[SOLVED] New Ryzen 5800x rig running hot

[Tangled in the Interwebs]

Hi everyone, I am a newbie to Tom's! My first question.

I recently built a new PC:
AMD Ryzen 7 5800X 8-Core Processor 3.80 GHz
Gigabyte X570 Aorus Master Motherboard
Crucial Ballistix DDR4-3200 64 GB RAM
Noctua NH-D15 chromax black, both fans mounted
1x exhaust fan, 1x door mounted side fan (~8" diameter), 3x front fans drawing air in

I am coming from Intel, so I am not sure what temps to expect. But, on idle the system is in the upper 60s to the mid 70s C, which seems hot.

What I have done so far:

• set aggressive fan curves that by the time CPU reaches 65, they go full blast. Still have these temps
• re-seated the Noctua... and used much more thermal paste than I initially used. I got the idea from reading an article showing that for Ryzen chips, using more paste shows better cooling. But after re-seating, the temp range is now slightly higher, tending more towards mid 70s C

I must admit, I am not sure how tight to turn the screws for the Noctua.

Right now, I only have Firefox and Outlook open while I write this post. CPU shows 67 C, but anything that tickles the CPU will send temps higher immediately.

Edit: using Ryzen Master software, I ensured that I am running the "Default" speeds - no overclocking, no performance boost. But still my speeds are running ~4600 MHz

How do I get my temps lower?
How tight is enough on the Noctua mounting screws?

Thanks!

 

[Tangled in the Interwebs]

That is a red herring.
An all core benchmark does not represent overall performance. Performance scales with how many cores are active at a time.
The cpu ratio of 44 will reduce performance when fewer cores are active.

Ok fair enough. But to call it a red herring usually indicates misleading purposefully - which I certainly did not do. I am naive, and I am not experienced enough with this stuff to mislead anyone.

However, your note is otherwise well-taken. So, while not exactly scientific in my approach, I extended my little test battery to single core Cinebench runs to see the difference. Turns out there is a better score for the default settings where the cores are free to jump beyond the 4.4 GHz range, which as you indicated should be expected.

The test may have been running a single core at a time, but it was certainly switching cores between each box being rendered, mostly between best and next-best cores. At times other cores took the lead, but I wasn't quick enough to catch those fleeting moments with a screenshot. Some screenshots for those interested below.

If you have suggestions for other, better, free, benchmarking/stresstest packages worthwhile to try I'd give one or two of them a go as well.

But honestly, I am grateful that you pointed me in the direction to resolve the heat issue that still gives a very decent performance result, without having to resort to expensive or time consuming fixes. Indeed, many thanks.

BTW, it does seem to take a while for the imgur pics to load in, but they do. Eventually. Not sure if its because I am at a novice status on this forum and all images need to be reviewed/sanititzed, or because of an imgur limitation.

Single core results @ Default settings - temps cycled between 69-72 C, score of 1419:

https://imgur.com/zq3SRr7

View: https://imgur.com/zq3SRr7


Single core results @ 44 CPU Ratio - temps remained at 55C, score of 1403:

https://imgur.com/a/c0TkM9G

View: https://imgur.com/a/c0TkM9G


Core cycling on single-core Cinebench tests, both on a default-settings run:
"best core"

https://imgur.com/ZaI1sE4

View: https://imgur.com/ZaI1sE4

"next-best core"

https://imgur.com/NqwDQh7

View: https://imgur.com/NqwDQh7