10700K on stock speeds and Noctua NS-D15S - Is Prime95 hitting 90C normal?

Joined
Jan 3, 2009
Messages
644
These are the core parts of my system:

https://pcpartpicker.com/user/Cyber_Akuma/saved/yWGfyc

I got the CPU as Open Box from Microcenter so I want to make sure there is nothing wrong with it, not planning to overclock it.

I am running a 10700K at stock speeds and motherboard settings, not overclocked. I tried to Google about this, but almost all the results were from people trying to overclock, not run at stock settings, and most were using an AIO.

After putting just the essentials together and running Memtext86+ for two passes successfully, I installed Windows on a temporary drive to perform CPU testing and benchmarking.

I installed MSI Afterburner to monitor temps and then ran Prime95 in testing mode to check stability and temperatures.

The temperatures seem somewhat high though considering it's not even overclocked.

When it started out, it would go between the mid- 60s and mid-80s:



Eventually though, it would peak at the 90s from time to time:



It seemed to more or less then go in a pattern, about 14 minutes of running in the mid-60s, and then 7 minutes of peaking at the 90s, over and over:



Right now, after having run for about 10 hours, I see that the max it has apparently hit is 93C (From when I am monitoring it manually it normally seems to peak at 90C, sometimes hits 91C for a split-second before going back down to 90C or lower)..... that's a mere 7C away from the limit... and this is on stock speeds and in a rather cool basement (The PC isn't going to stay there, I am just building it there for now). Though, it is also on an air cooler with one fan, but it's supposed to be one of the best coolers out there. So far all cores seem to more or less never go down from 4690MHz so I don't think it's throttling as I looked it up and 4.7GHz seems to be the standard Turbo across all cores at load. All the cores seem to more or less follow the same temperature pattern, with two of them running about 2-5C colder than the rest and a few sometimes having about 5 seconds of uneven jittering here and there, so I don't think the paste was applied unevenly.

Is this normal? It feels like this is way too high for stock speeds with how close it is to throttling, though I understand that Prime95 puts an extreme load on a CPU.... but wouldn't doing some tasks such as encoding x265 or hours of heavy file compression also put such a strain?

Normally in my testing after Prime95 has been successful for about 24 hours I do about 30-60 minutes of Furmark while Prime95 is running to ensure the motherboard/PSU can handle the strain of that power draw, but with the only GPU currently in the system being the CPU, I worry that just simply using CPU tasks alone is at it's limit, and then stressing the GPU on top of that could push it past it's throttle limit.

Also, looking at my cooler, I notice that the fan was placed higher than I thought. I thought I had placed it as low as it can go without touching the mount, but now that I look again it's not covering the bottom of the radiator while the top is blowing past the radiators and could likely be lowered a little more, not sure if just simply lowering the fan about 1-1.5 inches can make much of a difference though, how does this setup look?:







Also considering adding a second fan, mostly because I have the room for it in my case, motherboard headers, and budget as well as it's pretty quiet right now so noise isn't a concern, though I heard a second fan on this cooler only tends to only drop temps by about 2-3C.

So do I need to be concerned about these temps? Is this normal for a 10700K on stock speeds with the latest version of Prime95? Would it be pushing it too far if I added Furmark for 30-60 min on top of this current test? Anything else anyone can suggest or advice they can give about this?
 
I see a LOT of reports of Prime making these chips run hot as f**k. Try doing a real-world high load scenario like the video encode you mentioned and see how that goes.
 
What is the actual clocks under prime small fft's avx2. Actual stock should be close to actual base, so like 3.6-4. The pictures are to blurry and when I link them they don't pull up.
 
I'd also like to note that with the past couple generations at least of -K series Intel chips, "stock" on many/most motherboards might not actually be stock, because Intel allows/encourages mobo manufacturers to do all sorts of zany things with default settings in the name of out-of-box performance tuning. Also, if Prime is doing any AVX stuff (depending on the test selected), that will cause temps to soar on Intel chips. It's really a worst-case scenario temperature-wise.
Doing some real-world tests would be most illuminating.
 
The clock stays at 4690MHz the entire time.

To expand on what katekatekate said, it's probably applying to much voltage for multicore boost. It's definitely not within stock wattage limits, but that's nothing new since 8xxx series.

I'd probably use something like realbench for temps, or just lower the wattage limit. *I haven't used a recent intel board, so I don't know if that's an actual option*
 
is it normal? yes and the up/downs are normal, its p95 doing its thing.
edit: the "too much voltage" mentioned is very possible too. mobos love giving chips more than they need when on auto.
 
Also, if Prime is doing any AVX stuff (depending on the test selected), that will cause temps to soar on Intel chips. It's really a worst-case scenario temperature-wise.

I just used the default Blend test, would that have all AVX stuff enabled?

I had someone recommend a Small FFT test with all AVX settings disabled, is that what I should use to test temperatures? Or just blend with AVX disabled?

Also, would it still be safe to continue my original testing idea where after a roughly 19-24 hour mark of the current test I am doing now (It's at about the 14-15 hour mark right now) I would do 30-60 min of Furmark wile Prime95 is still running? Or would that be pushing the CPU (considering Furmark would be running on it's IGP) way too far and could damage it?
 
I just used the default Blend test, would that have all AVX stuff enabled?

I had someone recommend a Small FFT test with all AVX settings disabled, is that what I should use to test temperatures? Or just blend with AVX disabled?

Also, would it still be safe to continue my original testing idea where after a roughly 19-24 hour mark of the current test I am doing now (It's at about the 14-15 hour mark right now) I would do 30-60 min of Furmark wile Prime95 is still running? Or would that be pushing the CPU (considering Furmark would be running on it's IGP) way too far and could damage it?
yes.
that will give more normal temps since avx is off. thats what drives temps up. you could also set an avx offset in your bios to lower speed and temp a bit during avx tasks.
if youre at 15hrs of p95 its probably good. if you want to test the whole system in a realistic manor, use OCCT. run its 1hr test a few time and then if its good, get to gaming or whatever.
 
You're just seeing the effects of Intel deciding that turning off power limits isn't overclocking. That basically just lets a chip run at max turbo all the time. Intel doesn't consider it overclocking unless you run it faster than it can run in a stock configuration. Stock for a 10700 (non-k) is a 65W long term power limit, but it cant burst to 224W. The 10700k has a stock long term limit of 125W and the burst max is similar to the 10700, I just forget exactly what it is. Might be 224W too, might be 250W.

At any rate, most boards from MSI, Gigabyte, and ASRock will let the CPU have as much power as the board can handle by default. Asus tends to respect Intel defaults. Funny thing is there are "B" boards that'll do 250W sustained and "Z" boards that can't come close to that.

At any rate if you're topping out in the mid 90s in a Prime95 stress test and planning on running stock it's fine. 10xxx chips don't thermal throttle until they hit 100C. Prime95 throws a massive amount of AVX at a chip, which you will almost never see in any real world application unless you're running some sort of math crunching app that uses a ton of AVX... basically what prime is doing. If you want to limit the temps the chip hits just fiddle with Prime a bit and figure out which specific test gets it the hottest, then turn down the power limit in the BIOS until you're happy with your max temps. I'm pretty sure it's small FFTs with AVX enabled.

If you're overclocking you'll likely need to set an AVX offset, but for running in max turbo mode as long as it's useful (not officially overclocked according to Intel) you can just turn down the long term power limit a bit until you like your temps. If the heaviest load you'll throw at the machine is gaming I wouldn't worry about it. Games don't get a chip nearly has hot as Prime95 small FFTs with AVX on. Also, note that increasing the power limit doesn't affect any of the power saving features, so it'll still clock way down when idle.
 
Is this normal? It feels like this is way too high for stock speeds with how close it is to throttling, though I understand that Prime95 puts an extreme load on a CPU.... but wouldn't doing some tasks such as encoding x265 or hours of heavy file compression also put such a strain?



So do I need to be concerned about these temps? Is this normal for a 10700K on stock speeds with the latest version of Prime95? Would it be pushing it too far if I added Furmark for 30-60 min on top of this current test? Anything else anyone can suggest or advice they can give about this?
I will reinforce that Prime 95 will generate more heat than just about any actual workload. Especially now that it runs with AVX. It is designed to feed your CPU with a constant, consistent load which is exactly meant to max out your CPU. Most actual workloads do not feed your CPU such consistent work.

IMO Prime 95 is only useful for stability testing. I mean, its cool if you can get good temps while running Prime 95. But your temps while gaming or encoding video will be notably lower.

You can do the furmark thing. But again, that's checking stability. Your CPU will never be that hot while gaming.

Probably the most strenuous thing the average person could do with their CPU with a real workload, would be to play a demanding game and also stream to Twitch while doing the encoding for the stream on the CPU. And that still won't get as hot as Prime 95. But that would be a great stability test. As that will really stress multi-threading performance and put a lot of demand on your CPU, in general. Also, I've seen overclocked systems which seems stable, but then crash while encoding in Handbrake or something. And by that I mean the person has played all of their games and whatnot. Then they try some encoding and their CPU says "nope".

And yes, most motherboard manufacturers have special turbo tunings turned on by default. These modes make your CPU turbo more consistently and for longer. MSI in particular, doesn't even put a limit on their tuning. These modes may also ad voltage to the CPU. All mobos should have settings to use the stock Intel Turbo behavior and voltages. You will probably drop some heat, but also drop some performance in certain workloads. However, with a K CPU you aren't likely to notice the difference unless you are benchmarking. Because K CPUs have more aggressive turbo anyway.
 
Interesting thread.

I too have a stock 10700k that behaves identically in P95, except I'm using a EVO 212 BE RGB in push/pull as a stop gap. I was looking towards the Dark Rock Pro 4 or d15 as a upgrade, but now I'm not sure.

What really got me thinking that I needed better cooling, was the PSU test in OCCT. All 8 cores hit 99C instantly.
In gaming (Serious Sam 4, Metro Exodus) I avg. about 65-72C after 4hrs. And that's with none of the MSI auto OCing bullshit enabled.

It seems if I'm going to OC, a 240mm/360mm AIO is recommended to cool this monster.
 
Couple things I see. There is only 1 fan on the D15 and the chip is likely extending turbos as others have said because depending on the manufacturer, this is the "stock" settings. You might actually have to change settings off of "Auto" to get it to stay within the 125W limits for long term boost. The AVX load is going to heat up the chip more than normal. A max of 93C is probably what I would expect under these conditions even under "stock" loads.
 
You're just seeing the effects of Intel deciding that turning off power limits isn't overclocking. That basically just lets a chip run at max turbo all the time. Intel doesn't consider it overclocking unless you run it faster than it can run in a stock configuration. Stock for a 10700 (non-k) is a 65W long term power limit, but it cant burst to 224W. The 10700k has a stock long term limit of 125W and the burst max is similar to the 10700, I just forget exactly what it is. Might be 224W too, might be 250W.

At any rate, most boards from MSI, Gigabyte, and ASRock will let the CPU have as much power as the board can handle by default. Asus tends to respect Intel defaults. Funny thing is there are "B" boards that'll do 250W sustained and "Z" boards that can't come close to that.

At any rate if you're topping out in the mid 90s in a Prime95 stress test and planning on running stock it's fine. 10xxx chips don't thermal throttle until they hit 100C. Prime95 throws a massive amount of AVX at a chip, which you will almost never see in any real world application unless you're running some sort of math crunching app that uses a ton of AVX... basically what prime is doing. If you want to limit the temps the chip hits just fiddle with Prime a bit and figure out which specific test gets it the hottest, then turn down the power limit in the BIOS until you're happy with your max temps. I'm pretty sure it's small FFTs with AVX enabled.

If you're overclocking you'll likely need to set an AVX offset, but for running in max turbo mode as long as it's useful (not officially overclocked according to Intel) you can just turn down the long term power limit a bit until you like your temps. If the heaviest load you'll throw at the machine is gaming I wouldn't worry about it. Games don't get a chip nearly has hot as Prime95 small FFTs with AVX on. Also, note that increasing the power limit doesn't affect any of the power saving features, so it'll still clock way down when idle.

In real world testing with my 10700 (non-k), I didn't find it to actually draw 225W. I disabled all the turbo limits and it would only draw maybe 170W even under an AVX load with a Z490 board (as measured in the IXT...so who knows if that is accurate or not).

I'm guessing that might be because it wasn't pushing quite as high of a max clock as the "K" chip. I would get about 4.7Ghz all core with a BCLK of 102 and unlimited turbo.
 
Interesting thread.

I too have a stock 10700k that behaves identically in P95, except I'm using a EVO 212 BE RGB in push/pull as a stop gap. I was looking towards the Dark Rock Pro 4 or d15 as a upgrade, but now I'm not sure.

What really got me thinking that I needed better cooling, was the PSU test in OCCT. All 8 cores hit 99C instantly.
In gaming (Serious Sam 4, Metro Exodus) I avg. about 65-72C after 4hrs. And that's with none of the MSI auto OCing bullshit enabled.

It seems if I'm going to OC, a 240mm/360mm AIO is recommended to cool this monster.
make that a 280 or 360. your temps are well within limits now but better cooling is always a plus
 
So I think I screwed up my motherboard settings.

Remember when I said I was not overclocking? I had left my motherboard on default settings. I looked through the bios to try to make sure it was not set to overclock it by default, and didn't see any settings that appeared to be so, and any OC settings I did notice appeared to be off, so I left it like that. But apparently this "Multi Core Enhancement" feature that by default is enabled is just a fancy name for "auto overclock"? And it tends to compensate by using voltages that are way too high?

So was I actually overclocking without even realizing it?

Anyway, so that Blend test is over.

Was having longer and longer periods where the temperature stays in the 60s for some reason. It used to be for 14 minute periods and then 7 minute periods of up to 90C but then it started having several-hour periods of 60C with an occasional few-minute burst of up to 90C.

Anyway, so I ran Furmark for an hour just to make sure it remains stable with that, temperatures didn't seem to go up a single degree during that despite running it on the IGP. I was really hoping to run it during those segments where it goes to 90C just to make sure but it was not getting that hot again, so I just did it while it was 60C.

Max temp after about 24 hours was 94C, but that was a spike for about a second, I rarely saw it go beyond 90C during the times when temps were above 60C. CPU-Z was reporting the VCore being around 1.332 the majority of the time, but HWiNFO64 reported a min of about 1.28 to a max of about 1.32, with it normally being around 1.3. Are those normal voltages considering the whole Multicore Enhancement thing?



Moving on, as per many recommendations I got, I re-started Prime95 in Torture Test mode, set to Small FFT with AVX and AVX 2 disabled (Apparently my 10700K does not support AVX-512, seems that's only in current i9s and Xeons) and reset my monitoring software.

After about 8 hours now (Hey, I got sleepy) the results don't seem any different then the first 15 or so minutes of the test. Where temps hover around the 71-72C range nearly the entire time (had a spike to 78C for about a second) and VCore is between 1.2 to 1.26, generally around 1.25:



That.... looks normal to me, is it? If my motherboard really is auto-overclocking with aggressive voltages then that doesn't sound right.

I too have a stock 10700k that behaves identically in P95, except I'm using a EVO 212 BE RGB in push/pull as a stop gap. I was looking towards the Dark Rock Pro 4 or d15 as a upgrade, but now I'm not sure.

I was reading up on many coolers when I decided on the NH-D15S for my CPU, and the 212 did come up a lot, namely, on how starting with the 9th gen Intel CPUs tended to run too hot for the 212 to keep up anymore. I would definitely say that you should upgrade to something like a D15 or D15S if it would not cause any compatibility problems in your case.

In gaming (Serious Sam 4, Metro Exodus) I avg. about 65-72C after 4hrs. And that's with none of the MSI auto OCing bullshit enabled.

Those temps definitely look a bit too high for just gaming with MCE disabled, I think it would be a benefit to upgrade from that 212.

The majority of the time my Prime95 blend tests were in the low 60s with my D15S, and I haven't added a second fan yet.

Couple things I see. There is only 1 fan on the D15 and the chip is likely extending turbos as others have said because depending on the manufacturer, this is the "stock" settings. You might actually have to change settings off of "Auto" to get it to stay within the 125W limits for long term boost. The AVX load is going to heat up the chip more than normal. A max of 93C is probably what I would expect under these conditions even under "stock" loads.

Yeah, seems I did have some sort of auto-overclock enabled that I wasn't aware of. Though if it doesn't cause temps above 80C for anything outside of Prime95's Small FFT test is it harmful?

I am planning to upgrade the CPU in this system to a 11700K in a year or two, mostly for PCIe4, so I am not that put off by "losing the silicon lottery" as it were, was never planning to overclock this.

And yeah, the D15S comes with one fan, but is very similar in design to a D15 and lets you add a second 140 or 120mm fan, even comes with the mounting hardware to do so in the box.
 
Yes, MCE is essentially an auto overclocking. If you want to run within the dictated "stock" power limits you would have to disable it. Every motherboard is different though. On the ASUS board I have I can literally type in how much wattage I want the turbo to run at and the duration. I'm not sure what the Intel specs are for that off the top of my head but you can look it up and set it to exactly what it should be.
 
Yes, MCE is essentially an auto overclocking. If you want to run within the dictated "stock" power limits you would have to disable it. Every motherboard is different though. On the ASUS board I have I can literally type in how much wattage I want the turbo to run at and the duration. I'm not sure what the Intel specs are for that off the top of my head but you can look it up and set it to exactly what it should be.
Yep. In addtion to turning off MCE, there should indeed be somewhere you can manually determine the turbo settings.
 
I have a 10700K installed on a Gigabyte Z590 Ultra with a be quiet! Dark Rock Pro 4. Out of the box default motherboard settings resulted in 88C peak temp during Prime95 small 4k FFT.

Here’s what I changed to reach 79C peak with zero loss in performance per geekbench.

- Enhanced Multi-Core Performance = disabled
- CPU Internal AC/DC Load Line = power saving

That’s it. Voltage according to coretemp dropped from 1.25 to 1.13, max 160W power draw.

gaming temps are in the low 60s
 
Some mobos even have a spot where you can just set the TDP limit in increments of 1. Just gotta dig through all those settings and see what your options are. Everyone probably has a few ways to effectively lower power/heat.
 
Back
Top