NVIDIA will be announcing its GeForce RTX 4080, and RTX 4090 graphics cards sometime in late August, followed by a hard launch in September. The initial shipments of the next-gen GPUs are expected to hit stores the same month. This was confirmed by Greymon55, one of the more reliable sources on NVIDIA and AMD graphics cards. Furthermore, as already shared earlier, the RTX 40 series will be an absolute power hog, with the top-end SKUs drawing as much as 600W of power under load.
A power draw of 600W means a power supply of at least 1200W, and 1500W to be on the safer side. The GeForce RTX 4080 should be set with a 1000W power supply, but the RTX 4090/4090 Ti will require a 1200W PSU at the very least. The primary reason for this is that the Ada Lovelace microarchitecture powering the RTX 40 series GPUs will be a mild re-organization of Ampere on TSMC’s N5 (5nm) process node. Although the fundamentals will stay the same, the overall compute density (ALU count) will be ramped up quite significantly.
Much of the IPC and efficiency gains will come from TSMC’s 5nm process node. TSMC’s 5nm node is 16% faster than the 7nm node at the same power or 14% more efficient at the same performance, with a density gain of around 70%. This will allow NVIDIA to offer the kind of performance uplift required to challenge AMD’s multi-chiplet RDNA 3 design.
The RTX 4080 should feature 16GB of GDDR6X memory running at around 21Gbps, while the RTX 4090 should pack somewhere between 20-30GB of GDDR6X memory. In terms of specifications, we’re looking at an FP32 core count of up to 18,432. The AD102 flagship is rumored to feature 144 SMs distributed across 12 GPCs. That results in a raw compute gain of over 2.5x (90 TFLOPs) over the GA102, granted the core is running close to 2GHz.
As per industry rumors, NVIDIA has paid TSMC $1.64 billion last quarter to reserve its share of the 5nm pie, with another $1.79 billion to be paid in the first quarter of 2022. Overall, the chipmaker will be spending close to $10 billion to secure its 5nm supply for the RTX 4080, 4090, and their 40 series brethren.
According to Igor’s Labo, NVIDIA designed the RTX 3090 Ti as a playground for AICs to prepare for the upcoming RTX 4080, 4090 graphics cards. The GeForce RTX 3090 Ti which has a default TBP of 450W is designed to have a maximum power limit of 520W. Going by these numbers, it’d be safe to assume that the RTX 4080, and 4090 will also have a TBP of around 450W, with factory-overclocked variants pushing the 500W mark. That’s a bit below the 600W limit of the 12VHPWR power connector but considering safety standards it’s high enough.