Over the past few months there has been a constant drip-drip-drip of rumors about the upcoming GPU from AMD and Nvidia. The most surprising element of this rumor, at least towards the fence Nvidia, is the dramatic increase in power consumption. Nvidia’s next-generation Ada Loveless RTX 40-Series is reportedly in need Double The power of its current ampere card. Those numbers have dropped somewhat since then, but the latest rumors still indicate a 600W flagship GPU.
It’s hard to understand. While on the rumor mill’s AMD side, things have gotten much quieter. We know the company is moving to a chiplet-based GPU design for the first time, but will AMD also crank up the power level? According to an AMD engineer in a recent interview, yes, it will. By how much, he did not say, only that it is a new trend.
Sam Nafziger of AMD shared insights in an interview with Tom Hardware. He has been with the company for 16 years and is a Senior Vice President, Corporate Fellow and Product Technology Architect. He has been working on the company’s chiplet design for several years now, which was previously reserved for its Jane, Epic and Threadripper CPUs. With RDNA3, that design will now be applied to the GPU for the first time. Nafziger explained that it is not a binary question whether a chiplet should be made energy efficient Or More powerful. Although these two goals are at odds with each other, Nafziger explains that improving efficiency gives AMD more headroom to increase efficiency.
When Naffziger was asked if RDNA3 would follow Nvidia’s rumored strategy to increase energy levels, he replied that AMD has no choice but to follow the same path. “Performance is king, but even if our designs are more energy-efficient, that doesn’t mean that if you’re competing in the same thing, you don’t increase the energy level. It’s just that they have to be pushed much higher than us.” He was further asked what was the main motivation for graphics cards that suddenly needed a lot more power, and he said that it was due to the growing computing demand and the added power of compressed gains in node progress.
“It’s actually the fundamentals of physics that are driving it,” he said. “The demand for gaming and computer performance, if anything, is just accelerating, and at the same time, the underlying process technology is slowing down dramatically – and the rate of improvement is increasing. So the power levels will continue to rise. We’ve got a multi-year roadmap, but the trend is there. “
AMD hasn’t released much about RDNA3 yet, but it does put a few numbers in writing. One of them is that it expects more than 50 percent improvement in performance-per-watt. It’s a huge jump, but it’s achieved in the change of the previous generation. It was going from Polaris / Vega to RDNA, then to RDNA2. However, they date back to the days before GCN 14nm for Polaris and TSMC 7nm for RDNA. AMD will tap TSMC’s 5nm node for RDNA3, but power requirements are not yet known.
All that has been reported so far is that the company’s RDNA3 GPUs will have a seven-chiplet design. The company announced that it had “rebuilt” its computer unit. We also know that it will offer a “next-gen” version of its Infinity Cache. Still, Nafzigar was clear that although it would increase energy levels, it probably wouldn’t go to the same extreme as Nvidia. It boasts of its recent superiority over the company’s Ampere GPU. Clearly, this is a mantle AMD interested in retaining in the future.