Yesterday at around 5:00 pm, NVIDIA unveiled part of its upcoming line of graphics cards based on “ADA Lovelace”. Technically, this chip is exceptional with its 76 billion transistors, its 4nm TSMC engraving, shader reordering, its 90 TFLOPS, its 200 RT TFLOPS, its 1400 Tensor TFLOPS, and over 18,000 CUDA cores in the same GPUs. Suffice it to say that it is quite impressive compared to the previous generation Ampere.
Please note that the presentation video is still available at this link . In the end, three video cards were announced, as we thought, or almost because, to be honest, we were surprised. Jensen Huang provided an impressive amount of information and focused on this new generation’s performance this evening. But we’ll have to give us more on that score, because NVIDIA’s communication is more and more like Apple: loss of reference, confusion, and impressive numbers. These announcements followed extensive discussion in the editorial office as well as our Discord.
Three cards in the next two months, RTX 4090 first:
These are, of course, the top models presented, as in the previous generation. The idea could also be to allow RTX 3000 stock to continue to be sold.
Therefore, it was with this RTX 4090 that NVIDIA began its presentation. NVIDIA’s flagship model is equipped with an AD102-300 GPU with 16384 CUDA cores and a Boost frequency that can reach up to 2520MHz. It has 24GB GDDR6X memory with 1TB/s bandwidth and its TDP reaches 450W. The speed of this RTX 4090 should be 2-4 times faster than the RTX 3090 Ti.
This is the first map to be made available with the launch scheduled for October 12th. NVIDIA plans to offer Founders Edition in addition to partner cards. This version of NVIDIA will take up 3 slots wide and will be powered by a 16-pin power connector. Let’s go back to its price a little lower.
RTX 4080 16GB and also 12GB version, why?
At the same time, NVIDIA announced two RTX 4080s, and we must admit that we were very surprised. We were expecting the RTX 4080 and RTX 4070. It was this dual model that already caused a lot of buzz on social media last night. But let’s get back to the RTX 4080 16GB first. This card will have an AD103 GPU with no less than 9728 CUDA cores. 16 GB of GDDR6X memory will run at 22.5 Gb / s, and the TDP will reach 320 watts.
Where we expected the presentation of the RTX 4070 then, Jensen Huang presented the second RTX 4080, but with 12 GB. It was on this model that the reaction was the strongest, because, oddly enough, even if NVDIA still uses the term RTX 4080, it does not have the same GPU, since it is an AD104 GPU. This version has 7680 CUDA cores, GDDR6X memory chips and a TDP of 285W. So why not call it the RTX 4070? Please note that these two models are expected in November.
Above is a summary table compiled by our colleagues from Videocardz and includes three announced models. Now let’s move on to the price of this RTX 4090 and two versions of the RTX 4080.
Prices that will sting, especially in Europe!
Pricing announced during the conference is $1,599 for the RTX 4090 Founders Edition, $1,199 for the RTX 4080 16GB Founders Edition, and starting at $899 for the RTX 4080 12GB at partners. Indeed, the “12 GB” version will not be released in the Founders Edition. Suffice it to say straight out that it can be very hard to see an affiliate card being able to settle for $899. We were even able to finally find 12GB RTX 4080 cards from partners for more than the Founders Edition 16GB version.
Given the performance shown, the price of the RTX 4090 seems right to us, the price of the RTX 4080 16GB is pretty high, and the price of the RTX 4080 12GB is hard to justify as we’ve been waiting for the RTX 4070 version. But there’s a cold shower coming, again with a price conversion displayed in dollars , In Euro.
As we said, we are talking about the Founders Edition price for the two largest models. Suffice it to say that the partner cards map is still under threat of exploding. Last night, most of the partner brands presented their various models, but did not mention their price. Now, obviously, it is the performance that we will evaluate during our tests that will allow us to get a general idea and a clearer opinion.
Open questions about DLSS 3:
On the sidelines of this conference, NVIDIA also announced DLSS 3. This is very good news as it is still being developed to provide a performance boost. NVIDIA has released several slides demonstrating the performance of their new cards related to DLSS 3.
We’ll get back to DLSS3 in detail in another rundown, it’s time to get to grips with all of its new features, but also while waiting for answers to our questions such as: “Will DLSS3 be compatible with Ampere and Turing cards?” “. See you soon for more information and also for our tests of this new generation!