• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

10900k temps and core to core delta

Associate
Joined
6 Nov 2014
Posts
476
so I decided to grab a 10900k for my new build judge all you want
wink2.gif


Im busy playing with it and have found it very good for overclocking but at 1.345V 5.2ghz all cores if I run prime 95 avx small fit's it hits over 100 degrees instantly. Anyone else with one of these experiencing this just want to know its normal and not my waterblock mount???

Also I've noticed a core to core delta of a whopping 20 degrees
eek3.gif


Is this normal?

Just want to judge my experience against someone else to make sure everything is normal and I've not got any loop issues

Running Aida64 cpu, FPU, cache stress test 1 hr with Valley in background
Cpu package 93 degrees, core temps 74 74 78 76 86 93 89 80 81 73

Running 3 x 360 rads in a lian li 011 dynamic xl D5 pump
 
You will need to direct die to run small avx2 at 5.2. Derbauer has a DD frame coming out soon. Can't use the 9900k one due to die size differences.

Also running aida + valley is actually weaker than just running aida as the cpu is splitting the load.

If want to check the flatness of the IHS, you can do the spin test on a flat surface. Core to core shouldn't be beyond 7ish degrees give or take. 20c is a lot.
 
Yeah that's what I thought, at idle all core drop down to high 30's low 40's without any real difference in core delta

Does using konductanaught on cpu ihs void warranty?
 
so I decided to grab a 10900k for my new build judge all you want
wink2.gif


Im busy playing with it and have found it very good for overclocking but at 1.345V 5.2ghz all cores if I run prime 95 avx small fit's it hits over 100 degrees instantly. Anyone else with one of these experiencing this just want to know its normal and not my waterblock mount???

Also I've noticed a core to core delta of a whopping 20 degrees
eek3.gif


Is this normal?

Just want to judge my experience against someone else to make sure everything is normal and I've not got any loop issues

Running Aida64 cpu, FPU, cache stress test 1 hr with Valley in background
Cpu package 93 degrees, core temps 74 74 78 76 86 93 89 80 81 73

Running 3 x 360 rads in a lian li 011 dynamic xl D5 pump

Any variance in all core temps of more than 10c while under all core load is a clear sign there is a problem. Identifying it on the other hand is a hit and miss - first place I would start is to delid the CPU (the IHS may not be even), could also lap it but I'd just go straight for the delid, Derbauer says a delid will drop temps 7c anyway so you hit two birdies with one stone

 
I'm guessing 1.345V @ 5.2 is alittle to much more AVX stress loads, you can get away fine with AIDA64 FPU, muti cinebenchR20 runs ect but blender runs or OCCT small with avx is too much

maybe try 5.1 with 1.28/1.3vcore see if that's ok?

avx sometimes adds alittle more juice to things, keep an eye on vcore with cpuZ (HWinfo doesn't seem to show vcore atm?)
 
Would a 20 degree core delta be grounds for an rma? Most I have ever seen is 7 or 8 degrees

If it can be replicated at stock settings then I'd say yes the delta is too great and is worth an RMA. That's just my personal opinion, not sure how the vendor views it, but a 20c delta is ridiculous.
 
Im going to drain the loop and re-apply my thermal paste and block and see if that makes a difference.

I did the old spread it thinly and evenly method with my thermal grizzly might try the X method and see if that makes a difference
 
Ok guys so I thought YOLO and went for it and I delidded my 10900k and applied liquid metal and what a differnece!!!!!!

Same test as before aida64 with valley in the background my temps are now average:
Cpu package 76 core temps 65 65 68 67 68 70 66 68 62 64

Previous

Cpu package 93 core temps 74 74 78 76 86 93 89 80 81 73
No more core to core delta and roughly a 20 degree improvement in temps!!!!!

This is without letting my liquid metal and kryonaught cure so temps should improve further
 
Awesome stuff! Delidding seems to definitely be worth it on the 10th series even though Intel claims they've made improvements, their quality insurance checks can't be accurate enough if they let slip an issue that your 10900k had.
 
BRAVE MAN!
any pics?? I wouldn't mind having a go popping the heatspeader like with the Der8auer tool or maybe even in a vice (hit with a block of wood same as the 4770/4790k)
but the knife and the soft metal solder TIM contact of the die scares me alittle
 
Better to get quicksilver or just use liquid metal as it dissolves the indium solder, scrape off as much as you can with a credit card then apply a blob of liquid metal and spread it, then agitate it every 5 mins or so. After repeating for about 20 mins wipe off and only the plating remains which can be removed with a polish.

Takes more time but risk free.
 
The indium solder is so soft i could use my finger nails and then some 3000 grit sandpaper to finish it off.

AVX workloads are still way too much for this chip even delidded it hits 100 degrees prime 95 small fft.

I get whea errors in first 10 mins of prime 95 AVX small fft, occt avx2 with temps in high 90's/100+ so errors are likely temp related.

What would you guys say is gold standard for stability testing?

Does errors in avx mean my system will be unstable?

I've got it running 5.2 all core 1.35v 46 uncore

Done 4 hours realbench stress test
1hr prime 95 small fft in place (12/12) non avx
12 runs of cinebench r20
12 hours occt non avx
4 hours aida 64 cpu, fpu, cache

No issues yet
 
Last edited:
I use testmem5 with extreme profile for mem first to hammer my mem oc.

Then occt large/avx2 for a couple of hours. This is more effective with tuned mem as it speeds up avx workloads. Without mem tuning, it’s not as useful.

That combo has been rock solid for me. I’ve done everything from p95, x264, realbench, x265, blender, all the Aida variants, and so on.

Here’s an AIDA64 test piece I did not too long ago that might be of relevance: https://www.reddit.com/r/overclocking/comments/fgxxxq/understanding_aida64_stress_tests_on_a_9900k/

For reference, my pc stays on 24/7 and no crashes, blue screens, L0 correctable errors or such. 9900k @ 52/48x and 17/4200 1T mem on a XI apex.
 
so I decided to grab a 10900k for my new build judge all you want
wink2.gif


Im busy playing with it and have found it very good for overclocking but at 1.345V 5.2ghz all cores if I run prime 95 avx small fit's it hits over 100 degrees instantly. Anyone else with one of these experiencing this just want to know its normal and not my waterblock mount???

Also I've noticed a core to core delta of a whopping 20 degrees
eek3.gif


Is this normal?

Just want to judge my experience against someone else to make sure everything is normal and I've not got any loop issues

Running Aida64 cpu, FPU, cache stress test 1 hr with Valley in background
Cpu package 93 degrees, core temps 74 74 78 76 86 93 89 80 81 73

Running 3 x 360 rads in a lian li 011 dynamic xl D5 pump
My friends does same nothing to worry about.
 
Better to get quicksilver or just use liquid metal as it dissolves the indium solder, scrape off as much as you can with a credit card then apply a blob of liquid metal and spread it, then agitate it every 5 mins or so. After repeating for about 20 mins wipe off and only the plating remains which can be removed with a polish.

Takes more time but risk free.

I'm about to do the delid using that tool with my 10900k (5.1 at 1.30v, im hitting 90C with prime95 blend and a 360 aio). Its been awhile since i delidded, for the scraping off with the card, thats just the part that held the heat spreader on right? Or are you saying add liquid metal to the solder thats leftover and then scrape it off?

Any other tips on the general process. I ordered liquid metal so ill put that on the center etc.
 
Back
Top Bottom