One of many issues that caught our consideration throughout Nvidia’s GeForce RTX 40-series announcement was the corporate’s declare that it has already run its AD102 GPU at 3.0 GHz in its labs. This means the brand new GPU might be an overclocking champ with loads of headroom that AIB companions might faucet on factory-overclocked boards, in addition to guide overclocking. The RTX 4090 is ready to launch on October 12 and can seemingly be a part of the greatest graphics playing cards — a minimum of for these with deep pockets.
Nvidia’s AD102 chip packs 76 billion transistors and is made utilizing TSMC’s 4N fabrication know-how. Operating a GPU this advanced at 3.0 GHz is a giant deal, as usually chips that run at excessive clocks are significantly much less advanced. In the meantime, Nvidia’s GeForce RTX 4090 has a lift clock of 2520 MHz, so overclocking an AD102 GPU to ~3.0 GHz should not be too arduous for an excellent air or closed loop liquid cooler. Naturally, liquid nitrogen ought to allow the GPU to be pushed even additional.
The TSMC 4N fabrication know-how that Nvidia Ada Lovelace GPUs use has been custom-made for the corporate, presumably to assist with yields, course of window, energy, and clock speeds. Nvidia doesn’t disclose the peculiarities of TSMC’s 4N or the way it custom-made the corporate’s ‘default’ N4/4P node to realize its targets by way of complexity, efficiency, and energy. Contemplating the clock velocity potential of Nvidia’s AD102, Nvidia and TSMC seemingly optimized the manufacturing node each for complexity and frequency. Nonetheless, we have no idea whether or not TSMC applied libraries set for use for upcoming N4P or N4X fabrication course of applied sciences.
With its GeForce RTX 30-series graphics playing cards, Nvidia made thick three-slot cooling programs comparatively widespread on customized graphics card, however with its GeForce RTX 40-series boards it seems to be like the corporate will make such ultra-large coolers really widespread. Primarily based on the customized GeForce RTX 40-series boards we have now seen, they are going to all function subtle cooling programs that can most likely improve their overclocking potential. What number of of those boards will have the ability to hit 3 GHz stays to be seen, however it’s actually a chance for a minimum of among the fashions.
One of many issues we do not know with AD102-based graphics boards is how a lot energy a 3 GHz GPU frequency would require. Over the previous few months, we have heard rumors of flagship merchandise with 600W and even increased energy envelopes, however the reference GeForce RTX 4090 is barely rated for as much as 450W thermal graphics energy. Some rumors counsel that Nvidia’s companions would possibly allow an as much as 600W energy envelope for his or her premium choices. If that is true, we could quickly see some closely overclocked GeForce RTX 4090 boards.
Hitting a 3+ GHz frequency for a GPU is hardly a breakthrough, as AMD’s Navi 21 XTXH was overclocked to 3.20 GHz in mid-2021 and the world’s first GPU to interrupt the three GHz barrier was an Nvidia Maxwell GPU, circa 2015. However AMD’s Navi 21 is a chip that packs 26.8 billion transistors, made utilizing TSMC’s N7 know-how (which was presumably optimized for AMD’s wants), whereas Nvidia’s AD102 is sort of thrice extra advanced. And Maxwell required liquid nitrogen to hit that velocity, whereas it looks like Ada could attain that on air cooling.
We’re positively excited by seeing simply how far customized RTX 4090 playing cards push clock speeds, and the way a lot energy they require to take action. We must always have full opinions up on or barely earlier than the October 12 launch date.