To monoblock or not to monoblock?

[Abaidor]

I sanded my 7980XE die a few months back it did help drop some of the higher cores down guessing these are the ones that are lower and have less contact, I used 2000 grit wet sand paper as long as you take your time there is no risk just don't sand down to far and you be fine, the lowest point i stop at just to be safe, I took the risk but ill rather have stable temps so I think it was worth the risk.

My main concern was whether Liquid Metal could "penetrate" into the die after removing that protective layer on top of the die...(what der8auer was talking about).
Obviously yours is holding on but you are also using a different brand of liquid metal..I am using Thermal Grizzly but maybe the other one is worth a try while it is also less "liquid" if I remember well from some of your previous posts.

[tistou77]

Yeah Orange for Copper and I emailed them about one and ordered it that way, looking at the temps there are some cores with up to 8c difference on a high overclock that 8c really maters, maybe they should of just plated the bottom part might be worth seeing if sanding the top off helps.

Rockit Cool answered my question about the difference between Nickel and Copper

You can get a slightly better temp if you polish the Nickel plate.

The Nickel IHS will not equal the copper IHS. Temps will be 2 to 4c higher with the Nickel.

Why? Because Nickel is not such a good conductor of heat compared to Copper.

With the Nickel plating you have CPU die >> TIM >> Nickel plate >> Copper IHS >> Nickel plate >> TIM >> Water block.

With the Copper IHS you have CPU die >> TIM >> Copper IHS >> TIM >> Water block.

You can see that there are two more layers for the heat to travel through.

The benefit of the Nickel plated IHS is that it will not absorb the Gallium and the TIM will last longer, will not harden up and the copper will not oxidize.

It’s a compromise. Better temps or longer life??

For this reason we don’t switch all IHS to the Nickel. People expect to see a temp drop with the Copper.

Wholesale IHS of origin => Nickel => Copper

[LiveOrDie]

My main concern was whether Liquid Metal could "penetrate" into the die after removing that protective layer on top of the die...(what der8auer was talking about).
Obviously yours is holding on but you are also using a different brand of liquid metal..I am using Thermal Grizzly but maybe the other one is worth a try while it is also less "liquid" if I remember well from some of your previous posts.

I switched to Thermal Grizzly after running out of CL, as for the LM penetrating the die there is no proof that the LM will have any other affect that it doesn't already have which is just pitting, the LM pits the protective layer into the silicon anyways.

Rockit Cool answered my question about the difference between Nickel and Copper




Wholesale IHS of origin => Nickel => Copper

Yeah well I had my copper IHS on for 5 months and it did dry out and fuse to the copper, it didn't affect my temps at all so I don't care if it drys out if my temps are the same, also the nickle plated IHS also the LM drys to compared to the stock IHS which it doesn't so I hardly think its worth using.

[Abaidor]

I switched to Thermal Grizzly after running out of CL, as for the LM penetrating the die there is no proof that the LM will have any other affect that it doesn't already have which is just pitting, the LM pits the protective layer into the silicon anyways.

Yeah well I had my copper IHS on for 5 months and it did dry out and fuse to the copper, it didn't affect my temps at all so I don't care if it drys out if my temps are the same, also the nickle plated IHS also the LM drys to compared to the stock IHS which it doesn't so I hardly think its worth using.

I will keep watching you for the next months and see how it goes then. If I do it I want to have the money for a new CPU on the side first since the system is needed for work and I don't want to risk it. If only it wasn't the worst period of the year for hardware purchase on my side......gotta pay for schools, taxes and purchase other things for work as well right now...ah and Summer Holidays along with several other summer activities of course for 4 people.

[LiveOrDie]

I will keep watching you for the next months and see how it goes then. If I do it I want to have the money for a new CPU on the side first since the system is needed for work and I don't want to risk it. If only it wasn't the worst period of the year for hardware purchase on my side......gotta pay for schools, taxes and purchase other things for work as well right now...ah and Summer Holidays along with several other summer activities of course for 4 people.

No problem ill keep you updated if the CPU has any side affects ill keep mine clocked to 4.5Ghz 1.2v for manageable temps.

[Deathscythes]

Hello there,

I am new here, but I have been playing and experimenting with a 7980XE for quite a while and experimented a few things regarding its cooling so I think you guys may be interested =)

The 7980XE in question is binned by silicon lottery for 4.5GHz @1.125V cooled by a Supremacy Edge from EK (limited edition EVO)

As I want to get the maximum out of this chip, I went for direct die cooling using der8auer's direct die frame.
My first attempt was a complete failure :
I don't have any screenshot but some of the cores spiked to a whopping 104 °C in cinebench instantly while others sat at around 50°C.

I am not 100% sure what was the reason for such a brutal failure. I have 2 hypothesis:
1 ) I did let quite a bit of glue that SL used for sealing the chip, it was hard to remove and after freaking out for an hour delidding the CPU with a razorblade (rockit tool not working against SL glue) I didn't feeling risking slipping and potentially killing SMDs.

2) This one I am sure it has a massive effect as I retested it later on; the amount of liquid metal applied.

Weeks later i decided to fix the CPU and removed a lot of the glue that was on its PCB and reapplied the liquid metal. Originally i just "painted" the block and the die with LM. But this time after i was done doing that, I added some more. Enough to make it look more "wet" but without making a mess of it.

Here are the temperatures under prime 1344k AVX off after two hours for the chip running 4.9GHz @1.31V :
As I am running a 7980XE on the Omega I believe those temps are reported 5°C higher than reality.



I do not worry about maximums as those seemed only to be instantaneous during load transients.
However I must say I was quite frustrated by the core 12# and looked for ways to improve that.

A few people recommended me to sand my CPU block... Such a tedious process... but hell was it worth it!!

Here is another screenshot of prime running the exact same settings after the CPU block was lapped and reapplied a substantial amount of liquid metal.
(A previous test with a small quantity LM yielded poor results, I didn't take any screenshot)



Regarding the run being only 8min long against 2 hours previously it actually doesn't matter as the heat doesn't build up in the loop.
I measured 2°C between ambient and coolant after 2 hours of prime.
Overall everything is cooler after having sanded the block and the thermal disparity between cores has shrinked massively.
Great results here IMO.

I then tried something else; Increasing the pressure of the CPU block on the die.
When mounting it I couldn't help but notice how loose it felt even screwed to the max.
I therefore decided to increasing it by adding 2washers under each of the thumbscrews.

here are the results after a short test


Slightly better, I'll probably try to increase it further later on.

With that setup here is the OC I am currently running.
All cores at 5GHz mesh at 3.3GHz with 1.2 on the mesh and 1.355V on cores except a few who need closer to 1.38V

Here are what the thermals look like after hours of 1344k with the ambient temperature at 24°C


as for the Vcore :


Notice the discrepancy in terms of mesh frequency, I messed and forgot to set it to 3.3GHz in the test I took the first screenshot from, sorry about it.

I am going to explore another option to improve those thermals while staying in the realm of ambient cooling, a few weeks after computex.

I don't know how much you guys tested but hopefully those data bring something =)

[Abaidor]

Pretty much what I found out with the direct die frame too. Some cores sit lower than other so they don't get cooler with the same efficiency. Several factors are at play here such as:


-How "level" the direct die frame is with the die (tighten through its own screws)
-Whether the block is flat and whether it can "warp" even at the slightest degree to take the form of the die (my Velocity is relatively flat and passes the spin test with the included 2066 jet plate)
-The amount of liquid metal so that it effectively fills the voids of the "lower" cores to facilitate thermal transfer.


What LiveOrDie did, I mean lapping is certainly the most effective so at some point in the future I am going to do it too...when this is going to be is a matter of being able to replace the CPU should things go North...

Meanwhile, I am sitting tight @ 14X48 All cores with temps during my workflow and gaming @ 4K never exceeding 66c at the hottest core. I am fine with that. The only reason to keep testing different applications of LM in conjunction with CPU Block & DD Frame seatings is to possibly achieve even higher clocks but for now I am fine. A 9900K build is in my plans anyway for pure gaming and high clocks and I am slowly gathering the parts for it.

In any case, thanks everyone for the feedback.

[LiveOrDie]

Hello there,

I am new here, but I have been playing and experimenting with a 7980XE for quite a while and experimented a few things regarding its cooling so I think you guys may be interested =)

The 7980XE in question is binned by silicon lottery for 4.5GHz @1.125V cooled by a Supremacy Edge from EK (limited edition EVO)

As I want to get the maximum out of this chip, I went for direct die cooling using der8auer's direct die frame.
My first attempt was a complete failure :
I don't have any screenshot but some of the cores spiked to a whopping 104 °C in cinebench instantly while others sat at around 50°C.

I am not 100% sure what was the reason for such a brutal failure. I have 2 hypothesis:
1 ) I did let quite a bit of glue that SL used for sealing the chip, it was hard to remove and after freaking out for an hour delidding the CPU with a razorblade (rockit tool not working against SL glue) I didn't feeling risking slipping and potentially killing SMDs.

2) This one I am sure it has a massive effect as I retested it later on; the amount of liquid metal applied.

Weeks later i decided to fix the CPU and removed a lot of the glue that was on its PCB and reapplied the liquid metal. Originally i just "painted" the block and the die with LM. But this time after i was done doing that, I added some more. Enough to make it look more "wet" but without making a mess of it.

Here are the temperatures under prime 1344k AVX off after two hours for the chip running 4.9GHz @1.31V :
As I am running a 7980XE on the Omega I believe those temps are reported 5°C higher than reality.



I do not worry about maximums as those seemed only to be instantaneous during load transients.
However I must say I was quite frustrated by the core 12# and looked for ways to improve that.

A few people recommended me to sand my CPU block... Such a tedious process... but hell was it worth it!!

Here is another screenshot of prime running the exact same settings after the CPU block was lapped and reapplied a substantial amount of liquid metal.
(A previous test with a small quantity LM yielded poor results, I didn't take any screenshot)



Regarding the run being only 8min long against 2 hours previously it actually doesn't matter as the heat doesn't build up in the loop.
I measured 2°C between ambient and coolant after 2 hours of prime.
Overall everything is cooler after having sanded the block and the thermal disparity between cores has shrinked massively.
Great results here IMO.

I then tried something else; Increasing the pressure of the CPU block on the die.
When mounting it I couldn't help but notice how loose it felt even screwed to the max.
I therefore decided to increasing it by adding 2washers under each of the thumbscrews.

here are the results after a short test


Slightly better, I'll probably try to increase it further later on.

With that setup here is the OC I am currently running.
All cores at 5GHz mesh at 3.3GHz with 1.2 on the mesh and 1.355V on cores except a few who need closer to 1.38V

Here are what the thermals look like after hours of 1344k with the ambient temperature at 24°C


as for the Vcore :


Notice the discrepancy in terms of mesh frequency, I messed and forgot to set it to 3.3GHz in the test I took the first screenshot from, sorry about it.

I am going to explore another option to improve those thermals while staying in the realm of ambient cooling, a few weeks after computex.

I don't know how much you guys tested but hopefully those data bring something =)

What type of cooling are you using and what waterblock? moving to an omega does open up moving to a normal block rather than a mono but my CPU isn't a great clocker and I also paid what a 7940x cost for it so I really can't complain, but I'm interested in getting my temps better I would like to run higher clocks 4.7Ghz I would be happy with, but the fact is some CPUs just run hotter than others and need less voltage, even with direct die on this chip its temps are high for 1.27v, also what are your water temps like my water temp can get up to 38c when the CPU is under load.

[LiveOrDie]

Pretty much what I found out with the direct die frame too. Some cores sit lower than other so they don't get cooler with the same efficiency. Several factors are at play here such as:


-How "level" the direct die frame is with the die (tighten through its own screws)
-Whether the block is flat and whether it can "warp" even at the slightest degree to take the form of the die (my Velocity is relatively flat and passes the spin test with the included 2066 jet plate)
-The amount of liquid metal so that it effectively fills the voids of the "lower" cores to facilitate thermal transfer.


What LiveOrDie did, I mean lapping is certainly the most effective so at some point in the future I am going to do it too...when this is going to be is a matter of being able to replace the CPU should things go North...

Meanwhile, I am sitting tight @ 14X48 All cores with temps during my workflow and gaming @ 4K never exceeding 66c at the hottest core. I am fine with that. The only reason to keep testing different applications of LM in conjunction with CPU Block & DD Frame seatings is to possibly achieve even higher clocks but for now I am fine. A 9900K build is in my plans anyway for pure gaming and high clocks and I am slowly gathering the parts for it.

In any case, thanks everyone for the feedback.

The 9900K has bad temps die to it's thinker silicon and solider by the way.

[Deathscythes]

What type of cooling are you using and what waterblock? moving to an omega does open up moving to a normal block rather than a mono but my CPU isn't a great clocker and I also paid what a 7940x cost for it so I really can't complain, but I'm interested in getting my temps better I would like to run higher clocks 4.7Ghz I would be happy with, but the fact is some CPUs just run hotter than others and need less voltage, even with direct die on this chip its temps are high for 1.27v, also what are your water temps like my water temp can get up to 38c when the CPU is under load.

I did mention the CPU block
It's a supremacy edge from EK. IMO monoblocks are not an option for more aggressive overclocking as they don't really allow you to apply a lot of pressure let alone even pressure.
When I was still running the rampage VI extreme, I had the same CPU block and a VRM block from watercool.

Here is what the rig looked like back then:
https://i.imgur.com/uA5xIQ3.jpg

The coolant doesn't warm up or barely does. Running the Vardars at 1000 RPM, after 6 hours of prime at 5GHz the Delta between ambient and coolant was of 2°C placing the coolant at around 26°C.

I believe there is a way to dissipate more heat from the CPU. The thing is the motherboard gets absolutely SOAKED in heat. It's crazy. Without active cooling the dimms reach a whopping 60°C under prime or HCI because the CPU warms everything up whereas it's significantly cooler under RamTest. What I am going to try is replacing the direct die frame backplate by a waterblock that will cover a large portion of the back of the motherboard. I found a post of some guy who dropped 6°C on a 3930K just by adding cheap heatsinks and a crappy fan at the back of the socket. Considering that was achieved on a CPU that didn't generate anywhere near as much heat, imo this is very promising and something I can't wait to try with the aforementioned custom waterblock.

This is for after computex, the wait is going to be long but I should receive 2 matrix next week which should keep me busy :3