Why it matters: Nvidia’s Ampere’ successor is almost here according to leaks, and it will be hot and hungry. GeForce RTX 4000 series cards, codenamed Ada Lovelace, could consume 50 to 100W more than their predecessors across the board, with flagship models breaking new ground.
Veteran leaker kopite7kimi, who pinned Ampere’s specs weeks ahead of anyone else, has updated its predictions of Ada Lovelace’s power limits. It is important to note that power limits are not the same as TDPs or TBPs; they are an upper bound that only comes into play when overclocking is involved. However, they are still a good guideline for power consumption, just an overestimation.
Our power supplies come to torture with the flagship AD102 with a limit of 800W. One level down, the desktop AD103 has a limit of 450W, the AD104 of 400W and the AD106 of 260W. In the past, the xx106 was used by entry-level GPUs and the xx104 by mid-range GPUs, but Nvidia could change that.
As you’d hope, leaked laptop power budgets are tamer. Kopite reports that the mobile versions of the AD103 and AD104 have limits of 175W, and the AD106 a limit of 140W. However, since they are laptop parts, manufacturers have a lot of leeway to customize them anyway.
Back to the numbers, you’re probably still reeling: 800W. It’s doable with a pair of new PCIe 5.0 power connectors, each capable of handling 600W (if the card had 8-pin power connectors, six would be needed!). But there aren’t many power supplies with two of those ports and that much power on the market.
Cooling also becomes an interesting challenge. But aside from those practical concerns, is an 800W ceiling unreasonable for a GPU with 18,432 CUDA cores (according to the information from the February hack)?
In its fully unlocked state, the AD102 consumes 0.0434 W per CUDA core, which is only marginally more than the 0.0419 W per core consumed by the RTX 3090 Ti. As Steve noted in our RTX 3090 Ti review, that GPU is a power-guzzling monster, but theoretically the AD102 is just as terrible.
And remember, 800W is the (probable) hard limit, not the TDP. An earlier report said that Nvidia tested a 600W TDP for the 4090 that would improve the GPU’s efficiency and leave 200W as overclocking space.