22 cores Intel Xeon Delid + Direct Die

pedrofan

Weaksauce
Joined
Feb 10, 2016
Messages
67
Today I delided my E5-2696 V4 Intel Xeon processor.
It’s a massive silicon, with a tdp of 155w, capable of doing 6,723 points in Cinebench R20, still among the best multithreaded modern processors.
(It is faster than E5-2699 V4)

E02D9FBE-9CA5-430E-8136-19E3850F2910.jpeg


Thermal gains are insane, it’s dropped from 84º while rendering to 59º. This was my previous idle!

36DB0CCC-A7D2-4252-BAE8-5CA7252E3151.jpeg


It had a very thick indium soldering that I removed with an aliexpress delinding tool and a shaving razor.

991E14B8-990F-48F8-9D4B-49F9ADA4DE6F.jpeg


Direct die was possible as the main chip is slightly higher than the metal retention bracket. I used thermal grizzly kryonaut

C111258C-2EFB-43AB-A06C-53EA743D47E7.jpeg
AE00AACB-658E-4CEF-B535-37A728DE11D4.jpeg


For me it is insane to have such a powerful cpu inside a case (Dan A4-SFX) that can be stored inside a backpack and being carried anywhere, specially with this temps.

8421BAAE-3656-4201-98FC-2924DDCAB405.jpeg


All the tools that I used

6DD0481D-3B07-4D00-92F0-9BCEF2755E12.jpeg
 
Last edited:
did you double check contact/pressure without the ihs? any results?
 
did you double check contact/pressure without the ihs? any results?
I just tighted the screws the same amount of turns. It’s been working flawlessly for all the afternoon and night.

The chip is slightly higher than the mounting and it is getting good temps.
 
Its having perfect contact, with the asetek aio cooler. It’s such a big chip that it is almost the same surface than a his, so Inthink that the pressure is well distributed.
 
before/after temps? better oc? that kind of results, info.... never mind, didnt see the update....
 
Last edited:
Always wanted them to release some of these unlocked they would have been insane even with broadwell's shit overclocking.
 
He says in the post temps dropped 25C, which is interesting given the IHS is soldered. I might try this on a 2683v3 or something, I have a very small form factor build that runs at 90C right now. Wouldn't do this on a 2699v4 myself though, busting through the indium solder on a $500 CPU is harrowing especially when you can get the same performance out of a 3950x at the same price.
 
He says in the post temps dropped 25C, which is interesting given the IHS is soldered. I might try this on a 2683v3 or something, I have a very small form factor build that runs at 90C right now. Wouldn't do this on a 2699v4 myself though, busting through the indium solder on a $500 CPU is harrowing especially when you can get the same performance out of a 3950x at the same price.
yeah, none of that was there when i first started asking.
1625789214346.png
 
before/after temps? better oc? that kind of results, info....
Before 84º and after 59º!!! For a single cinebench run. Im pretty amazed by the result.

Im stressing it with a long run cinebench r20 test 500 seg (It has already peaked 6,723 points on cinebench R20)

DB5017E6-9036-4D5E-A197-56A77E9BC14C.jpeg


65ºfor a long stress test on this chip is like science fiction for me. I started with + 100ºC with my first thermal solutions.

Idling is also insane after delid, it is almost like ambient temp.

C0BDAD2C-8552-429B-B296-826F3E9F8FB2.jpeg
 
Always wanted them to release some of these unlocked they would have been insane even with broadwell's shit overclocking.
Always wanted them to release some of these unlocked they would have been insane even with broadwell's shit overclocking.
My goal was to use it inside A4-SFX with non audible fans. It’s been very loud untill now. I have just set the fan curve to an almost inaudible level wich is awesome!!
 
He says in the post temps dropped 25C, which is interesting given the IHS is soldered. I might try this on a 2683v3 or something, I have a very small form factor build that runs at 90C right now. Wouldn't do this on a 2699v4 myself though, busting through the indium solder on a $500 CPU is harrowing especially when you can get the same performance out of a 3950x at the same price.
I have a post about this setup (before the delid) with more temp data, etc.
before/after temps? better oc? that kind of results, info.... never mind, didnt see the update....


https://hardforum.com/threads/dan-h...50mm-heatsink.1925480/page-35#post-1043813328

Of course modern cpus are great, but I bought this cpu for 1300$ in ebay in 2017 and it is still rocking in comparison with modern cpus!! I think It’s been a very good investment, thats why I took the risk to delid it.
 
I have a post about this setup (before the delid) with more temp data, etc.



https://hardforum.com/threads/dan-h...50mm-heatsink.1925480/page-35#post-1043813328

Of course modern cpus are great, but I bought this cpu for 1300$ in ebay in 2017 and it is still rocking in comparison with modern cpus!! I think It’s been a very good investment, thats why I took the risk to delid it.
Yeah agreed, if you already have a 2699v4 there isn't much reason to move away from it, my various Haswell-EP machines serve duty as everything from workstations to HTPCs.

If you really want a challenge try stuffing 22 cores into one of these:

DSC_4857.jpg


I'm currently up to 14 cores but the thermals are...poor.
 
Yeah agreed, if you already have a 2699v4 there isn't much reason to move away from it, my various Haswell-EP machines serve duty as everything from workstations to HTPCs.

If you really want a challenge try stuffing 22 cores into one of these:

View attachment 373283

I'm currently up to 14 cores but the thermals are...poor.
I already solved it.
Dynatron t-318 + ty-100 fan pushing air (noctua a9x14 hs-pwm chromax can also work) + 2x 20x40mm fans pulling air (if you have more space you can try bigger models) + a plastic duct.

If you already have a heatsink try to change only the fans. The fan model and this air chanel is the key.

86AE9C94-C7B2-4379-84B7-392DA6260EDC.jpeg


It’s like a Turbo Charger XD
 
Last edited:
  • Like
Reactions: bwang
like this
I already solved it.
Dynatron t-318 + ty-100 fan pushing air (noctua a9x14 hs-pwm chromax can also work) + 2x 20x40mm fans pulling air (if you have more space you can try bigger models) + a plastic duct.

If you already have a heatsink try to change only the fans. The fan model and this air chanel is the key.

View attachment 373346

It’s like a Turbo Charger XD
Slick, thanks for the tip! I didn't think to duct pullers into the end of the heatsink, I bet that would help a lot.
 
Slick, thanks for the tip! I didn't think to duct pullers into the end of the heatsink, I bet that would help a lot.
They help a lot, only with the main fan It reached +100ºC in sustained load. I made a plastic duct arround them. We have the same motherboard, initially I put four pulling fans in both sides but the ones on the intel sticker side dont make much difference because i couldn’t fit a duct there and there are a lot of components in this zone.

This kind of heatsinks with closed bottom are made for horizontal blower fans like in 1080 FE gpu, because they are designed for slim rack servers. So they perform better with an horizontal air flow than a vertical flow.

Here you have the duct:

B5DE2111-B2CC-42E6-BA6A-6210FA19DEB4.jpeg

8B9BB2C9-83DD-4150-8692-9A71BE421BF1.jpeg

It looks that you have more horizontal space than me, you can try to use bigger fans, as the high performance small ones make a lot of noise at high rpm (rendering for example)


455A757B-6EA3-43A7-93D2-1142301CE321.jpeg


I’ve tryed A LOT of fans. The only ones that worked for me where Thermalright TY-100 and Noctura Chromax A9x14 (the black one, not the brown. The brown was too weak. They look the same but the chromax is more powerful)
 
  • Like
Reactions: bwang
like this
It's a shame that these CPUs are locked down. I bet that at 4GHz all-core, this chip would give the 10980XE a really hard time.
 
Back
Top