While Nvidia’s most recent Ada Lovelace series GPUs provided next-generation power and efficiency when they were introduced in late 2022, a few of their features were rather antiquated. People questioned at the time why the cards lacked DisplayPort 2.1 and next-generation PCIe 5.0, which AMD’s Radeon 7000 series already included. It now seems likely that Nvidia’s future design, code-named Blackwell, will make up for these mistakes.
The most infamous Nvidia news leaker, @kopite7kimi, has disclosed a number of fresh details on Blackwell, including the fact that Nvidia will overtake AMD by implementing DisplayPort 2.1 for Blackwell. This interface was a big talking point for AMD when it introduced its RDNA 3 GPUs last year. AMD said it could support 8K gaming as well as high-refresh 4K gaming at frame rates faster than Nvidia’s outdated DisplayPort 1.4a interface.
Although Nvidia is much behind AMD in incorporating DP 2.1 into its GPUs, considering the dearth of 8K gaming these days, no one seems to care too much. DP 2.1 offers several real advantages over DP 1.4a. Furthermore, even with the growing popularity of ray tracing, high-refresh 4K gaming remains a tiny niche, with only two GPUs capable of attempting it: the RX 4090 and the RX 7900 XTX. This is expected to stay that way until more games support ray tracing. Upscaling is also necessary when ray tracing is used, thus this is still a new frontier for gamers that hasn’t even approached standardization because of system requirements.
It is also predicted that Nvidia will use TSMC’s 3nm technology for Blackwell, as has always the case. Now, though, @kopite7kimi has confirmed the node, marking the first time this leaker has done so. Since TSMC offers a variety of FinFlex 3nm designs, it is unclear which version of TSMC N3 Nvidia is using. Not all of the variants are currently available online, and some are tuned for power, efficiency, or a combination of the two. All of the big tech giants, including Intel for its GPU tiles for Arrow Lake and the subsequent designs, are anticipated to use it in 2024, making this an extremely busy node for TSMC. Though it’s previously rumored that AMD will employ TSMC 4nm for its next-generation Ryzen CPUs, it’s uncertain if AMD will use in RDNA 4.
According to Videocardz, Nvidia is anticipated to introduce a PCI Gen 5 interface to match AMD’s GPUs in addition to implementing DP 2.1 and TSMC N3. Since PCIe 4.0 is still functional in 2023 and hasn’t yet materially hampered the performance of PCs using these SSDs and GPUs, Nvidia’s decision to choose this path looks well-timed. However, such will not be the case after 2024.
The GeForce version of Blackwell is anticipated to launch later in 2024, following the announcement of the HPC version by Nvidia. It’s still unclear if it will happen in 2024 or 2025 because Nvidia has said they’ll launch a year later than normal in 2025. We find it difficult to believe Nvidia’s roadmap won’t change given its unwillingness to give up any ground to AMD in this ongoing turf battle. But Nvidia is also starting to focus more on its AI hardware, so its next-generation products—which may include GeForce—may take a backseat to its data center products.