We are only talking about the flagship models of the NVIDIA RTX 4000 series for now, but 600W graphics cards are yet something that many of us simply cannot understand. The latest tweet from a trustworthy NVIDIA leaker kopite7kimi appears to confirm that the next-gen XX90 class GPU is to feature a TGP (total graphics power) of 600W.
One should mention that the cooling solution for next-gen GPUs is designed months in advance, maybe even before the graphics chips are prepared for mass production. Physical features, such as the GPU size and its position on the board, can be planned in advance.
Also, just because the cooler is capable of dissipating 600W of power, it does not mean that such configuration will be used by the GPU by default. The cooling capacity is usually defined as a maximum heat that can be generated by the GPU, for example when power limitations are modified by the user during overclocking.
For board partners, a 600W power limit will surely be a challenge, but it won’t be as complicated as some users might think. There are multiple air-cooled models like ROG STRIX 3090 from ASUS with max TGP already reaching as high as 480W. The graphics card still covers 3-slot and requires three fans, but this is not really that remarkable these days.
AMD graphic cards such as RX 6900XT TOXIC Liquid from Sapphire have a maximum TGP of 400W and already come with an external water cooling. This is the path that NVIDIA might take with their next-gen GPU, with liquid-cooling for its flagship models. AMD has already been doing this for many years now.
According to previous reports NVIDIA “Ada” GA102 GPU is targeting a TGP of 450W, 650W, and even 850W. The last number particularly might be overestimated or only refers to some type of special edition graphics card that we know nothing about yet.
Maybe all the RTX 40 series will use PCIe Gen5 power connector, so with a TGP of 600W, these cards should still be within the specs of a singular cable. The 16-pin power cord combined with the PCIe slot can deliver up to 675W of power.
Next-gen Flagship GPU Comparison (RUMORED)
| NVIDIA AD102 | AMD NAVI 31 | |
|---|---|---|
| Fabrication Node | TSMC N5 | TSMC N5/N6 |
| Architecture | NVIDIA Ada | AMD RDNA3 |
| GPU Package | Monolithic | Multi-Chip-Module (MCM) |
| Estimated GPU Size | ~600mm² | ~800mm² |
| Graphics Dies | 1 | 2 |
| GPU Mega Clusters | 12 Graphics Processing Clusters (GPC) | 6 Shader Engines |
| GPU Super Clusters | 72 Texture Processing Clusters (TPC) | 60 RDNA Workgroups (WGP) |
| GPU Clusters | 144 Streaming Multiprocessors (SM) | – |
| GPU FP32 Cores | 18432 CUDA Cores | 15360 Stream Processors |
| GPU Clock | ~ 2.2 GHz | ? |
| Memory & Bus | GDDR6X 384-bit | GDDR6 256-bit |
| Large Cache | 96 MB | 256-512 MB Infinity Cache |
| Rumored Power Consumption | 450~650~850W | 450~480W |
| Release Date | September 2022 | Q4 2022 |
| Rumored SKU | GeForce RTX 4090 | Radeon RX 7900XT |
| Expected Performance | 2x GA102 | 2.5x NAVI 21 |
