Stability tests for 10th gen 10900k/10850k?

edo101

Limp Gawd
Joined
Jul 16, 2018
Messages
480
I am aware of Cinebench and Realbench but are there any other tests to use for stability?

I have prime95 but I have been told trying to run AVX2 with Prime95 with 10 series is a waste of heat power and not indicative since most coolers can't handle it. I have a D15 black

What other tests can I use?
 
I am aware of Cinebench and Realbench but are there any other tests to use for stability?

I have prime95 but I have been told trying to run AVX2 with Prime95 with 10 series is a waste of heat power and not indicative since most coolers can't handle it. I have a D15 black

What other tests can I use?
why do you keep making new threads?!
 
I'd download the Intel Extreme Tuning Utility. It has a stability test baked in you could run. You can set the time to run as long as you want.

OCCT and IntelBurnTest are two others.
 
Realbench and Cinebench R20 / R23 are perfectly fine. One thing I found that's far more useful in this regard is Handbrake encoding tests. I've seen systems that Prime95 and other tests pass without issue crash using Handbrake.
 
why do you keep making new threads?!
Because the threads are all diffferent? And are centered around different topics?

Should I also use this thread to ask about my AIO, my air cooler or my upcoming memory OC? How would a bystander browing the forums know that? pendragon1
Realbench and Cinebench R20 / R23 are perfectly fine. One thing I found that's far more useful in this regard is Handbrake encoding tests. I've seen systems that Prime95 and other tests pass without issue crash using Handbrake.
Yes I just downloaded handbrake. Are there particular settings I should use to ensure I'm getting a good stability test? Does handbrake use AVX by default? I have no offsets Dan_D
 
Because the threads are all diffferent? And are centered around different topics?

Should I also use this thread to ask about my AIO, my air cooler or my upcoming memory OC? How would a bystander browing the forums know that? pendragon1

Yes I just downloaded handbrake. Are there particular settings I should use to ensure I'm getting a good stability test? Does handbrake use AVX by default? I have no offsets Dan_D

You can set the program to use specific features to an extent. Just tell it not to use NVENC or anything like that. Let it rely on the CPU instead of using GPU acceleration. Unless its QuickSync, that you can use as its part of the CPU.
 
You can set the program to use specific features to an extent. Just tell it not to use NVENC or anything like that. Let it rely on the CPU instead of using GPU acceleration. Unless its QuickSync, that you can use as its part of the CPU.
Right that's what I mean. What settings do you use to ensure max CPU usage. I've never used Handbrake before
 
Right that's what I mean. What settings do you use to ensure max CPU usage. I've never used Handbrake before

That's pretty much it. Just do a 4K encode of something. The longer the video the longer the test will run. I leave most settings on default for test purposes.
 
Because the threads are all diffferent? And are centered around different topics?
you are dealing with the same system, just carry on a single thread so people that are trying to helps dont have to jump all over the place.
you can you OCCT to do a whole system test.
 
you are dealing with the same system, just carry on a single thread so people that are trying to helps dont have to jump all over the place.
you can you OCCT to do a whole system test.
Hm I definitely get that. I just worry it gets too congested and anybody that might have input that isn't intrested in my AIO issue but could contribute to my stability test would miss it if you get what I mean
 
That's pretty much it. Just do a 4K encode of something. The longer the video the longer the test will run. I leave most settings on default for test purposes.
Dan_D Would using Quicksync help stress the CPU more?
 
handbrake.PNG

I hope this means it's using AVX
 
Realbench and Cinebench R20 / R23 are perfectly fine. One thing I found that's far more useful in this regard is Handbrake encoding tests. I've seen systems that Prime95 and other tests pass without issue crash using Handbrake.
Dan_D So Handbrake uses AVX for for H265 encoding?
 
I couldn't get Prime95 small fft to run without setting off CoreTemp overheat protection on my 10850k at stock unless I open the window. It's 9 F/ -12 C here... I'm using a three yar old Corsair H110i dual 140mm AIO and Thermal Grizzly Kryonaut paste applied two days ago. This CPU is pulling 260 watts according to CoreTemp and I haven't even attempted overclocking it yet. Failure to run Prime95 small fft is mildly concerning. Like having a car that can't go down a certain road. I may never travel that road but I'll always know I can't... I have been able to run a couple of the other Prime95 tests for 30 minutes. I'll have to try Cinebench tonight.

1611675683338.png
 
I couldn't get Prime95 small fft to run without setting off CoreTemp overheat protection on my 10850k at stock unless I open the window. It's 9 F/ -12 C here... I'm using a three yar old Corsair H110i dual 140mm AIO and Thermal Grizzly Kryonaut paste applied two days ago. This CPU is pulling 260 watts according to CoreTemp and I haven't even attempted overclocking it yet. Failure to run Prime95 small fft is mildly concerning. Like having a car that can't go down a certain road. I may never travel that road but I'll always know I can't... I have been able to run a couple of the other Prime95 tests for 30 minutes. I'll have to try Cinebench tonight.

View attachment 322869
make sure your pump and fans are at full tilt
 
260w heheh. Impressive. I have to beat the piss out of mine to hit 120w lol. But it does it under 85 so..

I would say ramp your pump up for sure, and your fans too because 260w.. Just breathe on it a little and your in threadripper territory :D

I used to think my old setup hated trees and polar bears.. yours does too :D

Still though.. 4.8GHz on 10 cores is pretty sick. I had to kick my 3770K right in the cores to do 4.8.

Edit:

Had to kick my X5690 in the cores to for 4800 too.
 
260w heheh. Impressive. I have to beat the piss out of mine to hit 120w lol. But it does it under 85 so..

I would say ramp your pump up for sure, and your fans too because 260w.. Just breathe on it a little and your in threadripper territory :D

I used to think my old setup hated trees and polar bears.. yours does too :D

Still though.. 4.8GHz on 10 cores is pretty sick. I had to kick my 3770K right in the cores to do 4.8.

Edit:

Had to kick my X5690 in the cores to for 4800 too.
The only tweaking I've done is setting XMP on the memory. Otherwise this thing is running bone stock. With Prime95 small fft and my water at 48c, still can't get the heat out of the CPU fast enough. My pump is set to Quiet and I'm not willing to go Extreme due to the sound long term, only two options. I'll try extreme tonight and see what happens. I've never even thought about neutering a CPU but I might have to figure out how to lower the voltage from stock. The aggressiveness of the Intel boost or thermal velocity, or whatever they're doing is pretty awesome. The heat and the quickness of the heat is hardcore. You start Prime95 and every core shoots up to nearly 95c. I've lived in the land of quad cores for so long this thing is like a carnival ride and I wasn't prepared.
 
Z490 motherboards like to overvolt CPU's. This is for MCE (Multicore enhancement) or the equivalent. Essentially, it's almost an automatic overclock on the board that runs these CPU's super aggressively. Nearly every motherboard maker does this. My Maximus XII Extreme tried to run my 10900K @ 1.51v. I simply capped my voltage at 1.25v and that resolved the issue. Now, with MCE it boosts to 5.3GHz on multiple cores. Before, it was hitting 94c at 4.8GHz or so on custom water cooling.
 
Z490 motherboards like to overvolt CPU's. This is for MCE (Multicore enhancement) or the equivalent. Essentially, it's almost an automatic overclock on the board that runs these CPU's super aggressively. Nearly every motherboard maker does this. My Maximus XII Extreme tried to run my 10900K @ 1.51v. I simply capped my voltage at 1.25v and that resolved the issue. Now, with MCE it boosts to 5.3GHz on multiple cores. Before, it was hitting 94c at 4.8GHz or so on custom water cooling.
Dan,
Is that just setting the voltage at a single value instead of auto? Then the CPU/MB will take care of itself to get what it can get within that voltage envelope? I suppose the direct question is, do I need to adjust other dynamic/automatic boosting settings to fall in line with that voltage setting?
edit: My assumption is that if I set the voltage to the Intel stock voltage, I should expect advertised boost rates at a minimum.
 
Dan,
Is that just setting the voltage at a single value instead of auto? Then the CPU/MB will take care of itself to get what it can get within that voltage envelope? I suppose the direct question is, do I need to adjust other dynamic/automatic boosting settings to fall in line with that voltage setting?
edit: My assumption is that if I set the voltage to the Intel stock voltage, I should expect advertised boost rates at a minimum.

1611698003058.png


All I did on mine was set the CPU Core/Cache Voltage to "Adaptive" and another menu will pop up allowing you to set limits. It will have a value for +/- signs (+ by default) and then you can set the adaptive voltage. The voltage should stay around this value. Your actual voltages could still go above that. Voltage caps can be set elsewhere in the BIOS but I wouldn't use them. In other words adaptive voltage isn't a voltage cap exactly. Basically, overcurrent values and other settings will allow the voltage past what you've set for adaptive voltage, but this setting will reign in what your motherboard is trying to do. Before I made the change my system was overheating and not boosting correctly. CPU-Z showed a voltage of 1.51v a lot of the time. This was also even with MCE set to auto or even disabled. The voltages used by default on Z490 boards are insane. Gamer's Nexus has a whole video about it.

The short version is, Intel has allowed motherboard makers to do whatever they want since AMD started kicking their asses. Basically, Intel's allowing motherboard vendors to one up each other and try and make themselves look better than the other guy in the benchmarks. Intel used to clamp down on this but doesn't seem to care anymore. No one follows Intel's voltage guidelines anymore. MCE, is a way for motherboard manufacturers to boost performance but its not about making things better for you, it's about winning benchmarks. I actually have MCE set to on to allow the CPU to boost more often and to higher levels.

1611699234072.png


As you can see, my CPU can still pull upwards of 1.50v+ but the average is lower than it was on "auto". You can also see the boost values. As you can see, running Cinebench R23 it will hit 5.3GHz. Although, that's not the average clock speed. You can also see the average voltages. Basically, much like AMD's platform, adaptive voltage doesn't mean your CPU can't use more than that and it will need upwards of 1.5v+ to boost to its maximum values.

Before I made the change to adaptive voltage, my system would hit 97c on some cores. It would lock up when stressed, etc. This is on full custom water cooling. On adaptive, my temps stay under 70c for the most part. It is also worth noting that all the overclocking guides I've seen for ASUS boards at least, show setting the voltage to adaptive mode the way I have here. In my own testing, it's been necessary to set this value. However, if you set the voltage caps (ASUS' BIOS offers this feature) then you won't ever see the boost clocks behave this way.

1611699419829.png


Hope this helps.
 
that 48c would probably be why, thats too high. your water temps should be 10-15c over room temp under load.
lol, I think he means his CPU is at 48C when using water :)
Most people dont exclusively state the water temp or have a way to measure it.
I've never heard of water hitting that high!
 
Last edited:
lol, I think he means his CPU is at 48C when using water :)
Most people dont exclusively state the water temp or have a way to measure it.
I've never heard of water hitting that high!
maybe but the h110i w/ icue/link shows water temp as "pump temp" so he'd know what his water temp is. he is running the pump at low speed and im assuming the fans too.
edit: it also looks like he has all core 4.8GHz w/ 1.31v, so he might need to tweak that a bit to bring temps down more.
 
  • Like
Reactions: Nenu
like this
The only tweaking I've done is setting XMP on the memory. Otherwise this thing is running bone stock. With Prime95 small fft and my water at 48c, still can't get the heat out of the CPU fast enough. My pump is set to Quiet and I'm not willing to go Extreme due to the sound long term, only two options. I'll try extreme tonight and see what happens. I've never even thought about neutering a CPU but I might have to figure out how to lower the voltage from stock. The aggressiveness of the Intel boost or thermal velocity, or whatever they're doing is pretty awesome. The heat and the quickness of the heat is hardcore. You start Prime95 and every core shoots up to nearly 95c. I've lived in the land of quad cores for so long this thing is like a carnival ride and I wasn't prepared.
if you get your aio set correctly and tweak the voltage on your cpu your temps will come down...
 
I couldn't get Prime95 small fft to run without setting off CoreTemp overheat protection on my 10850k at stock unless I open the window. It's 9 F/ -12 C here... I'm using a three yar old Corsair H110i dual 140mm AIO and Thermal Grizzly Kryonaut paste applied two days ago. This CPU is pulling 260 watts according to CoreTemp and I haven't even attempted overclocking it yet. Failure to run Prime95 small fft is mildly concerning. Like having a car that can't go down a certain road. I may never travel that road but I'll always know I can't... I have been able to run a couple of the other Prime95 tests for 30 minutes. I'll have to try Cinebench tonight.

View attachment 322869

That cooler probably can't handle prime95 with a 10 core CPU.
My 10900k @ 5 ghz gets up to 86C running FMA3 small FFT (15k fixed) prime95 with an Arctic Liquid Freezer II 360 with 3x Noctua 120mm industrial fans @ 3000 RPM.
 
I was able to get Small FFT to run for a bit with the Vcore set to adaptive instead of auto. Also the H110 set to extreme on pump and fans. It pulled 300 watts to get it done. The temps were creeping up but they didn't slam into 95c anymore. Not sure if was the Auto to Adaptive or the Extreme profile or both.
The machine will rarely see all cores maxed but that is allot. Gigabyte z490 Aorus elite ac motherboard with F20a bios. Going to play with it some more to see what I can get with the Quiet setting. This thing is annoyingly loud sitting here idling in extreme mode.

1611797140588.png
 

Attachments

  • Capture.JPG
    Capture.JPG
    217.8 KB · Views: 0
I was able to get Small FFT to run for a bit with the Vcore set to adaptive instead of auto. Also the H110 set to extreme on pump and fans. It pulled 300 watts to get it done. The temps were creeping up but they didn't slam into 95c anymore. Not sure if was the Auto to Adaptive or the Extreme profile or both.
The machine will rarely see all cores maxed but that is allot. Gigabyte z490 Aorus elite ac motherboard with F20a bios. Going to play with it some more to see what I can get with the Quiet setting. This thing is annoyingly loud sitting here idling in extreme mode.

View attachment 323491
does the new version of the h110i let you control pump speed based on cpu temp or just low, med, high? if so, set it to ramp to max at like 70c
 
does the new version of the h110i let you control pump speed based on cpu temp or just low, med, high? if so, set it to ramp to max at like 70c
The fans can be set low, balanced, extreme, or custom. The pump can only be set to quiet or extreme. With fans on low and the pump on extreme it's the loudest thing in the system. There might be a happy medium I need to find. My case is a little "custom" to keep the sound down, and to keep the cat from laying on or puking into the computer so that isn't helping anything.

Capture.JPG


edit: I tried a few Handbrake passes using 4k and 1080 but couldn't get CPU use over about 65%. Max of about 80c on the hottest core. Voltage of 1.37 max per core.
 
Last edited:
I was able to tame the beast enough. Got Prime95 to run fine at stock with the cooler set to extreme. Then I changed the multipliers to all core 4.9ghz and still good but temps hovering around 80c. Left the 4.9 setting in place and set the vcore to 1.30v and that seemed to take a few degrees off. Bought Cyberpunk on Friday and 20 hours in by Sunday night, from level 0 to 15ish, no problems. I had HWinfo runnings the entire time and no core was over 65% usage and no core broke 75c with the game running for almost 12 hours straight, game paused for the occasional break. That's with the H110i set to silent on fans and pump. H110i temp (water temp?) did not rise above 36c with the room at 20c ambient.

Unfortunately every workload seems to generate "all core" multipliers. Even web browsing and basic desktop stuff in W10 I really haven't seen the 5.2 clocks but for a blip. Still... going from a 7700k @ 5ghz to a 10850k @ 4.9 and keeping roughly the same noise level is pretty awesome. When I eventually start loading up all 10 cores I'll just have to manually switch cooling profiles or upgrade. Still need to see if 5ghz all core is in the cards and maybe get into the weeds on overclocking.
 
I was able to tame the beast enough. Got Prime95 to run fine at stock with the cooler set to extreme. Then I changed the multipliers to all core 4.9ghz and still good but temps hovering around 80c. Left the 4.9 setting in place and set the vcore to 1.30v and that seemed to take a few degrees off. Bought Cyberpunk on Friday and 20 hours in by Sunday night, from level 0 to 15ish, no problems. I had HWinfo runnings the entire time and no core was over 65% usage and no core broke 75c with the game running for almost 12 hours straight, game paused for the occasional break. That's with the H110i set to silent on fans and pump. H110i temp (water temp?) did not rise above 36c with the room at 20c ambient.

Unfortunately every workload seems to generate "all core" multipliers. Even web browsing and basic desktop stuff in W10 I really haven't seen the 5.2 clocks but for a blip. Still... going from a 7700k @ 5ghz to a 10850k @ 4.9 and keeping roughly the same noise level is pretty awesome. When I eventually start loading up all 10 cores I'll just have to manually switch cooling profiles or upgrade. Still need to see if 5ghz all core is in the cards and maybe get into the weeds on overclocking.
ochadd you must have a really good 10850K or that water cooler is beast. At 5.0Ghz all core with HT, I need 1.33V Vcore with no AVX offset. I tried using a EK 360MM AIO supposed to be tied with the AF II in top perfomrnace and my heat was very bad. I might try again in the future but the performance was so bad it was worse than my Noctua U14 and DH15 so I just stuck with the DH15. the DH15 cannot handle Prime95 no AVX well. even with AVX encoding, my temps reach high 80C
 
ochadd you must have a really good 10850K or that water cooler is beast. At 5.0Ghz all core with HT, I need 1.33V Vcore with no AVX offset. I tried using a EK 360MM AIO supposed to be tied with the AF II in top perfomrnace and my heat was very bad. I might try again in the future but the performance was so bad it was worse than my Noctua U14 and DH15 so I just stuck with the DH15. the DH15 cannot handle Prime95 no AVX well. even with AVX encoding, my temps reach high 80C
I've been using it for a couple weeks now and I'm impressed. Have not made a push for anything beyond 4.9 ghz yet. Lack of time and this CPU is already overkill for the gaming I do. Maybe that delta of 100 mhz between us is the difference. For me the tangible difference isn't worth it if I have to go to extreme cooling profile and deal with the noise. The 1080 ti is likely my bottleneck until GPU supply comes back.
 
Prime95 small FFT FMA3/fixed works okay too. Watch for CPU Cache L0 errors at the bottom of HWinfo64.

5 ghz.
(1.260v set, LLC6)

prime95_fma3_5ghz.jpg
 
I've been using it for a couple weeks now and I'm impressed. Have not made a push for anything beyond 4.9 ghz yet. Lack of time and this CPU is already overkill for the gaming I do. Maybe that delta of 100 mhz between us is the difference. For me the tangible difference isn't worth it if I have to go to extreme cooling profile and deal with the noise. The 1080 ti is likely my bottleneck until GPU supply comes back.
ochadd Yeah my 1080 Ti is also my bottleneck at 4K right now. If only miners weren't a thing
 
ochadd you must have a really good 10850K or that water cooler is beast. At 5.0Ghz all core with HT, I need 1.33V Vcore with no AVX offset. I tried using a EK 360MM AIO supposed to be tied with the AF II in top perfomrnace and my heat was very bad. I might try again in the future but the performance was so bad it was worse than my Noctua U14 and DH15 so I just stuck with the DH15. the DH15 cannot handle Prime95 no AVX well. even with AVX encoding, my temps reach high 80C
Couple days ago I set the multiplier to 50x to get me to 5 ghz. After a couple nights of Cyberpunk it's been stable but my voltages really shot up. Average went from 1.28 to 1.36 just adding 100 mhz all core. Peak core temp was 82 c over a couple hour gaming session. I'll probably dial it back down to 4.9 ghz once spring comes or I get some dust buildup. Computer has yet to crash on me or fail to boot.
 
I've been using Handbrake to stress test my Asus z490-E/10850K with memory running XMP profile (4000MHz at 1.4v). Been using the H265 2160p60 preset and encoding a +20GB 4K 3840 x 2160 file. Has been running it rock stable at LLC lvl 2. I'm almost disppointed it hasn't crashed haha because I expected it would due to an issue I had with CPU internal errors with unraring certain archives showing up in HWiNFO64 but I'm no longer going to allow that HW monitor to rule my life lol. Though my cpu runs at nearly 100% using Handbreak at 4k encoding, temps never go over 74c (using a new EVGA 280mm liquid cooler... my first foray into liquid cooling after using HSFs and air cooling all my life) and cores typically only run at around 4200 MHz at 125 watts. Am I doing it right or is liquid cooling just so much better? Would expect cores to be running higher & hotter.
 
To me, failing any test is a fail. That said, I've read that with 8th to 10th gen Intel chips, their cache structure makes them more likely to crash out in certain things other than Prime. I believe there was at least one game that needed to be patched due to actually crashing at stock speeds (but passing underclocked). So my opinion is, yes you must be able to pass Prime. A computer that fails a math test is not good. If it fails one, eventually it will fail another. BUT you must also test beyond Prime. That's true on any platform, but this one especially. There are a lot of good suggestions in this thread.
 
The fans can be set low, balanced, extreme, or custom. The pump can only be set to quiet or extreme. With fans on low and the pump on extreme it's the loudest thing in the system. There might be a happy medium I need to find. My case is a little "custom" to keep the sound down, and to keep the cat from laying on or puking into the computer so that isn't helping anything.



edit: I tried a few Handbrake passes using 4k and 1080 but couldn't get CPU use over about 65%. Max of about 80c on the hottest core. Voltage of 1.37 max per core.

Does that box have more vent holes than just the gap at the back? You have plenty of real estate to punch a bunch of holes on the sides with hole saws/augers for more airflow.

Temperature, wattage, and voltage all go hand in hand. See how low you can get that voltage at the desired speed, every CPU will do better at different speeds. You might hit a wall at 4.8, or 5, or somewhere else that takes a lot of extra voltage to cross, where it might not be worth it due to the extra thermal load. The more cores you have the harder to reach high speeds and the hotter it'll run too.

If 5ghz @ 1.28V worked, set it to that and stress test, see how far you can dial back the voltage. I prefer to use offset instead of adaptive when dialing in the "sweet spot", you can then use that number to switch back to adaptive later on.

Also check your other voltages as they can affect temps, and when using XMP for higher ram speeds the motherboard could be pushing them very high on auto. VCCIO and VCCSA (System Agent), run them as low as you can stable, 1.2V should be "good enough" for most people, if not try 1.25V or somehwere between depending on how fine of control you get. If you can get down to 1.15V on each even better.

When setting up my x299 system, IIRC the VCCSA was something crazy like 1.45V and added a bunch of heat to my stress tests. If you left them on auto and they went a lot higher that 1.15-1.2, you'll see a significant drop in temps right there alone.
 
Back
Top