• 23.10.2018, 13:42
  • Registrieren
  • Anmelden
  • Sie sind nicht angemeldet.

 

Kaine

Junior Member

Sonntag, 24. Juni 2018, 16:23

When the processor is well soldered to ihs, it is best not to remove it. Heat is easier to extract using a larger surface. It is better to polish the ihs to improve the contact.

As for the current capacity to cool 250w, it all depends on what you mean by "easily" cooling.

For a cpu under burn at 70°, something really effective should make a delta of 10-20° (max) with water. Something like 50-60°

Water out of waterblock, pipes and radiator(s) must be hot, let's see the current heat transfer waterblocks:



The hotter the water (of course keeping a minimum flow) the higher the performance. With 36.5 ° as delta for the best, there is still a lot of room for a real performance evolution!

One must think about transferring the heat throughout the block before looking for a way to cool it.

Just as the cpu must be welded to the IHS, the top and bottom parts of the waterblock should be "welded".

As we have to open it to clean it, it's impossible to have a block in one block. The research should be articulated in this direction.

Dieser Beitrag wurde bereits 12 mal editiert, zuletzt von »Kaine« (24. Juni 2018, 16:45)

Sonntag, 24. Juni 2018, 17:01

Out of curiosity, to you have anything to do with "deeper" engineering ?
Verkaufe Twinplex; Twinplex Pro

Kaine

Junior Member

Sonntag, 24. Juni 2018, 17:23

Please develop, I have a doubt about the meaning of the question.

Sonntag, 24. Juni 2018, 18:51

Since gpu blocks can keep the delta between water and chip under 10 Kelvin the problem lies deeper. Starting with temp sensors in the CPUs.

Kaine

Junior Member

Sonntag, 24. Juni 2018, 19:55

You think it's the sensor that does not give the right result? This is possible to some extent, but let's not forget that the gpu contact surface is much larger than the CPU surface.

Sonntag, 24. Juni 2018, 19:57

I guess the next problem could be the IHS. Because on GPUs the waterblock has direct contact but I'm not aware of how it works out if you place the CPU-block direct on the DIE.

Kaine

Junior Member

Sonntag, 24. Juni 2018, 20:46

The best would be the waterblock directly soldered to the dies. No ihs, no thermal paste ...

Montag, 25. Juni 2018, 07:42

That is correct and you can buy a block that fits directly onto desoldered Intel CPUs. Tempwise that one is good but it's lacking usability.

Kaine

Junior Member

Montag, 25. Juni 2018, 15:27

I think Intel is forced to compete with Amd, increase the number of cores and thus restart a solder dies at IHS.

We can therefore expect to have to dissipate a large heated surface homogeneously.

It suits us well but will surely completely change the situation regarding the best waterblock.

Montag, 25. Juni 2018, 21:44

Judging by I9 and TR results not much is about to change.

Kaine

Junior Member

Samstag, 30. Juni 2018, 02:42

Now the 4 dies will be used in AMD threadripper2 (2990x 32/64) the temperature will be distributed more homogeneously to the IHS (maximizing center cooling will not be the most efficient) :



A manufacturer has already developed something that I find good and that should work well on the processors in general (although the shape is rectangular, I think we can make a slot longer than what is already done for a "square" waterblock.), I let you judge:





Intel will be forced to solder also to catch up, a lot of power not mastered by the process of engraving (but this is only my opinion).

Dieser Beitrag wurde bereits 9 mal editiert, zuletzt von »Kaine« (30. Juni 2018, 04:14)

Samstag, 30. Juni 2018, 13:04

Intel is soldering any HEDT processors, their just not soldering their mainstream line (8xxx)
Question is: Will Cofeelake be soldered?

For the new TR design, I though that AC said that they wanna design a new cooler for it (since they will probably cool Epyc cpus too)

Ähnliche Themen