To monoblock or not to monoblock?

[LiveOrDie]

They sent me a new package (Fedex information)
I guess it's the nickel version, they have not yet responded to the email

I told you that theres no nickel version lol.

[LiveOrDie]

After finally ditching the monoblock and going with Direct Die cooling + an EK Velocity my temps are better than ever and I get 14 Cores @ 4.8Ghz

These are LOAD temps on continuous loops of Cinebench R20. During normal usage or gaming they are much lower.

Attachment 79508Attachment 79509Attachment 79510





Some almost idle temps

Attachment 79511

for direct die your temps are not that much better than mine and that probably down to your WC setup not vs copper ihs vs direct die mate.

[Abaidor]

for direct die your temps are not that much better than mine and that probably down to your WC setup not vs copper ihs vs direct die mate.

First of all it is really hard to make proper contact and it took a lot of effort. But it works better than it ever did before.

Based on your specs:


I am 14 Cores @ 4.8Ghz - 1.28V (and 1 core 1.31v , 1 core 1.29v)
You are 10 Core @ 4.6Ghz - 1.245v

Big difference in heat output! / VRM 45C / GPU (2075 Core / 12000 Mem) 36C and this is under full stress. During normal usage my temps are funny low while under 4K gaming they never exceed 45C average.

If I load my 14X 4.7Ghz profile at 1.24v max the temps will drop significantly and it works with no AVX offset. But as it stands now I can afford to go 4.8GHz and could even push some cores further but I am honestly tired with stress testing for now.


Direct die is and will always be better than IHS which is an extra thermal layer in the middle. The problem is to make contact with the block but if you manage it there is no comparison.


Look at these idles - lol the cold wall the rad is on does work since some cores go below ambient.

[LiveOrDie]

First of all it is really hard to make proper contact and it took a lot of effort. But it works better than it ever did before.

Based on your specs:


I am 14 Cores @ 4.8Ghz - 1.28V (and 1 core 1.31v , 1 core 1.29v)
You are 10 Core @ 4.6Ghz - 1.245v

Big difference in heat output! / VRM 45C / GPU (2075 Core / 12000 Mem) 36C and this is under full stress. During normal usage my temps are funny low while under 4K gaming they never exceed 45C average.

If I load my 14X 4.7Ghz profile at 1.24v max the temps will drop significantly and it works with no AVX offset. But as it stands now I can afford to go 4.8GHz and could even push some cores further but I am honestly tired with stress testing for now.


Direct die is and will always be better than IHS which is an extra thermal layer in the middle. The problem is to make contact with the block but if you manage it there is no comparison.


Look at these idles - lol the cold wall the rad is on does work since some cores go below ambient.

Maybe learn to read might help yeah? , my 7940X needs 1.3v for 4.7Ghz it isn't the best clocker and my watercooling loop can't handle the heat, it all depends on your setup really.



This is a mates 7940x with a stock IHS delidded he has a crazy setup, lets see you pump these volts lol?

[Abaidor]

Maybe learn to read might help yeah? , my 7940X needs 1.3v for 4.7Ghz it isn't the best clocker and my watercooling loop can't handle the heat, it all depends on your setup really.



This is a mates 7940x with a stock IHS delidded he has a crazy setup, lets see you pump these volts lol?

LOL - I would swear it said 7900X on your PC Specs tabs - Did you just change it or was I BLIND!

Anyway, still I only have two cores hitting not even your worst temps..the rest and the average are much better but ok I will give you that I have a bigger RAD (more than double and external on the coldest wall of the room) and dual D5s..

Overall, I still prefer it since I have cores that stay really cool no matter the stress still absorbing package hit from the others. The problem is that each die is different and not flat so each installation will behave differently.

I don't see the point of argument here to be honest, if Direct Die cooling has good contact nothing beats it - pure physics! In my case I have achieved satisfactory contact because of the convexity of my die..but the potential is obvious when you have cores below 60C under stress.

If the core was perfectly flat and making contact was easy nobody would even bother with an IHS as GPUs don't either....The IHS is there because of this and to protect the die from cracking or ****ty contact with bad installations. Other than that it's a an extra layer hindering thermal transfer.

Since your setup works enjoy it..I took a shot at it and it is working too so I am not touching it unless it goes south.

The only thing I am going to do is add another big MO-RA 420 and maybe passive to simply overwhelm any heat output my chips pump out while my fans never spin above 500rpm. I can even get away then with passive cooling at low loads.

[LiveOrDie]

LOL - I would swear it said 7900X on your PC Specs tabs - Did you just change it or was I BLIND!

Anyway, still I only have two cores hitting not even your worst temps..the rest and the average are much better but ok I will give you that I have a bigger RAD (more than double and external on the coldest wall of the room) and dual D5s..

Overall, I still prefer it since I have cores that stay really cool no matter the stress still absorbing package hit from the others. The problem is that each die is different and not flat so each installation will behave differently.

I don't see the point of argument here to be honest, if Direct Die cooling has good contact nothing beats it - pure physics! In my case I have achieved satisfactory contact because of the convexity of my die..but the potential is obvious when you have cores below 60C under stress.

If the core was perfectly flat and making contact was easy nobody would even bother with an IHS as GPUs don't either....The IHS is there because of this and to protect the die from cracking or ****ty contact with bad installations. Other than that it's a an extra layer hindering thermal transfer.

Since your setup works enjoy it..I took a shot at it and it is working too so I am not touching it unless it goes south.

The only thing I am going to do is add another big MO-RA 420 and maybe passive to simply overwhelm any heat output my chips pump out while my fans never spin above 500rpm. I can even get away then with passive cooling at low loads.

Direct die is better but the differences in temps are just from error in mounting the same with any mount, its not that the die isn't flat you should always try a few times to get the best temps if doing so, its harder when your doing a delid because most people glue the IHS back, all my gear is in my case I only run 1 pump with my 1080ti WB, monoblock and 2 rads but there not think ones and they only have fans on the one side but it does the job for my 4.6Ghz temps stay under 80C which I'm happy with, also doesn't help its summer here and room temp is 33C.

[Abaidor]

Direct die is better but the differences in temps are just from error in mounting the same with any mount, its not that the die isn't flat you should always try a few times to get the best temps if doing so, its harder when your doing a delid because most people glue the IHS back, all my gear is in my case I only run 1 pump with my 1080ti WB, monoblock and 2 rads but there not think ones and they only have fans on the one side but it does the job for my 4.6Ghz temps stay under 80C which I'm happy with, also doesn't help its summer here and room temp is 33C.

The die IS NOT flat, period. Did measure, did a spin test too and there are lengthy discussions about it on 7980XE elsewhere. I am not even debating it.

I tried more than 30 times (spent days on it).

You know what, next time I take it out I will sand like der8auer did for exactly the same reason....the die being convex.

Here is the video where he also provides measurements showing that the die is not flat

https://www.youtube.com/watch?v=tnd2LO0IBic


*EDIT

Crap! Not actually recommended (lapping) since he is not sure of long term damage...here goes my hope of a perfectly flat die..

[tistou77]

As der8auer and Wizery have explained, it's very difficult to have a good mounting with the Direct Die,
If the temperatures are so different is that the mounting is not good
They also said that some had spent hours without getting there (+ 15° difference), and them, at the first mounting it was good (less than 10° difference)

Me, I tried a few times, temperature difference of more than 15°, I gave up and sold the Direct Die
I believe it's the same for jpmboy and he resold it

[Abaidor]

As der8auer and Wizery have explained, it's very difficult to have a good mounting with the Direct Die,
If the temperatures are so different is that the mounting is not good
They also said that some had spent hours without getting there (+ 15° difference), and them, at the first mounting it was good (less than 10° difference)

Me, I tried a few times, temperature difference of more than 15°, I gave up and sold the Direct Die
I believe it's the same for jpmboy and he resold it

True but as I said previously the 15c difference is not what matters....for example if all my cores are 45C and the hottest are 60C who cares? The only thing that matters is the peak temperature of the hottest cores. As long as it is not bad, and it is not bad or worse in my case than what others get with the IHS, I am keeping it while my other cores are cooler. It is damn cool in normal workloads, 4K gaming anyway.

I would lap the CPU in the middle (convex) and all this would be history but the truth is that nobody knows if it is safe to use liquid metal then.

The only other thing that could improve surface contact is making the waterblock concave...I believe mine warped a bit too after 30 different mounting efforts and that is why it is even working because most of the time certain cores were not even making contact. It is my CPU die that is quite convex, it even spins naked, that is why.


*EDIT*

I had forgotten about this CPU block but the way it works could be the answer to making perfect contact with the convex skylake-x die.
You can change the base of the cold plate with the adjustment screws from convex to concave with this so that the corners make perfect contact. The problem is A) I don't like its looks vs the Velocity D-RGB B) it costs quite a lot to get one just for testing.

[LiveOrDie]

The die IS NOT flat, period. Did measure, did a spin test too and there are lengthy discussions about it on 7980XE elsewhere. I am not even debating it.

I tried more than 30 times (spent days on it).

You know what, next time I take it out I will sand like der8auer did for exactly the same reason....the die being convex.

Here is the video where he also provides measurements showing that the die is not flat

https://www.youtube.com/watch?v=tnd2LO0IBic


*EDIT

Crap! Not actually recommended (lapping) since he is not sure of long term damage...here goes my hope of a perfectly flat die..

der8auer is a waste of space and full of ****.....