10875H testing conclusions

[Light-Yagami]

Hello everyone,

 

In the past months, I've posted multiple topics on this chip, because I got it with my Dell XPS 9700. I guess this also applies to similar systems. I hope you enjoy this read, I mostly do these so I have somewhere to look up stuff when I forget details. Enjoy.

 

What 10875H is good at: 

1. Idle power consumption: When simply browsing the web or writing a document, doing simple tasks that do not require video decode or number crunching, this CPU consumes less than 5W , usually between 2 and 4W. If the rest of the system is power efficient as well, this can result in system power usage of under 8 watts, meaning 10+ hour battery life on systems with big batteries. (but do not expect average system power of less than 10W - because nobody edits word documents all day)

 

2. Video decode compared to Nvidia: Chip needs around 4-7W to decode a 1080p video, most of that budget used for iGPU. Nvidia (rtx 2060) requires around 15W to do the same job, or even more depending on the video. Margins increase with resolution.

 

3. Frequency response up to 30W: It only looses around 25% performance when on battery power (3200pts vs 4100pts in CB R20), because it usually doesn't need to go above 25W to do most tasks. Of course, opening heavy programs or doing multiple things at once will see produce high spikes in power, but that's not the average. You will find it hard to push this CPU above 15W average for the most part, unless you're doing more serious work. CPU reaches 60% of its capacity at 27W, which is the amount of power to sustain base clock of 2.3ghz at full load. to go another 1.9 ghz higher, the claimed 4.2 all core turbo, you need another 83W (115W in total). 

 

What 10875H is bad at:

 

1: High performance applications: As soon as you start pushing it, demanding more from the chip, it looses efficiency. To gap the performance difference between it and AMD chips, Intel has mad this CPU draw high amounts of power that very little cooling systems can sustain. My Dell XPS has a liquid metal mod, which completely eliminates any issues - but it's important to say that drawing 75W sustained for video compression and alike just to stay in 10% range of 4000 AMD series is.. a dire last resort attempt at squeezing as much from the CPU as possible. (CPU is happy to draw more than 110W is you're able to cool it, and scales as you would expect from a desktop version chip, with better efficiency. It is a better bin after all)

 

2: Disabled voltage tuning: Intel is pushing too much voltage through this chip, and there's no way to prevent it. There is also aggressive stepping regarding how much power you need to achieve a certain frequency. This means that sometimes, just to go a few hundred mhz higher, you need a substantial amount of power, even though that is most likely not needed. See how the graph breaks its continuity in my frequency response testing.

 

Graph is for CB R23 testing. The chip usually didn't draw the pre-set amount of power, so don't pay too much attention to that. It's just to show that sometimes, I needed to add much more power just to get it to run faster, because there's an aggressive frequency stepping present. If I'm mistaken here, feel free to correct me.

 

3. Does NOT achieve frequencies that Intel claims it does: Intel claims up to 5.1 on a single core, but even with my liquid metal mod, I've never seen it go above 4.9ghz. Even in single core testing, it never went above 4.87ghz, with temperatures in mid 60s - way lower than Intel requires for turbo boost to engage the extra few hundred mhz. I believe this to be false marketing.

 

4. Weak iGPU: Apart from video decoding and quick sync applications, the GPU is worthless. Nothing more to say. It is not capable of driving a 4k display efficiently. Reducing screen resolution to 1080p sees massive power reduction for the iGPU, upwards of 3 watts for 1080p video playback. It's completely fine if you don't do watch video with higher screen resolution though.

 

My gripes with it:

 

Intel states the base clock so low it's virtually impossible for this chip to run out of its operating range. At 45W, this chip runs at around 3.3ghz or higher at full load (Aida64, CB Rxx). Intel has protected themselves so much that this chip can perfrom more than 30% differently from machine to machine and is still within the operating range. That is just cowardly from Intel.

 

Conclussion: 

 

Intel 10th gen is still very good in terms of single core operations, and will feel just as snappy as the latest from AMD - but it will lack behind in multicore and high power usecases. That's why - if you have to buy a 10th gen CPU, only buy a 10875H or 10885H equiped laptops. They are the only one who will stand the test of time, for now.

 

Thanks for reading.

 

 

 

 

[tikker]

22 minutes ago, Light-Yagami said:

3. Does NOT achieve frequencies that Intel claims it does: Intel claims up to 5.1 on a single core, but even with my liquid metal mod, I've never seen it go above 4.9ghz. Even in single core testing, it never went above 4.87ghz, with temperatures in mid 60s - way lower than Intel requires for turbo boost to engage the extra few hundred mhz. I believe this to be false marketing.

The 5.1 GHz is only attainable through their "Thermal Velocity Boost", which for 10th gen mobile CPUs means it has to be below 65 C for it to boost that high. Temperatures mid-60s will thus likely mean in your case you never see this speed. It's probably more meant for a very short intense burst and not sustained load. I don't think  this is false advertising, but I agree it is extremely vague as nowhere is this clearly documented for the consumer.

[Light-Yagami]

1 hour ago, tikker said:

The 5.1 GHz is only attainable through their "Thermal Velocity Boost", which for 10th gen mobile CPUs means it has to be below 65 C for it to boost that high. Temperatures mid-60s will thus likely mean in your case you never see this speed. It's probably more meant for a very short intense burst and not sustained load. I don't think  this is false advertising, but I agree it is extremely vague as nowhere is this clearly documented for the consumer.

I've checked my HWinfo64. Out of 8 cores, two reached 5.1ghz in the last 24 operating hours. Others have maximum frequencies recorded at 4.9.

 

This is like saying a car has 500 horsepower and you brag about it to everyone, yet it can only produce them in very niche scenarios, rarely enough that you can say never. It's worthless either way. And so close to false advertising it's practically leaning over the line. 

 

Thanks for the reply though. Glad you read the post

[tikker]

38 minutes ago, Light-Yagami said:

I've checked my HWinfo64. Out of 8 cores, two reached 5.1ghz in the last 24 operating hours. Others have maximum frequencies recorded at 4.9.

 

This is like saying a car has 500 horsepower and you brag about it to everyone, yet it can only produce them in very niche scenarios, rarely enough that you can say never. It's worthless either way. And so close to false advertising it's practically leaning over the line. 

 

Thanks for the reply though. Glad you read the post

In laptops the chip will run hot already because of the form factor, so indeed a less common occurence. I guess it gives a small boost when starting a program or something.