According to reports from domestic media on March 24th, NVIDIA has informed its partners that the supply of existing RTX 40 series GPUs will be significantly reduced to pave the way for the upcoming RTX 50 series graphics cards.
It's been reported that as the RTX 50 "Blackwell" architecture GPU is set to launch in the latter half of 2024, NVIDIA has notified its partners that the shipment volume of GeForce RTX 40 "Ada" architecture GPUs will be greatly reduced.
The report also suggests that in the Korean and Asia-Pacific markets, the prices of RTX 40 series have already begun to decline. It is expected that prices in the US and EU markets will also start to drop in the coming months.
The latest buzz reveals that the powerhouse GB202 chip of the RTX 50 series is set to utilize the TSMC 4NP process, identical to that of the B100/B200, steering clear of TSMC's 3nm tech. However, it's crucial to note that neither the 4N nor the 4NP processes are truly 4nm as they might seem at first glance. Essentially, these are advanced versions of the 5nm process, custom-tailored for Nvidia, offering a bit more punch than the standard 5nm tech.
Furthermore, a significant boost in the L1 cache for the GB202 is on the cards. While the exact capacity remains under wraps, it's guaranteed to far exceed the mere 128KB found in its predecessors, the GA102 and AD102. This enhancement means we can expect a notable jump in the throughput of each SM array.
Moreover, the RTX 50 series is also poised to introduce PCIe 5.0 and DisplayPort 2.1, among other features.
