Question MSI z690-a pro wifi + 12900kF + ddr5 + kraken x73 360mm cooler, high temps? bios settings? Fails after 5 hours p95 blend

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.

markm75

Distinguished
Mar 21, 2007
51
1
18,535
Im using this combo with the ram being: CORSAIR Vengeance DDR5 32GB (2x16GB) DDR5 5200 (PC5-41600) C40 1.25V (32gb total), ‎CMK32GX5M2B5200C40

I have the lga 1700 bracket, everything is tight, cpu was done two different times with two different pastes/methods, same results. 2nd time now with arctic mx5 and did the spatula spread method.
Pump is running near 2800, cooler fans at their max rpm (2000ish).

In the bios i have only changed the P core and E core values.. set P to51x on all of them, set E to 39x, set the ring to 39x, set avx to -2.. set voltage to 1.30, LLC to 2.
I didnt turn off anything else or change other settings.. let turbo i think it is, to auto etc. Didnt set SA or Dram voltages, left on auto, set ram to XMP and auto on frequency (i think gear 2), 5200mhz.

All that said, p95 blend will fail after 5 hours. temps still randomly spike to 100c on 2 cores, though most of the time its around 87C.. if i do cinebench it spikes to round 100c again, usually avg around 89c.. score ~26,700 package load seems to avg 235-285 during cinebench, however i think it spiked to 350 at one point.. its nowhere near these wattage with P95 blend though

I'd have like to have thought the voltages are too high for this 360mm x 3 cooler, but i've seen 1.35 mentioned with similar configs and 95c, so unsure.

Any thoughts or suggestions, or maybe i need to tweak some bios settings.. for now i've lowered the ram to 4800 and i'm retesting blend to see why it errored at 1.30.
I ran into this with the 10900k with this same cooler, so im thinking worst case i try another brand? (suggestions?).. i had to delid the 10900k, then my worst temps were 89-92c.
I might redo things after doing the washer mod next
 
Last edited:

markm75

Distinguished
Mar 21, 2007
51
1
18,535
Do you really have a problem that is worth fixing?
Are you a P95 overclocking competitor?
It seems that your current gaming and rendering runs well.
Even if you hit 100c. momentarily, the cpu does not throttle and you keep running.

I do agree that the thermal grease pattern you showed looks strange and offset.
But, not any die bending is evident.

I might think there was too much residual and a bit less paste would be in order the next time you apply some.
Paste is an insulator, compared to metal to metal contact.
But it is much better than air.
Use as l1ttle as possible.

Thats a valid point. And yeah in looking back at least at that pic i took before i redid it, it did look like there was a bit too much.

I guess im just out to solve the mystery more than anything, but running with 241watt boxed cooler mode, i run fine, under 95 in the worst tests at least, but at that mode whether its avx -1 or avx -3, it always runs at 4888 mhz for avx loads during cinebench, i guess neutered by the 241 watt limit, but keeping temps in check.

I was fairly certain others with 360 style coolers were under at least upper 80s no 90s on 4096 watt mode (though they dont list that wattage), i assume its the only way they are getting their avx clocks high enough to get into the 28.5k+ r23 range.

I guess thats the only thing im trying to clarify, wattage limit 241, means neutered avx frequencies of 4888 no matter what.
Ill accept that if so, maybe later change the paste one more time, toss out the generic lga 1700 backplate i got off amazon when the real nzxt one arrives and retest for fun.
 

markm75

Distinguished
Mar 21, 2007
51
1
18,535
So i undid the washers, cleaned it up, tried again this time with the cool master PL360 setup. No change, still throttles (with 4096w or watercooler setting) to 100c in short time.

Specifically here are some results. 5.2ghz/4ghz/4ghz unless otherwise noted.. and case side panel off:

I tried LLC4, set volts to 1.34 in bios, did cinebench it was 96C in 20sec, seemed to be 1.29v then bsod (lower voltages didnt work), this was fixed no offsets.

I then did a standard auto defaults.. with the two cores at 5.2ghz, auto llc, auto volts.. left it on water cooler mode again.. 94C+ easily in 20 sec in cine (1.30v load) and p95 small ffts (disabled avx or not doesnt make a difference i dont know why) but it was 100c within 30 sec too. Despite this too, the avx load in cinebench was still capped at 4888.

Even when i left things at my offset setting of 0.02 neg.. and -2 avx and the 4096 limit, it was 100c within 30sec or 289 watts with p95 small ffts even at 1.22v

Whats strange is that the system is not stable at 1.34v in bios (likely 1.29 load) at 4096 wattage limit, whereas at 241watt limit, with the offset 0.02 neg, its stable in all tests and doesnt exceed 94c on any test (or maybe thats not that strange i dont know).
So i either have a really bad chip since i cant replicate others success (with 4096 limits) or something else is off here.

Last redo of my old cpu cooler the x73 kraken it appears i had nailed the cpu paste spreading, so i dont think thats an issue at this point (my old x73 is back in place)

With my neg offset 0.020 and 241 watt limit, my system peaks (with case side off) at 93C during one of my astro render sequences, i think ill consider this a win for now and not try anything further.. except maybe a component y cruncher to be sure the system is truly stable.
 

markm75

Distinguished
Mar 21, 2007
51
1
18,535
Last night at the 0.02 i ran y-cruncher component all night.. somewhere around iteration 28 and 7 hours into the run, the pc went to sleep, still not sure how in the middle of a run it could do that (i should have had sleep off).

Im callin things stable as is.
However the highly curious in me might grab another chip and test it. Id like to be able to run 4096w limit without hitting 100c as others have been able to.. or even 93C on my day to day render software. Seems to be 78-82C for gaming (flight sim mostly, vr) at this 0.02 negative offset, 5.2/4/4 and 5200mhz ddr5 (VDDQ and ram volts at 1.35)

I guess based on comments here and comments elsewhere though, it really must just be the silicon lottery effect and why i run so hot, not the cooler/not contact now.

Update after another eve of real world usage:
So far, as is, i'm averaging 83-87C for gaming (avx games, msfs etc) and the one program that integrates an image still averages 89 to 93 (depends on room temp obviously).
When the program integrates the image its averaging 1.29 to 1.30v, unlike when p95 small fft ran it was 1.15
I also lowered the ecore to 3.9 (so now running 5.2/3.9/4 ring), that gave me 1-2C drop on heaviest load. (still 0.020 offset)

**blender 1.21v 83c max all 3 tests monster 184 junkroom 105 class 83 204 watt actual max (still at 241w limit, 5.2/3.9/4.0)
 
Last edited:

markm75

Distinguished
Mar 21, 2007
51
1
18,535
It must be a headache bouncing between here and TechPowerUp - I agree with some of the posts there.
What's hurting things the most is likely the plastic backplate and lopsided retention pressure.

Nah i really dont think this is the case at this point, after putting on that completely different aio, having the same results, taking it off, seeing that the past was evenly spread even with that one, i think i got rid of the pressure issue.

I think its just luck of the draw.. my chip needs more juice to use the 4096 watt power limit mode and too much so.
 

markm75

Distinguished
Mar 21, 2007
51
1
18,535
Update:
So i opted to try another chip to try my luck. I found that it was largely the same. Same temperature issues with even auto and 4096 power limits etc. It may have been a smidge less hot (2c) at the same 020 offset i was using originally though, nothing to write home about.

So i put my old chip back in, i've now forgotten whether i still had the washer mod in play or not, i think last round i took them out, but made so many changes i got lost lol

At any rate, the old -0.020 offset i had set before didnt work, i had errors in prime95 in short time on blend even, so i think either i got lucky with my previous 8 hour error free period or i wasnt as stable as thought.

I ended up adjusting ram to 1.38v instead of 1.35v (i think it was the ram that was unstable, as i set that to 3600 temporarily and did a p95 small fft for 4 hours no errors, whereas small fft gave me errors within 1 hour before).

I also ended up going from llc4 to llc5 and going with 0.025 negative offset. P95 blend over many hours (16), was a max of 94c with the case side off. P95 small fft would heat to about 89c with the case side off within 30sec for reference. Im at the 241watt power limit as well.

*I did set the ring to auto, im not sure if it was this and/or llc 5 that lead to better stability at even 025 vs 020. Ecore still at 3.9

At any rate, all stable, all good, still a max of 82c+ for msfs but thats ok, can always add throttle points for the 90c image rendering that i do if it ever goes higher.
 
Last edited:
Jul 23, 2022
1
1
10
I have a similar issue, but I am able to clock my cpu at 4.9ghz at 1.25v with 253w allowance, in cinebench my cpu will get up to about 94-95c but will not thermal throttle, so I found a happy median for now. I ordered one of those bend free brackets and will install that after it comes in to see how it effects lower temps with a little more overclock.

But out of the box enabling XMP, my MSI z690 godlike would ramp the thing up stock in cinebench my cpu would max out at 5200mhz and thermal throttle straight to 100c on a Corsair h170i 420mm radiator.

So I'm living with 4.9ghz until the frame thing comes in and trying it again. I mostly use the rig for gaming and while gaming, the CPU will settle around 55-65c
 
  • Like
Reactions: markm75

markm75

Distinguished
Mar 21, 2007
51
1
18,535
Well, i've switched ram and that threw a monkey wrench to my stable settings in some way.

I picked up this ram F5-6400J3239G16GA2-TZ5RS 1.40v C32 6400mhz (up from 5200mhz 1.25 i think it was with the corsair). There were some ok gains in VR with the faster speeds (3-4 fps at best).

At any rate, i thought i had a stable set of settings, just went with auto on volts for the ram (had tried 1.45 initially and something wasnt right). I did initially start here with offset and changed my -0.020 to AUTO, thinking maybe i could get away with that (dumb i know :D ) . Ram volts set to auto still and llc at 5

I ran prime95 blend with its default settings (the avx checkbox to disable was not checked), it hit a max peak of 97c, but averaged 68 to 85) for some 12 hours no issues.. Ran cinebench no issues, gamed no issues. I run my astro imaging render tool and before it could even get 3/4 the way through or very hot at all, it bsod with 0x101 error (usually means needs more vcore).

So im back to the drawing board. Im not shocked an offset and auto would fail randomly but i am shocked i could go 12 hours p95 no issue but my render tool bombs (avx is at -2 and i dont think it was using avx). I think maybe it vdrooped in one of the cycling rendering points where the cpu volts just got too low (even though p95 smallest ffts im at 1.165 or so and fine for 2 hours)

So i may put everything back at system defaults and try again (with vcore auto and llc auto), if i bomb on auto with the simple render again, something is really off.
 

markm75

Distinguished
Mar 21, 2007
51
1
18,535
Well, i've all but given up on this board/combo..

Recap at least as of now.. its the msi z690-a wifi board + 12900kf + gskill 16gbx2 6400 cl32 ram at 6400 (latest bios), wattage at 241 for power limit. washer mod. mx4 paste, x73 kraken

I've been now using adaptive + offset.. current setting 1.34v -0.010 offset
Ram xpm
Auto turbo ratios on both p and e and auto on ring (i think the 6400 mhz ram forced me into this setting, before it was pretty stable at 5.2 fixed and 3.9 or mostly stable).

Prime95 blend without disabling avx still spikes to 100c (before i swear it never topped 95c, but that was at just "offset" and -0.015 with 5200mhz corsair ram), though voltages similar under load which puzzles me (around 1.17v).. P95 blend hits 10 hours no errors now, thanks to higher 1.34v bios setting.

However, msfs upon exiting gave me an error about memory not being able to be read etc, i think means the memory is unstable still.
I'm doing a memtest to 8000% to see if any errors (probably 9 hours, so far 1.5 in no errors).

I may switch to the unify z690 x series board, i've heard its more stable.
I know i shouldnt need 1.34 to be stable at 5.2 (which in this case is auto so only a few cores 5.2, if i set to 5.2 fixed all core it will crash frequently on tests)
 
Last edited:

markm75

Distinguished
Mar 21, 2007
51
1
18,535
I have a similar issue, but I am able to clock my cpu at 4.9ghz at 1.25v with 253w allowance, in cinebench my cpu will get up to about 94-95c but will not thermal throttle, so I found a happy median for now. I ordered one of those bend free brackets and will install that after it comes in to see how it effects lower temps with a little more overclock.

But out of the box enabling XMP, my MSI z690 godlike would ramp the thing up stock in cinebench my cpu would max out at 5200mhz and thermal throttle straight to 100c on a Corsair h170i 420mm radiator.

So I'm living with 4.9ghz until the frame thing comes in and trying it again. I mostly use the rig for gaming and while gaming, the CPU will settle around 55-65c
did you get that bracket installed yet, curious if it helped.

i've basically gotten to the same point, override 1.265 (LLC5, llc6 i can probably use and drop 2c), everything on auto, ram at 6400, my astro render pulls in around 92c, i think cinebench was about 96c, smallfft max power is 99 to 100c, 241 watt limiter. I think p95 small ffts under load was something like 1.14, i forget.

I CANNOT get adaptive to reduce the nominal windows voltage, its always near 1.38 to start, i think from reading its a probably with adaptive. Offset even at -0.020 or -0.015 im unstable (at least at 5.1/3.9/4.0).

Maybe with the bracket which should be here soon i could think about trying for 5.1/3.9/4 at least, for now, auto.
 
Last edited:

markm75

Distinguished
Mar 21, 2007
51
1
18,535
The other day i installed the bracket. It brought temps under max load down about 4C. I was happy with that.

I now run at 1.255 adaptive (no offset) LLC6 (6 was the sweet spot), still neutered with box cooler mode and 241 watt max load (i did find that cinebench would throw an error if i set this to 4096/water cooler, so i guess 1.255 isnt enough volts for that mode). I run the main p cores at Auto, so around 4.9ghz most of the time.

I left the e cores at 40x and ring 41x. P95 stable for 12 hours. Max normal game temp of 83C, render max temps are 89-92c i think (these both being with the case side on, so about 5c higher than if i left it off). I could try for 5 ghz all core, but the higher volts would result in overall higher load temps and not worth it (i even tested 5.2ghz briefly, there was no change in msfs performance).

With the 1.255 adaptive,windows vcore can show as high as say 1.322. Cinebench tho under load is around 1.20, i think p95 max sfft was around 1.14-1.16.

passmark cpumark is roughly 44976 (5ghz was 45900, single thread 4205)
Cinebench is around 27680 (at least in my blank os, my main one still has lower numbers), i believe it was 28k for 5.0ghz.
 
  • Like
Reactions: shady28

shady28

Distinguished
Jan 29, 2007
425
297
19,090
That's an impressive setup, especially for a relatively inexpensive DDR5 motherboard.

There's a site overclock.net that you might want to check out the forums, they have ones specific for Intel and Intel motherboards. Nothing wrong with Tom's but that site has a different and pretty knowledgeable crowd that will speak your lingo more fluently and provide better feedback.