Question 13700k on air cooler Cinebench results question

[mjbn1977]

This weekend I put my new 13700k build together. And with new builds I always have paranoia and panic that I didn't apply the thermal paste correctly. I am using a Dark Rock Pro 4 and the CPU is now a little larger and rectangular.

So, when I run Cinebench R23 in Mulitcore I noticed that I get a score a little bit under the expected value compared to what I saw in reviews, around 29500. Most review show it more around 31000. I notice that one of the performance core hits 100C (core 7) within a few seconds and two others will be around 98-99C (core 3 and 5). The rest of the performance cores stay around high 80s and low 90s during Cinebench multicore. By the way, at this point the CPU is running at stock, no changes to power limits, voltages, or multipliers. I know that 100C for this processor is not unsually, but still a little baffled by 10c difference between cores....

1. I wonder if other 13700k owner with air tower cooler have similar experiences and if it is just normal.
2. Can it be that core 7, 5, and 3, are on the same side of the die and for whatever reason when I installed the Darkrock cooler, the thermal paste was applied uneven and misses to cover the area over those cores?
3. Do I assume correct, that the lower multicore score is due to the fact that I am not using a AIO cooler and my cooling is causing earlier throttling? Or should I reach a higher score with this processor on stock setting with an 250W tower cooler?
4. Or is it just silicon lottery and 29500 is just expected on some of the silicone?

Thanks.

 

[Phaaze88]

the CPU is running at stock

This doesn't mean anything, because 'stock' varies between motherboards, sadly.
10C difference between hottest and lowest P-cores is FAR from unusual.

Take info from the following, and picture the 13700K as somewhere in between those, but more towards the i9:

Spoiler

View: https://www.youtube.com/watch?v=dNFgswzTvyc

 

[Roland Of Gilead]

This weekend I put my new 13700k build together. And with new builds I always have paranoia and panic that I didn't apply the thermal paste correctly. I am using a Dark Rock Pro 4 and the CPU is now a little larger and rectangular.

So, when I run Cinebench R23 in Mulitcore I noticed that I get a score a little bit under the expected value compared to what I saw in reviews, around 29500. Most review show it more around 31000. I notice that one of the performance core hits 100C (core 7) within a few seconds and two others will be around 98-99C (core 3 and 5). The rest of the performance cores stay around high 80s and low 90s during Cinebench multicore. By the way, at this point the CPU is running at stock, no changes to power limits, voltages, or multipliers. I know that 100C for this processor is not unsually, but still a little baffled by 10c difference between cores....

1. I wonder if other 13700k owner with air tower cooler have similar experiences and if it is just normal.
2. Can it be that core 7, 5, and 3, are on the same side of the die and for whatever reason when I installed the Darkrock cooler, the thermal paste was applied uneven and misses to cover the area over those cores?
3. Do I assume correct, that the lower multicore score is due to the fact that I am not using a AIO cooler and my cooling is causing earlier throttling? Or should I reach a higher score with this processor on stock setting with an 250W tower cooler?
4. Or is it just silicon lottery and 29500 is just expected on some of the silicone?

Thanks.

Hey there,

So, the 13th gen Intel are designed to run hot. To keep the crown as fastest consumer chips (and with AMD having to do the same) to get more performance, they increased the power budget, and now all but the mid tier CPUs (12400/13400/7600x etc) run much hotter than we are used to.

Under a heavy load, you have to expect hot temps. However, the most important thing here, is does that cause the CPU to throttle!? If you are not throttling and your CPU roughly hits it's all core boost, then you have nothing else you can do, as this is all Intel give you. You can check if the CPU is throttling by using HWInfo (with sensors) and running CB23 alongside it. If it's less than the expected all core, then you need to look at other cooling.

With that said, stronger coolers can reduce these temps (all core full load), and most definitely for gaming, you would not expect temps that high. In terms of tower air coolers, prob the Noctua DH15 is one of the best, and performs as well as a lot of 240 AIOs. AIOs can also be really good at keeping temps down a notch, but really you need a very good one to outperform the Noctua air.

To answer your questions:

1. Yes, more and more. We see it here in the forums. Normal enough.
2. Yes, some cores can perform slightly faster than others. Known as the silicon lottery. Each chip is different. On yours it could be 7,5 and 3, and on another chip, other cores might be slightly faster.
3. No. This could be a whole load of factors, as @Phaaze88 pointed out, the mobo manufacturers settings, what revision bios you are running etc.
4. Yes, possibly. This is how it is sometimes. You'd expect a little variance, maybe like 3-5%.

Lastly in terms of thermal paste application, and the temp difference between cores - No, the difference in temps per core has a +- of 10c or so. Again, this is whole dependant on the individual chip, but is accepted as a tolerance. Spreading the thermal paste evenly does play apart of course, but not as much in terms of per core variance.

 

[mjbn1977]

The motherboard I am using is the MSI Z690 MPG Edge Wifi with the latest firmware. I didn't touch any of the bios settings.

So, most likely its the silicone and not my paste. Glad I don't have to tear it apart and reapply paste....LOL

 

[Roland Of Gilead]

The motherboard I am using is the MSI Z690 MPG Edge Wifi with the latest firmware. I didn't touch any of the bios settings.

So, most likely its the silicone and not my paste. Glad I don't have to tear it apart and reapply paste....LOL

Well, see how it goes. If you feel like the temps are too high, or more importantly when gaming, that the fans aren't 100% whacked out of it (Because who wants to hear that when gaming!! Right?), then you can think about some big AIO to tame that beast.

But, seriously, enjoy the new system. The 13700k is a sterling CPU. Top notch.

 

[mjbn1977]

Well, see how it goes. If you feel like the temps are too high, or more importantly when gaming, that the fans aren't 100% whacked out of it (Because who wants to hear that when gaming!! Right?), then you can think about some big AIO to tame that beast.

But, seriously, enjoy the new system. The 13700k is a sterling CPU. Top notch.

Oh, I have no issues with sound. I had a build with the Dark Rock before and use all Silent Wings fans in the case. Its super silent. No fan noise when gaming.....temps in the 65 to 75 depending on the game. Haven't tested and tweaked too much yet.

I was just wondering about the Cinebench multicore test....seemed about 5% lower than what most reviews show.....

 

[Darkbreeze]

12th and 13th Gen Intel CPUs are NOT designed to be run at temperatures of 100°C and higher, and I don't care what any of the reviewers have to say about it. They are wrong. ALL of them.

The Intel spec says Tjuncture is 100° which means that is the temperature at which it starts throttling, actively, and I know this is true because I've just worked a thread where there was visual proof of his 12600k on Z690 configuration with a maximum core temperature of 100°C and HWinfo showing YES for thermal throttling. That means, that at 100°C it has already exceeded recommended thermal spec and is throttling to remain below the thermal envelope it was designed for. We know this is an Intel specification because if it were up to motherboard manufacturers, especially on their flagship chipsets, they've hope to bump things up even a bit further if they felt their VRMs could handle it in order to deliver a little more performance.

Plus, when I had another 12th Gen on my bench, which was a 12700k, I briefly saw the same throttle behavior at 100°C due to the fact that I had pulled a Cheech and Chong and had forgotten to go back and fully tighten down the heatsink on it after initially getting the threads all started and getting called away for a bit. Came back, forgot I hadn't tightened them down, went through the rest of the process and then during first testing I immediately saw that, noted the HWinfo sensor info, shut down, tightened up, went back in and then saw nothing over 85°C with a properly installed NH-U14S when I ran stress.

I'm not saying anybody here is wrong, because I've read some of those reviews and I know it's the reviewers saying this, and possibly some Intel marketing people as well, but I really don't believe any of the engineers from Intel are saying this and if they are then I'm saying they are wrong too, or else they've passed the wrong information on to board manufacturers and the HWinfo dev, who I've never seen be wrong on spec related issues in the past, must all be wrong too. I will be much happier when Computronix completes his 12th Gen testing and posts that updated information to the Intel temperature guide but until then, I'm going to lean towards fudging specs a bit closer to the almost universal thermal spec that has existed for the last 11 Generations prior of recommending no higher than 85°C for longevity sake.

In fact, the 3rd Gen Ivy bridge parts have a 105°C Tjmax, which is higher than the data sheet indicates for 6th through 13th Gen Core processors which all show a TJmax of 100°C, and those STILL had a recommended maximum, all of them, including the 3rd Gen parts, of no higher than 85°C with a preferred maximum of 80°C. So what's being said is in direct contradiction to what the data sheets indicate and what motherboards and HWinfo say.

And for the record.

 

[Darkbreeze]

Guess what, he DID update it. And the TJmax specification is the same for all 6th through 13th Gen CPUs. So is the recommended maximum core temperature.

How To - Intel CPU Temperature Guide

Update: December 17th, 2024 Preface The topic of processor temperatures can be very confusing. Conflicting opinions based on misconceptions concerning terminology, specifications and testing leaves users uncertain of how to properly check cooling performance. This Guide provides an...

forums.tomshardware.com

 

[Roland Of Gilead]

No one in this thread has said that the chip is running at 100c or beyond, apart from you that is. We are simply discussing that these procs run hot, which they do.

Also, again, no one suggested TJ max is anything as it wasn't mentioned. But we can find that out directly from Intel's website with a very simple google search for those who do not know.

Suggesting all review sites and the data obtained are simply wrong, and you're right does not make it so.

I will happily read reviews based on hard data from multiple websites with extensive experience benching these chips.

We also gave the OP advice on how to determine if his chip is throttling.

 

[Roland Of Gilead]

Oh, I have no issues with sound. I had a build with the Dark Rock before and use all Silent Wings fans in the case. Its super silent. No fan noise when gaming.....temps in the 65 to 75 depending on the game. Haven't tested and tweaked too much yet.

I was just wondering about the Cinebench multicore test....seemed about 5% lower than what most reviews show.....

The different review sites may not be using the same mobo or ram type or power restrictions. I guess there will be some variation between reviews.

 

[mjbn1977]

12th and 13th Gen Intel CPUs are NOT designed to be run at temperatures of 100°C and higher, and I don't care what any of the reviewers have to say about it. They are wrong. ALL of them.

The Intel spec says Tjuncture is 100° which means that is the temperature at which it starts throttling, actively, and I know this is true because I've just worked a thread where there was visual proof of his 12600k on Z690 configuration with a maximum core temperature of 100°C and HWinfo showing YES for thermal throttling. That means, that at 100°C it has already exceeded recommended thermal spec and is throttling to remain below the thermal envelope it was designed for. We know this is an Intel specification because if it were up to motherboard manufacturers, especially on their flagship chipsets, they've hope to bump things up even a bit further if they felt their VRMs could handle it in order to deliver a little more performance.

Plus, when I had another 12th Gen on my bench, which was a 12700k, I briefly saw the same throttle behavior at 100°C due to the fact that I had pulled a Cheech and Chong and had forgotten to go back and fully tighten down the heatsink on it after initially getting the threads all started and getting called away for a bit. Came back, forgot I hadn't tightened them down, went through the rest of the process and then during first testing I immediately saw that, noted the HWinfo sensor info, shut down, tightened up, went back in and then saw nothing over 85°C with a properly installed NH-U14S when I ran stress.

I'm not saying anybody here is wrong, because I've read some of those reviews and I know it's the reviewers saying this, and possibly some Intel marketing people as well, but I really don't believe any of the engineers from Intel are saying this and if they are then I'm saying they are wrong too, or else they've passed the wrong information on to board manufacturers and the HWinfo dev, who I've never seen be wrong on spec related issues in the past, must all be wrong too. I will be much happier when Computronix completes his 12th Gen testing and posts that updated information to the Intel temperature guide but until then, I'm going to lean towards fudging specs a bit closer to the almost universal thermal spec that has existed for the last 11 Generations prior of recommending no higher than 85°C for longevity sake.

In fact, the 3rd Gen Ivy bridge parts have a 105°C Tjmax, which is higher than the data sheet indicates for 6th through 13th Gen Core processors which all show a TJmax of 100°C, and those STILL had a recommended maximum, all of them, including the 3rd Gen parts, of no higher than 85°C with a preferred maximum of 80°C. So what's being said is in direct contradiction to what the data sheets indicate and what motherboards and HWinfo say.

And for the record.

Based on everything you said, I should be good then. I mostly using the computer for gaming and during gaming the CPU temps not even going over 75C on any cores, most sit around 70C or less. The situation with a few cores hitting 100 is only in full load multicore rendering benchmarks. But I don't really use the computer for those task. Maybe for a few minutes compiling an updated aircraft in MS flight simulator but that is pretty much it. Again, I mostly wondered about the Cinebench Multicore score. I know that the CPU runs hot and might throttle, but just wanted to ease my thermal paste paranoia...LOL

 

[Roland Of Gilead]

Your gaming temps are fine:

Intel Core i7-13700K Review - Great at Gaming and Applications - Temperatures | TechPowerUp

 

[Darkbreeze]

Based on everything you said, I should be good then. I mostly using the computer for gaming and during gaming the CPU temps not even going over 75C on any cores, most sit around 70C or less. The situation with a few cores hitting 100 is only in full load multicore rendering benchmarks. But I don't really use the computer for those task. Maybe for a few minutes compiling an updated aircraft in MS flight simulator but that is pretty much it. Again, I mostly wondered about the Cinebench Multicore score. I know that the CPU runs hot and might throttle, but just wanted to ease my thermal paste paranoia...LOL

Correct. So long as you are staying below 85°C, maybe in this case a person could even moderately safely allow that envelope to expand to 90°C, regardless of what it is you are doing, then you are perfectly fine running in that thermal window. Only if you are seeing temps more closely approaching 95°C or higher should you be worried UNLESS you are seeing those temps while doing things that shouldn't be pushing the cores that high. Then you might have something going on you need to look at.

 

[mjbn1977]

So, just a quick follow up. I tested around a little bit. During Cinebench R23 mulitcore I run into throtteling on 2-3 cores. So the mulitplier goes down to 51 instead of 53. I think that explains the lower score of 28500 to 29500. Still good, I guess, considering I am on a tower cooler. I noticed that I can set the type of cooling solution as a setting in the MSI Bios, can choose from Box cooler (231Watt max), Tower Cooler (264 Watt max) and Liquid AIO (4080 Watts max??). I tried all three of them. No difference in score on Tower Cooler and Liquid AIO, but on Box cooler setting I only got 20000 score in R23 multicore....really low multipliers. But temps under 90C all the way through the test. I also just out of curiosity, I changed the temp limit to 105C for one test run in the Bios. That actually gave me a 500 points higher R23 multicore score. But its not worth it. Again, mostly a gaming system with temps at stock with Tower Cooler setting in the Bios absolutely fine during gaming. So, I guess I will keep it at that.

One strange thing that I noticed: Core 7 is the thread that always runs the hottest. Hits 100C instantly on full loads, while other cores don't even hit 90C. But whenever I run Cinebench R23 mulitcore, its always core 7 that is used for that task (70% load) and core 5 (30% of the load). I remember that I read that Turboboost 3.0 is supposed to pick the best performing core for those situations, which it is clearly not? Is there I way to have the processor pick other cooler operating threads???

 

[Roland Of Gilead]

So, just a quick follow up. I tested around a little bit. During Cinebench R23 mulitcore I run into throtteling on 2-3 cores. So the mulitplier goes down to 51 instead of 53. I think that explains the lower score of 28500 to 29500. Still good, I guess, considering I am on a tower cooler. I noticed that I can set the type of cooling solution as a setting in the MSI Bios, can choose from Box cooler (231Watt max), Tower Cooler (264 Watt max) and Liquid AIO (4080 Watts max??). I tried all three of them. No difference in score on Tower Cooler and Liquid AIO, but on Box cooler setting I only got 20000 score in R23 multicore....really low multipliers. But temps under 90C all the way through the test. I also just out of curiosity, I changed the temp limit to 105C for one test run in the Bios. That actually gave me a 500 points higher R23 multicore score. But its not worth it. Again, mostly a gaming system with temps at stock with Tower Cooler setting in the Bios absolutely fine during gaming. So, I guess I will keep it at that.

It looks like the different cooler settings on your mobo are putting in lower power limits than normal, which is why the temps drop.

One strange thing that I noticed: Core 7 is the thread that always runs the hottest. Hits 100C instantly on full loads, while other cores don't even hit 90C. But whenever I run Cinebench R23 mulitcore, its always core 7 that is used for that task (70% load) and core 5 (30% of the load). I remember that I read that Turboboost 3.0 is supposed to pick the best performing core for those situations, which it is clearly not? Is there I way to have the processor pick other cooler operating threads???

That's very unusual for the CBR23 Multitest, that the cores are loaded different. What you are looking for is it's all core boost frequency there. So for these chips with beefy cooling they can run at 5.5ghz all core on the P Cores. 5.1 all core is pretty low in comparison. The cooler isn't sufficient for higher loads with faster clocks, but as you've said, is fine for gaming. Not many games will put a load on your CPU like CB would. The Intel Thread Director helps with allocation to the right core. So, if your happy with temps in the 65-75c range for gaming (which is expected), then be happy out and enjoy the system.

At least you know now, your CPU is limited by your cooler, and you can extract more performance if you need it down the line.

 

[mjbn1977]

It looks like the different cooler settings on your mobo are putting in lower power limits than normal, which is why the temps drop.



That's very unusual for the CBR23 Multitest, that the cores are loaded different. What you are looking for is it's all core boost frequency there. So for these chips with beefy cooling they can run at 5.5ghz all core on the P Cores. 5.1 all core is pretty low in comparison. The cooler isn't sufficient for higher loads with faster clocks, but as you've said, is fine for gaming. Not many games will put a load on your CPU like CB would. The Intel Thread Director helps with allocation to the right core. So, if your happy with temps in the 65-75c range for gaming (which is expected), then be happy out and enjoy the system.

At least you know now, your CPU is limited by your cooler, and you can extract more performance if you need it down the line.

Sorry, I misspoke (miswrote???). I meant in Cinebench Singlecore it always loads the core that runs the hottest. in Cinebench Multicore it loads all threads even at 100%. I can make a screen shoot tonight. Then you can see what I mean. I will post one of each from hardwareinfo64 from single and multicore run....

 

[Roland Of Gilead]

Sorry, I misspoke (misworte???). I meant in Cinebench Singlecore it always loads the core that runs the hottest. in Cinebench Multicore it loads all threads even at 100%. I can make a screen shoot tonight. Then you can see what I mean. I will post one of each from hardwareinfo64 from single and multicore run....

Ah, yes. Exactly. This is typical. Because it's the best core for highest clockspeed. The Intel thread Director, along with Windows Schedular assign most single core tasks to the best core. This is the design. Not unexpected.

 

[Darkbreeze]

Ah, yes. Exactly. This is typical. Because it's the best core for highest clockspeed. The Intel thread Director, along with Windows Schedular assign most single core tasks to the best core. This is the design. Not unexpected.

This is correct.

 

[mjbn1977]

Ah, yes. Exactly. This is typical. Because it's the best core for highest clockspeed. The Intel thread Director, along with Windows Schedular assign most single core tasks to the best core. This is the design. Not unexpected.

Wait wait wait.....do I understand it right.....the core that runs the hottest is the best? shouldn't it be the one that can run the highest clock the coolest?? I mean, if I do the gaming mode overclock on my bios all clocks have no issue in running on 5.4 Ghz. But again...the only one of those that runs hits 100C is also the one that is used for single thread? So my hottest core is the best according to the binning??

 

[Darkbreeze]

It's running the hottest BECAUSE it's getting used more than any other core.

 

[Roland Of Gilead]

Wait wait wait.....do I understand it right.....the core that runs the hottest is the best? shouldn't it be the one that can run the highest clock the coolest?? I mean, if I do the gaming mode overclock on my bios all clocks have no issue in running on 5.4 Ghz. But again...the only one of those that runs hits 100C is also the one that is used for single thread? So my hottest core is the best according to the binning??

hehe! Yes, you would almost think so.

So, the best core typically runs at max speed and as long as the other cores aren't' at the same level, then temps don't mean as much. The boost algorithm uses any thermal headroom to boost the clocks as far as they go, while staying within its thermal envelope. If this is a single core process, then yes, that core will nearly always be used.

If in an all core situation, your CPU will boost to it's best speed (across all cores), but again within a thermal envelope for all the P cores to work.

The algorithm can switch between the best single core clocks, and the best all core clocks on the fly! It's really ingenious how it works. Essentially the CPU does all these calcs/variations on the fly, as your system goes along on it's route.

Often, the best results for CPU's and GPU's can just be to let them run with their boost systems enabled. Take away all the headaches of OC'ing this CPU or GPU.

With that said, if you want to OC your CPU, take a look at @Darkbreeze ' excellent review on how to OC your CPU. Great guide for beginners and experienced OC's to look at. It's a guide I often reference here: (1) CPU overclocking guide and tutorial for beginners | Tom's Hardware Forum (tomshardware.com)

 

[Darkbreeze]

If ALL cores are running at a steady state load (Requires running a steady state stress utility like Prime95, Heavy load, Intel processor diagnostic tool) with a 100% load on all cores and that core is running hotter than ALL other cores, and is being used as the best core, then I'd say there is something wrong in that equation. Bad CPU, need to update BIOS, need to install latest chipset driver, need to update windows, something.

 

[Phaaze88]

I remember that I read that Turboboost 3.0 is supposed to pick the best performing core for those situations, which it is clearly not? Is there I way to have the processor pick other cooler operating threads???

Did you go into bios and see which cores are binned? They should be marked by something.
At least, on my X299's bios, I can see which 2 they are if I go in there and look into per core frequency settings - Cores 2 and 3 have an * next to them.

 

[mjbn1977]

Did you go into bios and see which cores are binned? They should be marked by something.
At least, on my X299's bios, I can see which 2 they are if I go in there and look into per core frequency settings - Cores 2 and 3 have an * next to them.

Let me check tonight.

 

[mjbn1977]

If ALL cores are running at a steady state load (Requires running a steady state stress utility like Prime95, Heavy load, Intel processor diagnostic tool) with a 100% load on all cores and that core is running hotter than ALL other cores, and is being used as the best core, then I'd say there is something wrong in that equation. Bad CPU, need to update BIOS, need to install latest chipset driver, need to update windows, something.

I didn't test it in Prime95, only with Cinebench. But the core than runs by far the hottest in the Cinebench Multicore bench is also the one used in the Singlecore bench. That is what is confusing me.

Which Prime95 should I use to identify the hottest core, the version with or without AVX?