Jensen Huang, CEO of Nvidia, revealed the GPU giant’s RTX 40-series flagship graphics cards at GTC today.
Powered by Nv’s Ada Lovelace miniature and TSMC 4nm process, the RTX 4090 and 4080 are said to offer more than twice the performance of the previous 3090 TI and 3080 TI.
Starting things off with the RTX 4090, the unit bears more than a passing resemblance to the 3090 TI Huang was comparing during his keynote presentation. The card features 24GB of GDDR6X memory, including a 384-bit interface, and 450W TDP as the 3090 TI.
Moreover, the specifications of the new card indicate that it will offer significantly higher performance, presumably thanks to architectural and process node improvements. 4,090 packaged 16,384 CUDA cores, up from 10,752 in the 3090 TI, while at the same time, Nvidia pushed the base clocks from 1.67GHz to 2.23GHz.
In addition to more CUDA cores and faster clocks, the card – and the 4080 for that matter – also features Nvidia’s third-generation real-time ray tracing cores and fourth-generation Tensor cores, as well as support for Nvidia’s DLSS 3 AI supersampling technology.
When Nvidia claimed that the units would offer twice the performance of their predecessors, it was based on a carefully selected selection of games, which included Microsoft Flight Sim and Warhammer 40K Darktide.
The chip designer also claimed performance was four times greater when trying out a pre-release version of Cyberpunk 2077 using next-generation ultra-fast AI technology. This doesn’t mean that these cards are aimed only at gaming: as GPU hardware, they can speed up a lot of other workloads.
But this allure will not come cheap. Nvidia is claiming an MSRP of $1,599 per card, although it’s still cheaper than the 3,090 TI when it shipped this spring with an MSRP of $1,999.
Of course, what with the dwindling world of GPU-based crypto-mining, thanks to changes in Ethereum, and all this economic uncertainty causing people to buy less stuff, customers might have a better chance of finding one of these 4,090 cards at this suggested price when it arrives. In mid-October. However, we are not holding our breath.
Be careful not to buy error 4080
Things get a lot more interesting when it comes to Nvidia’s little boss, or rather the little heads: 12GB and 16GB 4080s.
Nvidia compared the RTX 4080 series cards favorably with the outgoing 3080 TI, touting poor gaming performance on the Ampere-based card.
This is despite the 3080 TI keeping a tight CUDA core lead at 10240 versus the 16GB 4080 of the 9,728 CUDA cores. The gap increases further in the case of the 4080 12GB version, which features 7680 CUDA cores, about 2,000 fewer than its predecessor.
And that’s not the only way Nvidia has apparently hampered the 4080. While the cards use the same GDDR6X memory as their predecessors, Nvidia has lowered the memory interface to 256-bit on the 16GB model and 192-bit on the 12GB version. Up to 10GB 3080 had a 320-bit interface.
Nvidia appears to have made up for this by making architecture changes and boosting the 4080 clock speed, which now beats at 2.51GHz on the higher-end 16GB model, and 2.61GHz on the 12GB model, all while claiming it did so while using less power. Yay, TSMC?
The 12GB version claims a TDP of only 285W, while the 16GB version is rated to consume 320W. That’s roughly 9-19 percent less power than the 3080’s 350W TDP.
If the Nv numbers are to be believed – and some initial specs suggest that the 4080 12GB at least outperforms the 3080 and 3080 TI on FP32 performance and memory bandwidth – then the 4080 Series should outperform the 3080 TI despite its lower base CUDA count and display memory bus .
Just beware that there is a small gap between the 4080 and 16GB and 12GB models. Both 12GB and 16GB 4080s will launch in November with a suggested price of $899 and $1199, respectively.
Nvidia waits on mid-tier Series 40 cards while AMD prepares RNDA 3
As with previous years, Nvidia sticks to its flagship cards out of the gate. The GPU giant should start rolling out 40-series updates on the 3070, 3060 and 3050 cards over the next year. But for now, the mid-tier Nvidia Series 30 cards are here to stay.
Meanwhile, we’ll have to wait a few more weeks to see how Nvidia-powered Ada Lovelace cards stack up against AMD’s RDNA 3 architecture. office last month, we know it will be based on TSMC’s 5nm manufacturing process.
And if AMD pursues a similar partner to launch RDNA 2 in 2020, we won’t be getting any mid-tier cards from AMD out of the gate either. ®
note: Nvidia also announced a Lovelace-based RTX 6000 GPU, aimed at workstations that do heavy graphics and similar computations.
#Nvidia #reveals #RTX #watch