The Snapdragon 865 was fabbed on TSMC’s N7P process (second
gen 7 nm), but Qualcomm switched to Samsung foundries when ordering the
Snapdragon 888. The Snapdragon 888+ is built on the same Samsung 5 nm node,
it’s just a higher-binned part with the Cortex-X1 clocked at 2.995 GHz (up from
2.84 GHz) and the NPU cranked up to 32 TOPS (up from 26 TOPS).
Things will look different next year, says Ice Universe –
the Snapdragon 895 will be fabbed by Samsung at its 4 nm node, while the
Snapdragon 895+ upgrade will be made at TSMC’s 4 nm foundries instead.
Both are refinements of the respective 5 nm nodes of the
foundries, however, the performance advantages and potential power savings of
the new processes have not been publicly announced yet.
Chances are that the Snapdragon 895 and 895+ will follow a
similar timetable as the two 888 versions this year – the vanilla version
launches in December/January, the Plus comes half a year later. If nothing
else, the semiconductor shortage will have (hopefully) cleared up by then.
Qualcomm’s foundry choices last year may have been
restricted by Apple buying a large share of TSMC’s 5 nm supply for both its
Apple A14 and M1 chipsets. Next year, however, the company will reportedly
split the orders and get 4 nm A15 chips for iPhones and 3 nm chips for iPads.
One way or another, there is enough production capacity available at TSMC’s
foundries that even MediaTek ordered 4 nm chipsets (in fact, those may be the
first 4 nm chips to hit the market, but they will be pricey).
Samsung is expected to use its own 4 nm foundry to make the
Exynos 2200, which should be the first chip with an AMD RDNA 2 GPU. Qualcomm
picked Samsung’s 4 nm foundry for its X65 and X62 5G modems, so it’s only
natural for the Snapdragon 895 (which will have an integrated X65) to use the
same process.
0 comments:
Post a Comment