While the MX450 (Turing architecture) is getting old, NVIDIA would replace it with an MX550 based on Ampere.
MX550: 14% more performance than the MX450?
The information comes straight from the ITHome site which found and shared a result under Geekbench 5 with what seems to be an MX550.
Indeed, a laptop from Lenovo appeared on Geekbench, using an Alder Lake P named i7-1260P (4 performance cores and 8 efficiency cores ), accompanied by an NVIDIA card with 16 Compute Units and only 2 GB of video memory. According to the site, we would be in the presence of the MX550 equipped with the Ampere architecture that would replace the old MX450 still based on Turing. With 16 Compute Units (or SM at NVIDIA), this card would have 2048 FP32 cores against 896 for the MX450, an increase of over 120%!
Of course, as we saw on the RTX 3000, the increase in FP32 cores is not directly reflected on the performance in benchmarks or in games. Under Geekbench, the card reaches 39124 points in OpenCL against 28741 points for the MX450. A performance increase of about 13.6%. This card should be unveiled during CES 2022 which will take place on January 4th. It will compete with the Rembrandt APU from AMD or the next entry-level graphics cards from Intel.
