Requested proof of Thermalright quality control issues

[Notlag]

ya ok after 3 heatsinks, with the same results... its something else becides the heatsinks... namely the chip your using, get a different chip and stop complaining about the thermalright heatsinks man. I mean have you even tried the heatsink on a different cpu?

Ok...

The first two Thermalright heat sinks were nearly tied with a Maxorb. They were only 1C better. Rotating the heatsink only improved temps by 2C, so at most they were 3C better than what should have been a FAR less capable Maxorb.

How does the CPU explain that? Does it magically decrease its power consumption when a Maxorb is mounted on it?

The third heatsink is marginally better. One could argue it is getting closer to the performance it should have and could blame the CPU for that, but the CPU can not account for the performance of the first two heatsinks relative to the Maxorb.

 

[Metallica]

Yah, I'll give it some thought. The bigger issue to me is being without a heatsink for the duration and having to put everything off longer. Even if my heatsink ended up doing as well as the test TR U120X in your OCPulse review, I'd be at a complete loss to explain it:

1. Actual air flow within my case is a non-issue because the side panel was removed. Even if the side panel was not on, my case does have negative pressure and the rear fans are close-by to exhaust hot air from the case.
2. High fan inlet air temperature is no longer an issue because I rotated the heatsink (no longer pulling air over the GPU and NB)
3. Have tried various amounts and methods of applying the thermal compound (which isn't rocket science)
4. Air flow over the heatsink is good - I'm using a fairly high output fan that is well regarded (ie - not like a generic fan or TT or something)
5. My original heatsink performed better after lapping the CPU, so I know the CPU was not lapped badly
6. Mounting pressure is more than adequate - I've tried using washers to increase mounting pressure w/ no change

Despite this, I'm still looking at delta's of 35C with an E4300 @ 9 x 333MHz, 1.325v

And this is with the 3rd one too? Something seems off ... did they send you a brand new one? Or did they "fix" your old one?

Just being that i got mine, the base looked horrible, i have the "crappy mount" and I'm missing the washer .. and mine is still doing great ..

Hard to think that 3 in a row were duds .. there must be something we are missing ... This makes me wanna see the results on another machine even more ... you should consider Spawnes offer You have your ORB in the meantime right? As they are performing the same right?

 

[Notlag]

Hard to think that 3 in a row were duds .. there must be something we are missing ... This makes me wanna see the results on another machine even more ... you should consider Spawnes offer You have your ORB in the meantime right? As they are performing the same right?

The Orb performs the same as the first two (well, it was 1C worse ). The third heatsink is 3C to 4C better than the first two, and rotating it gained another 2C, which puts it 6C or so worse than the current heatsink. So... The Orb is sufficient if I don't want to push my clocks further. It's definitely not going to cut it though if I overclock farther or up my voltage (I mean, my votlage is only 1.325v right now).

I'd be more tempted to just go w/ what I got and lap it (in which case I wouldn't send it off because I'd have resigned myself to what it will manage) or order a Tuniq and see how it does. If I got the Tuniq I could send out the U120X. And it wouldn't be a total waste as I could put the Maxorb and Tuniq (or U120X) in my other PC and my gf's. Those two are Pentium D's which run hot even at stock speeds.

Btw, what did your final core temps end up at?

 

[muddocktor]

To me, it's sounding like Intel did a rare bad job on their interface between the processor's IHS and slug instead of the U-120 eX, Notlag. Maybe they did a bad solder job when they installed the IHS onto the processor package. And since you lapped the processor's IHS, no warrantee from Intel now.

 

[Metallica]

The Orb performs the same as the first two (well, it was 1C worse ). The third heatsink is 3C to 4C better than the first two, and rotating it gained another 2C, which puts it 6C or so worse than the current heatsink. So... The Orb is sufficient if I don't want to push my clocks further. It's definitely not going to cut it though if I overclock farther or up my voltage (I mean, my votlage is only 1.325v right now).

I'd be more tempted to just go w/ what I got and lap it (in which case I wouldn't send it off because I'd have resigned myself to what it will manage) or order a Tuniq and see how it does. If I got the Tuniq I could send out the U120X. And it wouldn't be a total waste as I could put the Maxorb and Tuniq (or U120X) in my other PC and my gf's. Those two are Pentium D's which run hot even at stock speeds.

Btw, what did your final core temps end up at?

Me? The one's in my sig are current and accurate. I'm gonna lap my HS soon, i have some other things I would like to work on (making it look a little nicer ..) When i do, I'll post my new temps.

To me, it's sounding like Intel did a rare bad job on their interface between the processor's IHS and slug instead of the U-120 eX, Notlag. Maybe they did a bad solder job when they installed the IHS onto the processor package. And since you lapped the processor's IHS, no warrantee from Intel now.

Yea. You should try it on another C2D. I'm just thinking there HAS to be something else other than the Heatsink itself.

 

[Spawne32]

To me, it's sounding like Intel did a rare bad job on their interface between the processor's IHS and slug instead of the U-120 eX, Notlag. Maybe they did a bad solder job when they installed the IHS onto the processor package. And since you lapped the processor's IHS, no warrantee from Intel now.

Well there is always reseating the IHS, but thats only if hes willing to take the risk, one of the other admins on my site did it a couple times, he can give pointers, i know that the different in height between the contact area of the die to the IHS internally has been different between soldered and glued IHS's

 

[nd4spdbh2]

im 99.9% sure its NOT the ultra 120s or the max orb... its the fact the heat is inefficiently trasfered between the core an the IHS... resulting in 2 opposite end of the spectrum heatsinks preforming the same... and 3 of the same heatsinks doing the same thing (ya ya its a bloddy fawkin 3*c diff who cares)

I say you pop the lid in your e4300 and call it a damn day...

but you will probably continue to presist and say its a ****ty damn thermalright... so be it.

 

[muddocktor]

Well there is always reseating the IHS, but thats only if hes willing to take the risk, one of the other admins on my site did it a couple times, he can give pointers, i know that the different in height between the contact area of the die to the IHS internally has been different between soldered and glued IHS's

I didn't know that Intel was back to gluing the IHS on some of their processors now. Which series have the glued IHS? If he has a glued one, that could make for quite a difference in cooling ability if the tim job inside isn't good. And if it is a glued IHS, it isn't that difficult to remove an IHS safely if you are careful and use a very thin razor blade for the cutting of the glue around the edge of the IHS. I did that with an old Winchester A64 3000 I had replaced with a DC Opty last year for giggles. It survived with no problems.

 

[Spawne32]

I didn't know that Intel was back to gluing the IHS on some of their processors now. Which series have the glued IHS? If he has a glued one, that could make for quite a difference in cooling ability if the tim job inside isn't good. And if it is a glued IHS, it isn't that difficult to remove an IHS safely if you are careful and use a very thin razor blade for the cutting of the glue around the edge of the IHS. I did that with an old Winchester A64 3000 I had replaced with a DC Opty last year for giggles. It survived with no problems.

E2xxx and E4xxx as far as i know, ive had a couple of both and they were all glued.

 

[muddocktor]

That's nice to know, thanks. I have an E4400, but since the board it is in isn't a very good overclocking board, I have no heat problems with it. But if in the future I upgrade to a better board I'll keep that in mind if my temps get high.

 

[nd4spdbh2]

well they all have ruber epoxy around the perimeter of the IHS... but its the type of thermal interface that differs... The IHS could be soldered onto the core, or like older P4's just have some AS5 like TIM inbetween the ihs and core.

as for taking the IHS off that is soldered on, over at xs a guy came up with an ingenius way of fist cutting ALL of the rubber epoxy around the perimeter of the IHS, then sticking 4 razor blades one in each corner, making a diamond like shape, then suspending the chip by the blades, then heating the center of the chip with a torch, and the pressure supplied by the blades pops the cpu off of the solder.

heres the link.

http://www.xtremesystems.org/forums/showthread.php?t=116246&highlight=ihs+removal


NOTLAG - take that IHS off your chip (or get another chip known to run @ a good temp with an ultra 120x), only then could you officially say if thermalright sucks in my book. But untill then... THERMALRIGHT IS THE BEST OUT THERE!

 

[Spawne32]

oh wow i was unware they soldered the cores directly to the IHS, but i have seen several E2xxx and E4xxx series cpus with the lids popped and all that was transfering the heat from the die to the IHS was some as5 looking paste like you said, however i would be willing to bet it probably doesnt perform nearly as well as AS5

 

[nd4spdbh2]

oh wow i was unware they soldered the cores directly to the IHS, but i have seen several E2xxx and E4xxx series cpus with the lids popped and all that was transfering the heat from the die to the IHS was some as5 looking paste like you said, however i would be willing to bet it probably doesnt perform nearly as well as AS5

ya some are soldered some rnt.

 

[Immortal_Hero]

NOTLAG - take that IHS off your chip (or get another chip known to run @ a good temp with an ultra 120x), only then could you officially say if thermalright sucks in my book. But untill then... THERMALRIGHT IS THE BEST OUT THERE!

I as well will find your results more creditable if you try another chip.

 

[Notlag]

To me, it's sounding like Intel did a rare bad job on their interface between the processor's IHS and slug instead of the U-120 eX, Notlag. Maybe they did a bad solder job when they installed the IHS onto the processor package. And since you lapped the processor's IHS, no warrantee from Intel now.

I considered that possibility, but two piecse of evidence points to the contrary:

1. Poor IHS / die interface would explain high temps, but not the Maxorb tying with the U120X's

2. If I compare TAT temps with the CPU temps reported through the mainboard, the temperature difference does not vary much between idle and load. If there was a poor die / IHS interface, then the difference between the core temps and CPU temp should grow noticably larger as power disipated increases. Admittadly, the temps reported through the mainboard are not the exact temp of the surface of the IHS, so the measurement may be closer to the core temp than the IHS temp. I'd need to run w/o a heatsink to confirm the accuracy of those temps.

Because of the question about reporting on the 2nd point, it's not 100% conclusive. I'll have to think about whether I want to lower my speed / voltage and boot w/o a heatsink to measure the surface temp via a reliable surface temp probe and compare it to what is reported through the sensors.

 

[Notlag]

If I had another Core 2 Duo I'd give it a whirl. Unfortunately, I don't - my other box is a Pentium D with entirely different performance characteristics. Again, if the die / IHS interface is bad it can explain the overall high temps, but wouldn't cause the Maxorb and U120X to perform the same unless something was wrong w/ the U120X.

Consider how heat is transmitted from one medium to another given the thermal resistance of the interface and the amount of thermal power being dissipated:

Temperature Difference = Thermal Resistance * Watts Disipated

Like resistors, thermal resistance adds in series, so if the difference in resistances between a good U120X and the Maxorb was 9C for the wattage I'm dissipating, even if the die / IHS interface was bad, the difference would still be 9C (though the absolute temps would be higher w/ the poor die / IHS interface).

The exception to that is that not all heat is transmitted through the IHS. Virtually all, because the thermal resistance of that interface is much lower than the thermal interface of the package itself, and like electricity, heat takes the path of least resistance. If you up the resistance of the die / IHS interface, the heat will try to find other paths to dissipate, which will result in a lower percentage of thermal power transferring through the heatsink, and therefore lower differences between the Maxorb and U120X. However, that difference should be small, perhaps even negligible - even w/ a poor die / IHS interface, that will still be the far far less restrictive path for heat (versus transfering through the rest of the package).

 

[Notlag]

I as well will find your results more creditable if you try another chip.

I can price E6320's and E6420's, but I'm reluctant to drop more money on another CPU for testing purposes. This one works well enough, and if the heatsink isn't great shakes (or has an issue w/ this particular CPU) is that really worth another $200? I can try my Pentium D, but the chips are going to behave very differently to begin with.

 

[nd4spdbh2]

If I had another Core 2 Duo I'd give it a whirl. Unfortunately, I don't - my other box is a Pentium D with entirely different performance characteristics. Again, if the die / IHS interface is bad it can explain the overall high temps, but wouldn't cause the Maxorb and U120X to perform the same unless something was wrong w/ the U120X.

Consider how heat is transmitted from one medium to another given the thermal resistance of the interface and the amount of thermal power being dissipated:

Temperature Difference = Thermal Resistance * Watts Disipated

Like resistors, thermal resistance adds in series, so if the difference in resistances between a good U120X and the Maxorb was 9C for the wattage I'm dissipating, even if the die / IHS interface was bad, the difference would still be 9C (though the absolute temps would be higher w/ the poor die / IHS interface).

The exception to that is that not all heat is transmitted through the IHS. Virtually all, because the thermal resistance of that interface is much lower than the thermal interface of the package itself, and like electricity, heat takes the path of least resistance. If you up the resistance of the die / IHS interface, the heat will try to find other paths to dissipate, which will result in a lower percentage of thermal power transferring through the heatsink, and therefore lower differences between the Maxorb and U120X. However, that difference should be small, perhaps even negligible - even w/ a poor die / IHS interface, that will still be the far far less restrictive path for heat (versus transfering through the rest of the package).

USE YOUR DAMN PENTIUM D, and compare the ultra 120x to the maxorb...

that will be an EVEN BETTER situation as the pentium d definely puts out more heat.

and please please ohh please dont get into thermal resistance stuff again... if the IHS contact from the core to the IHS is bad... donesnt matter what cooler you have, ull just have bad temps.

 

[Notlag]

USE YOUR DAMN PENTIUM D, and compare the ultra 120x to the maxorb...

that will be an EVEN BETTER situation as the pentium d definely puts out more heat.

and please please ohh please dont get into thermal resistance stuff again... if the IHS contact from the core to the IHS is bad... donesnt matter what cooler you have, ull just have bad temps.

Yes, it will be better because it puts out more heat, but it also requires I re-run all of my previous tests because the absolute temps and deltas will not be comparable between the two.

As for the thermal resistance stuff, I think the electricity analogy explained it fairly well. Fact is, if the die / IHS interface is bad you'll have high temps, but if one cooler works better than another, it should still work better and the temps will reflect that. This should be fairly clear. If the U120X is supposed to beat the Maxorb, it should beat the Maxorb regardless of how much power is dissipated through it, provided that the amount of power dissipated doesn't change between tests. The die / IHS interface isn't going to be bad for the U120X, then magically get better for the Maxorb. So it doesn't explain the Maxorb / U120X comparison w/ the 1st and 2nd heatsink. The 3rd one, because it performs somewhat better, is more up for debate.

 

[Karadhas]

It doesn't matter if he's trying to cut a PB&J sammich or a wooden desk, a razor sharp scalpel should still cut better than a wedge of cheese. If the scalpel is cutting things with the same difficulty as the cheese something is wrong with the scalpel.