29 minutes ago, flycln said:

Bad thermal compound application. 
 

I’ve never seen such huge differences when all cores are at 100% load. 

Could be but I doubt it. If I recall correctly I mounted the heatsink twice and on both occasions I got the same result..

The waterblock is mounted direct to die with Coolaboratorys liquid pro.

The is a tiny amount in there. only enough to fill sliver of a space between probably less than 1/3 of a sq. inch die

and block. I think it just a case of some cores are better than other.

I might take it off and take another look though.

20 hours ago, F737NG said:

Hmm... Affinity Mask, you say...? 🤔

https://www.avsim.com/forums/topic/612746-new-p3d-53-affinity-mask-calculator/?do=findComment&comment=4713056

Give it a try. It isn't required to run P3D on Core 0 any more.

 

Well that was easy and it worked. I have HT enabled in the BIOS but on the last physical core I have it disabled so in effect I have 19 cores.

That websites calculator allowed me to take that into account and make a custom JobScheduler for my .cfg with main thread on core18.

Took a test flight and its working perfectly with RTSS showing the main thread is on Core18.

 

Previously, I had HT turned off on physical Core0 and physical Core9 to accommodate both MSFS and P3D. Now both apps put the main thread on logical core18 which is physical core9

Great!

21 hours ago, flycln said:

Bad thermal compound application. 
 

I’ve never seen such huge differences when all cores are at 100% load. 

You where right!

Funny how moving a PC can make something stand out. I had to move my system out of the basement to a room up stairs for reno's

Its about 6 or 7 degrees cooler in the basement and when I was seeing temps in the mid 80's it got me suspicious.

Anyway, the CPU is a delided i9 10900k direct to die cooler using a Rocket fish kit and H115i Pro.

I was getting about 80c under load. I though this was OK because that seemed to be the expected temp

with a H150i pro and the IHS in the sandwich. I was always puzzled by the temp spread but I never questioned weather the

Block was seated well. I was very careful installing it. But when I took it off this morning it was obvious that one spot on the was

very tight against the block and rest not so much. Here is where it's at now:

 

10 mins + into OCCT LInpack with AVX.

 

I lapped the Waterblock. I think it was slightly concave. The die with the rocket fish kit installed is only a tiny bit proud of the kits top plate.

What I did helped but I think there is rome for improvement. I also think there is a little bump in die and I also think the rocket kit top plate could be interfering with

a tight contact with part of the die. I'll do further work but for now I am happy to see that after being initially Blah about the results of direct to die. I see its really very good

73c at 5.2ghz on all cores. 🙂

How did you lap the waterblock?

5 minutes ago, pgde said:

How did you lap the waterblock?

Piece of glass, started at 400 grit then 600, 800, 1000, 2000.

coolaboratories liquid pro sets after a while and leaves a stain on the water block.

I sanded enough to just about remove the stain and the way the stain gradually got smaller

and then disappeared was the give away that the surface started out as concave.

After completing a OCCT linpack test I see that the hottest core is 84c and coldest is 71c

So still quite the spread. 71c max on a i9 10900K at 5.2ghz is pretty darn good I think.

That's the same core that was the coolest pre remounting the block and could see from the

pattern in the tim that that spot on the die was pressed hard against the waterblock compaired

to the rest to the die. I think I'm going lap the die and sand down the Rocket Fish Kit top plate because

I think its slightly proud of the die surface. I'll do the latter first.

 

Thanks FBW!

So if anyone is interested. I've fixed the issues.

67c is lowest and 75c is the highest and they are right next to each other.

bank of america drive thru hours