Why doesn't increasing the power in my UEFI settings increase the performance of my i3-10105F, but it does for my i7-11700F?

[Joshua5684]

This is a follow-up to this post.

 

I found in my UEFI settings that I could increase the power limit, so I did and tested in Death Stranding:

i3-10105F at stock power and at max power—94 FPS

i7-11700F at stock power—103 FPS

i7-11700F at max power—113 FPS

 

I was disappointed to see that increasing the power does not increase performance for the i3 as I was hoping I could do that and flip the i7. I compared them on Intel's database and found that the only seemingly relevant difference is Turbo Boost Max Technology 3.0, but that shouldn't account for the difference because K-series CPUs, such as the i5-10600K, will use more than the default power despite not having TBMT 3.0. Why doesn't increasing the power increase performance? This is an important question because if I were to buy a different CPU, such as an i5-11400, I'd want to know if it would be able to use more power.

[Tetras]

How are you measuring the performance exactly?

 

In the majority of tasks (especially when gaming), the i3 would not be limited by the power limits, whereas the 11700 would be limited by the power limits.

 

Rocket Lake was notoriously power hungry, so the i5/i7 non-K CPUs tend to get a bigger performance boost (and across a wider variety of apps / load levels) from raising the limits than previous gen CPUs.

[PDifolco]

They're all non overclockable chips, so they run at max clocks close to each other, increasing power won't give much if anything

You only got a measly +10% "generational increase"
You may see a real difference only in CPU heavy games that need more than 4 cores and overwhelm the 10105

[Joshua5684]

6 minutes ago, Tetras said:

How are you measuring the performance exactly?

FPS in Death Stranding standing in the same spot for each test.

[porina]

If you look at the actual power consumption compared to the power limit, I'd guess the 10105F isn't hitting that since it only has 4 cores, and the 11700F is mainly because it has more cores. 6700k used around 90W running Prime95, which is worst case. 10105F is on a more efficient process and lower base clock. Combined with non-worst case workloads I can see its power usage being lower.

 

7 minutes ago, Tetras said:

Rocket Lake was notoriously power hungry

Efficiency wise they were comparable to Comet Lake and earlier (Skylake), at similar operation points on the curve. The only time Rocket Lake really chewed power was if you used AVX-512. However it also did a LOT more work so efficiency remained about the same. It was the only mainstream desktop CPU at the time to have it, and to this day remains the only Intel mainstream desktop CPU to have ever officially supported it.

[Joshua5684]

20 minutes ago, PDifolco said:

You may see a real difference only in CPU heavy games that need more than 4 cores and overwhelm the 10105

Death Stranding not only runs my i3 at 100%, but it also runs my i7 at max power at 100%, which is so strange. It has double the cores and faster cores, but only increases performance by 20%. How is that possible while still being CPU-limited?

[Haswellx86]

The clock frequency is the real deal. Compare the clocks between stock and power limits lifted. Also report the power usage.

[Joshua5684]

18 minutes ago, Haswellx86 said:

The clock frequency is the real deal. Compare the clocks between stock and power limits lifted. Also report the power usage.

i3 is 4.2 all core at just ~40 even when power limit is increased. I don't remember the i7 clocks, but it went from ~65 W to ~120 W.  For what it's worth, the i3 will go up to 50 W in Cinebench.

 

It looks like the i7 wasn't worth the upgrade, but I'm still scratching my head as to why doubling cores, doubling power, more than doubling L3 cache and +700 MHz only nets a 20% increase in FPS despite still being at 100% usage.

[PDifolco]

1 hour ago, Joshua5684 said:

Death Stranding not only runs my i3 at 100%, but it also runs my i7 at max power at 100%, which is so strange. It has double the cores and faster cores, but only increases performance by 20%. How is that possible while still being CPU-limited?

That's pretty weird, are you sure the i7 is utilized 100% ?

And are you sure you aren't GPU limited ?

[Joshua5684]

10 minutes ago, PDifolco said:

That's pretty weird, are you sure the i7 is utilized 100% ?

Confirmed with task manager and AMD Adrenaline, yes.

11 minutes ago, PDifolco said:

And are you sure you aren't GPU limited ?

My RX 5600XT usually sits around 70% at 1080p, max settings.

[Tetras]

1 hour ago, Joshua5684 said:

FPS in Death Stranding standing in the same spot for each test.

Gaming isn't the best case for testing lifting the power on these CPUs, but in any case, the rocket lake CPUs need more than 65 watt, so you will see a FPS difference, but the 10th gen i3 does not.

 

1 hour ago, porina said:

Efficiency wise they were comparable to Comet Lake and earlier (Skylake), at similar operation points on the curve. The only time Rocket Lake really chewed power was if you used AVX-512. However it also did a LOT more work so efficiency remained about the same. It was the only mainstream desktop CPU at the time to have it, and to this day remains the only Intel mainstream desktop CPU to have ever officially supported it.

I don't know about efficiency at the same voltage, or in general applications, but in the context of gaming specifically, from what I can recall the Skylake-based i3/i5 CPUs would struggle to break 65 watts in most scenarios, whereas a power unlocked 11400 (or higher) can easily break through 65 watts.

 

Skylake was generally very good on power consumption when gaming, even on the worst case/least efficient 10th-gen CPUs. E.g. a 10400 can use half the power of the (unlocked) 11400, but they weren't doing double the FPS. Though, that was in testing before cards like the 4090/5090 existed, so maybe 11th gen will pull away a lot more now.

[podkall]

They're both 65W CPUs, but one has more cores and higher rated boost clock.

 

So I don't see anything weird here. The faster and higher core count CPU has more assets, so it's more likely to be faster when you lift power targets or limits.

 

 

It's like you have performance limiter on a slow car and a fast car, that's set to like 120mph, and then you turn the limiter off and only fast car goes faster. Maybe the slow car can't go above 120mph in the first place.

 

• It's like if I asked you to run as fast as possible, but not more than 30mph.
• Then if I told you to run as fast as possible, but now you no longer need to be slower than 30mph, and wondered why you're not going any faster.

 

Perhaps get https://www.hwinfo.com/download/, and see if the i3 even draws more than 65W and look at the same thing with the i7.

 

You can use program like Cinebench R23/R24 to really tell these CPUs to go as 100% as they can on all cores, and see if the i3 even needs more power and affects the Cinebench scores.

[Joshua5684]

7 minutes ago, podkall said:

Perhaps get https://www.hwinfo.com/download/, and see if the i3 even draws more than 65W and look at the same thing with the i7.

 

You can use program like Cinebench R23/R24 to really tell these CPUs to go as 100% as they can on all cores, and see if the i3 even needs more power and affects the Cinebench scores.

1 hour ago, Joshua5684 said:

i3 is 4.2 all core at just ~40 even when power limit is increased. I don't remember the i7 clocks, but it went from ~65 W to ~120 W.  For what it's worth, the i3 will go up to 50 W in Cinebench.

My concern is more so why the i7 is still at 100% utilization in Death Stranding despite only increasing performance by 20%. It doubled the cores over the i3.

[Haswellx86]

9 hours ago, Joshua5684 said:

My concern is more so why the i7 is still at 100% utilization in Death Stranding despite only increasing performance by 20%. It doubled the cores over the i3.

Really? I don't know about Death Stranding but your i7 which has 8 cores is always at 100%? That game is this demanding? Are you sure nothing else is running in the background?

 

Would still be nice if you could report the i7 clocks but yeah they should be higher anyways.

 

9 hours ago, Joshua5684 said:

My RX 5600XT usually sits around 70% at 1080p, max settings.

So that's on the i7? And what was it like on the i3?

 

I checked the game in some videos and it is actually quite demanding. Someone's 11700K did go often over 80% usage and it ran at 4.4 GHz (the 11700K should be configured 100 MHz higher than your 11700F). But they were also getting over a hundred FPS on GPUs like the 3080. Your clearly doesn't.

 

Also show us the Cinebench results of both.

[podkall]

11 hours ago, Joshua5684 said:

My concern is more so why the i7 is still at 100% utilization in Death Stranding despite only increasing performance by 20%. It doubled the cores over the i3.

Because it's working.

 

1 hour ago, Haswellx86 said:

Really? I don't know about Death Stranding but your i7 which has 8 cores is always at 100%? That game is this demanding? Are you sure nothing else is running in the background?

 

Would still be nice if you could report the i7 clocks but yeah they should be higher anyways.

The game isn't 2013

 

[Joshua5684]

6 hours ago, Haswellx86 said:

Would still be nice if you could report the i7 clocks but yeah they should be higher anyways.

 

Also show us the Cinebench results of both.

That would require me to switch my CPU again, which I don't really want to cuz I think I've decided not to keep the i7.

[Joshua5684]

5 hours ago, podkall said:

Because it's working.

That is the crux of my question because if everything is working properly, I'm not gonna keep a $180 CPU for +20% FPS. Are you sure it's all working?

[Haswellx86]

8 minutes ago, Joshua5684 said:

That would require me to switch my CPU again, which I don't really want to cuz I think I've decided not to keep the i7.

So you didn't own that CPU and now you are returning it?

8 minutes ago, Joshua5684 said:

That is the crux of my question because if everything is working properly, I'm not gonna keep a $180 CPU for +20% FPS. Are you sure it's all working?

I doubt everything was working fine. At least if we got the Cinebench results we could understand something.

[podkall]

14 minutes ago, Joshua5684 said:

That is the crux of my question because if everything is working properly, I'm not gonna keep a $180 CPU for +20% FPS. Are you sure it's all working?

20% is a lot, did you mean +20? Also you're going to base your pick/choice based off of one game out of thousands that exist, including other programs?

[Joshua5684]

1 hour ago, Haswellx86 said:

So you didn't own that CPU and now you are returning it?

No, I do own it, but if I don't decide to keep it, I'll resell it. It's just that switching CPUs is a hassle and I'm running low on paste. If there are any more tests that you absolutely need me to do to come to a conclusion tho, I will.

1 hour ago, Haswellx86 said:

I doubt everything was working fine. At least if we got the Cinebench results we could understand something.

The i3 got 321 and the i7 got over 600 (don't remember exactly) in Cinebench 2024. The i7 also initially went up to ~120 W then thermal throttled down to ~90 W and sustained it for the rest of the run (I'm using the stock Intel cooler).

[Joshua5684]

42 minutes ago, podkall said:

20% is a lot, did you mean +20?

+20% of 94 FPS is also approximately 113 FPS.

43 minutes ago, podkall said:

Also you're going to base your pick/choice based off of one game out of thousands that exist, including other programs?

Death Stranding is the only program I use that is CPU bottlenecked.

[podkall]

1 hour ago, Joshua5684 said:

+20% of 94 FPS is also approximately 113 FPS.

Technically yes, because 100FPS is 1% per 1 fps

 

1 hour ago, Joshua5684 said:

Death Stranding is the only program I use that is CPU bottlenecked.

100% utilization doesn't always mean bottleneck

[Joshua5684]

1 hour ago, podkall said:

100% utilization doesn't always mean bottleneck

It does when my

20 hours ago, Joshua5684 said:

RX 5600XT usually sits around 70% at 1080p, max settings.

[podkall]

33 minutes ago, Joshua5684 said:

It does when my

Yeah that still doesn't mean it's a bottleneck, or a concern to have.

 

Worse would be neither components hitting 100%, and it happening more games than just Death Stranding.

[Haswellx86]

4 hours ago, Joshua5684 said:

No, I do own it, but if I don't decide to keep it, I'll resell it.

You are really straight up going to resell your upgrade? No, wait hold on.

 

4 hours ago, Joshua5684 said:

The i3 got 321 and the i7 got over 600 (don't remember exactly) in Cinebench 2024. The i7 also initially went up to ~120 W then thermal throttled down to ~90 W and sustained it for the rest of the run (I'm using the stock Intel cooler).

https://www.cpu-monkey.com/en/benchmark-intel_core_i7_11700f-cinebench_2024_multi_core They got 807 and so if your CPU thermal throttles, that is your fault. What were your temps like in the game? It must be thermal throttling as well.

 

https://www.cpu-monkey.com/en/benchmark-intel_core_i7_11700f-cinebench_2024_single_core

https://www.cpu-monkey.com/en/benchmark-intel_core_i3_10105f-cinebench_2024_single_core

The single core is supposed to be 30% higher which is quite a lot.

 

As I said I saw someone running a 11700K with a 3080 and they weren't hitting 100% on the CPU (only 100 MHz higher than 10700F) and their GPU was well totally utilized.

22 hours ago, Joshua5684 said:

My RX 5600XT usually sits around 70% at 1080p, max settings.

Was that with the i3 or the i7? If on the i3 you must be GPU limited on the i7.

 

You must check your OS and background tasks if anything is eating CPU. And your cooling is also to blame. As I said, I wanted to see the clocks of the i7.

 

A 10105F to an 11700F is definitely a good upgrade.