Graphics playing cards are virtually unimaginable to buy as of late because of the ongoing chip scarcity, however that’s not stopping Nvidia. The corporate reportedly plans to announce its GeForce RTX 40-series GPUs someday subsequent 12 months.
We already knew the following collection of graphics playing cards can be primarily based on the Ada Lovelace GPU structure however now a brand new DigiTimes (through Wccftech) report claims a number of Taiwan factories, TSMC and ASE Know-how amongst them, will assist Nvidia carry chips to market.
If this report is correct, Nvidia will likely be shifting methods for its RTX 40-series playing cards. The corporate at present faucets Samsung’s 8-nanometer fab course of to make its GeForce RTX 30-series GPUs, and makes use of TSMC’s 7nm node for knowledge middle chips. Not solely would Nvidia be shifting to a extra superior course of, however it will rely completely on TSMC’s 5nm tech for its upcoming product traces.
Talking of which, Nvidia is anticipated to launch two new GPU architectures: Ada Lovelace for its GeForce RTX 40 GPUs constructed for gaming and Hopper for knowledge facilities and different enterprise-related purposes.
Whereas latest studies counsel the RTX 40-series chips will arrive in 2022, Nvidia can be anticipated to launch refreshed RTX 30 processors subsequent 12 months within the type of new variants of current chips (12GB RTX 3080, 16GB RTX 3070 TI, and 12GB RTX 3060 are rumored). It wouldn’t make sense to launch new 30-series chips with out giving them room to breathe earlier than the RTX 40 playing cards arrive to steal the present, so we’d put our chips on Ada Lovelace and Hopper launching in late 2022 on the earliest.
Once they do arrive, the RTX 40 chips are rumored to be able to 64 teraflops (TFLOPs) of energy. For reference, the Xbox Sequence X, probably the most highly effective console of this newest technology, has 12 TFLOPs of computing energy whereas the PS5 reaches 10.3 TFLOPs. Oh, and the present top-of-the-line RTX 3080 chip will get as much as solely 34 TFLOPs, about half of what the RTX 40 playing cards are rumored to hit. Together with the smaller node measurement (5nm down from 8nm), Ada is claimed to be packing as much as 18,432 CUDA cores, up from 8,704 within the RTX 3080. In accordance with Tom’s {Hardware}, this might end in as much as 71% sooner efficiency over chips primarily based on the Ampere structure.
Nvidia’s GeForce RTX 30 are extensively thought-about a few of the finest graphics playing cards for avid gamers and artistic execs alike, with the one direct competitors coming from AMD and its Radeon playing cards. However none of that has mattered for the previous few years as a result of the chip scarcity brought on by the coronavirus pandemic has introduced GPU manufacturing to a screeching halt. Until you purchase a pre-built PC or a gaming laptop computer, getting your palms on a brand new GPU requires both plenty of persistence or a great deal of cash.
At this level, I can’t assist however suppose that any efficiency positive factors supplied by Nvidia’s upcoming RTX 40-series playing cards will likely be overshadowed by the dread of determining the right way to purchase one.