FSB. That'd be absurd (and hypocritical! Sometimes all I do is SuperPi 1M ). But when you're settling on a "final" OC, make sure you do more than run Orthos.

 

If you want to know a bit more about what happens when you don't cool your CPU correctly, read this article. It's old, but the principle still applies.

 

EDIT: Whoops! Didn't notice the third page... I guess I could have replaced much of that with "yeah, what he said."

Edited June 10, 2007 by Bleeble

[Systemlord]

There's no set volts-per-GHz guide. It's different for every CPU. I've seen people get 3.6GHz with only 1.4 Vcore while others need more than 1.5 for a stable OC. You can use someone else's OC (ideally many OCs) to get a ballpark figure for where you should be and go from there.

 

I'll generally set my OC and run some kind of stability test. If it fails, I bump the voltage up a bit and test again. If it passes, I increase the FSB and start the process again. I have to ask, did you just pick 1.56V or do you need that for stability? Whatever you do, make sure you don't put your Vcore too high or you'll greatly shorten the life of your CPU. That said, one thing a lot of people do is set a high(er) Vcore and keep increasing the speed until it becomes unstable. After that they take smaller steps.

 

On a side note, a single stability test is simply not enough if you really want a stable system. When I was OCing this CPU, I was able to get Orthos to run both Blend and Small FFTs at a given speed for over 24 hours each. But when I went to play Quake 4, the game would freeze after only a few minutes of play. After a bunch of messing around, it turns out my Vdimm was too low. I had another similar situation where I got really busy and left Orthos running for 73 hours. That has to be stable, right? Apparently not. Rome: Total War froze after a couple hours of play several times. Same issue. I'm not saying you should run three days of testing every time you increase your FSB. That'd be absurd (and hypocritical! Sometimes all I do is SuperPi 1M ). But when you're settling on a "final" OC, make sure you do more than run Orthos.

 

.

 

I have had my Vcore set to auto since day one as I didn't know any better, wish I had. I ran Orthos for over 11 hours and my gut tells me that the more the maryer. I have heard people running two different stress tests at once, is that ok? So I guess I need to run as many different stress tests as I can. Tonight I'll be lowering my Vcore step by step untill it becomes unstable then raise it up one or two setting in the Vcore. My temps now @1.56vcore with TAT is no higher than 66c Tjunction. :thumbs-up:

[road-runner]

I think someone that has a Asus board like you have posted here the other day that the vcore voltages automatically went up as they overclocked with it set on auto. I could be wrong but it sure seems like someone did. The reason I say that is 3.6Ghz on auto is kind of hard for me to believe that auto was putting out stock volts.

[Bleeble]

I have heard people running two different stress tests at once, is that ok? So I guess I need to run as many different stress tests as I can. Tonight I'll be lowering my Vcore step by step untill it becomes unstable then raise it up one or two setting in the Vcore. My temps now @1.56vcore with TAT is no higher than 66c Tjunction. :thumbs-up:

Generally, the only time I would run two stress tests is if the program is not multi-threaded or does not stress the CPU 100%. Programs like SuperPi and Prime95 will only stress one core, so you'd have to run two to get 100% load. Orthos and TAT, however, stress both cores.

 

Orthos is fine for stress testing. But when you want to make sure that your settings are really stable, try looping 3dmark while running TAT or Orthos. I say both because 3dmark won't load your CPU 100%.

 

You could probably just drop your Vcore down to 1.45V and just go from there instead of slowly inching your way down from 1.56. Either way works, but this is probably faster.