11700k help

[Zero83]

Hi all.

 

So I got my new pc up and running. I am using Aida64 to stress my cpu. With CPU only, I am getting temps barley above 50 degrees, as soon as I introduce FPU test as well, they shoot up to the 90's. I am running a noctua dh15. Should I redo my thermal paste?

[porina]

It's normal. FPU test is by far more stressful. It is kinda like a lite version of Prime95. It depends a bit on the mobo too. I've tried the same a 11700k on two mobos. First under a D15 on Gigabyte Z490, I got throttling with boost to 4.6 GHz. Switch to Asus B560 only boosting to 4.2 GHz with "only" an Arctic 34 eSports duo, didn't hit throttling.

 

Generally speaking, even though it is running hotter, it is doing more work in that state. 

[Zero83]

5 minutes ago, porina said:

It's normal. FPU test is by far more stressful. It is kinda like a lite version of Prime95. It depends a bit on the mobo too. I've tried the same a 11700k on two mobos. First under a D15 on Gigabyte Z490, I got throttling with boost to 4.6 GHz. Switch to Asus B560 only boosting to 4.2 GHz with "only" an Arctic 34 eSports duo, didn't hit throttling.

 

Generally speaking, even though it is running hotter, it is doing more work in that state. 

So should I worry that my pc throttles with the FPU? Is there any other good stress test I can do to make sure my PC is stable?

[Zero83]

Just now, Zero83 said:

So should I worry that my pc throttles with the FPU? Is there any other good stress test I can do to make sure my PC is stable?

I will also note I am not OC

[Stahlmann]

7 minutes ago, Zero83 said:

So should I worry that my pc throttles with the FPU? Is there any other good stress test I can do to make sure my PC is stable?

I like to use CinebenchR23 on loop. It resembles more of a realistic 100% load scenario.

[porina]

6 minutes ago, Zero83 said:

So should I worry that my pc throttles with the FPU? Is there any other good stress test I can do to make sure my PC is stable?

It's been a long time argument that loads like this don't represent the majority of use cases. Especially since you're not overclocking, I wouldn't worry about it. The thermal throttle is unlikely to hit in other use cases.

 

Now if you do happen to have workloads that do stress it that much, then you might want to consider adjusting settings in bios to help tame it. For example, setting a lower power limit. Most mobos, especially enthusiast overclocking enabled ones (regardless if you overclock or not) use aggressive settings to get max performance. Reality is even a small power limit wont affect most things and will only cap those that do stress that hard. Note unlimited power setting is not considered overclocking by Intel.

[Zero83]

6 minutes ago, porina said:

It's been a long time argument that loads like this don't represent the majority of use cases. Especially since you're not overclocking, I wouldn't worry about it. The thermal throttle is unlikely to hit in other use cases.

 

Now if you do happen to have workloads that do stress it that much, then you might want to consider adjusting settings in bios to help tame it. For example, setting a lower power limit. Most mobos, especially enthusiast overclocking enabled ones (regardless if you overclock or not) use aggressive settings to get max performance. Reality is even a small power limit wont affect most things and will only cap those that do stress that hard. Note unlimited power setting is not considered overclocking by Intel.

Thanks for the advice. The only thing I am going to be doing is heavy gaming. Stahlmann suggested using cinebench to simmulate. My pc at idle is running between 22 and 25 c

[Zero83]

29 minutes ago, Stahlmann said:

I like to use CinebenchR23 on loop. It resembles more of a realistic 100% load scenario.

So I am running cinebnch now on loop and so far my temps havnt gone abobe 65. Is this more akin to what I would get gaming?

[Stahlmann]

1 minute ago, Zero83 said:

So I am running cinebnch now on loop and so far my temps havnt gone abobe 65. Is this more akin to what I would get gaming?

Gaming temps will be lower. Cinebench will load up your CPU with a realistic 100% load scenario. It's pretty comparable with blender rendering for example. Gaming doesn't put as much load on your CPU and thus your temps will be lower.

 

With 65°C in cinebench you're golden.

[Zero83]

7 minutes ago, Stahlmann said:

Gaming temps will be lower. Cinebench will load up your CPU with a realistic 100% load scenario. It's pretty comparable with blender rendering for example. Gaming doesn't put as much load on your CPU and thus your temps will be lower.

 

With 65°C in cinebench you're golden.

Ok great to hear. I am going to let it run for 6 to 8 hours and see what I end up with

[Stahlmann]

6 minutes ago, Zero83 said:

Ok great to hear. I am going to let it run for 6 to 8 hours and see what I end up with

1 hour ought to be enough. If temps don't increase by then, then won't in 8 hours.

[Mister Woof]

I like RealBench.

 

It will stress both CPU and GPU in a more realistic load.

 

Cinebench is good for CPU, but temps will be lower because there's no GPU waste heat.

 

RealBench for me generates higher CPU temps because of 300+ watts of GPU waste heat, which you will see in gaming that Cinebench won't capture.

 

In most games it doesn't really matter too much because they won't really stress CPU that much, but some games can.

 

You could also try A64's test to stress both CPU and GPU, which it's probably better than just a CPU test alone, but A64s stress test isn't as good at identifying CPU OC instabilities.

 

For that, a couple hours of RealBench and Cyberpunk 2077 usually will find anything.

[Zero83]

1 hour ago, Mister Woof said:

I like RealBench.

 

It will stress both CPU and GPU in a more realistic load.

 

Cinebench is good for CPU, but temps will be lower because there's no GPU waste heat.

 

RealBench for me generates higher CPU temps because of 300+ watts of GPU waste heat, which you will see in gaming that Cinebench won't capture.

 

In most games it doesn't really matter too much because they won't really stress CPU that much, but some games can.

 

You could also try A64's test to stress both CPU and GPU, which it's probably better than just a CPU test alone, but A64s stress test isn't as good at identifying CPU OC instabilities.

 

For that, a couple hours of RealBench and Cyberpunk 2077 usually will find anything.

Thanks. I will try that. Honestly it scared me when Aida64 hit 98c then immediatly throttled. Thats why im questioning if I did the thermal paste right as I did a pea size dot, but struggled with the noctua picking it up and readjusting until I got it

[Mister Woof]

1 hour ago, Zero83 said:

Thanks. I will try that. Honestly it scared me when Aida64 hit 98c then immediatly throttled. Thats why im questioning if I did the thermal paste right as I did a pea size dot, but struggled with the noctua picking it up and readjusting until I got it

It depends on the default settings on your mobo - it could just be that your motherboard is applying a ridiculous amount of voltage.

 

 

[Zero83]

1 hour ago, Mister Woof said:

It depends on the default settings on your mobo - it could just be that your motherboard is applying a ridiculous amount of voltage.

 

 

Well here is another update. I just ran realbench stress test for the past several hours and my temps never went above 67c. Too bad I can hold that with P95 or Aida 64. I did forget to mention that right now I only have my one fan on the noctua set in the middle as a push and one top exhaust. I still have not hooked up my front and side intake and rear exhaust all in a pull in cool and push out hot config

[Mister Woof]

P95 is going to get you higher temps than RealBench, but then again P95 isn't realistic at all.

 

I haven't tested aida64 in a long time, but when I last tested it, it was quite tame. Perhaps it's been updated since then.

[porina]

19 hours ago, Zero83 said:

Ok great to hear. I am going to let it run for 6 to 8 hours and see what I end up with

Generally speaking longer runs like that might only be used if you're looking to diagnose something. If it is just for confidence, shorter times can be enough.

 

16 hours ago, Mister Woof said:

It depends on the default settings on your mobo - it could just be that your motherboard is applying a ridiculous amount of voltage.

As written earlier it isn't just voltage, but that could be related to the clock. I have a Z490 that boosts Prime95 to 4.6 GHz, and a B560 that goes to 4.2 GHz, with very different temps. While 4.6 GHz is the maximum all core turbo clock, I don't think that takes account of AVX offset and I don't know if there is an official value for that.

 

14 hours ago, Mister Woof said:

P95 is going to get you higher temps than RealBench, but then again P95 isn't realistic at all.

P95 is realistic for P95 like uses, if that is not too circular of an argument. It is a real workload not just a stress test designed to cook a CPU. Do agree that it does not represent the majority of software out there.

 

P95 is all the more extreme on Rocket Lake, because it is the first mainstream CPU to include AVX-512 which P95 can use. I forgot the exact number now, but it does in the ball park of over +40% more work per clock compared to previous CPUs (in P95), which plays a large part in the increased power usage. If you're just using it to heat up a CPU, you don't see that performance increase and this has led many to bad conclusions.

 

14 hours ago, Mister Woof said:

I haven't tested aida64 in a long time, but when I last tested it, it was quite tame. Perhaps it's been updated since then.

My Aida64 is out of date as I decided not to keep up with the sub, but they say they keep updating it for newer CPU features.