Originally Posted by tcclaviger
Just a datapoint for the masses:
I've had this 1680v2 for a long time, its been flogged over 1.5 volts, seen 1.45v daily for a year, and generally abused quite heavily. It has been cooled well the whole time though.
It's just NOW starting to show signs of degrading. I had to back off from 4.6 to 4.5 daily @ 1.372v (after vdroop), to get stability. So thats like 4 or 5 years getting the balls beat off it and it dropped 100mhz.... lmao, amazing little monster.
Not a chance you'll get 4.5 with a 212. Think more like 4.3ish.
At 4.5 with my monster loop I see 60ish in prime95, but that's with liquid metal TIM, was near 70 with Kryonaut (mount was probably contributing tbh).
Temperatures don't matter much tbh, this chip is voltage limited not thermal limited. In the dead of winter with the water near freezing, it still takes 1.5v ish to get anywhere near 4.8.
LN2 isn't going to make the same difference as it does on some other chips, sure you'll get more, but it'll be limited by the whopping 25MB unified L3 on 1st generation 14nm.
Any X79 board (aside from the weird chinese crap produced recently) with a monoblock will handle a heavily OCd 1680v2. Without monoblock...I'd suggest keep package power around 160watts max.
The power phase thing, it's massively thrown around by people who understand the electronics often (thanks youtube) but don't actually understand what all that cool tech info translates to when pushing a chip to it's limits. Small story below on why you NEED to water cool the VRMs if you're going to load your octocore at 200 package watts for extended periods of time:
I first bought a Rampage 4 Gene, going for a small form factor build when it came out. I threw in my 3960x ES and proceeded to take cracks at records at the time, sure enough, my 3960x ES was an absolute monster, scaling linearly up to 5.1 ghz!! Getting there though took a good chuck of vcore as one would expect, and drew a lot of power. Chip was water cooled and I was doing this in December, with window open, so room ambient was mid 40s f.
Hit it with a realtime priority intel burn test and.... smoke, sparks, and the magic smell. One of the MOSFETs had melted the solder under it and slid downwards on the board, it did autoprotect and immediately hard shutdown :O Like an idiot, I hit the power button to restart it after a minute of head scratching, sure enough, it still worked (lmao). Head straight to the ROG forums posted a picture and Shamino direct swapped my board out for one in his lab via DHL 2 day express, prepaid return provided. He wanted it to inspect and identify the failure mode, as it was one of the first sold in the US, and as far as I know the first pushed that hard, so I left all bios settings as they were and sent it to him to probe and prod.
I have had 4 boards that would VRM throttle when pushed really hard and the VRMs were air cooled, every single one stopped doing it when I threw monoblocks on them instead of CPU only blocks. The old "monoblocks are for show" is 100% NOT true if you intend to actually flog your CPU, but, there are not many people out there who do, most people do some crazy OC get some scores and back way off to a middling OC for daily.
NOW: that was a lower power part than a 1680v2 when pushed to it's limits by a good bit. All these people buying ryzen boards, eschewing monoblocks and going with just CPU blocks, are blessed the Ryzen 1k/2k/3k are essentially tirds for OCing beyond what AMD has specced them to. The majority of the non-halo boards out there could not handle 200watt+ draws, air cooled, in a case for extended periods of time.
Hi, I reference the first part of your post and your own OC exploits - ! - I got a Corsair H150i - triple for a test bench - and did a test OC of 4.6 Ghz - @ 1.340_v_tested_in_P95_29.4 __temps reached_77_hottest_core_20k-points in passmark & 2550 single threaded