Ryzen 2700x de-lid. (I did it a while ago...)

[ShrimpBrime]

I don't think that will work. I think you will end up with worse temperature.

You need to cover the whole circumference with silver to cover the losses bonding between the silver/copper.

It works well in place of the removed IHS plate.
Been using it for about 10 years now.

 

[xtreemchaos]

great post, I found the most stable overclock on my 2700x is 4.2 1.36v it gets cb15 1950. I have run it at 4.350 and it runs good enough to play games but crashes in cinnibench no matter how much volts. on boost it hits over 4.3 but I think that's not all core but the voltage is 1.450 or more. change pic the othier was not underload.

 

[ShrimpBrime]

Runs just about the same as mine, but I need a little more v-core than you.
1.412v @ 4.1Ghz. (Stable)
More than this requires me to disable SMT, 4.2ghz is scetchy all 16 threads. Sometimes passes stress test, sometimes does not.
8 cores only....-->
1.450-1.5v bench/game stable at 4.4Ghz x4 cores.
1.450v-1.475v stable at 4.3ghz IBT.
1.4250v-1.450v stable at 4.2ghz IBT.

Will seemingly run lower v-core with performance and pbo activated, but cores jump all around. I clock this chip manually almost always.

 

[xtreemchaos]

mine run really ruff until I updated the aorus gigbite bios to f40 thats the first pci4 bios and after that there was a marked improvement in clocks both cpu and ram , im quite happy at 4.2 it keeps the temps to around 70c when its working hard.

 

[ShrimpBrime]

Not bad. Think Im the latest Asus bios. Clocked the same on 2 different boards both with different bios versions. Memory was a different story.
Now at 3200 cas 14-16-16-35 1.3750v corsair.

 

[xtreemchaos]

I have hyperx pred in the rig it will run great @3600 after the bios update but for some reason the game fallout 4vr don't like it it makes it stutter so I have it @3200 and it runs fine it only happens with that game I think its down to bad coding from Bethesda.

 

[ShrimpBrime]

Here's a screen shot for the memory where I've gotten it to so far. It booted 16-17-17-42 on auto timings.

 

[ShrimpBrime]

Here's a better picture of the Ryzen 3 1200. You can see where all of the solder is lapped from the silicon, a little remains at the ends.


This next one is a Ryzen Athlon 200GE processor which was not soldered. The 220GE looked identical.



And here's another pic of the 2700X straight after the de-lid. You can see there's quite a bit of solder to remove.

 

[Jism]

Those unused pads solder capacitors to them if you are skilled enough. It may give you enhance performance. Get the capacitors from old dead CPUs.

Wait what?

They wont offer enhanced performance as in IPC or something, but perhaps a more stable OC. However with Ryzen your already pretty much maxed out unless you start going exotic (below 20). A tec is usefull for that but an awefull requirement of power (and heat you need to take away).

Edit: You need to stay away from voltages above 1.37V for the 2700x, esp when you start burning it with IBT or so. Your chip will degrade in matter of months or even worse weeks!

 

[ShrimpBrime]

Ive degraded a Ryzen 200ge Athlon already. It was above 1.6v to do so.
PBO OC state 3 and 4 go near 1.6v read more like 1.589v but above 1.6v if enable LLC to maximum setting.

The TeC did pretty well. I did hav to increase the plate size and very not practicle for daily use. Had a idle temp of about -30c and 4ghz cinbench just above 0. I would not load duration with a TEC unless it had the capability.

No I will not be modifying the cpu by trying to solder it.

 

[ShrimpBrime]

 

[delshay]

Wait what?

They wont offer enhanced performance as in IPC or something, but perhaps a more stable OC. However with Ryzen your already pretty much maxed out unless you start going exotic (below 20). A tec is usefull for that but an awefull requirement of power (and heat you need to take away).

Edit: You need to stay away from voltages above 1.37V for the 2700x, esp when you start burning it with IBT or so. Your chip will degrade in matter of months or even worse weeks!

If you get a meter & measure the resistance it will have links to either vcore, memory, SOC, ect, ect. VCORE is what you are interested in, plus I want to see users show they skill in soldering.

IMPORTANT: Use a ESR meter before attempting this task.

EDIT 2: Spot on. From what I can tell by doing this mod on older processors you only get a more stable processor, it does nothing to enhance overclocking.

Undervolt & OC (same speed) is more stable, in other words, you have less errors "at the same clock speed". It's a reduction in errors, but does not get rid of all the errors when Undervolting/OC.

 

[ShrimpBrime]

Cold plate to naked die 4ghz.


TEC with small plate.


TEC with large plate.

 

[delshay]

Have you experimented with regulation. ie connect it between 3.3v & 12v rail for around 9v, or even try connecting it to 5v & 12v rail for 7v. I know this can be dangerous but it's something I want to try to stop the frosting & be used for daily use.

EDIT: Probably better to build a proper regulated circuit board off the 12v rail.

@Jism

If you want to see "real world performance changes" to your processor underclocking/overclocking then you need to be looking at modifying MOSFETS.
Pay extreme attention to the turn on/off & "rise & fall" times. Outright massive amount of Amps of the MOSFET is not always the best way to go.

Test here has shown switching characteristic of the MOSFET has out performed higher amperage MOSFETS in my old classic PC set-up. Gain extra 4MHz overclock & around 3Mhz when undervolted.

 

[ShrimpBrime]

Well no I have not regulated in this way. I can regulate via end cooling solution. Can run the TEC warmer, used different sized plates, or even stack plates.
At 12v the TEC is actuallty under powered.
Have to be careful using a TEC straight to cpu because if the TEC gets heat soaked, the cpu will overheat with it.

Ryzen chips, if set up just right, could be TEC cooled for daily sub ambiebt temps I suppose.
Maybe a rheostat and 15v power supply.
I did the above out of curiousity and think to have nice results.
By running that Ryzen chip cold, I was able to use significantly less voltage at the same clock speeds.

So at any rate, with temps below 20c on a cinebench run and frozen at idle, The voltage actual during use was only 1.188v.
One thing I discovered is that this processor does not like v-core too low, as it starts drooping the reference clock nearly 1mhz.
Never the less, I find this chip to be a decent leaker. It would be pretty good under LN2 I think.