Nvidia’s GPU Technology Conference (GTC) is underway. During Lecture by CEO Jensen HuangNvidia next-gen details funnel architecture were revealed. While it’s an AI and data center focused GPU, it does give us some hints of what we can expect from gaming-oriented Nvidia. ada lovelace GPU architecture, expected to be released in late 2022.
The H100 is a big step up from the current flagship A100. The full GPU contains 80 billion transistors or 26 billion more than the A100. It is built on a custom TSMC 4nm process. It supports up to 80GB of HBM 3 memory offering up to 3TB/s of bandwidth.
The H100 supports PCIe 5.0 and NVLink to connect multiple GPUs together. It can deliver 2,000 TFLOPS of FP16 and 1,000 TFLOPS of TF32 performance, triple that of the A100. Hopper introduces a new set of instructions called DPX. It is designed to accelerate performance in fields as varied as disease diagnosis, quantum simulation, graph analysis and routing optimizations.
The complete H100 GPU includes 18432 CUDA cores and 576 Tensor cores. This compares to the A100 with 8192 and 512 respectively, although for now not all cores are unlocked, presumably to maximize yields. Core clocks are also unfinished. Despite being manufactured on such an advanced node, the SXM version of the H100 comes with a TDP of 700W. That’s right, seven. hundreds. watts.
The H100 is set to be a card monster, but is it relevant for PC gamers? The answer is more or less. The H100 is all about computing performance and not graphics, but we can take some information and use it to predict what the game version will look like.
The move to a custom 4nm TSMC node is an important step up from Samsung’s 8nm process used for the RTX-30 series. It is likely to be used for RTX-40 series cards as well. Also noteworthy is support for PCIe 5.0. While on its own it is not expected to offer any real performance benefit over PCIe 4.0, it could very well be over PCIe 3.0, which is still widely used in many gaming systems.
But perhaps the biggest nugget of all is the astonishing 700W TDP of the high-end configuration. Just look at the VRM of that card! 700W for a data center product is something that can be managed, but if we can get something like this for a flagship RTX 4090 we would be shocked. sadly, Rumors of sharp increases in energy consumption continue to surface. Even 500W is a leap and means four slot graphics cards could become the norm, at the top of the market anyway.
Nvidia is still working on the H100. If its key features are shared with the RTX 40 series, it’s fair to say that the high-end cards will be hot and power-hungry, but packed with technology and much faster than the RTX 3090 (and the RTX 3090 Ti will soon be released). AMD will compete with its RDNA3 based cards and it’s turning into a hell of a battle, with overall performance clearly being a priority for both companies over energy efficiency. We can’t wait!