Overclock.net banner
Cancel
Create thread New Forums

My 5800X seems to be running *exceptionally* hot. What can I do?

Jump to Latest
1 - 20 of 88 Posts
C

coredumperror

· Registered
Joined
·
19 Posts
Discussion starter · #1 · (Edited)
So I have a 5800X in a Gigabyte Aorus Elite X570 board. I'm new to Ryzen, having previously owned an Intel I7-6700K. While I have a decent amount of experience with overclocking that CPU, I'm realizing that said experience doesn't seem to translate all that well to Ryzen. Or else something is very wrong...

My biggest issue is that even when not overclocking (using bone stock BIOS setting except XMP), my 5800X hits TJmax (90C) and starts thermal throttling after just a few minutes of stress testing of any kind. This is rather disturbing to me, as I have a Noctua NH-U12A on there, which worked fantastically for keeping my rather poorly binned 6700K cool under stress tests during heavy overclocking. With my 6700K OC'd to 4.6Ghz at 1.42v, my temps capped out at 80C during 15 minutes of SmallFFTs in Prime 95 v26.6 (so no AVX).

For comparison to my previous CPU, this 5800X hits 90C after ~2 minutes of SmallFFTs, then sits at 90 as the system throttles down to avoid getting hotter. Hitting TJmax under non-OC loads certainly seems like it shouldn't be happening.

I've already tried re-seating my cooler. My first attempt did turn out to be bad, as I did not use nearly enough paste, and only ~90% of the IHS got covered as the paste spread out from the dot I placed in the middle. I used an X pattern the second time, and my temps at idle are now around 35C (they used to be in the low-mid 40s), which I'm told is pretty typical. But my temps under load are still just as bad.

What can I do to mitigate this heat issue? How can I even begin overclocking if I'm already thermal throttling at stock settings? Would it make sense to buy an AIO cooler to cut down my temps a bit, perhaps?

And I should mention that if I can solve this heat issue, my ultimate goal is a solidly stable 24/7 overclock. I'm not interested in competition, or performance without stability. I want rock-stable performance that I can use for the next five years without issue (I tend to build new PCs every 5 years).
 
#2 ·
the fact you could cool a quad core is irrelevant, you don't have a quad core now

restrictive case/bad airflow? high ambient temperature? cooler/IHS overly concave/convex?
 
Save
Reply Quote Rep+ Rep+
Discussion starter · #3 ·
Oh shoot, I should have mentioned that, lol.

This system is running the same case, and the same room, as my OC'd 6700K. I have a Phanteks P400A case, and my room does not get any hotter than 75F (24C). It was actually noticably cooler than that when I was testing, as we've had a bit of a cold spell here in Los Angeles over the last few days. So my condo's indoor temp was around 65F (18C).

I'm not really sure how to determine if my IHS is overly concave or convex, though.

I also have the latest BIOS for my motherboard (F31e).
 
#4 ·
You need to check if the MB is overclocking the CPU or overvolting. It should not get that hot at stock.
 
Save
Reply Quote Rep+ Rep+
#6 ·
Stock boost is mostly constrained by temperatures and power. Thermal density of these parts is very high, especially for an eight-core, single CCD part. Reaching maximum temperature before any package power, current, or voltage limits are hit is exactly what I'd expect for Prime95 SmallFFTs without extremely good cooling.

I have a NH-U12A on my 3900X and I had to lap both the CPU IHS and the cooler to get them to sit flush, which knocked about 5-6C off my temps at the same load and fixed OC, as well as allowed for another 25-50MHz out of PBO, but it didn't reduce PBO temperatures (they still sit at about ~85C peak load under PBO) because it was boosting higher.

Holding a straight-edge over the bare CPU and looking for a gap should let you know if the IHS is appreciably concave. Same can be done to the cooler base (which is intentionally convex, but usually won't match an AM4 IHS).

Anything that overrides the default power, thermal, or boost clocks is an OC. Even PBO is technically an overclock. If you want to make sure only stock boost is used, manually disable PBO.
 
Save
Reply Quote Rep+ Rep+
Discussion starter · #7 ·
Did you use the Noctua paste for your lapped 3900X / NH-U12A setup, or a different paste? I'm using the Noctua stuff, but I'm almost out at this point, after doing a number of remounts on my previous system, and now one on my one. I have some Thermal Grizzley Kryonaut that I haven't used yet, though, in case that stuff works better than Noctua's paste.

I'm going to go into my BIOS and double check that everything that might be an OC is disabled, and then do some more thermal testing. I'll post that here once I've got the data.
 
#9 · (Edited)
I have a 5800X under a custom loop and it hit 90c at stock with a 30c ambient with small ftt so this is perfectly normal for this CPU.
How are your temps in Cinebench 20? custom loops in ~20-25c ambient are getting ~65-75c while high end air\AIO are seeing ~75-85c.
Early BIOS seem to be feeding the CPU more v than it needs but mostly AMD was happy to let the CPU suck up as much power as it could to boost as high as possible and let the users cooling and type of load dictate the final clock speed after throttling rather than something else limit the clock speed.

The Cores are such a tiny hot spot it is hard for a HSF to get the heat out of that small corner of the IHS quick enough.
Which makes me wonder if premium paste like Kryonaut may see more than the typical 3c benefit over Noctua NT-H1 on the 5800X
 
Save
Reply Quote Rep+ Rep+
#11 ·
coredumperror said:
Did you use the Noctua paste for your lapped 3900X / NH-U12A setup, or a different paste?
Click to expand...
I've used several different TIMs on this combination. Didn't make much difference before being lapped and made even less of a difference afterwards.

I settled on Cooler Master MasterGel Maker, which is less than a degree better than NT-H1, mostly because I used it last in my comparison and didn't care to take it apart again.
 
Save
Reply Quote Rep+ Rep+
Discussion starter · #13 · (Edited)
OK, here's my results with stock BIOS, XMP enabled, PBO Disabled (it was enabled by default before). I also messed with my fan curves a bit, though I may have messed up a little, as I noticed that my NH-U12A's fans were not quite hitting 2000 RPM, which is their max (they were only getting to 1850). This is all as reported by Ryzen Master.

Idle: 39-40C

R20 Multicore - Run 1:
Temp: 70 instant, quickly 87, soon 89, eventually 90.
Speed: 4,490 Mhz. Dropped ~20 MHz after hitting 90C from thermal throttle.
Score: 5923
Peak Core Voltage: Started at 1.400 as the test began. As temps rose, it dropped lower, before hitting 1.370 after temps reached 90C. It then slowly dropped as low as 1.360 before the test ended.

R20 Multicore - Run 2 (immediately afterward):
Temp Max: 90C instant
Speed Max: 4,490 Mhz. Dropped up to 50 MHz from thermal throttle.
Score: 5875
Peak Core Voltage: Started at 1.370 and dropped as low as 3.60.

I'm gonna go reseat my cooler again, with Kryonaut this time (mostly because I'm out of the Noctua stuff), and just spread it out myself. Will report back on results (and if my first reseat also happened to get bad spread). Will also check concavity/convexity with a razorblade, as suggested by Blameless.

EDIT: Tried again with fans definitely hitting 100% RPM. The only effect was that it took a few more seconds to hit 90C, and the maximum downclock was a little bit higher. Though my R20 scores didn't appreciably change, so that's probably wishful thinking.
 
Discussion starter · #14 ·
=== RESEATED CPU COOLER WITH KRONAUT ===

Idle: 36C

R20 Multicore - Run 1:
Temp: Instant 80, 85 soon, eventually hit 89
Speed: ~4530 start, dropped to 4500 by the end
Score: 5970
Core Voltage: 1.375 fairly steady

R20 Multicore - Run 2:
Temp Max: 86 instant, hit 90 and started throttling for the last half of the test
Speed Max: 4500 start, ~4485 when throttling
Score: 5887
Core Voltage: 1.375


Small FFTs in Prime 95 26.6:
Instantly 90C. Clocks dropped to 4400, vcore ~1.31. Clocks kept dropping slowly.
After 3 minutes, temps still 90, clocks at 4360, vcore 1.30.


Blend in Prime 95 26.6:
Instantly 71C, clocks 4560, vcore 1.385.
After 1 minute: 77C, clocks 4540, vcore 1.378
At 1:30 temps suddenly spiked to 88-89, clocks 4490, vcore 1.373
At 2:20 temps dropped back to 77-78, clocks 4535, vcore 1.377

This temp cycle repeated approximately once every two minutes from then on (~1 min at high temps, ~1 min at lower temps). I'm guessing this is some aspect of the Blend test, though I don't recall ever seeing temp spikes when doing that test on my 6700k.
 
#15 ·
redhat_ownage said:
im gonna bet xmp is adding a stupid amount of voltage somewhere.
Click to expand...
The only voltage apart from DRAM that XMP increases is SOC to 1.1V(and VDDG/VDDP since they're derived from SOC) which shouldn't be an issue since most of us are around 1V or over anyway when using high-speed memory.

VDDG can increase core temps quite a bit. On some 3600s you had to increase it in order to hit 1900MHz FCLK without audio crackle. Perhaps his motherboard is setting VDDG too high automatically. If SOC is 1.1V due to XMP, VDDG will be 1.05V. Quite high. Most chips needed 0.900 or 0.950 at most. Perhaps manually setting VDDG to .950 will help.
 
Save
Reply Quote Rep+ Rep+
#17 ·
coredumperror said:
So I have a 5800X in a Gigabyte Aorus Elite X570 board. I'm new to Ryzen, having previously owned an Intel I7-6700K. While I have a decent amount of experience with overclocking that CPU, I'm realizing that said experience doesn't seem to translate all that well to Ryzen. Or else something is very wrong...

My biggest issue is that even when not overclocking (using bone stock BIOS setting except XMP), my 5800X hits TJmax (90C) and starts thermal throttling after just a few minutes of stress testing of any kind. This is rather disturbing to me, as I have a Noctua NH-U12A on there, which worked fantastically for keeping my rather poorly binned 6700K cool under stress tests during heavy overclocking. With my 6700K OC'd to 4.6Ghz at 1.42v, my temps capped out at 80C during 15 minutes of SmallFFTs in Prime 95 v26.6 (so no AVX).

For comparison to my previous CPU, this 5800X hits 90C after ~2 minutes of SmallFFTs, then sits at 90 as the system throttles down to avoid getting hotter. Hitting TJmax under non-OC loads certainly seems like it shouldn't be happening.

I've already tried re-seating my cooler. My first attempt did turn out to be bad, as I did not use nearly enough paste, and only ~90% of the IHS got covered as the paste spread out from the dot I placed in the middle. I used an X pattern the second time, and my temps at idle are now around 35C (they used to be in the low-mid 40s), which I'm told is pretty typical. But my temps under load are still just as bad.

What can I do to mitigate this heat issue? How can I even begin overclocking if I'm already thermal throttling at stock settings? Would it make sense to buy an AIO cooler to cut down my temps a bit, perhaps?

And I should mention that if I can solve this heat issue, my ultimate goal is a solidly stable 24/7 overclock. I'm not interested in competition, or performance without stability. I want rock-stable performance that I can use for the next five years without issue (I tend to build new PCs every 5 years).
Click to expand...
I see you say only 90% of the IHS got covered. There are many methods how to apply thermal paste with every review recommending different one.
What i do is, I apply a line on one side around 2mm from edge and than using credit card or any straight plastic and pull it to the other end just a 1-2mm short of edge.
With that IHS is covered all but 1-1,5mm in all edges, cooler pressure will squeeze the rest to the edge, with no leaking over. I made crude sketch using MS Paint.
I find this the best method. You can check video below showing how paste spreads.

2465585
2465586



 
#18 ·
HowYesNo said:
I see you say only 90% of the IHS got covered. There are many methods how to apply thermal paste with every review recommending different one.
What i do is, I apply a line on one side around 2mm from edge and than using credit card or any straight plastic and pull it to the other end just a 1-2mm short of edge.
With that IHS is covered all but 1-1,5mm in all edges, cooler pressure will squeeze the rest to the edge, with no leaking over. I made crude sketch using MS Paint.
I find this the best method. You can check video below showing how paste spreads.

View attachment 2465585 View attachment 2465586


Click to expand...
If your cooling setup relies on a proper TIM application, your cooling is definitely not sufficient.
 
Save
Reply Quote Rep+ Rep+
Discussion starter · #19 ·
Yeah, I looked into this after realizing I'd screwed up that paste job the first time. When I applied the Kryonaut just now, I spread it out much like what you suggested. Though I like that strategy of starting with a line on one side and using a credit card. That'd likely be much better than the spread job I got with the little spatula that came with the Kryonaut.

your cooling is definitely not sufficient.
Click to expand...
I mean, it's a Noctua NH-U12A. That's a pretty decent cooler. Not the best air cooler on the market, but reasonably close.

That said, I did just order an Arctic Liquid Freezer II 360 AIO. When it arrives, hopefully I'll at least be able to not thermal throttle with stock settings...

I'm also considering just returning this 5800X to Micro Center and replacing it with a 5600X, since I've heard those run a lot cooler. Assuming they have any stock, of course.
 
#20 ·
coredumperror said:
Yeah, I looked into this after realizing I'd screwed up that paste job the first time. When I applied the Kryonaut just now, I spread it out much like what you suggested. Though I like that strategy of starting with a line on one side and using a credit card. That'd likely be much better than the spread job I got with the little spatula that came with the Kryonaut.



I mean, it's a Noctua NH-U12A. That's a pretty decent cooler. Not the best air cooler on the market, but reasonably close.

That said, I did just order an Arctic Liquid Freezer II 360 AIO. When it arrives, hopefully I'll at least be able to not thermal throttle with stock settings...

I'm also considering just returning this 5800X to Micro Center and replacing it with a 5600X, since I've heard those run a lot cooler. Assuming they have any stock, of course.
Click to expand...
The 5600X might run into the VDDG problem and actually be hotter. My 3600 was hotter than my 3700X and that should be impossible.
 
Save
Reply Quote Rep+ Rep+
1 - 20 of 88 Posts
About this Discussion
87 Replies
31 Participants
Last post:
Section31
Overclock.net
posts
28M
members
562K
Since
2004
The premier forum for overclocking experts and enthusiasts. Discuss hardware optimization, custom builds, benchmarking, cooling solutions, and pushing the boundaries of computing performance. From beginner guides to extreme overclocking, join our technical community to master system tuning.
Full Forum Listing
Explore Our Forums
Intel CPUs Hardware News NVIDIA PC Gaming Water Cooling