Originally Posted by ivanlabrie
Originally Posted by Belial
Woke up to this, this morning when testing a 'conservative' 4.6ghz @ 1.4vcore.
p95 custom 3200ram, worker #3 stopped with 1 error at 19 hours 51 minutes.
Max temp 70*C.
Can you take screenshots of your bios settings? There are a few cpu related pages...post em here. I seriously doubt you can't run 4.6 with 1.4v really.
Okay, I did the following to try things:
PWM phase Control: eXtreme Performance
Vcore Voltage Response: Fast (auto/standard/fast)
Vcore LLC: Extreme (turbo is recommended for ambient oc, extreme for ln2)
I also turned up my ram to 1866mhz with it's XMP timings (cl7-10-etc...turns out 2133 CL7 doesnt work, i think when i got it to work before it automatically turned down the timings... or maybe not, im not sure, it just wouldnt boot when i manually set them, ill check later if it'll boot 2133 xmp with auto for timings instead of having manually put them in, w/e).
Anyways, it's really helped out with stability, I think, despite Sin0822 and others saying these features don't mean anything. It's not much, but for example, [email protected]
, which is instant blue screen, can boot and do a bit of p95 before failing. [email protected]
, I've been running 20 minutes now ([email protected]
is crash in like 5 min).
I'll keep laying with it but it seems messing with those settings helped (i think the ram is just whatever). I mean the VRM on this ud5h is so cool anyways (i got 3 temp sensors in it, its hard to put temp sensors on this vrm heatsink because it's so wide you can't slip one under the heatsink like you can on the ud3h, but i did stick a few inbetween some of the middle/top chokes which tend to be the hottest, which usually run at temps higher than mosfets as i understand it, and they are running 42*C max.
I guess [email protected]
isn't really impressive either (then again I haven't tried 5ghz at 1.4vcore yet with these new pwm settings) but already lasting 20+ minutes on a setting that should be instant bsod is great.
Will post pics though.
Originally Posted by justanoldman
Are the cores lined up in order on the die, as is core 0 at the top and core 3 at the bottom? Why are the middle ones (#1, #2) always the hottest?
Yes. It's like this. The right side of this image is at the 'top' of the chip, the gpu at the bottom. That's why people are always screwing up their IMC, have ram issues, etc, when they screw up a delid and scratch the pcb - the edge of the IHS that's closest to the die is the top part. I wish people who scratched their die would say where exactly they ended up scratching, but I imagine that's what they are hitting. It'd be no problem if you scratched up the gpu ^^.
Cores 1-2 can be hottest for many reasons. They are in the middle so maybe they get heated by the stuff around them. Some cores actually get more of a workload, or a higher/lower workload (you might think p95 or ibt is p95/ibt balanced load on all cores, buts even if your increase priority on p95, it's not - background applications is running on some of the cores, which may make them run cooler, or hotter, than the cores taking a full load). I used to think on an old quadcore that 2 of the cores were faster (they would complete more fft passes in 15 minutes in p95, after 24 hours they would be a good couple fft lengths ahead), but then I realized that those two cores were where background apps were focused on.