A machinist lapped an AMD Ryzen 16-core CPU within a nanometer

erek

erek

[H]F Junkie
Joined
Dec 19, 2005
Messages
10,894
Courtesy of Jelly

""A contributing factor to my project was the knowledge that silicon temperature is a crucial factor that limits the turbo frequencies of Zen 2 processors and when I installed a second radiator without noticing any significant temperature differences from before, I started looking at alternatives solutions," Penrowe said.

Penrowe claims he did not have the time to test the temp results of his efforts and make a bunch of graphs, but teases at the end of the video that someone else now in possession of the CPU will be doing that for him. So sometime in the hopefully near future, we'll find out if it can in fact run without any thermal paste."

https://www.pcgamer.com/a-machinist-lapped-an-amd-ryzen-16-core-cpu-within-a-nanometer-of-its-life/
 
I've done this with mulit grit sandpaper before. Pretty solid results. -4-5c Also, use gloves, managed to shave off some of the layers of skin on my thumb without knowing. :)
 
Krenum said:
I've done this with mulit grit sandpaper before. Pretty solid results. -4-5c Also, use gloves, managed to shave off some of the layers of skin on my thumb without knowing. :)
Click to expand...

I'm also not convinced that machining a CPU will yield much better results then doing it with sandpaper on a flat surface like glass
 
mnewxcv said:
Why is this news? People have been lapping cpus since the ihs has been on processors. I'm guessing the inefficiency between the die and the ihs is greater than the inefficiency of the ihs to cooler contact. Seems not worth it to me.
Click to expand...
Maybe for Intel CPUs lol.
Ryzen is soldered
 
mnewxcv said:
Why is this news? People have been lapping cpus since the ihs has been on processors. I'm guessing the inefficiency between the die and the ihs is greater than the inefficiency of the ihs to cooler contact. Seems not worth it to me.
Click to expand...
Exactly.

Things he could have done differently that would have been time better spent on improving performance:
1. Machined OFF the headspreader.
2. Desoldered the heatspreader
3. Used sub-zero cooling.

Rockenrooster said:
Maybe for Intel CPUs lol.
Ryzen is soldered
Click to expand...
He probably had a bubble like this soldered on chip:
https://pcper.com/2017/03/delidded-...is-using-solder-with-ihs-on-ryzen-processors/

...or he was already maxed out with the first radiator.
 
I mean... There's normal lapping with sandpaper and a flat surface, and then there's OCD level lapping like in this article. I've never lapped something so good that it would pick up the cpu with the heatsink like shown in the video in the article. Sure, lapping has been around for ages, but this is still pretty damn impressive.
 
mnewxcv said:
Why is this news? People have been lapping cpus since the ihs has been on processors. I'm guessing the inefficiency between the die and the ihs is greater than the inefficiency of the ihs to cooler contact. Seems not worth it to me.
Click to expand...

sleepeeg3 said:
Exactly.

Things he could have done differently that would have been time better spent on improving performance:
1. Machined OFF the headspreader.
2. Desoldered the heatspreader
3. Used sub-zero cooling.

He probably had a bubble like this soldered on chip:
https://pcper.com/2017/03/delidded-...is-using-solder-with-ihs-on-ryzen-processors/

...or he was already maxed out with the first radiator.
Click to expand...
He's a machinist, this was more out of curiosity to see if lapping to the point you could ring the surfaces together, how would that affect the cooling? He's not out to set a world record, he's attempting to answer his own curiosity.
 
IcePickFreak said:
He's a machinist, this was more out of curiosity to see if lapping to the point you could ring the surfaces together, how would that affect the cooling? He's not out to set a world record, he's attempting to answer his own curiosity.
Click to expand...
That's fine, and all well and good. I'm not trying to cast stones or disrespect any enthusiast or craftsman. I do various experiments in my various hobbies as well, but it still doesn't make for an interesting article. That being said, he didn't have time to check results even. The story wasn't ready to be published either way IMO.
 
i'd think too close to a conductive surface like a heatsink would allow for quantum tunneling or just classic conducting by not having enough insulating material (if it was truly around 1nm).

The real story would be testing it. given the urgency in reporting the lapping ...wonder why it's taken more than 1 day for some initial results to be posted.
 
  • Like
Reactions: Jelly
like this
Darth Ender said:
i'd think too close to a conductive surface like a heatsink would allow for quantum tunneling or just classic conducting by not having enough insulating material (if it was truly around 1nm).

The real story would be testing it. given the urgency in reporting the lapping ...wonder why it's taken more than 1 day for some initial results to be posted.
Click to expand...

I doubt we would hit a point of quantum tunneling but you could definintly make it thin enough to have conductivity issues if you do go far enough down. Ya dont need all that silicon tho. People have cut grooves into the cpu die to run water over it before
 
Last edited:
Imagine doing this with a PGA chip instead of LGA!
Edit: Don't have to imagine, watch the video like I clearly didn't
 
Last edited:
Jelly said:
I mean... There's normal lapping with sandpaper and a flat surface, and then there's OCD level lapping like in this article. I've never lapped something so good that it would pick up the cpu with the heatsink like shown in the video in the article. Sure, lapping has been around for ages, but this is still pretty damn impressive.
Click to expand...

I've lapped a CPU IHS good enough to pick up the cooler with it. Gets kind of annoying because you have to slide it off.

The trick to cut down on time for the initial lapping is to use an electric orbital sander. A lot of the CPUs I have lapped I have used and orbital sander for the whole process. Gets it plenty flat and with the higher grit you get almost the same mirror finish as you do by doing it by hand.
 
  • Like
Reactions: Jelly
like this
Krenum said:
I've done this with mulit grit sandpaper before. Pretty solid results. -4-5c Also, use gloves, managed to shave off some of the layers of skin on my thumb without knowing. :)
Click to expand...
I've lapped both my Ryzen 7 1700 and 2700X and as I type this out on my 1700 running Linux overclocked to 3.7Ghz the current temp is 30C. I also end up sanding some skin off and end up bleeding. Here's my Vega 56 reference cooler lapped to a mirror.

ODZ23PJ.jpg
 
DukenukemX said:
I've lapped both my Ryzen 7 1700 and 2700X and as I type this out on my 1700 running Linux overclocked to 3.7Ghz the current temp is 30C. I also end up sanding some skin off and end up bleeding. Here's my Vega 56 reference cooler lapped to a mirror.

View attachment 269635
Click to expand...

Only when your blood, sweat and tears have been poured into your craft can you be truly proud of your work :p

I remember cutting mosaic strip glass tile for my kitchen reno....my blood is in those walls.
 
cdabc123 said:
I'm also not convinced that machining a CPU will yield much better results then doing it with sandpaper on a flat surface like glass
Click to expand...

From the limited testing I've done and with only second hand knowledge of liquid metal I can still say that going #timless isn't just superior to thermal paste but also comfortably beats the piss out of liquid metal.
The only that might be comparable is direct die cooling.

To put things in perspective, actual lapping of a CPU allows me to achieve a flatness error that is less than 1% of the best that sandpaper can do.
 
Last edited:
  • Like
Reactions: erek
like this
You must log in or register to reply here.
Back
Top