Nvidia has announced its latest RTX 40 series GPUs, as well as a creator focussed GPU, the RTX 6000 at its GTC 2022 keynote. The new GPUs are based on the Ada Lovelace architecture and seem to be in line with the recent leaks, at least for the RTX 40 series. 

The announcement included the upcoming RTX 4090 and RTX 4080 in addition to an RTX 6000 GPU. The RTX 4090 comes in at 82.6 teraflops, more than double the compute power of the previous flagship RTX 3090 Ti. As for the RTX 6000, Nvidia claims anywhere between two to four times improvement in performance in enterprise workflows. 

The RTX 4090 arrives first, launching on October 12 with a staggering price tag of $1,599. The RTX 4080 will follow later in November coming in two variants coming in at $899 for 12GB and $1,199 for 16GB.

As for the RTX 6000, the card will be available starting in December. While we don’t have the official pricing yet, considering the prices of the latest RTX cards and the fact that the older RTX A6000 still retails for around $5,000, it’s expected that Nvidia will follow suit. 

In the News: Windows 11’s first major update brings these 6 additions

RTX’s latest punch

The RTX 4090 is the crown jewel of Nvidia’s new Ada architecture coming in with 24GB of GDDR6X VRAM, 16.384 CUDA cores, 1,321 Tensor cores, 191 RT cores and 83 Shader cores. As for clock speeds, the card has a base clock of 2.23GHz which can boost up to 2.52GHz. Nvidia claims that the new card can be anywhere from two to four times faster than the RTX 3090 Ti. 

Below the 4090 are two variants of the RTX 4080, with the major difference being more than just the VRAM. The 12GB RTX 4080 comes with 7,680 CUDA cores, 639 Tensor cores, 92 RT cores, 40 Shader cores and a base clock of 2.31GHz that can boost up to 2.61GHz. 

The final gaming focussed card in the announcement, the 16GB RTX 4080 is more powerful than its lower VRAM variant across the board. The card comes with 9728 CUDA cores, 780 Tensor cores, 113 RT cores, 49 Shader cores and a base clock of 2.21GHz that can go as high as 2.51GHz. 

Both the RTX models however don’t have as much of a difference when it comes to power consumption. The company says the 12GB RTX 4080 will require around 700W while the beefier 16GB variant should be content with 750W. As for the RTX 4090, 850W from your PSU will be going to the GPU. All this means that the newer ATX 3.0 power supplies will be crucial to running the cards properly. All three cards also use the PCIe Gen-5 16-pin connector.  

Ada Lovelace, the third-generation of Nvidia’s RTX architecture promises major improvements in Ray Tracing, partly in thanks to a new DLSS 3 AI. The company also claims that DLSS 3 is coming to more than 35 games and apps starting in October. 

While benchmarks and test reports will only be available once the cards hit the market, they do seem to have a significant edge over Nvidia’s current offerings at least on paper. 

Specifications	GeForce RTX 4090	GeForce RTX 4080 (16GB)	GeForce RTX 4080 (12GB)
Price	$1,599	$1,199	$899
VRAM	24 GB GDDR6X	24 GB GDDR6X	24 GB GDDR6X
Memory Bandwidth	384-bit	256-bit	192-bit
Base Clock	2.23 GHz	2.23 GHz	2.31 GHz
Boost Clock	2.52 GHz	2.51 GHz	2.61 GHz
CUDA Cores	16.384	9728	7,680
Tensor Cores	1,321	780	639
RT Cores	191	113	92
Shader Cores	83	49	40
Required System power	850W	750W	700W

Better performance for creators with the RTX 6000

When it comes to raw horsepower, the new RTX 6000 absolutely dominates the RTX 4090. For starters, the GPU comes with a staggering 48GB of GDDR6 ECC memory. It has 18,176 CUDA cores, 568 Tensor cores and 142 RT cores. For those unaware, these numbers are almost double what the Ampere architecture-based RTX A6000 currently offers. 

The maximum power draw for the card however comes in at 300W, most likely due to Nvidia’s tendency to lower the power consumption on workstation-grade GPUs to enhance their lifespan.

As mentioned before, Nvidia claims performance improvements of as much as two to four times compared to the older A6000. The card will also perform thrice as fast in terms of video encoding and supports virtualization with Nvidia’s virtual GPU software, allowing multiple remote users to share resources and work off one card. 

All this power and performance however will not come cheap. While we won’t know the exact price until the card becomes available later in December this year, workstation-grade cards from Nvidia are notoriously expensive. The previous generation Ampere-based RTX A6000 is still retailing around the $5,000 mark, and the new card is expected to follow suit. 

Specifications	RTX 6000
VRAM	48GB GDDR6 ECC
CUDA Cores	18,176
Tensor Cores	568
RT Cores	142
Max Power Consumption	300W

In the News: YouTube moves from fixed fee to 45% revenue share with Shorts creators