Meanwhile M2 at this point is already 1.5 years old. And btw end of 2024 / early 2025 M4 will come out. And will wipe the floor with the puny Intel and AMD iGPUs.
Even Qualcomm’s Adreno desktop chipset in upcoming X elite is hell of a lot faster than 780m
That’s a lot of Europeans
but not Brits
So it mistook 760M as 780M.
Pretty impressive that Intel is already competing with AMD’s best mobile IGP. Intel would’ve even been ahead if they actually brought the 192 EU Meteor Lake IGP die like they originally marketed. We never got a reason why it went from 192 EU to 128 EU, but there was that rumor that Intel was supposed to use some N3B for MTL but that clearly didnt happen.
Really excited for this competition. Hopefully AMD brings some good RDNA 3.5 designs for 2024, because Intel will probably bring out a 192 EU Battlemage on N3E die for Arrow Lake mobile late in the year.
I’m guessing the gains from tsmc 3nm were not as good as everyone hoped. We see this in the new Apple socs. Intel probably couldn’t get power draw low enough with n3b to make it work in a portable environment. The rumors so far are that this chip is 25-35w. It’s already pushing the upper limits of what is ok, at least for hand helds.
Trading blows with old M2 as the M3 is already out? Is that supposed to be an accomplishment?
Pretty sure a intel igpu in haswell is better than apple anything cpu. because you don’t have to forced into a trash upgrade path and software set. :/ rant
intel is easily the superior option here but not for what you think. It has av1 encoding, and is lower power. we are going to get some meteor lake soldered on board options for mini pc’s and mini nas systems. 5W idle 20~25w boxes that can route 10g/run the nas/run pfsense/encode av1 all in one box.
trash upgrade path and software set
What software set are you being forced into on Mac?
Apples M2 is gonna see a lot of competition next year.
Chip released mid-2022 going to see a lot of competition in 2024 isn’t really saying a lot.
Hope this new iGP sorts out some issues with asymetric memory configs. In my experience with the iris xe 80CU version, going from 4gb soldered + 4gb DIMM to 4gb soldered + 16GB DIMM meant a non-trivial loss of performance. It seems the flex memory (8GB dual channel, 12GB single channel) config doesnt play well with how the GPU uses the memory channels.
Its a shame because its a pretty capable iGP but it is very prone to OEMs handicapping it with asinine soldered RAM configs that inhibit it from drawing all its potential.