[SOLVED] 10900 vs 10900k power draw

Feb 5, 2021
38
0
30
Hello,

I don't OC but I also don't mind paying a bit more for the ~5 percent faster speed the K offers.

But why does K use 125w? That's an additional 60w compared to non-K version, while I don't see such huge difference in speeds or number of cores/threads.
 
Last edited:
Solution
So basically when rendering something for hours (let's say a 3D scene that'll be rendered by CPU) the chart you posted means most of the time both CPUs will be using 225-250w is that right?
If the mobos follow the TDP and tau rules the K version will run 250W for 56 sec while the non k will run 224W for 28sec.
After that they will go down to 125 and 60 respectively which means 3.7 under heavy loads for the one and 2.7 for the other.
That's a 40% difference and is going to be substantially more than just 5%
If you keep the non k version running at PL2 all the time it will be about 5% different but then the TDP also will not be 60W!

It takes a few seconds at very low usage to reset TAU so if you render something that takes a...

Lutfij

Titan
Moderator
If you're not going to overclock, you don't need the K suffix processor, you also won't need a Z series chipset or higher spec'd rams. You might want to compare the turbo clocks for both processors side by side as well as their base frequencies, you'll see why the higher TDP comes from with relation to the clocks.
 
Feb 5, 2021
38
0
30
If you're not going to overclock, you don't need the K suffix processor, you also won't need a Z series chipset or higher spec'd rams. You might want to compare the turbo clocks for both processors side by side as well as their base frequencies, you'll see why the higher TDP comes from with relation to the clocks.

Left 10900K
Right 10900


Clockspeed3.7 GHz2.8 GHz
Turbo SpeedUp to 5.3 GHzUp to 5.2 GHz

While Turbo speed is pretty close to each other, Clockspeed has noticeable difference. I assume those numbers are without OC-ing so why the non-K version is suggested to people who do not OC? An additional 0.9 GHz doesn't count much?

Also 3.7 vs 2.8 doesn't explain the significant difference between 65w and 125w... K version is basically using twice as much power while I don't see such difference in the specs. Where does the additional 60w in K version go?
 
Last edited:

Phaaze88

Titan
Ambassador
View: https://imgur.com/YwqWtBb


Power Limit 1 is the limit for all core at the base clock.
Power limit 2 is limit for all cores for those turbo boosts.

Tau is how long the cpu is allowed to run at PL2 before it's forced back down to PL1, but depending on the workload, it will either:
-sit at PL1 until the task is complete.
-reset before Tau expires, thus extending PL2 duration.
 
Feb 5, 2021
38
0
30
Power Limit 1 is the limit for all core at the base clock.
Power limit 2 is limit for all cores for those turbo boosts.

Tau is how long the cpu is allowed to run at PL2 before it's forced back down to PL1, but depending on the workload, it will either:
-sit at PL1 until the task is complete.
-reset before Tau expires, thus extending PL2 duration.

Hi man,

Although I'm somewhat old I've never been into hardware details so I apologize if my questions sound frustrating.

So basically when rendering something for hours (let's say a 3D scene that'll be rendered by CPU) the chart you posted means most of the time both CPUs will be using 225-250w is that right?

So essentially I'm trying to figure out if both use similar amount of power when under full load for long times. If yes, I get the K. If non-K uses half or 2/3 of the power, I'd ignore the 5% improvement and get non-K CPU.
 

Phaaze88

Titan
Ambassador
So basically when rendering something for hours (let's say a 3D scene that'll be rendered by CPU) the chart you posted means most of the time both CPUs will be using 225-250w is that right?
It depends on the application. I'm not familiar with the work you do, but mentioned the 2 things what would happen; it's going to be one or the other.
Perhaps someone familiar with that can chime in?

So essentially I'm trying to figure out if both use similar amount of power when under full load for long times. If yes, I get the K. If non-K uses half or 2/3 of the power, I'd ignore the 5% improvement and get non-K CPU.
Obviously they don't. The K uses a fair bit more power.

I mean, if you want the best out of the box performance and not have to tinker with anything: 10900K.
Best have a good cooler for it - well, for both really.
 
Last edited:
So basically when rendering something for hours (let's say a 3D scene that'll be rendered by CPU) the chart you posted means most of the time both CPUs will be using 225-250w is that right?
If the mobos follow the TDP and tau rules the K version will run 250W for 56 sec while the non k will run 224W for 28sec.
After that they will go down to 125 and 60 respectively which means 3.7 under heavy loads for the one and 2.7 for the other.
That's a 40% difference and is going to be substantially more than just 5%
If you keep the non k version running at PL2 all the time it will be about 5% different but then the TDP also will not be 60W!

It takes a few seconds at very low usage to reset TAU so if you render something that takes a break every minute or so it might always stick to max.

But the main issue is that almost no mobo will stick with intel recommendations, they will use whatever settings so nobody can answer your question, chances are that the mobo will use way higher settings causing much higher power draw for either CPU.
You have to spend the time to check and setup things the way you want them.
 
Solution
Feb 5, 2021
38
0
30
If the mobos follow the TDP and tau rules the K version will run 250W for 56 sec while the non k will run 224W for 28sec.
After that they will go down to 125 and 60 respectively which means 3.7 under heavy loads for the one and 2.7 for the other.
That's a 40% difference and is going to be substantially more than just 5%
If you keep the non k version running at PL2 all the time it will be about 5% different but then the TDP also will not be 60W!

It takes a few seconds at very low usage to reset TAU so if you render something that takes a break every minute or so it might always stick to max.

But the main issue is that almost no mobo will stick with intel recommendations, they will use whatever settings so nobody can answer your question, chances are that the mobo will use way higher settings causing much higher power draw for either CPU.
You have to spend the time to check and setup things the way you want them.

This is the point I wanted to get to.
You're saying that at 3.7 and 2.7 the K will perform 40% better than non-K.
This would justify the doubling of power usage for K and I'd be fine with it.

But then again, I wasn't able to find any benchmarks where the performance difference between the two would get anything near 40% argh!
 
This is the point I wanted to get to.
You're saying that at 3.7 and 2.7 the K will perform 40% better than non-K.
This would justify the doubling of power usage for K and I'd be fine with it.

But then again, I wasn't able to find any benchmarks where the performance difference between the two would get anything near 40% argh!
It's not going to be 40% since the power used and thus the clocks depend on how heavy the workload is so the 10900 only running at 2.7 is going to be very rare.
Here you can see some decent differences when both are at stock as well as the non-k being faster than the K model at stock with the non-k always running at max turbo.
https://www.techpowerup.com/review/intel-core-i9-10900/6.html
 
Feb 5, 2021
38
0
30
It's not going to be 40% since the power used and thus the clocks depend on how heavy the workload is so the 10900 only running at 2.7 is going to be very rare.
Here you can see some decent differences when both are at stock as well as the non-k being faster than the K model at stock with the non-k always running at max turbo.
https://www.techpowerup.com/review/intel-core-i9-10900/6.html

Based on those charts in your link (another page) the power draw from 10900K would either be around 50w (idle) or around +250w to 400w when working. (single threaded not considered). I wonder what the 125w TDP on CPU specs mean then...
 
Based on those charts in your link (another page) the power draw from 10900K would either be around 50w (idle) or around +250w to 400w when working. (single threaded not considered). I wonder what the 125w TDP on CPU specs mean then...
The page is only giving you the peak, highest number, remember how we said several times that the 10900k will reach up to 250W for 56 sec if a software can use that much?! That's the only thing they are measuring, it's not an average.

The 400W number is while gaming so it includes the power draw from the GPU, all the numbers are for the whole system and not for the CPU alone.

I wonder what the 125w TDP on CPU specs mean then...
If you boot the system or have it running at idle for any other reason for about 1 min and then run the CPU for the full 250W then immediately after the 56sec boost you will already be at only 150W average if the workload keeps running for another minute it will be restricted to 125W so the average can only decrease from the already low 150W.
 

Karadjgne

Titan
Ambassador
TDP is Thermal Design Power. That's wattage the cpu puts out as heat. Because the heat in watts is so close to the power used, TDP was informally adopted as power draw as well.

To that end, the power draw in watts as a TDP stated number refers to the amount of power draw at base clocks, no boost or turbo under a 100% cpu load using all cores.

In affect, TDP is P1. So a 10900k after a 56 second Tau at P2 boost levels should drop to 125w P1/TDP. A 10900 should drop to 60w P1/TDP.

That's assuming the motherboard bios is using Intel guidelines for power limits. Only Asus does. Gigabyte, MSI, ASR all have changed the power limits, do not operate within Intel guidelines. This was done for the sole purpose of claiming the motherboard gets better performance. A sales gimmick.

https://www.gamersnexus.net/guides/...erboards-with-default-settings-for-your-build
 
Feb 5, 2021
38
0
30
The page is only giving you the peak, highest number, remember how we said several times that the 10900k will reach up to 250W for 56 sec if a software can use that much?! That's the only thing they are measuring, it's not an average.

The 400W number is while gaming so it includes the power draw from the GPU, all the numbers are for the whole system and not for the CPU alone.


If you boot the system or have it running at idle for any other reason for about 1 min and then run the CPU for the full 250W then immediately after the 56sec boost you will already be at only 150W average if the workload keeps running for another minute it will be restricted to 125W so the average can only decrease from the already low 150W.

Oh quite right. I missed the whole system label.
I have decided on the 10900K. I was going to pickup a superNOVA 1000w psu for the new parts but now I'm not sure if I'll need it. Do you think my own 750w is enough for this system?

10900K
RTX 3080
64GB
HDD, SSD and all the other stuff included.

Also do you think a I should also pickup a cooler for the 10900K ? I've been running my 4790K on stock cooler for +5 years doing heavy work.

TDP is Thermal Design Power. That's wattage the cpu puts out as heat. Because the heat in watts is so close to the power used, TDP was informally adopted as power draw as well.

To that end, the power draw in watts as a TDP stated number refers to the amount of power draw at base clocks, no boost or turbo under a 100% cpu load using all cores.

In affect, TDP is P1. So a 10900k after a 56 second Tau at P2 boost levels should drop to 125w P1/TDP. A 10900 should drop to 60w P1/TDP.

That's assuming the motherboard bios is using Intel guidelines for power limits. Only Asus does. Gigabyte, MSI, ASR all have changed the power limits, do not operate within Intel guidelines. This was done for the sole purpose of claiming the motherboard gets better performance. A sales gimmick.

https://www.gamersnexus.net/guides/...erboards-with-default-settings-for-your-build

Excellent point sir!
I'm picking up B460 HD3 for the 10900K. I don't do OC or any tweaking so I wanted to save some $ on that end. But now that you said only Asus respects those guidelines, do you think I should pickup a cheap Asus board or B460 HD3 is fine?
 
Oh quite right. I missed the whole system label.
I have decided on the 10900K. I was going to pickup a superNOVA 1000w psu for the new parts but now I'm not sure if I'll need it. Do you think my own 750w is enough for this system?
Especially since you don't O/C I would say yes, way more than enough, although a PSU from the haswell era is quite old but if you trust it...
You will need a cooler with the 10900k since it doesn't come with one anyway, I don't know if the stock intel cooler from the 4790K would be enough or will even fit.
 
Feb 5, 2021
38
0
30
All the talk about how these turbo/boost up (relates to power draw) depends on this, what cooler do you use. What budget for a cooler and what case do you have?

Get the K or KF model, with a good cooling setup, and then, never be forced to wonder afterwards if you should have. :)

Especially since you don't O/C I would say yes, way more than enough, although a PSU from the haswell era is quite old but if you trust it...
You will need a cooler with the 10900k since it doesn't come with one anyway, I don't know if the stock intel cooler from the 4790K would be enough or will even fit.


Thanks guys! :)
I don't have budget issues for cooler, but I also don't care about the looks of it or minor differences so I'd prefer to get one that's cheap but also would put me on the safe side. What do you recommend?

Also side question,
Does the fastest available Gen4 NVME SSD perform noticeably slower in Gen3 slot or difference is not noticeable in terms of percentage? I watched 3080 benchmarked at Gen4 vs Gen3 and difference was usually around 1% only.
 

Phaaze88

Titan
Ambassador
So even with the, well an, intel stock cooler you will get a sustained 3.7-3.9Ghz which is the rated base clock, on the slowest core at 100degrees, but still.
That's not thermal throttling as he wrongfully claims since that's the expected behavior. Being able to get even higher clocks with better cooling is called overclocking but the lack of this is not thermal throttle.

Even the hyper 212 that he shows, with all default settings as the OP is going to use is going to be way enough, linus got 69degrees at 4Ghz after the boost duration so that would be decent.