Graphics cards are practically impossible to purchase these days due to the ongoing chip shortage, but that’s not stopping Nvidia. The company reportedly plans to announce its GeForce RTX 40-series GPUs sometime next year.
We already knew the next series of graphics cards would be based on the Ada Lovelace GPU architecture but now a new DigiTimes (via Wccftech) report claims several Taiwan factories, TSMC and ASE Technology among them, will help Nvidia bring chips to market.
If this report is accurate, Nvidia will be shifting strategies for its RTX 40-series cards. The company currently taps Samsung’s 8-nanometer fab process to make its GeForce RTX 30-series GPUs, and uses TSMC’s 7nm node for data center chips. Not only would Nvidia be moving to a more advanced process, but it would rely entirely on TSMC’s 5nm tech for its upcoming product lines.
Speaking of which, Nvidia is expected to launch two new GPU architectures: Ada Lovelace for its GeForce RTX 40 GPUs built for gaming and Hopper for data centers and other enterprise-related applications.
While recent reports suggest the RTX 40-series chips will arrive in 2022, Nvidia is also expected to release refreshed RTX 30 processors next year in the form of new variants of existing 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 without giving them room to breathe before the RTX 40 cards arrive to steal the show, so we’d put our chips on Ada Lovelace and Hopper launching in late 2022 at the earliest.
When they do arrive, the RTX 40 chips are rumored to be capable of 64 teraflops (TFLOPs) of power. For reference, the Xbox Series X, the most powerful console of this latest generation, has 12 TFLOPs of computing power while the PS5 reaches 10.3 TFLOPs. Oh, and the current top-of-the-line RTX 3080 chip gets up to only 34 TFLOPs, about half of what the RTX 40 cards are rumored to hit. Along with the smaller node size (5nm down from 8nm), Ada is said to be packing up to 18,432 CUDA cores, up from 8,704 in the RTX 3080. According to Tom’s Hardware, this could result in up to 71% faster performance over chips based on the Ampere architecture.
Nvidia’s GeForce RTX 30 are widely considered some of the best graphics cards for gamers and creative pros alike, with the only direct competition coming from AMD and its Radeon cards. But none of that has mattered for the past few years because the chip shortage caused by the coronavirus pandemic has brought GPU production to a screeching halt. Unless you buy a pre-built PC or a gaming laptop, getting your hands on a new GPU requires either a lot of patience or loads of money.
At this point, I can’t help but think that any performance gains offered by Nvidia’s upcoming RTX 40-series cards will be overshadowed by the dread of figuring out how to buy one.