Question 5900x overheating problem with Neptwin RGB cooler

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Jan 13, 2022
70
0
30
0
Just reading and in my opinion you used FAR too much paste. Take the cooler off, clean off the cooler, then clean off the top of the cpu. I’d probably use a q tip dipped in alcohol to clean around the socket.

I have a 5900x as well, I am using this case, with 6 fans installed.


For a cooler I’m using the vetroo v5 like this one. But I’ve got 2 fans on it and using a push pull configuration.

https://www.newegg.com/vetroo-v5/p/13C-00F3-00002

Also using Arctic silver mx4 like you. When you apply paste, put a dot about the size of a small pea or a grain of rice in the center, use your finger to spread the paste in a thin layer over the cpu surface.

Mount the cooler like normal and see what you get. With this setup my idle temp is in the 20s usually. Under a full load of cinebench r23 or a 10 pass run of Intel burn test I’m staying right at 70-72 under load. In fact I did almost that well using the same method with a wraith prism stock cooler. So your big thing is airflow.

However when you use as much paste as you used, you are insulating the cpu and slowly cooking it. You just need enough to let it make good contact with the cooler and fill in any gaps. If you put too much on your temps will actually go up and could harm the cpu.

In summary….

  1. Clean your cpu, cooler and possibly socket and remove all paste(at least most)
  2. Apply a lot less paste and mount the cooler again
  3. Check the airflow setup in your case as well the more cool air you get flowing the less heat that will build up and make it easier for your cpu cooler to do it’s job.
Yeah man it was too much thermal paste, but now I know that it was too much :D ,just been trying different methods to see if thermal paste application actually matters that much on the temps themselves, which turns out not much, only too small or too much amount of paste would actually gonna make a big difference, but then I read that for ryzens you should use 5 dot method, since the heatspreader is actually bigger than the one on the intel's, so now I'm trying to do decide which cooler to use, I actually thinking of going with d15s.
 
Jan 13, 2022
70
0
30
0
Most important part is middle, if it doesn0t cover edges is not as much of a problem. If whole surface is not covered by dot in the middle it just means it wasn't enough.
Well temps during gaming lingering from 70C-to-76C~ , funny thing though, cinebench, OCCT, didn't even passed through high 70's ,but once the prime95 started, it started to hit up to the 80's.
 
Well temps during gaming lingering from 70C-to-76C~ , funny thing though, cinebench, OCCT, didn't even passed through high 70's ,but once the prime95 started, it started to hit up to the 80's.
Those temps are fine and prime95 is stability and error checking test. not for temps. Temps under your practical load and usage is real value. The rest comes under "Who cares"
 
Jan 13, 2022
70
0
30
0
Those temps are fine and prime95 is stability and error checking test. not for temps. Temps under your practical load and usage is real value. The rest comes under "Who cares"
So what temps are actually considered safe for 5900x , cause I can't really decide if I should improve my cooling ?
 
Last edited:
Jan 13, 2022
70
0
30
0
Tjmax is 90c so anything under that will not even limit the boost or throttle down. As for practical limit, temps that don't force cooling to make much noise.
[/QUOTE
I see. I read that for lower end cpu's like 5600x and lower, having temps close to their Tjmax, is gonna degrade them faster, is that actually true ?
 
Jan 13, 2022
70
0
30
0
Only if high voltage is the cause oh heat. That combination is worst for electron migration and even in that case it would have to run like that for years while in normal usage that might happen only few % of time.
Oh, I see, I could have swear that I heard the same thing about undervolting, as it draws more power or something liket that, thus degrading the chip faster, or is that incorrect ?
 
Jul 7, 2022
102
67
160
0
Oh, I see, I could have swear that I heard the same thing about undervolting, as it draws more power or something liket that, thus degrading the chip faster, or is that incorrect ?
No undervolting also decreases current as well. It’s a product of the formula of ohms law where Voltage = current * resistance. Since resistance doesn’t change when you undervolt, current must decrease as well. Since power = voltage * current, undervolting results in lower power usage.
 
Yeah man it was too much thermal paste, but now I know that it was too much :D ,just been trying different methods to see if thermal paste application actually matters that much on the temps themselves, which turns out not much, only too small or too much amount of paste would actually gonna make a big difference, but then I read that for ryzens you should use 5 dot method, since the heatspreader is actually bigger than the one on the intel's, so now I'm trying to do decide which cooler to use, I actually thinking of going with d15s.
Everyone has their own method. I have used the spread method since at least socket am2, if not 939 and before. But the spread method has always worked well for me. Seems to give just the right amount of paste. On cooler, a lot of folks like water. Personally I’m using the black version of this.

https://vetroo.com/products/v5?variant=43428661887210

Keep in mind I’m using 2 fans in a push pull configuration, and 6 case fans as well. But my temps on the cpu are topping out at about 71 or 72 under load.
 
So what temps are actually considered safe for 5900x , cause I can't really decide if I should improve my cooling ?
AMD has told us it's normal and expected for 5000 series CPU's to hit 90C in use. But that doesn't mean more robust cooling isn't worth the effort. Better cooling means it will hit higher clocks in boosts and hold higher mid-range clocks longer even if still hitting 90C. The way the boost algorithm seems to work it is like overclocking your CPU.

The best article I've seen on thermal paste application:

The effects are the same so I am confident it applies to CPU's as well.
 
Last edited:

Karadjgne

Titan
Ambassador
Bah. Misconceptions.

A 5900x rarely ever hits 90°C, it actually rarely ever hits 80°C. Only a single core does, which is not representative of the entire cpu. It takes a full core load like blender or Prime95 to do so. Otherwise, when you open Opera, the cpu is not hitting 85°C, only 1-2 cores are hitting that temp. Very big difference between 2 cores at 85° and 10 others at 40° to the power and heat output of all 12 cores at 85°C.

Software reporting only shows the hottest core at any given time period. And that's totally dependent on exactly when the report is read. If a read is every 256ms, that's 4 reads per second. Spikes last @ 10ms ±. So if the workload produces 40 70 55 55 in that second, and the reported read is on the third read, you'd see a temp of 55. Exact same load of 40 70 55 55 reported on the second read would show a temp of 70 for that core. Reports are generally updated every 3 seconds. Any faster than that and your eyes would see nothing but the digital '88' blur as the numbers would blend together.

So core temps can be anywhere from 40ish to 90ish at any given time but you'll only see exactly what temp is read and reported every 3 seconds. To see actual core temp variation for any time period requires a line graph, which is semi-useless since cores are not static, windows will assign different tasks to different cores, at different times. So the only thing shown is the plot of whichever core happens to be hottest at every read, leading you to believe the cpu is running 85° in total when running Opera, but in reality it's the full mix of all the cores hot points only and not the actual temp of the entire cpu.

It's in this where Ryzen Master temps are better suited as they average the temps per core, every 3 second time period. So you'd not see lows or highs with 40 70 55 55, just the average of 55 for the core.

Ryzen aren't Intel. With an intel at idle, all the cores are chopped to low power state, but all remain active. So an 8 core Intel is going to divide the idle workload amongst all 8 cores, so no one core is overloaded and the hottest core is only running at @ 32° at any given point. The other cores sitting at 30-31°C.

Ryzens don't work that way. At idle all the cores are effectively parked except for one. Naturally that'll be the hottest core, as it carries the entire idle workload. That core might see spikes upto 70°ish, but being the hot core, that's what gets reported. Doesn't mean that the cpu is running 40-70° at idle, the cpu is closer to 30° at idle, just the single in-use core is hitting those temps. A Ryzen also rotates the cores, so no one core is overly used. Core0 is first, then every 1-3 seconds it's rotated down in sequence, 1-2-3 etc. If you open HWInfo64 to core temps section and let the pc go to idle, the rotation and temps become very obvious.

But are still governed by exactly when each core is read compared to reported.

Temps are reflective of load and power used. But there's not a cooler on the planet effective enough to instantly affect those temps. You've got workloads happening in nanosecond time periods, and that temp being read directly from the core at that instant. That temp must then travel out of the silicon, through the TIM, through the IHS, through the paste, into the coldplate to be absorbed. That takes time, a lot longer process than the change in workloads. So workloads are going to show spikes in temp, it's unavoidable, it's the difference between when the load actually hits and when it's absorbed by the coldplate. The cooler can't affect that temp any faster.

Some coolers are more effective and more efficient than others, so will mitigate the extent of range a workload is allowed to hit. The cheapo stock 4pin Intels are not efficient, so allow a larger range on the workload, whereas Noctua is very efficient, so that range is considerably more limited. The Intel might allow the load to produce 90° startup spike, the Noctua would limit that to 70° etc.

Numbers for use purposes, not exact definition.

What it boils down to is temps should be taken with a grain of salt, especially idle temps. The exact number is not nearly as important as the average, unless the exact number is well into dangerous territory. If your load temps are generally in the 80's for a heavy, consistent load, you are good. The occasional spike to 90's, no worries. Constant loads in the 90's, get a more effective cooling solution. Ryzens see the highest boosts, on more cores, for longer time periods when the cores in use are between @ 60-80°C
 
Last edited:

ASK THE COMMUNITY

TRENDING THREADS