Nvidia GeForce RTX 4090 promises up to four times the performance of its predecessor
The new RTX 40 series is here. Nvidia CEO Jensen Huang has presented the GeForce RTX 4090, RTX 4080 16 GB and 4080 12 GB. The cards promise plenty of performance and will be available from 12 October.
The new Ada Lovelace architecture offers up to 76 billion transistors. The GPUs are manufactured at TSMC in the 4N process. The new architecture introduces 3rd generation ray tracing cores and 4th generation tensor cores. In addition, Lovelace supports the DLSS 3 (Deep Learning Super Sampling) upscaling technology. The AI-supported technology is supposed to generate entire frames and thus massively accelerate gameplay. According to Nvidia, the technology overcomes CPU performance limitations in games by enabling the GPU to generate entire frames independently.
Nvidia did not reveal much about the specifications at the keynote. The RTX 4090 comes with 24GB of GDDR6X memory and is said to be two to four times faster than an RTX 3090 Ti, depending on the application. The new flagship will arrive on 12 October. As for the price, Nvidia quotes 1599 US dollars.
The RTX 4080 comes in two variants: With 12 or 16 GB GDDR6X memory. It is supposed to be two to four times faster than the RTX 3080 Ti. According to Nvidia, the 16 GB version will cost 1199 US dollars, the 12 GB version 899 US dollars. The card will be available in November.
An RTX 4070 is missing from the line-up for the time being. The RTX 3080, 3070 and 3060 continue to be actively promoted by Nvidia.
The more exact specifications can only be guessed at based on the limited details from the keynote and information from previous leaks. Here are the specs of the three models without guarantee:
| GeForce RTX 4090 24GB | GeForce RTX 4080 16GB | GeForce RTX 4080 12GB | |
|---|---|---|---|
| GPU | AD102-300 | AD103-300 | AD104-400 |
| Cuda-Kerne | 16384 | 9728 | 7680 |
| Boost-Takt | 2520 MHz | 2505 MHz | 2610 MHz |
| Speicherbus | 384-bit | 256-bit | 192-bit |
| Speichergeschwindigkeit | 21 Gbps | 22,5 Gbps | 21 Gbps |
| Speicherbandbreite | 1008 GB/s | 720 GB/s | 504 GB/s |
| TGP | 450 Watt | 320 Watt | 285 Watt |
Jensen Huang spoke at the keynote of up to 18 000 Cuda cores in the RTX-40 series. If you look at the number of presumed Cuda cores in the table, that leaves a future RTX 4090 Ti open. Whether this is true and whether Nvidia's performance promises are accurate will be shown by the first independent reviews.
You can watch the whole keynote in the following video. The part with the gaming graphics cards starts right at the beginning.
Little side note: Two hours before the announcement, competitor AMD announced via Twitter that the new graphics cards with the RDNA3 architecture are coming on 3 November. The graphics card market will therefore change significantly in the coming months.
From big data to big brother, Cyborgs to Sci-Fi. All aspects of technology and society fascinate me.