RTX 4080/4090: 5 things Nvidia didn’t reveal in the keynote
Many questions remained unanswered in the Nvidia keynote on the RTX 4000. We dug through the available information and found what we were looking for.
Nvidia has introduced the RTX 4090 and RTX 4080.
Nvidia has finally revealed the big secret and presented the first cards of the new RTX 4000 generation. Check out the specs and insane pricing in our keynote recap. But as befits Nvidia, CEO Jensen Huang left many questions unanswered when presenting the RTX 4090 and RTX 4080, also with regard to pure performance and especially gaming performance. That’s why we’ve taken a close look at the specs available, and we’re giving you five key details that you didn’t learn during the presentation.
RTX 4080: One name, two very different GPUs
If you followed the keynote, you already know that Nvidia is releasing two RTX 4080 models. The designations suggest that they only differ in terms of the built-in video memory. But the “16 GB RTX 4080” and the “12 GB RTX 4080” are actually two completely different cards – with very different performance.
RTX 4080 is not equal to RTX 4080.
As you can see from the table above, the 12GB RTX 4080 not only has less memory, but also a significantly narrower 192-bit memory interface and 21 percent fewer CUDA cores. The clocking of the is a bit higher, but the computing power will still be significantly below the 16 GB variant, regardless of the installed memory. Confusing for consumers, the model designation is the same.
12 GB RTX 4080 probably without Founders Edition
If you scroll a little further in the technical specifications, you will notice that although Nvidia specifies the specific dimensions for the 16 GB variant of the RTX 4080, the 12 GB model only says “Differs depending on the manufacturer”. This suggests that Nvidia will not be making its own Founders Edition of the 12GB RTX 4080.
RTX 4000 supports AV1 encoding
AV1 is one of the most important compression methods for media professionals, thanks to good image quality with low bandwidth requirements. Intel integrated AV1 processing into the upcoming Arc graphics cards before AMD and Nvidia, and now Team Grün is following suit. The data sheets reveal that the RTX 4090 and RTX 4080 will also support AV1 encoding (AV1 decoding was already supported).
“GeForce RTX 4090 and RTX 4080 graphics cards feature two of our new 8th generation NVIDIA encoders (NVENC), which now support AV1 encoding, opening up new possibilities for live streamers, video editors and video conferencing participants.”
says the official announcement for the RTX 4000.
DLSS 3 only for RTX 4000 – older GPUs get nothing
It is obvious that last generation GPUs will not include new generation hardware encoders. But Nvidia’s much-touted DLSS 3 scaling feature won’t be available on the RTX 20 and 30 series cards either, although DLSS 2.0 and 2.1 have been backported to both. Nvidia’s GPU generation comparison page only lists DLSS 3 for the RTX 4000 series, while the older RTX generations are still listed with DLSS 2.
DLSS 3 is reserved for the RTX 4000 series
Nvidia also confirmed this in a statement to The Verge : “DLSS 3 is powered by the new fourth generation Tensor Cores and Optical Flow Accelerator on the GeForce RTX 40-Series GPUs” and will therefore not be available for older RTX generations, Nvidia spokesman Benjamin Berraondo told the website.
NVLink for GeForce is coming to an end
The days of multi-GPU setups with GeForce GPUs are finally over. The SLI technology was discontinued some time ago, but the successor “NVLink”, which was at least still supported by the RTX 3090, no longer plays a role for the RTX 4000. The data sheet specifically states that the RTX 4090 and RTX 4080 do not support NVLink.