"We started with inference in the MI300 generation," he says, and "we will systematically build out training capability in MI355, both on the silicon as well as the software side.
"It's all culminating in our MI450 generation, which we're launching next year, where that is, for us, our no asterisk generation, where we believe we are targeting having leadership performance across the board, any sort of AI workload, be it training or inference."
Indeed, Norrod reckons MI450 will be AMD's EPYC moment for GPUs, referencing the success of the company's third-gen Zen CPU architecture for servers. "The third generation of EPYC CPUs is the one where we targeted having no excuses," Norrod says, "where it was the best CPU for any x86 workload period full stop. We're trying to view and plan for MI450 to be the same. It will be, we believe, and we are planning for it to be the best training, inference, distributed inference, reinforcement learning solution available on the market."
If you're wondering whether that just means beating Nvidia's current AI GPUs, Norrod is clear that MI450 will best not just the current Nvidia Blackwell architecture, but its next-gen Rubin GPUs, too. "We're going to be there at the same time as Vera Rubin, and we're going to be there with that part that's fully performing," he says.
Now, the MI450 isn't a GPU that's ever going to be used for gaming. However, AMD has said it plans to unify it AI and gaming GPUs into a single architecture known as UDNA and slated to succeed AMD's current RDNA cards, including the Radeon RX 9070 XT.
The biggest gaming news, reviews and hardware deals
Keep up to date with the most important stories and the best deals, as picked by the PC Gamer team.
Put it this way, if UDNA has the edge on Nvidia Rubin for AI, you'd have to give it a very good shot at taking the fight to RTX Rubin (or however th next-gen architecture is denoted) in the gaming GPU market. Norrod also says AMD wants to significantly up its market share in graphics cards.
"We aspire to be a meaningful portion of the market. What that means is if you're not strongly into the double-digit percentage, say, 20% of the market, you're not a meaningful player, and we certainly aspire to get to be a meaningful player as an intermediate step and then, of course, continue to grow over time."
Will AMD's next-gen gaming GPUs benefit from the investment the company is making in AI? (Image credit: Future)
Again, that's for AI GPUs. But with the technological unification will surely come some broader parallels for AMD's GPU tech. And gven AMD's gaming GPU share has fallen to just 6% in the last available figures, 20% would be a huge step forward, that's for sure.
As for when UDNA will arrive, Norrod says MI450 will be launching "about a year from now." Gaming GPUs have their own, separate road maps, but whether they go by the name UDNA or AMD sticks with it RDNA nomenclature for gaming and calls it RDNA 5, we're pretty much desperate to see AMD's next-gen gaming GPUs appear some time in 2026.
If AMD can get in early with a new gaming architecture that benefits from the huge investment the company is surely putting in AI GPUs, we could finally have some serious competition on the PC.
Jeremy has been writing about technology and PCs since the 90nm Netburst era (Google it!) and enjoys nothing more than a serious dissertation on the finer points of monitor input lag and overshoot followed by a forensic examination of advanced lithography. Or maybe he just likes machines that go “ping!” He also has a thing for tennis and cars.
You must confirm your public display name before commenting
Please logout and then login again, you will then be prompted to enter your display name.
