It’s still a long time before AMD and Nvidia’s next-gen graphics cards launch, and most details are still sparse. But if I know one thing, it’s that the relative power consumption will increase. RTX 40 (opens in new tab) and rDNA 3 (opens in new tab) The graphics card is set to push the graphics card power consumption to a higher level. While high-end TDPs of 450W or higher will gain pre-launch attention, it’s not immune to what looks increasingly like mid-range and low-end cards.
Tom’s Hardware (opens in new tab) Published an enlightening interview with Sam Naffziger, AMD senior vice president, enterprise researcher, and product technology architect. Naffziger makes some interesting points. One of the key passages states that AMD believes its architecture is more power efficient than Nvidia.
“Performance is king,” Naffziger said, “but even if our designs are more energy efficient, that doesn’t mean you won’t increase power levels if your competitors are doing the same thing. It’s just that they have to push them more than we do. much higher.”
The last point is interesting. “It’s just that they have to push them much higher than us.” If we take this quote at face value, firstly, this pretty much confirms that the RDNA 3 card will have a higher total board power level, and secondly, AMD believes that Nvidia’s card will have a higher power consumption level. It will be interesting to see how the two companies act and react to each other’s designs.
If we recall that AMD is claiming more than 50% performance per watt (which is certainly an upper bound estimate or a SKU-specific estimate), it’s possible, if unlikely, that the RX 7900 XT will end up with twice the 6900 XT. Of course, it depends on how aggressive AMD wants to be. The closer the chip gets to its frequency limit, the effect of power scaling decreases exponentially. Is it worth adding 10% or 15% or 20% more power through the card to extract the last few percent of performance?
Oh, and there’s a lot we didn’t know about RDNA 3 at those internal meetings. We know it will be a chiplet-based architecture, the chips will be fabricated using TSMC’s 5nm node, and will have an optimized Infinity cache, and possibly more.
Back on the topic of power, the days of the 300W flagship seem to be over. RTX 3090 Titanium (opens in new tab) As a taste of the future we can look forward to. PCIe 5.0 power supplies with 16-pin power connectors can support up to 600W, suggesting that GPU power consumption will only increase in the coming years. Let’s quote this quote from Tom’s Hardware’s interview with Naffziger: “The demand for gaming and computing performance, if anything, is accelerating…so power levels will only continue to rise.”
As the world faces complex issues such as energy supply, inflation, and climate change, the trend of ever-higher GPU power consumption is not good. Of course, not all GPUs have ridiculous power levels. There will be demand for high-quality and affordable cards that can push good frame rates, although I hope the performance-at-all-cost mentality doesn’t forget to at least put some emphasis on power efficiency.
We’re going to get great performance from AMD and Nvidia GPUs, but there has to be a limit or resistance from consumers. where does it end? 600w, 800w? Performance is king, but don’t completely ignore everything else.