Nvidia unveils RTX 4090 but it’s the 4080 to watch out for • The Register

Nvidia CEO Jensen Huang today unveiled the flagship RTX 40 series graphics cards from its GPU giant at the GTC.

Powered by Nv’s Ada Lovelace microarchitecture and a 4nm TSMC process, the RTX 4090 and 4080 are said to offer more than double the performance of the previous flagship 3090 TI and 3080 TI.

Starting things off with the RTX 4090, the unit bears more than a vague resemblance to the 3090 TI that Huang was comparing it to during his main presentation. The card features the same 24GB of GDDR6X memory, including a 384-bit interface and 450W TDP as the 3090 TI.

Beyond that, the specifications of the new board suggest that it will provide substantially higher performance, presumably thanks to improvements in the architecture and process nodes. The 4090 packs 16,384 CUDA cores, compared to the 3090 TI’s 10,752, while at the same time Nvidia pushed the base clocks from 1.67 GHz to 2.23 GHz.

In addition to more CUDA cores and faster clocks, the card – and the 4080 for that matter – also includes Nvidia’s third-generation real-time ray tracing cores and fourth-generation Tensor cores, as well as support for DLSS 3 AI. Nvidia oversampling technology.

When Nvidia claimed that the units will deliver roughly double the performance of their predecessors, it was based on a carefully selected game selection, which included Microsoft Flight Sim and Warhammer 40K Darktide.

The chip designer further claimed that the performance was four times higher when he tested a pre-release version of Cyberpunk 2077 using its next-generation AI supersampling technology. That’s not to say that these cards are only meant for gaming – as GPU devices, they can accelerate many other workloads.

But that oomph won’t come cheap. Nvidia claims an MSRP of $ 1,599 for the card, though it’s still cheaper than the 3090 TI when it shipped this spring with an MSRP of $ 1,999.

Of course, with the declining world of GPU-based cryptocurrency mining, thanks to the changes in Ethereum and all this economic uncertainty leading people to buy fewer things, customers may have a better chance of finding one of these 4090 cards at that price. suggested when it arrives in mid-October. That said, we’re not holding our breath.

Be careful not to buy the wrong 4080

Things get a little more interesting when it comes to Nvidia’s mini-bosses or rather the mini-bosses: the 12GB and 16GB 4080s.

Nvidia favorably compared the RTX 4080 series cards to the upcoming 3080 TI, advertising twice the performance in games compared to the Ampere-based card.

This is despite the 3080 TI having a narrow advantage of the CUDA core at 10,240 against the 9,728 CUDA cores of the 16 GB 4080. The gap still widens in the case of the 12GB 4080 variant, which has 7,680 CUDA cores, about 2,000 fewer than its predecessor.

And that’s not the only way Nvidia has apparently brought the 4080 to its knees. Although the cards use the same GDDR6X memory as their predecessors, Nvidia has reduced the memory interface to 256-bit on the 16GB, 192-bit model. on the 12 GB version. The 10GB 3080 also had a 320-bit interface.

Nvidia seems to have compensated for this by making architectural changes and increasing the clock speed of the 4080, which now reaches 2.51 GHz on the 16GB high-end model and 2.61 GHz on the 12GB one, all while stating of having done so while using less energy. Yes, TSMC?

The 12GB version claims a TDP of just 285W, while the 16GB version consumes 320W. That’s about 9-19% less power than the 3080’s 350W TDP.

If Nv’s numbers are to be believed – and some preliminary specs indicate that the 12GB 4080 at least beats the 3080 and 3080 TI in FP32 performance and memory bandwidth – then the 4080 series should outperform the 3080. TI despite the small number of CUDA cores and the width memory bus.

Just be aware that there is a bit of a gap between the 16GB and 12GB 4080 models. Both the 12GB and 16GB 4080s will launch in November at a suggested price of $ 899 and $ 1,199, respectively.

Nvidia looks forward to the mid-tier 40 series cards as AMD prepares RNDA 3

As in previous years, Nvidia is keeping its flagship cards out of the gate. The GPU giant will presumably begin rolling out 40-series updates for the 3070, 3060, and 3050 cards in the next year. But for now, Nvidia’s mid-tier 30 series cards are here to stay.

In the meantime, we’ll have to wait a few more weeks to see how Nvidia’s Ada Lovelace-based cards perform with AMD’s RDNA 3 architecture. While little is known about AMD’s upcoming consumer cards beyond just a few seconds of gameplay demonstrated during the launch of the Zen titan’s Ryzen 7000 desktop last month, we do know that it will be based on a 5nm TSMC manufacturing process.

And if AMD follows a similar partner like its RDNA 2 launch in 2020, we won’t even have mid-tier cards from AMD out the gate. ®

PS: Nvidia also announced the Lovelace-based RTX 6000 GPU, which is aimed at workstations that perform heavy graphics work and similar calculations.

Leave a Reply

%d bloggers like this: