It was yesterday, that NVIDIA unveiled part of its future range of graphics cards architected around “ADA Lovelace”. Technically, this chip is out of the ordinary with its 76 billion transistors, its 4 nm etching by TSMC, the Shader Execution Reordering, its 90 TFLOPS, its 200 RT TFLOPS, its 1400 Tensor TFLOPS and more than 18,000 CUDA Cores in a single GPU. This is quite impressive compared to the previous generation Ampere.
The video presentation is still available by following this link. In the end, three graphics cards were announced as we thought they would be, or almost, because we have to admit, we were surprised. Jensen Huang has delivered during this evening a quantity of information often impressive and focused on the performance of this new generation. But on this point, we will have to give us more because the communication of NVIDIA looks more and more like Apple: loss of reference, mix and spectacular figures. These announcements followed a large discussion within the editorial staff but also on our Discord.
Three cards in the next two months, first the RTX 4090 :
Of course, these are high-end models that have been presented as it was already the case for the previous generation. The idea is perhaps to continue to sell off stocks of RTX 3000.
So it is with this RTX 4090 that NVIDIA started its presentation. The flagship model of NVIDIA has a GPU AD102-300 with 16384 CUDA cores and a boost frequency that can reach 2520 MHz. It has 24 GB of GDDR6X memory with a bandwidth of 1 TB/s and it sees its TDP reaching 450 watts. The speed of this RTX 4090 should be 2 to 4 times faster than the RTX 3090 Ti. This is the first card that will be available with a launch date of October 12. NVIDIA plans to offer a Founders Edition in addition to the partner cards. This NVIDIA version will be 3 slots high and will be powered via a 16-pin power connector. We’ll come back to the pricing a little lower.
A 16 GB RTX 4080 but also a 12 GB version, why?
At the same time, NVIDIA announced two RTX 4080 and we must admit that we were very surprised. We were expecting one RTX 4080 and one RTX 4070. It is this dual model that has already caused a lot of ink to flow on social networks last night. But first let’s get back to the RTX 4080 16 GB. This card will have an AD103 GPU with no less than 9728 CUDA cores. The 16 GB of GDDR6X memory will be clocked at 22.5 Gbps and the TDP will reach 320 watts. Where we expected the presentation of the RTX 4070, Jensen Huang presented a second RTX 4080 but with 12 GB. It is on this model that the reactions were the strongest because strangely, even if NVDIA still uses the term RTX 4080, this one does not have the same GPU since it is an AD104 GPU. This could lead to confusion for a future buyer between the 16 GB and 12 GB version. This version has 7680 CUDA cores, GDDR6X memory chips and a TDP of 285 watts. So why not name it RTX 4070? You should know that both models are expected in November.
Above, a summary table made by our colleagues of Videocardz and which includes the three models announced. Now let’s see the price of this RTX 4090 and the two RTX 4080 versions.
Prices that will sting, especially in Europe!
The prices revealed during the conference are $1599 for the RTX 4090 Founders Edition, $1199 for the RTX 4080 16 GB Founders Edition and from $899 for the RTX 4080 12 GB at partners. Indeed, the “12 GB” version will not be produced in Founders Edition. We might as well say that it will be very difficult to find a partner card that can match the $899 price. We could even find RTX 4080 12 GB cards more expensive than the 16 GB Founders Edition.
In view of the performance displayed, the price of the RTX 4090 seems correct, that of the RTX 4080 16 GB quite high and that of the RTX 4080 12 GB difficult to justify as we were expecting an RTX 4070 version. But the cold shower comes once again with the conversion of the prices displayed in dollars into euros.
As we said, for the two biggest models we are talking about the price of the Founders Edition. We might as well say that the price of the partner cards is likely to explode again. The majority of the partner brands unveiled their different models last night, but without making any reference to their price. Now, it is obviously the performance that we will evaluate during our tests that will allow us to have an overall view and a clearer opinion.
Some outstanding questions about the DLSS 3:
On the sidelines of this conference, NVIDIA also announced the arrival of DLSS 3. This is a very good news since this one is still evolving in order to offer a performance gain. NVIDIA has proposed several slides highlighting the performance of its new cards associated with DLSS 3.
We’ll come back to the DLSS3 in detail in another brief, while we’re trying to assimilate all the new features and while we’re waiting to find the answer to our questions such as: “Will the DLSS3 be compatible with Ampere and Turing graphics cards? See you soon for more information and for our tests of this new generation!