T.I.M battle – Old vs. New

My friend Ramiro did a versus between new and old TIM's and he has something new in his methodology: GPU's. One GTX480 with 4 TDP's and one HD5870 with 3 TDP's.
Have fun.

LE:Ramiro's conclusions on short:

ARCTIC SILVER 5
A nice behavior compared to latest TIM's. If you have a decent TDP there is no reason to ditch your good ol' friend AS5.

ARCTIC COOLING MX-4
The goal to produce a MX-3 performance with better viscosity was achieved. Top performance TIM recommended both for casual users and enthusiast users.

PROLIMATECH PK-1
Is one of his favorite TIM because is easy to apply and it has very good performance.

THERMALRIGHT CHILL FACTOR 3
Optimal viscosity and top performance. Silver Arrow and HR-02 come with it and kudos to TR for that.

ARCTIC COOLING MX-3
Top performer among traditional TIM's. Tricky to spread but again a top TIM.

ARCTIC COOLING MX-2
Still OK for casual users who doesn't want to squeezes the last °C from their CPU's.

Bottom line, MX-2/AS5 are still OK for majority of users and new TIM's performs practically almost the same.

What no Ceramique?

Main goal of his article was testing MX-4.
But he pick what he has around to compare them with MX-4.

burebista said:

Main goal of his article was testing MX-4.
But he pick what he has around to compare them with MX-4.

Click to expand...

I know, I was just yankin' your chain

Nice list tho

I didn't realize MX-4 came out.

I have an unopened tube of GC-Extreme here; I wished he had tested that as well.

I was going to get MX-2 again as I've heard MX-3 is hard to work with. I read some reviews and decided to try something new.

jason4207 said:

I have an unopened tube of GC-Extreme here; I wished he had tested that as well.

Click to expand...

He tested it one year ago with very good results.

Wait a second here... TDP is essentially the cards output. That is the same across the family of cards right? So how did he receive cards with different TDPs? Is that term being used wrong by me or this guy doing these reviews?

He seemed ot have used different stress testing applications/games for each TDP? What does that show...nothing to me. It needed to be the same program across the board.

I know this was a lot of effort but either I missed the point or this testing is utterly pointless between the different TDP's.

Wow.. Didn't know MX-4 came out.. Just opened a tube of GC-Extreme since I just ran out of MX-3..

And like ED mentioned, I'm confused with the use of the term TDP there..

Look at TDP as heat generated. From stock to OC and overvolt.

GTX480

TDP1 – 700mhz/0.99v/stalker/1300rpm/140mm

TDP2 – 700mhz/0.99v/furmark/1300rpm/140mm

TDP3 – 800mhz/1.138v/unigine/1300rpm/140mm

TDP4 – 700mhz/1.10v/combustor/1300rpm/140mm

HD5870

TDP1 – 850mhz/1.125v/stalker/1300rpm/140mm

TDP2 – 850mhz/1.125v/furmark/1300rpm/140mm

TDP3 – 1000mhz/1.250v/furmark/1300rpm/140mm

i7 920

TDP1 – 3500mhz/1.25vcore/1.35vQPI/prime95 large fft/10 min/800rpm

TDP2 – 4000mhz/1.37vcore/1.45vQPI/prime95 large fft/10 min/1600rpm

TDP3 – 4300mhz/1.47vcore/1.55vQPI/prime95 large fft/5 min/6000rpm

AH.. I see what he did now (Doh?) LOL

I guess, but thats like calling Vcore VID. VID = stock volts, and TDP = stock wattage/heat output or at least its how its generally used.

So that still begs the question, why different test methods for each TDP? One test cannot be used with another rendering this significantly less valuable to anyone really.

The term may not be used exactly right, but the point seems to be how the TIMs hold up to different levels of heat. Each TDP represents a different heat load.

Ahh, yes. Thanks for the clarity jason4207! But with different tests on each load, doesnt that make the results only comparable for each "TDP"? I guess I just dont see the point in that. I oulw have rather seen each TIM tested under a specifc heat load each time.

Sorry to be a negative nancy on this bureb...

EarthDog said:

Sorry to be a negative nancy on this bureb...

Click to expand...

No problem man, criticism is always welcome.
I've showed your thoughts to Ramiro and when he'll have some time he promise me an answer here so you'll have an explanation straight from horse's mouth.
Thanks jason4207 for clarification. Sorry for my English, sometimes is hard for me to put my Romanian thoughts in English text.

I guess my concern is, how does he know that Stalker or anything other than Furmark or Combuster is putting up that TDP consistently?? How does he know the TDP of the card mixed in with the system power draw? Gaming benchmarks are too variable in their load to show anything more than peaks.

I would have loved to have seen something like this...

Pick a test (most common is Furmark)

Furmark at stock clocks and voltages.
Furmark at max overclock and stock voltage.
Furmark at max overclock with max voltage.

That way the test was the same (furmark) and you can compare at varying but consistent outputs since hte test used is the same and the only difference was clocks speeds and voltages through the testing and be consistent across the gamet (sp?).

Looking forward to his thoughts!

A shoot in the dark here. I guess he want to show us a difference between daily use (gaming) and synthetic tests (Furmark/Kombustor).
And also you can see from graphs that for ATI it really doesn't matter what TIM you use because naked chip makes a perfect contact with Spitfire base and you a have a very thin TIM layer.
But for nVidia things looks different. Ramiro thinks that both IHS and Spitfire's base were concave so on the middle on IHS TIM layer was thick. That's why (he thinks) he saw those differences at extreme TDP (erm, extreme heat).
But let's wait and see his explanations.

Testing on GPU's is not the best comparison. Shin-Etsu 7783D not tested ?

I want to have a lot of situations, from where all class of users can take a pick.

Normal users do not play furmark, do not overvolt/overclock, so i have to show the diferences between the TIM's for this users.

I have low TDP/High TDP, bench class programs, games, etc... for a complex methodology. I have thik layer of TIM, thin layer of TIM, large contact surfaces, little contact surfaces. I have GF100, Radeon 5870, i7 920, evem more, i have huge amount of heat generated on specific area (nehalem type, with his small die), i have huge area with huge heat (GF100 on soft voltmod), i have huge die with not so much heat (gf100 on stock volt and stalker for example).

So i study the comportament of TIM's in various conditions, to have "The big picture", this is all about, (mind that for every category of tim i have equal conditions, for testind corectitude)

At 1.1v it was not possible to "play" furmark on GF100 (give it a try) the temperature was not stabilizing, so i have to find other programs (less stressing) to rise the temp and TDP.

The "radeon 5870 case" is relevant. Due to direct die contact and very thin TIM layer even the silicon paste gave good results.

Regarding TDP, higher temps means higher TDP, simply as that, because the test was done with the same cooler/same rpm/same settings for each category. TDP is not some "very measurable" number in cooling. Energy transformed in heat is not disipate all by the cooler, some part is disipated in pcb...etc.

The thermal design power (TDP), sometimes called thermal design point, represents the maximum amount of power the cooling system in a computer is required to dissipate. For example, a laptop's CPU cooling system may be designed for a 20 watt TDP, which means that it can dissipate up to 20 watts of heat without exceeding the maximum junction temperature for the computer chip. It can do this using an active cooling method such as a fan or any of the three passive cooling methods, convection, thermal radiation or conduction.

Click to expand...

Central processing unit power dissipation or CPU power dissipation is the process in which central processing units (CPUs) consume electrical energy, and dissipate this energy both by the action of the switching devices contained in the CPU (such as transistors or vacuum tubes) and by the energy lost in the form of heat due to the impedance of the electronic circuits.

Click to expand...

The main ideea, and the fact that count: do not matter what program is used, with bigger temp comes bigger TDP. What is really important in a test is to have equal conditions on specific comparrison, for example GF100 1.1v combustor - all the tim's was tested with same conditions on this category (ambient/cooler/fan/mainboard .....) I hope you understand, (i have terrible english, sorry)


More simple, do not compare programs, compare numbers because the numbers i get are from same conditions at every category, wich was created to simulate a lot of conditions.

Testing on GPU's is relevant for some of users, and there are test on nehalem with more than one tdp, to cover all kind of user classes.

All this metodology lead to one simple conclusion for common users: even AS5 is ok for them, or silicon paste in case of 5870 (to be more specific, in case of direct contact silicium die/cooler base, and thin layer of TIM)

Enthusiasts always use the best available so for them i have test with a metodology created to increase diferences between tim's, such as gf100 tests with overvolt , or nehalem tests with 1.5v.

Sorry for my english.

Normal users do not play furmark, but i include furmark tests. At 1.1v it was not possible to "play" furmark on GF100 (give it a try) the temperature was not stabilizing, so i have to find other programs (less stressing) to rise the temp and TDP.

Click to expand...

But the point is not to test temperatures 'in game' or in 'user situations'. The point is to test the TIM's effectiveness. And to accurately test it is not in a game situation where the load varies.

And as far as Furmark on Fermi, I have run it, at 75% fan and hit 83C... See my review on the GTX470 for my thermal results (stock paste).

More simple, do not compare programs, compare numbers because the numbers i get are from same conditions at every category, wich was created to simulate a lot of conditions.

Click to expand...

This is what Im saying...one cannot compare numbers between each of the tests b/c you changed the testing environment(app/game you used to reach high temps) AND the speeds + voltages of the card. You could have gotten away with that method (game testing + other) if you tested consistently. Too many variables changed between methods to extrapolate much relevant information for me.

EDIT: Have you seen how scientific the testing done by JoeC here with his product is? That is one of if not the best models I have seen on TIM testing. Now, I understand you may not have hte resources available quite like he does, but TIMs performance is generally so close today that even a slightly inconsistent mount or slightly poor application can skew results tremendously.

The point is to test the TIM's effectiveness. And to accurately test it is not in a game situation where the load varies.

Click to expand...

I have test the effectiveness, check the nehalem tests.

The load do not warry in my tests, i use the stalker bench, wich is the same for all tims tested. This tests are for a little part of testing, check the article.

And as far as Furmark on Fermi, I have run it, at 75% fan and hit 83C

Click to expand...

I am talking about gtx 480 at 1.1v. This card have very high temps on furmark even with stock volts and 100% cooler. 1100mv are not GTX 480 stock volts.

I can't see in your review furmark setting (here some tests with furmark in xtreme burning mode)

Always post screens with testing, so users can se clearly the settings used.
Try furmark with high resolutions and extreme mode.

This is what Im saying...one cannot compare numbers between each of the tests b/c you changed the testing environment(app/game you used to reach high temps)

Click to expand...

As long as you use the same TDP/programms/setting for one category of comparisons, the results are ok to compare.

Take a look here:

http://lab501.ro/racire/arhitectura-hdt-intre-dezamagire-si-miracol

In a cooler test you don't change settings (voltages, rpm etc) from one category to another? You do all comparrisons at the same voltages and settings?

I explained to you what TDP means, and what counts in a test, and the fact that in one TDP category i use the same conditions for each TIM, for equal conditions.

Think this way: every category is a separate "mini" article, but put together at the end, it is so hard to understand a complex metodology???