Quote from: Pleh on January 24, 2023, 12:11:33We need proof first, they had 4 generations to get to a starting point.... The total combined cache is on the avg side for todays standards but the 8 channel 64gb ram limit seems like big potential. AV1 support which is good but no info on h265 decoding which is a staple of Intel and Apple M. We also still dont know how well windows handles drivers and games on these chips yet and its unlikely they will be optimized for them any time soon. Also unsure if Adreno based GPU will even be competitive if it intends to compete w/ M series, RDNA2/3, RTX3050/3060 like performance so external GPU support will probably be pretty valuable for this chip.

H.265 support is obvious, Snapdragons have had that for years.

Where does it say anything about "8 channel" memory?

What a terrible name "8cx" is. What inebriated monkey in Marketing thought of that?

8cx = 8 Crap eXtreme

We need proof first, they had 4 generations to get to a starting point.... The total combined cache is on the avg side for todays standards but the 8 channel 64gb ram limit seems like big potential. AV1 support which is good but no info on h265 decoding which is a staple of Intel and Apple M. We also still dont know how well windows handles drivers and games on these chips yet and its unlikely they will be optimized for them any time soon. Also unsure if Adreno based GPU will even be competitive if it intends to compete w/ M series, RDNA2/3, RTX3050/3060 like performance so external GPU support will probably be pretty valuable for this chip.

I wonder if there's a reason for going tri-cluster instead of the 12-core cluster DynamIQ now allows.

I think the price of the laptop might be the most important factor here. Sure it might beat the m3 or whatever, but if the price is higher or the same as the macbook air, people will buy the macbook air

When they will starts to introduce a single memory interface (removing DRAM) they will have a better APU.
DRAM is a memory from the past that is still used nowadays when must be abandoned for a single shared memory based on giant cache like Mi300 that showed how performances can increase easily without useless DRAM.
Think a 8GB of HBM3 (like for 8905G GH AMD APU) used for 8cxgen4 removing 4200Mhz LPDDR5X how can increase in performances? At least 3x!

The question is if this feature the Nuvia based cores (Oryon) or if this feature regular ARM Cortex?

Like, Qualcomm is able to make this chip since more than a year ago, when ARM released the Cortex X2.