GTC Nvidia CEO Jensen Huang unveiled his GPU large’s flagship RTX 40-series graphics playing cards at GTC immediately.
Powered by Nv’s Ada Lovelace microarchitecture and a TSMC 4nm course of, the RTX 4090 and 4080 are stated to supply greater than twice the efficiency of the earlier 3090 TI and 3080 TI flagships.
Kicking issues off with the RTX 4090, the unit bears greater than a passing resemblance to 3090 TI Huang was evaluating it to throughout his keynote presentation. The cardboard options the identical 24GB praise of GDDR6X reminiscence, together with a 384-bit interface, and 450W TDP because the 3090 TI.
Past that, the brand new card’s specs recommend it is going to ship considerably larger efficiency, presumably because of architectural and course of node enhancements. The 4090 packs 16,384 CUDA cores, up from 10,752 within the 3090 TI, whereas concurrently, Nvidia has pushed base clocks from 1.67 GHz to 2.23 GHz.
Along with extra CUDA cores and quicker clocks, the cardboard — and the 4080 for that matter — additionally options Nvidia’s third-gen real-time ray-tracing cores and fourth-gen Tensor cores, in addition to assist for the Nvidia’s DLSS 3 AI supersampling tech.
When Nvidia claimed the items will ship about twice the efficiency of their predecessors, it was basing that on a rigorously cherry-picked choice of video games, which included Microsoft Flight Sim and Warhammer 40K Darktide.
The chip designer additional claimed efficiency was 4 occasions larger when making an attempt out a pre-release model of Cyberpunk 2077 utilizing its next-gen AI supersampling tech. That is to not say these playing cards are solely geared toward video games: as GPU gadgets, they’ll speed up loads of different workloads.
However that oomph will not come low cost. Nvidia is claiming an MSRP of $1,599 for the cardboard, although it is nonetheless cheaper than the 3090 TI when it shipped this spring with an MSRP of $1,999.
After all, what with the dwindling world of GPU-based cryptocurrency mining, because of the Ethereum adjustments, and all this financial uncertainty inflicting individuals to purchase much less stuff, prospects could stand a greater likelihood of discovering one in every of these 4090 playing cards at that recommended value when it arrives in mid-October. With that stated, we’re not holding our breath.
Watch out to not purchase the unsuitable 4080
Issues get a bit extra fascinating in terms of Nvidia’s mini-boss or quite mini-bosses: the 12GB and 16GB 4080s.
Nvidia in contrast the RTX 4080-series playing cards favorably to the outgoing 3080 TI, noting twice the efficiency in video games over the Ampere-based card.
That is regardless of the 3080 TI holding a slim CUDA core lead at 10,240 versus the 4080 16GB’s 9,728 CUDA cores. The hole grows even bigger within the case of the 4080 12GB variant, which options 7,680 CUDA cores, roughly 2,000 fewer than its predecessor.
And that is not the one means Nvidia has seemingly kneecapped the 4080. Whereas the playing cards use the identical GDDR6X reminiscence as their predecessors, Nvidia has lower the reminiscence interface all the way down to 256 bits on the 16GB mannequin and 192 bits on the 12GB model. Even the 10GB 3080 had a 320-bit interface.
Nvidia seems to have made up for this by making architectural adjustments and rising the 4080’s clock velocity, which now tops out at 2.51 GHz on the higher-end 16GB mannequin, and a pair of.61 GHz on the 12GB one, all whereas claiming to have carried out so whereas utilizing much less energy. Yay, TSMC?
The 12GB model claims a TDP of simply 285W, whereas the 16GB model is rated to devour 320W. That is roughly 9-19 p.c much less energy than the 3080’s 350W TDP.
If Nv’s numbers are to be believed – and a few preliminary specs point out the 4080 12GB at the very least does beat the 3080 and 3080 TI on FP32 efficiency and reminiscence bandwidth – then the 4080 sequence ought to outperform the 3080 TI regardless of the lowered CUDA core rely and reminiscence bus width.
Simply beware that there is a little bit of a niche between the 4080 16GB and 12GB fashions. Each the 12GB and 16GB 4080s will launch in November at a recommended value of $899 and $1,199 respectively.
Nvidia waits on mid-tier 40-series playing cards whereas AMD preps RNDA 3
As with earlier years, Nvidia is sticking with its flagship playing cards out of the gate. The GPU large will presumably start rolling out 40-series updates to the 3070, 3060, and 3050 playing cards over the following yr. However for now, Nvidia’s mid-tier 30-series playing cards are right here to remain.
In the meantime, we’ll have to attend a couple of extra weeks to see how Nvidia’s Ada Lovelace-powered playing cards stack up towards AMD’s RDNA 3 structure. Whereas little is thought about AMD’s upcoming shopper playing cards past a couple of seconds of gameplay demoed in the course of the Zen titan’s Ryzen 7000 desktop launch final month, we do know it’s going to be based mostly on a 5nm TSMC manufacturing course of.
And if AMD follows an analogous associate as its RDNA 2 launch in 2020, we’ll will not be getting any mid-tier playing cards from AMD out of the gate both. ®
: Nvidia additionally introduced the Lovelace-based RTX 6000 GPU, which is geared toward workstations doing heavy graphics work and comparable computation.