Hi, first post on the forums so go easy please.
I bought a 3900x with an x570 taichi at launch (upgraded from an 8370fx) and its been an absolute champ from a performance perspective compared to my old system, however cooling in my mind at least has been a bit problematic. I know a video has been posted about TIM application however no matter what I do I seem to get deltas between the CCD’s of about 10°C. Is this normal? Would using liquid metal on the IHS close the gap and provide more consistent results?
No the chip is soldered already.
What temps are you getting? What is your case and fan setup?
Just to clarify, I was not thinking about doing a de-lid, just using the liquid Instead of a regular TIM. The cooler is a H115i pro with stock fans (got some of the noctua 3000rpm high pressure fans in the post) running 1.3 volts 4.3GHz. At the moment ccd1 is 27°C and ccd2 is 38°C at idle
Not much you can do about that to be honest. its probably a silicon quality difference. If you have reapplied thermal past and no major difference after the re-apply. I highly doubt its anything you can change that would improve it.
I can check my 1950x when I get home and see if there is a major difference between the chips but its not exactly a 1-1 comparison.
Are you hitting TJMax or just were confused as to why there was a difference.
@SoulFallen have you see this a lot?
Fair enough, I was thinking that the ccd’s should have a smaller delta and maybe it was an application issue, in my mind i was thinking that the use of liquid metal might close the gap because your basically painting a thin layer on the ihs and cooler allowing for full coverage. At 1.3V its for the most part fine even in prime with small fft’s (does get toasty), go to 1.35V and your looking at thermal shutdown.
Pretty normal and no @Barrow117 liquid metal is not going to lower the delta, It’s because AMD uses 2 dies in the 3900x silicon lottery is gonna make the deltas larger than on a single chip CPU
1 Like
you also have to remember 1 chiplet is full and the 2nd is not the same core count (8 core and 4 core)
I Knew that there was a difference in binning quality but I wast aware that the core count on the chips was different, thought it was essentially 2 3600 and a 3600x chip on one substrate
no amd has always used as many full chips as possible as that allows a single task the highest chance not to hit 2 different chips (or at least that is what makes sense to me)
Ignore the 2 that was typo
the difference in binning quality is huge, so much so that 1 chiplet will boost 200mhz higher than the other. 1 chip is Hercules the other is a potato chip.
Thanks for the heads up, glad I don’t have to mess about with liquid metal lol
As far as I know that is pretty much the case, yes. That also answers your question. 
Just a thought before putting this topic to bed, would locking all the cores to the lowest common denominator and fixing the voltage equalise the temps ?
Are you referring to silicon quality or core count
Unlikely. Because of the difference in silicon quality.
Both.
Lowest boost freq across all cores
Sauce (kind of a departure compared to how they did TR 1st gen)
Potentially, no guarantee.
The difference in clockspeed is only the result. A lower binned chip will still need more voltage to reach the same clocks.
That is why AMD always states that TR gets the better binned chips.