Nvidia’s next-generation graphics cards are rumoured to be power hogs. We addressed power rumours, possibilities, and likelihoods in depth in a recent “all we know” post about Nvidia Ada Lovelace and the GeForce RTX 40 Series. Kopite7kimi(opens in new tab) released some numbers earlier today that suggest some of our higher estimations were too conservative.
Before going any further, it’s important to understand the difference between graphics card power restrictions and TDPs. The power limit will be set by the RTX 40 Series card’s BIOS and will be the maximum wattage the GPU can run at by design. High-end cards with the greatest coolers may be able to run close to the power limit; they may also provide higher power restrictions and be pushed far beyond by severe overclockers’ tinkering. On the other hand, TDPs for mainstream consumer cards will almost certainly be substantially lower than the GPU’s highest potential power limit.
According to our best understanding, between 132 and 140 SMs are enabled (from a maximum of 144). Previous rumours suggest that RTX 4090 cards will have TDPs of less than 600W. However, according to Kopite7kimi, the underlying GPU will have a power restriction of 800W.
The AD102 GPU’s power restriction, which suggests high TDPs, is thankfully an exception in the Ada Lovelace series. Moving down to AD103, this second-tier GPU will most likely be found in the next GeForce RTX 4080 and other cards. According to today’s leak, the power restriction is substantially reduced to 450W. Current-generation GeForce RTX 3090 and RTX 3080 reference cards feature maximum power constraints of 400W and 370W, respectively, set by the BIOS.