Originally Posted by Alastair
I wouldn't say temps are holding me back at all. At least not from a CPU frequency perspective. As 5GHz is very easy to attain and stabilize.
It's the CPU-NB clock that is throwing me off. Anything over 2700 is a crapshoot. Evn with 1.47V applied. Core temps under load like Prime or IBT don't break 60C. And generally sit around 55c or less.
Socket temps are another story because I know I riding the limit of what this motherboard is capable of by conventional cooling means (VRM fan and backside socket fan.) socket temps will hit 70s under stress test. And at 5GHz will actually break 80C socket. But funnily enough it doesn't throttle at those temps. Because my VRM temps are only at about 80C or so. I've noticed throttling of I turn my VRM fan down and it breaks 110C VRM it throttles.
Wanna know something really neat?? (You already know I de-lid soldered processors a many) and with that mentioned let me give you my temperature experiences and how I KNOW that Cpu temps, be it socket or core, make a large difference on clock speeds.
So what you experience above with a soldered chip, I also experienced before any de-lid (holds true with Phenom II processors I've de-lidded) having very high socket temps like yours but keep the cores cool.
WIth FX processors, the opposite happened. Had nice low socket/board temps, but the core temps where a little higher (not much within 10c).
So what is happening, is the cooler, cpu, and board get a little heat soaked and you get higher temp readings there. But it doesn't make sense that the cores are lower in temp, but that's because a lot of the heat is directed away through the cooling device, it's a straight shot up and naturally temps seep out and down through the boards PCB combining with the soaking VRM heat through PCB tracings. Thus you get a higher socket temp.
WIth Phenom II processors and high TDP, I was able to run the socket and core temps just about the same usually within a couple degrees.
FX de-lidded the cores where a little higher and socket temps where lower.
The naked core was having heat removed straight through to the cooler (water block) instead of heat soaking solder and IHS plate plus the TIM before the cooler (water block) and was less combined with the VRM heat through the tracings.
Once the PCB gets warmed up which is a slow process because it's not thermally conductive, it takes longer to cool off. The PCB being warm keeps the tracings warm and that's why FX users always see a higher socket temp. Kind of weird how it works, but I swear by the teens of de-lidded Phenoms Athlons and FX and Ryzen processors I've done.
I can talk more about Ryzen De-lid temp experiences if so requested, but here I'll try to stick to FX in this thread.
So any how, I did FX experiments with TEC cooling. It was an issue running 8 cores and naked chips. So to hit beyond 5Ghz and a TEC while de-lidded, I cheated and shut down cores. But never the less.... Point is how low I was able to run the voltage at -5c to 0c and actually benchmark a single thread. IMO that's pretty good.
Heres the specs on that:
FX-8350 @ 5426.77Mhz TEC cooled, idle -6c in ss and 0c at single core load.
V-core @ only 1.464v - Used Asus M5A78L-MX Plus 4+1vrms. PiMod time of 17.223s.
So yea, cpu temps really do hold an FX behind big time. In fact the Leakage goes away A LOT. Not just a little. So if your at 50c, you'll never see 5.6Ghz like I was able to accomplish at even 0-10c even only two cores on ambient air or liquid. (I've tried)
You guys get these colder, believe me, you'll see some pretty amazing accomplishments.