My very rough estimates of 30 series laptop performance

[Mnky313]

Disclaimer:  THS IS ALL SPECULATION and is probably not going to reflect actual mobile 30 series performance so don't expect this to be accurate.

So using the info I can find an math I decided to try and get a rough guess at 30 series laptop performance.

I used the same TDP for the current 20 series cards (highest for each class, I know this skews the data a bit but I don't have specific numbers for each TDP of maxq/p cards.)
All of the clock speed info was gathered from notebookcheck and the Time Spy/Firestrike results are from the UL results page.

 

Obviously a lot of these results are from cards that are overclocked and even stock ones will run higher than boost clock when properly cooled and running below TDP.
There are also plenty of factors that aren't being counted like the maturity of the architecture and how hard nvidia is pushing to get these clocks among many many other things.

 

Here are my results (yellow is all the info calculated base off of other provided values)

There are some weird things I noticed like how 3070q beats 3080q in Time Spy (this is because the TDP difference between 320W and 90W is much larger than 220W and 80W)
This is similar to my guess before of the 3070 max-p out performing the 3080 max-q (just like how the 2070s p beats the 2080s q) as well as both the 3080 q and 3070 q being beaten by the 2080s desktop and p. with the 3070 p being faster than the 2080s p by a bit and the 3080p obviously being the fastest.

Other things to note, I used the TDPs for previous mobile 20 series cards. It's very possible nvidia could raise these numbers allowing for more power draw (it's especially likely they raise the max-p TDP of 3080 to reflect the much higher power draw of the desktop card.
Here it is with higher TDPs:

 

There are also several things that impact there numbers that nvidia could completely change so again, DO NOT EXPECT THIS TO BE ACCURATE and I only did it to see if nvidia left TDPs alone how would the mobile 30 series stack up.

Feedback is encouraged, I would love to see what you guys have to say about these numbers and how else I could try and make them more accurate. (I know Time Spy and Firestrike Ultra are far from a complete collection showing real world performance but it's what I had easily available and without gathering results from hundreds of benchmarks individually it's hard to get something that's not just 1 card from 1 person's system running a handful of games...)

 

here's the excel sheet as well if you wanted to manually enter new TDPs (you should only need to change TDPs, everything else changes automatically)

30 Series Mobile Estimates.xlsx

[lexusgamer05]

Wow, looks like you put a lot of work into this. Looking really good.

 

I'd like to compare your calculations to real 3070/3080 Max-Q/P results when they launch, that'll be interesting to see

[Mnky313]

The latest WAN show talked about the rumours from TechpowerUP and noticed that these numbers might not be as far off as even I thought. Though TechpowerUP does have higher performance than these estimates.

Also to touch on Linus's quote: "it could even get to the point, in fact it probably will, where you could have like a chonkier laptop design with an RTX 3070 MAX-P that will be more performant than a thin laptop with a 3080 MAX-Q even though the thick one is cheaper."

 

this has already pretty much happened, the difference between a 2080 Super Max-Q and a 2070 Super Max-Q is barely any... Let alone the 2070 Super Max-P which is actually faster...