Nvidia GeForce RTX 50 Blackwell GPUs rumored to come back with TSMC 3nm and DisplayPort 2.1 — AMD already presents the usual on the present Radeon RX 7000 sequence
Nvidia’s GeForce RTX 40-series GPUs are already among the many best graphics cards available on the market. Nonetheless, rumors of the next-generation GeForce RTX 50-series (Blackwell) GPUs are already brewing, even when we might not look too eager on this improve path. Our pleasant neighborhood leaker, kopite7kimi, with a stable monitor file, has offered particulars about Ada Lovelace’s successor.
Based on the leaker, the GeForce RTX 50-series graphics card will reportedly function TSMC’s 3nm course of node. It is sensible because the present GeForce RTX 40-series GPUs are on the 5nm course of node, so logically, the following step can be 3nm. With the 3nm course of node, TSMC guarantees a efficiency uplift of as much as 15% over the 5nm course of node on the similar energy degree and variety of transistors.
Relating to energy financial savings, the foundry claims that the 3nm course of node might decrease the facility consumption by as a lot as 30%. However what’s spectacular about TSMC’s 3nm course of node is that the die sizes are estimated to be round 42% smaller than the 5nm course of node. Subsequently, it might be attention-grabbing if the development interprets to Blackwell’s silicon. As an illustration, the AD102, which powers the flagship GeForce RTX 4090, is a huge die that checks in at 609 mm². GA102, which Nvidia used for the prior GeForce RTX 3090 Ti, was on Samsung’s 8nm course of node and measured 628 mm²; due to this fact, the silicon did not get a lot smaller. However this time, Nvidia’s allegedly sticking with TSMC, making the 5nm to 3nm transition extra attention-grabbing.
The GeForce RTX 50-series graphics playing cards purportedly include DisplayPort 2.1 assist. Nvidia is late to the sport since AMD has already embraced DisplayPort 2.1 on the chipmaker’s current Radeon RX 7000-series merchandise, just like the Radeon RX 7900 XTX. Nvidia’s present GeForce RTX 40-series graphics playing cards solely have the DisplayPort 1.4a connection.
AMD’s implementation varies between the corporate’s client and workstation lineups. The Radeon RX 7000 sequence has UHBR13.5 assist, that means entry to 54 Gbit/s of bandwidth; in the meantime, the W7000 sequence has the whole UHBR20 trait for 80 Gbit/s of most bandwidth. Sadly, kopite7kimi was obscure in his remark and did not specify which commonplace Nvidia might use on Blackwell.
Different rumors declare the GeForce RTX 50-series product will function the PCIe 5.0 interface, which is overkill for mainstream graphics playing cards. What’s seemingly is that the Blackwell will proceed to make use of the 16-pin energy connector, though it will most likely rely upon the revised 12v-2×6 model.
This 12v-2×6 commonplace is very anticipated to be a a lot safer interconnect as Nvidia has stayed largely quiet concerning the 12VHPWR meltdown kerfuffle on the expense of its shoppers, aside from usually blaming person error. Opposite to Nvidia’s conclusion and some others within the {industry}, considered one of {many professional} restore technicians who get several damaged RTX 4090 per thirty days expressed the principle root of the trigger is the connector, followed by third-party adapters as a big contributor whereas stressing person error is among the potentialities if not others. It is at all times a good suggestion to be aware of such industry-wide oversight. Will probably be appreciated if Nvidia and others who undertake this commonplace stress on ATX 5.0 spec to make protected third-party cables, extensions and adapters.
At this stage, nothing is for certain about Blackwell SKUs. After all, there are rumblings about more cores or GDDR7 memory on the GeForce RTX 50 sequence, however at this level, take all the pieces with a bucketload of salt.
