Thermalright Ultima 90 **pics*temps**

[MaDNe55]

So the day finally arrives after my long 5 day wait. My Ultima 90 was delivered

Installation was pretty much a pain. I pretty much had to unplug everything to take the mobo out and install the backplate. I have to admit... I fudged up and installed the brackets the wrong way so the HS was facing up... the wrong direction. I even told myself not to do that when I Was reading the online instructions a few days ago. Luckily, I managed to swap the brackets in the right direction without having to take the mobo out again. Haha! Lucky, Lucky. Anyways, here it is mounted with a Scythe Flex F model:

The HS base was very flat. As far as I could tell it was near perfect. The finish was semi-mirror like. It had a nice haze to it and there were almost microscopic finish lines. The "nail test" hardly revealed any ridges nor did it make sound... and my nails are a bit long right now. I still have a Nexus 92mm fan to add to the other side. I am going to do that tomorrow when I get my Zalman mfc2 fan/temp controller and attempt 3.6 ghz.

TR Ultima 90
Current Stats:

e6750 @ 3.06ghz:
IDLE: 25C
LOAD: 43C (orthos @ 10 mins)

e6750 @ 2.66ghz:
IDLE: 22C
LOAD: 40C


Stock Intel HSF:

e6750 @ 3.06ghz
IDLE: 32C
LOAD: 58C

e6750 @ 2.66ghz
IDLE:28C
LOAD: 56C

As one would determine from those numbers; I am VERY pleased!!! I consider these numbers to be a HUGE difference and I am sure I will have no problem getting up to my goal of 3.6ghz.

I made a mistake when I was making the screenshots while idleing. I didnt realize it at the time but when using MWsnap to take screens the cpu would spike taking the temps with it and I didnt get accurate readings. My mistake. I did remeber what they were however

I will post more pics later as well as more data. I will be able to give ambient temps tomorrow when my Zalman mfc2 arrives. I will also install the Nexus 92mm as a pull to see what kind of temp difference it will really make. My prediction is that it wont make much of a difference but we shall see.

 

[bing]

Thanks for sharing !

If you running at stock speed, you might try lower vcore instead of leaving it at AUTO setting, you'll get much more lower temp.

 

[MaDNe55]

Thanks for sharing !

If you running at stock speed, you might try lower vcore instead of leaving it at AUTO setting, you'll get much more lower temp.

Thanks for the tip. I ran it on stock speed for 10 minutes just to capture temps. I am now running at 3.06ghz which I was running with the stock intel HSF. Tomorrow I shoot for 3.6ghz

 

[Metallica]

Hmm .. your temps seem .. odd ... at least to mine they do.

I you were tunning 22c Idle, then im not that far off idle # stock .. mine were about 16. - 18c idle ..

I guess we'll see with the higher O/C ...

I'm also interested in what it's gonna take to get to 3.6GHz stable.

For me i need 1.48v vcore ... Let's see if thats what you need too.

 

[UnrealAlex]

Thats about what he'll need; I'm stable at 3.4 at 1.418, load temps are about 50C
His temps seem correct to me, I think around 40 is where I was at at 3.00GHz as well.
Keep us posted, I really want to know if the 92mm helps =P Also if you're bored you can try and see what kind of temps you'll with just a push 92mm xD

 

[Trottel]

What will you do if 3.6 does not have enough stability?

 

[Metallica]

What will you do if 3.6 does not have enough stability?

3.6GHz shouldn't be too hard ...

 

[MaDNe55]

What will you do if 3.6 does not have enough stability?

Play Quake wars @ 1680x1050 full detail at 3.4ghz.


Cheers

 

[MaDNe55]

Hmm .. your temps seem .. odd ... at least to mine they do.

I you were tunning 22c Idle, then im not that far off idle # stock .. mine were about 16. - 18c idle ..

I guess we'll see with the higher O/C ...

I'm also interested in what it's gonna take to get to 3.6GHz stable.

For me i need 1.48v vcore ... Let's see if thats what you need too.

What exactly seems odd? I am getting 15C difference LOAD from stock HSF. If you are getting 30/50 at 3.6.. I am predicting that I will be AT or within 2-3C of your temps based on the current numbers.

 

[bing]

Suggestion folks, better start the Intel cpu temp lingo in "how many degrees to max tjunction" instead of the absolute temp reading cause its meaningless, and usually will lead to confusion when discussing it.

Click my sig on "Using the Coretemp the RIGHT way" to straight this issue out.

 

[MaDNe55]

Suggestion folks, better start the Intel cpu temp lingo in "how many degrees to max tjunction" instead of the absolute temp reading cause its meaningless, and usually will lead to confusion when discussing it.

Click my sig on "Using the Coretemp the RIGHT way" to straight this issue out.

All's that does is take my temp and subtract it from my Tjunction. So 100 tjunction - 25C = 75C to tjunction remainining.

I'm not sure how reversing the numbers makes it more accurate. Does it have something to do when some people's Tjunction displaying as 85C?

 

[bing]

Agree, still the same result, but it will be confusing for newcomer or worst misleads them when they deal with cpu temp, by embracing this kind of reading method, I really "hope" it will reduce the confusion. Maybe I'm hoping too much !

It doesn't matter on that tjunction either on 100 or 85 or even 50 C, cause you will never can get an accurate temperature anyway, it is the relative/distance to that max tjunction thats matter.

About that "real" maximum tjunction temperature, nobody knows and even Intel doesn't care about it.

That 85 or 100 numbers actually came from attitude like this -> "If it is too hot just use 100 or if it looks too cold use 85". See how ridiculous is that ?

How about -> "Use what ever tjunction temp what you 'want' to see & believe to make your CPU temperature looks bragging cool". LOL

Hope you understand what I'm up to !

Sorry for thread jacking, will stop now !

 

[MaDNe55]

Ok I found a killer write up on everything you wanted to know about CPU temps and more. You can find it here.


But here is an excerpt to answer our current question of why CoreTemp is wrong (and right).

"Note 1: Core Temp is an excellent utility, however, it has a fundamental flaw in terminology, which creates confusion in the CPU temperature community, by obscuring the distinction between temperature and specification. Core Temp shows Tjunction 85c (or 100c), which is an incorrect term. The proper expression is Tjunction Max 85c (or 100c), which is the term defined by Intel as shown above in the Specifications section, and as represented below:

Tjunction = Core temperature
Tjunction Max = Shutdown

(A) Junction Temperature is a thermal measurement because it scales, thus the term TJ, or Tjunction, which is synonymous with Core temperature.

(B) Maximum Junction Temperature is a specification because it does not scale, thus the term TJ Max, or Tjunction Max, which is synonymous with Shutdown.

(C) ~ 5c below Tjunction Max Throttling is activated. If Tjunction Max is reached, Shutdown occurs, which is either 85c or 100c, and is determined by Stepping. "

So I guess some C2Ds can run hotter than others... Metallica wasnt your TJ "MAX" 85c? I've always wondered how CoreTemp made some people's tjmax 85c and others 100c.

So anyways... the TJmax in reality is what is important because you dont want to get close to it otherwise your CPU is gonna be hurting. However the tjunction(Core temp) is still the same regardless of TJMAX and therefore for practical benchmark purporses I think it is still better to report the Tjunction or core temp.

 

[bing]

That THG guide is misleading IMO cause that 85c and 100c are baseless, no reference or supporting fact/documents that leads to that numbers.

They could be 95C or even 110C !

With the current silicon wafer that has millions of silicon junctions, it is almost impossible to have a single temperature reading that can represent "properly" the whole silicon die like this HERE.

That is why Intel sort of burn in this max. tjunction level into the CPU after they tested during the testing or binning process at factory to determine that level and it could be different from CPU to CPU.

For example and simple analogy, lets say there is a cpu die when loaded, 99.999999% of it's surface capable of withstand up 100C without any error , "BUT" there is a single "weak" spot in that die that will go crazy/error when it reached 80C, that is where Intel mark that max. tjunction for that particular die.

Now the at the other die, 100% of it's surface can operate without error upto 85C, but once it crossed the 86C, it starts to go crazy. Then on this cpu, the max tjunction will be set at that 85C !

Check at the Intel section for C2D Thermal sticky thread, there is a reference will help you to understand that and also fyi, at C2D silicon die, Intel put lots of thermal sensor (more than 1) througout the silicon surface.

Now don't you agree that Intel doesn't need to know what is the true/real temperature of that CPU nor care about it ?

 

[Metallica]

Ok I found a killer write up on everything you wanted to know about CPU temps and more. You can find it here.


But here is an excerpt to answer our current question of why CoreTemp is wrong (and right).


So I guess some C2Ds can run hotter than others... Metallica wasnt your TJ "MAX" 85c? I've always wondered how CoreTemp made some people's tjmax 85c and others 100c.

So anyways... the TJmax in reality is what is important because you dont want to get close to it otherwise your CPU is gonna be hurting. However the tjunction(Core temp) is still the same regardless of TJMAX and therefore for practical benchmark purporses I think it is still better to report the Tjunction or core temp.

Yea, mine reads 85c Tjunction .. which is wrong ...

That's why i assume + 15c to my readings .. i get 15c idle / 35c load @ 3.6GHz

 

[MaDNe55]

Yea, mine reads 85c Tjunction .. which is wrong ...

That's why i assume + 15c to my readings .. i get 15c idle / 35c load @ 3.6GHz

Where oh where does it say that if your TJMAX is off then your tjunction is off? If TJMAX is attained according to stepping then the 85C or 100C in coretemp is not reading that from any sensor on the cpu. Alls that number is is the max temperature you can reach before cpu shutdown.

 

[MaDNe55]

That THG guide is misleading IMO cause that 85c and 100c are baseless, no reference or supporting fact/documents that leads to that numbers.

They could be 95C or even 110C !

Ok there has to be some offical intel documentation that gives a standard TJMAX rating to all cpu's based on model and stepping. There are no "dynamic standards"...

With the current silicon wafer that has millions of silicon junctions, it is almost impossible to have a single temperature reading that can represent "properly" the whole silicon die like this HERE.

That is why Intel sort of burn in this max. tjunction level into the CPU after they tested during the testing or binning process at factory to determine that level and it could be different from CPU to CPU.

For example and simple analogy, lets say there is a cpu die when loaded, 99.999999% of it's surface capable of withstand up 100C without any error , "BUT" there is a single "weak" spot in that die that will go crazy/error when it reached 80C, that is where Intel mark that max. tjunction for that particular die.

Now the at the other die, 100% of it's surface can operate without error upto 85C, but once it crossed the 86C, it starts to go crazy. Then on this cpu, the max tjunction will be set at that 85C !

Check at the Intel section for C2D Thermal sticky thread, there is a reference will help you to understand that and also fyi, at C2D silicon die, Intel put lots of thermal sensor (more than 1) througout the silicon surface.

Now don't you agree that Intel doesn't need to know what is the true/real temperature of that CPU nor care about it ?

Well they dont have to care if they expect the people who buy their products to be sheep and only run at stock speeds with stock HSF. But then again there are so many factors that determine CPU temp. If I live in florida during the dead of summer and my A/C breaks I'd better know how hot I can run my CPU. So Intel should have a duty to set standards for TJMAX. Everyones ambient temperature wont be an ideal 22C.

I read that sticky and it basically says the same thing as the THG thread. I'm still not understanding why you want to read your temps backwards from TJMAX when you yourself say that the TJMAX # is worthless. Also, maybe you could tell us how coretemp is getting the TJMAX #s?

CoreTemp may not be as accurate as intels "trench" method. But who is going to do that? Coretemp is using the same calculations to get the temp on everyones computer. So although it may be off it is a standard point of reference that everyone uses and is accurate in that sense. I.E. Metallicas u120x is going to get better temps than my ultima 90... Coretemp cant distinguish who's heatsink dissipates more heat so why are his temps better? Because coretemp is accurate in a sense

Also we are all keeping a general rule of thumb not to go over 60C load. This is obviously very far from TJMAX on newer cpu's so there shouldnt be a problem if we adhere to those standards.

But im not sure what metallica is talkin about by adjusting his temps based on coretemps assumed TJMAX

 

[Rattle]

i'm interested in one of these coolers, I will have the same mobo and same PSu as you BTW. Anyway you can ramp that cpu to 3.6 or so and run orthos for 30minutes for me ?
1 fan then 2 would be awesome if you can.

thanks

 

[MaDNe55]

i'm interested in one of these coolers, I will have the same mobo and same PSu as you BTW. Anyway you can ramp that cpu to 3.6 or so and run orthos for 30minutes for me ?
1 fan then 2 would be awesome if you can.

thanks

Thats tonights plan

p.s. I used to live in FL and MA but I am now in a better place... Southern California.

 

[Rattle]

never been out that way but i would like to check it out.
I love it here where I am now, i went back to MA in may after 4 years in FL and i was moved back down to tampa area by september. Hate it up there and I was there all my life.

cool man good luck with the results, i have a ninja rev b i scored for $23 shipped to hold me over lol. Anyway I look forward to seeing how it does tonight.