News I tested Intel's Meteor Lake CPUs on AI workloads: AMD's chips sometimes beat them

[rluker5]

Nice try. 165U, 155U, 135U, and 125U all have a max turbo TDP of 57W, and maximum assured power of 28W (what they used to call "configurable TDP-up"). The "U" chips from both companies can be directly compared. You have to check what specific devices are running out of the box, can be changed to, and are capable of cooling, but these are obviously the same class of chip, except for Intel kneecapping the graphics and potentially using more power.

Now Intel has two more SKUs: the Core Ultra 7 164U and Core Ultra 5 134U. These are probably the models that have LPDDR5/x memory integrated on the package. They have a 9W TDP and 15W cTDP-up, but still have a max turbo TDP of 30W.

Typically the H versions of MTL would run at 28w sustained and the U series at 15w sustained. For short term loads of what has typically been 28 seconds or less they can run more. There can also be exceptions like in mini pcs where higher TDPs are chosen.

Regardless, if one checked the TDP of a MTL chip and it were a chip made for 28w, like an H series, and it were running at 28w then it would be comparable to an AMD chip designed for 28w running at a TDP of 28w.

So long as the intended and actual TDPs are the same then it is a fair comparison.

If you were to compare a 28w U series AMD chip running at a TDP of 28w to a 15w U series MTL chip running at 15w TDP then it wouldn't be a fair comparison, even though both chips were U series.

 

[bit_user]

Typically the H versions of MTL would run at 28w sustained and the U series at 15w sustained. For short term loads of what has typically been 28 seconds or less they can run more. There can also be exceptions like in mini pcs where higher TDPs are chosen.

Regardless, if one checked the TDP of a MTL chip and it were a chip made for 28w, like an H series, and it were running at 28w then it would be comparable to an AMD chip designed for 28w running at a TDP of 28w.

So long as the intended and actual TDPs are the same then it is a fair comparison.

If you were to compare a 28w U series AMD chip running at a TDP of 28w to a 15w U series MTL chip running at 15w TDP then it wouldn't be a fair comparison, even though both chips were U series.

It's too bad the time axis isn't labelled on these plots, because the multithreaded tests in Phoronix' testing typically look like this:

​

Source: https://www.phoronix.com/review/intel-core-ultra-7-155h-linux​

​

The Core Ultra 7 155H is typically plateauing well above 28 W. This is the CPU's self-reported package power - not laptop wall power. NotebookCheck tested the exact same laptop and reported the wall power figures, which are much higher.

Anyway, in test after test, what we see is the "Avg" scores of the Intel laptop measuring a few W above the Ryzen 7 7840U laptop. The exceptions to this seem to be single-threaded tests, where the averages are around 12 W or so.

I know some of these tests can run for minutes, yet none of the plots show the Intel CPU dropping to a lower plateau. So, I assume these plateaus around 36 W or so are at the sustained power value.

 

[rluker5]

It's too bad the time axis isn't labelled on these plots, because the multithreaded tests in Phoronix' testing typically look like this:

​

Source: https://www.phoronix.com/review/intel-core-ultra-7-155h-linux​

​

The Core Ultra 7 155H is typically plateauing well above 28 W. This is the CPU's self-reported package power - not laptop wall power. NotebookCheck tested the exact same laptop and reported the wall power figures, which are much higher.

Anyway, in test after test, what we see is the "Avg" scores of the Intel laptop measuring a few W above the Ryzen 7 7840U laptop. The exceptions to this seem to be single-threaded tests, where the averages are around 12 W or so.

I know some of these tests can run for minutes, yet none of the plots show the Intel CPU dropping to a lower plateau. So, I assume these plateaus around 36 W or so are at the sustained power value.

It looks like the standard long term power limit is being adhered to in Windows:
https://www.ultrabookreview.com/wp-content/uploads/2025/12/stress-cinebenchr15-perf-battery.png
from: https://www.ultrabookreview.com/66513-asus-zenbook-14-ux3405-review/#battery-life
But this is from the pre release bios. The new one improves performance and efficiency -from: https://wccftech.com/intel-core-ult...ce-efficiency-improvements-with-updated-bios/
I really like the monitoring slides included in the ultrabookreview for some example tasks. They provide a lot of details. It is a shame HWinfo does not have numbered scales on their graphs though.

As far as the NotebookCheck review you mentioned: their power limits were "55/45 watts" for the Acer and "50/33 watts" for the Asus. I'm guessing this is some factory configured short term/long term power limit maximum and the TDP may have been relaxed for pre release review units with different pre release bioses. Maybe some of the current off the shelf units are capable of higher than standard TDPs? I wish more of the mobile devices I've owned were, but so far only a Dell Venue 11 pro has been and that had it's sustained TDP raised from 4.5w to 6w.

Either way the TDPs are factory configurable and the TDP used should be taken into account and equalized for comparisons of performance, power consumption and efficiency.

But to be honest I do not know if the internal power measuring tools are better in Windows or Linux. I'm just familiar with the ones in Windows.

Also, since it seems Windows has some say as to which tiles get used for what, there may be significant variances in performance and power consumption depending on the Windows power plan settings used.