Overclock.net banner

1 - 20 of 22 Posts

·
Registered
Joined
·
88 Posts
Discussion Starter #1 (Edited)
Hello,

Recently I built a new system around the Ryzen 2700X. I have weird problems, but things seem stable. Coming from the 960T, it's been rocky, but fun, but also mysterious.
Specs are as follows:
- Ryzen 2700X @ stock cooled by Noctua UHS-12
- ASUS CHVII Hero Wifi (X470 board) - stuff set to auto besides memory
- 16GB Corsair 3000Mhz @ 15-17-17-35-53 (Slightly loosened from DOCP Profile)
- MSI Gaming X GTX 1060 6GB
- WD Blue 500 GB M.2 SSD + Old 2TB sata drive
- Seasonic FOCUS PLUS 750W PSU
- Stuffed into a Fractal Design R5 with 2 intake fans pulling?, 1 exhaust out back, 2 on top exhausting, and 1 on bottom intaking

Installed Windows fine, had a tough time getting it to POST with the memory from DOCP. It chooses when it wants to work initially. But once it's set to defaults, booted once, then changed to 3000 it works.

It can run prime95 blend just fine for an hour, gets to maybe 65C-67C with spikes up to 80C or so at ambient temps of 24C, haven't tested any longer than that. On small ffts, it gets hot. Uncomfortably hot.. like 85C hot. I know the max is like 95C before throttling, but it still seems hot, especially for a nocuta cooler. Same temps, maybe a little hotter on intel ibt. But the kicker is AIDA64. About 30 seconds after I run AIDA64 on default settings for stability tests, it fails citing a system error. Has anyone else experienced this?

I'm wondering what the issue could be, but given that only 1 program is truly failing, I wonder if it's because of the newness. But the temps.. Are the temps supposed to be that hot under small fft conditions? It truly seems odd that it spikes so easy, but then again it is an 8 core processor, when I was used to 4. I swear with the previous generation, I've read about temps in the maybe 65s, low 70s when stress testing, even with AIDA64...Am I crazy or is this normal and I need to adjust to this new platform? The stock cooler, got up to 87C, which is crazy compared to the noctua cooler because there doesn't seem to be much difference. I've reseated this thing twice and it all seems the same with temps. Even ran RAM at stock (no DOCP) to see if AIDA64 would proceed.. nope.
 

·
Registered
Joined
·
1,115 Posts
take the side panel off and see if it cools better (significantly so) my 1700 overclocked on a gammaxx 400 would run about 65C. However, how hot does it get when you use the computer normally? also try turning the top fans off and see what happens to the temp.
 

·
Registered
Joined
·
88 Posts
Discussion Starter #3
take the side panel off and see if it cools better (significantly so) my 1700 overclocked on a gammaxx 400 would run about 65C. However, how hot does it get when you use the computer normally? also try turning the top fans off and see what happens to the temp.
With the side panel off, small ffts seems to cap off at 77C. I guess not bad, but I was expecting better stock. Dunno if this indicates a contact issue with the cpu, but the heatsink for the cpu is barely warm as in I can touch it and not have a problem being burned..Does this indicate a contact issue or is it just effective at cooling? I could try turning the cooler 90 degrees and have the fan pull air through. Turning off top fans yields similar temps as I was running without those initially and got the fans in hopes of reducing temps.


Using the computer normally is fine, the temps are more spikey, though. Opening a web browser could make the cpu die temp spike to 50C sporadically sometimes a little higher. Doesn't get hot while gaming for obvious reasons.

Can the AIDA64 thing be explained at all, though? I find it strange.
 

·
Registered
Joined
·
245 Posts
Was the thermal paste applied from the factory to the noctua? Also back in the day I believe the recommended application of thermal grease was the card method. You needed to cover the entire heatspreader with paste. Intel application was different as it was the rice or pea method where you just place the grease in the center and it would be enough.

Sent from my SM-N950U using Tapatalk
 

·
Registered
Joined
·
88 Posts
Discussion Starter #5
Was the thermal paste applied from the factory to the noctua? Also back in the day I believe the recommended application of thermal grease was the card method. You needed to cover the entire heatspreader with paste. Intel application was different as it was the rice or pea method where you just place the grease in the center and it would be enough.

Sent from my SM-N950U using Tapatalk
No, it came with thermal paste and I removed the plastic. I first seated with the pea method using noctua paste and have successfully used that with the last noctua cooler I had with the 960T and it worked fine. Temps were still high. Taking off the cooler showed that it smooshed down good. Then I reseated with the line method using AS5, putting a thin line across the IHS so when it is cranked down it smooshes on both sides. I could try the card method, but I'm unsure what it is before, I don't think my technique is that bad. Is it even possible that it isn't seating right?
 

·
Registered
Joined
·
245 Posts
No, it came with thermal paste and I removed the plastic. I first seated with the pea method using noctua paste and have successfully used that with the last noctua cooler I had with the 960T and it worked fine. Temps were still high. Taking off the cooler showed that it smooshed down good. Then I reseated with the line method using AS5, putting a thin line across the IHS so when it is cranked down it smooshes on both sides. I could try the card method, but I'm unsure what it is before, I don't think my technique is that bad. Is it even possible that it isn't seating right?
Possible. When you take off the HS does it appear even? I know its probably going to be impossible to tell. Also how does that HS connect? I'm assuming special clamps designed for am4. I put 2 computers together for work and I had to remove the mounts on the mobo and use the backplate to mount the wraith coolers(?) On the ryzen 2400G.

You could also just smoosh it all around with your finger. Either way I'd say go for 100% coverage.


Sent from my SM-N950U using Tapatalk
 

·
Registered
Joined
·
245 Posts
Also I would like to add that on the 2400G I play around with currently at work, there is a spike in temps. From say 28C to 36C at random. Its definitely loading something causing the temperature spikes but it goes down just as quick. Could be a bug but idk.

Sent from my SM-N950U using Tapatalk
 

·
Registered
Joined
·
88 Posts
Discussion Starter #8 (Edited)
Possible. When you take off the HS does it appear even? I know its probably going to be impossible to tell. Also how does that HS connect? I'm assuming special clamps designed for am4. I put 2 computers together for work and I had to remove the mounts on the mobo and use the backplate to mount the wraith coolers(?) On the ryzen 2400G.

You could also just smoosh it all around with your finger. Either way I'd say go for 100% coverage.


Sent from my SM-N950U using Tapatalk
Eh, kinda appears so, kinda not. The Noctua uses their special brackets so I remove mine. Works well I think.
I've attached pictures of the smooshes after removing.

Also, I've remounted the cooler with the brackets so that it is rotated 90 degrees. Under small ffts, it is 80C now. Dunno how...
 

Attachments

·
Registered
Joined
·
245 Posts
Yea that paste needs to cover the whole spreader but idk how much that will help. What's the fan speed set to on the cpu cooler?
Is that paste hard or like gooey? If hard then it maybe old. I'm not sure if it has a self life or whatnot but it possible.
Are you getting it tightened down enough? I dont want you to break it but it does need to be relatively tight.

The purpose of the paste is not to add a layer between the HS and the spreader but to fill in imperfections and ridges and valleys to maximize the connection from processor to HS. In a perfect world with perfect surfaces you wouldn't need paste.

Also what is the ambient temp in the room?
Can you screenshot some idle temps and then some gaming temps?
What is the idle GPU temp?

I should be getting my 2700x today along with my new mobo and ram. I will be using the stock cooler for now so I will report back here what I find and see if we can assist further.

Sent from my SM-N950U using Tapatalk
 

·
Registered
Joined
·
245 Posts
Prime95 is a stress test but I shy away from it. I have heard of it killing intel processors back a few years ago. Not sure why but I think the series of processor was the 59xx series? I think it was x99 chipset.

Sent from my SM-N950U using Tapatalk
 

·
Registered
Joined
·
3,326 Posts
the thermal paste isn't covering the whole IHS
It doesn't need to cover the whole heat spreader. The die is much smaller and in the middle of the heat spreader.
 

·
Totally Tubular
Joined
·
1,508 Posts
It doesn't need to cover the whole heat spreader. The die is much smaller and in the middle of the heat spreader.
The die is much smaller, true, but the heat spreader spreads heat over its entire surface. There is currently a bunch of that surface that is not having its heat sunk. It likely isn't a huge deal, but it might make a few degrees difference if the spread was better.
 

·
Registered
Joined
·
3,326 Posts
The die is much smaller, true, but the heat spreader spreads heat over its entire surface. There is currently a bunch of that surface that is not having its heat sunk. It likely isn't a huge deal, but it might make a few degrees difference if the spread was better.
Covering the rest of that heat spreader would do nothing but make a mess as it squirts out the side into the socket area.
 

·
Registered
Joined
·
5,374 Posts
Firstly you aren't the only one having a hard time with Aida64 and that's ok because it's not a very good stress test anyway, GTA5 is harder to run as far as stability is concerned.
http://www.overclock.net/forum/13-amd-general/1687673-aida64-failed-but-other-test-pass.html
Temps are the real gremlin IMO.
Secondly that looks like a very restrictive case but I could be wrong there. The fan layout seems a little odd though.
Try either pointing the cooler back and only allow the front intakes and rear exhaust to run or try pointing the cooler up with front and bottom intake and top exhaust only. 180° front/bottom intake with 180° back/top exhaust seems to me like it could be circulating air in a loop, that may not be the case but to test all you really need to do is use fan software or change them in the BIOS. Since rotating the cooler made a difference I'm pretty sure this is part of your problem. I'd also try some testing with the front open.
It might not be an airflow issue though, use HWinfo to monitor your vcore while Prime95 is running. I'm not sure what AVX does to Ryzen CPUs but you can find out by either disabling AVX in P95 or just use v26.6, monitor your vcore difference between versions. If Small FFT results in 1.2v and Small FFT with AVX results in 1.25+ I'd say you have a similar situation to Intel with AVX forcing gobs of voltage in adaptive mode. I don't have a Ryzen CPU yet so I haven't done any reading on how it responds to P95's AVX.
Thirdly I've never noticed any difference in any way I've ever applied paste, I've smeared it, dotted it and have left the corners naked and never noticed much difference. If there was one I couldn't tell and I doubt it's enough to put you in the 80s.

I didn't take this as an overclocked system but if your vcore is getting way too high you might try setting the proper voltage and clocks manually, then just run the same version and settings on P95 to see if it's still happening. It's not uncommon for your CPU to be running a much higher voltage than needed for stock performance so it's definitely something you should check on.
Good luck, hopefully this is an easy fix.
EDIT: If you aren't overclocking (the boost already does a great job) there isn't much reason to care what your Small FFT temps are as long as your games or other applications are running cool, Small FFT (with or without AVX) is unnecessarily intense and hot for most systems. Personally I don't see much use in stress testing a new build until you're overclocking something or you have crashes in normal workloads so you can narrow down the culprit. Or maybe if you are trying to dial in your case fan layout and speeds but blend gets plenty warm enough to test this.
I can understand running Blend after you get your CPU and ram running at rated speeds but Small FFT fits in the cache alone so it's going to be hotter and it'll only really stress the core.
 

·
Registered
Joined
·
3,326 Posts
The die is much smaller, true, but the heat spreader spreads heat over its entire surface. There is currently a bunch of that surface that is not having its heat sunk. It likely isn't a huge deal, but it might make a few degrees difference if the spread was better.
The heat spreader really only has one function. To protect the die from damage incurred when mounting a heatsink. AMD and Intel started putting "heat spreaders" on so people would stop chipping the sides of the die itself. Before heat spreaders you just applied a tiny amount TIM to the die itself . I am sure you are aware of this. 95% of the heat spreader doesn't not even contact the cpu.
 

·
Totally Tubular
Joined
·
1,508 Posts
The heat spreader really only has one function. To protect the die from damage incurred when mounting a heatsink. AMD and Intel started putting "heat spreaders" on so people would stop chipping the sides of the die itself. Before heat spreaders you just applied a tiny amount TIM to the die itself . I am sure you are aware of this. 95% of the heat spreader doesn't not even contact the cpu.
Yes, but it still spreads the heat around -- it's even in the name. If you don't contact the entire surface, more heat must be transferred through a smaller area. That results in less efficient heat transfer, and higher temps.

You want maximum surface area contact to the heat spreader, just like you want maximum surface area in an air cooler, and in a radiator's fins.
 

·
curmudgeon
Joined
·
5,930 Posts
Spreading a thin layer of TIM all across the contact surface is most effective at showing how flat the cpu surface is. All of my FX cpus are low in the center and high on the edges, resulting in a puddle of TIM in the center and metal to metal contact on the corners.



What I see is that the TIM seems to be thickest at the top so the cooler might not be making full square contact with the cpu.
 

·
Registered
Joined
·
88 Posts
Discussion Starter #19
Yea that paste needs to cover the whole spreader but idk how much that will help. What's the fan speed set to on the cpu cooler?
Is that paste hard or like gooey? If hard then it maybe old. I'm not sure if it has a self life or whatnot but it possible.
Are you getting it tightened down enough? I dont want you to break it but it does need to be relatively tight.

The purpose of the paste is not to add a layer between the HS and the spreader but to fill in imperfections and ridges and valleys to maximize the connection from processor to HS. In a perfect world with perfect surfaces you wouldn't need paste.

Also what is the ambient temp in the room?
Can you screenshot some idle temps and then some gaming temps?
What is the idle GPU temp?

I should be getting my 2700x today along with my new mobo and ram. I will be using the stock cooler for now so I will report back here what I find and see if we can assist further.

Sent from my SM-N950U using Tapatalk
The paste is fine, for good measure I even used the paste that is newer than the AS5 I had laying around. This time I spread a thin layer on the heatsink using a credit card, so it definitely is covering it all. This resulted in a 1-2C difference vs. putting some in the center and tightening down the cooler. The noctua cooler mounting in particular tells you to tighten until it stops. It requires almost 0 torque, so I am reliably getting it tight each time, unless the heatsink surface is warped (or maybe mobo).

Ambient temp is 23C.

Idles are the low and highs are when I was gaming, just look in the min and max portion of HWINFO screenshots. As you can see, Tdie gets up to 52.5C, so not even hot. It just can fluctuate, jumping from 30C to like 45C in seemingly seconds.. GPU idles low 30s and gaming 60s in Celsius. Attached in post


Firstly you aren't the only one having a hard time with Aida64 and that's ok because it's not a very good stress test anyway, GTA5 is harder to run as far as stability is concerned.
http://www.overclock.net/forum/13-amd-general/1687673-aida64-failed-but-other-test-pass.html
Temps are the real gremlin IMO.
Secondly that looks like a very restrictive case but I could be wrong there. The fan layout seems a little odd though.
Try either pointing the cooler back and only allow the front intakes and rear exhaust to run or try pointing the cooler up with front and bottom intake and top exhaust only. 180° front/bottom intake with 180° back/top exhaust seems to me like it could be circulating air in a loop, that may not be the case but to test all you really need to do is use fan software or change them in the BIOS. Since rotating the cooler made a difference I'm pretty sure this is part of your problem. I'd also try some testing with the front open.
It might not be an airflow issue though, use HWinfo to monitor your vcore while Prime95 is running. I'm not sure what AVX does to Ryzen CPUs but you can find out by either disabling AVX in P95 or just use v26.6, monitor your vcore difference between versions. If Small FFT results in 1.2v and Small FFT with AVX results in 1.25+ I'd say you have a similar situation to Intel with AVX forcing gobs of voltage in adaptive mode. I don't have a Ryzen CPU yet so I haven't done any reading on how it responds to P95's AVX.
Thirdly I've never noticed any difference in any way I've ever applied paste, I've smeared it, dotted it and have left the corners naked and never noticed much difference. If there was one I couldn't tell and I doubt it's enough to put you in the 80s.

I didn't take this as an overclocked system but if your vcore is getting way too high you might try setting the proper voltage and clocks manually, then just run the same version and settings on P95 to see if it's still happening. It's not uncommon for your CPU to be running a much higher voltage than needed for stock performance so it's definitely something you should check on.
Good luck, hopefully this is an easy fix.
EDIT: If you aren't overclocking (the boost already does a great job) there isn't much reason to care what your Small FFT temps are as long as your games or other applications are running cool, Small FFT (with or without AVX) is unnecessarily intense and hot for most systems. Personally I don't see much use in stress testing a new build until you're overclocking something or you have crashes in normal workloads so you can narrow down the culprit. Or maybe if you are trying to dial in your case fan layout and speeds but blend gets plenty warm enough to test this.
I can understand running Blend after you get your CPU and ram running at rated speeds but Small FFT fits in the cache alone so it's going to be hotter and it'll only really stress the core.
Same, I've never noticed a difference in application method. It even didn't cover the corners of the IHS in the CPU and temps were as low as can be with my previous build, however the cooler was much bigger. I'm not entirely sure what the stock voltage for Ryzen 2700X should be. CPU can range the voltage anywhere from 1.25v to 1.35v when all cores are in use, and can go higher if overclocking some cores over others. It's not an overclocked system, I don't intend to as Ryzen seems to be doing fine. But yeah, I haven't any odd behaviors at all or crashes, so I think I'm just not used to it failing AIDA64 out of the box. Prime95 doesn't fail, so I think I'm in the clear. Thanks for all the advice, just wasn't aware that these new CPUs can run that hot... only 10C or so from throttle. Since gaming only gets it up to 50ishC, I'd say I'm in the clear.
 

Attachments

·
Registered
Joined
·
224 Posts
I agree that the tim does not need to be spread with a finger, a credit card, a spatula, or anything like that: the pea method has always served my noctua coolers well on FX and Ryzen mounts alike. The heat is going to be conducted where the paste lies.

If you wanna see the Noctua owl standing in for the Virgin Mary, I invite you to play "spot the cores!" below:

 
1 - 20 of 22 Posts
Top