A listing appeared in the EEC (Eurasian Economic Commission) confirms that NVIDIA is working on the RTX A4500, a professional graphics card from high performance With which the green giant will cover the gap between the RTX A5000 and the RTX A4000, two graphics cards that we can also consider as high-end.
The RTX A4500 will use the GA102 core, which means that it will have the same base as the RTX A5000. This nuance is important, as the RTX A4000 uses the GA104 graphics core. We know, for now, that the RTX A4500 should have a lower configuration than the RTX A5000, but higher than the RTX A4000, which allows us to deduce that it will have more than 6,144 shaders but less than 8,192 shaders. It is also possible that it has those 8,192 shaders, the same as the RTX A5000, and that it differs from that only by having a smaller amount of graphic memory.
Right now there is no specific information beyond what has been said, so both options are open. However, I think the RTX A4500 is more likely to have a configuration that will generally fall somewhere between the two GPUs. If this is true, that graphical solution should have 7,680 shaders, a 256-bit or 320-bit bus and would be accompanied by 16 GB of GDDR6 graphics memory. Your TBP will be around 200 watts.
In terms of power, we don’t expect much of a difference from this versus the RTX A5000. This reaches 27.8 TFLOPs, so the RTX A4500 should be in the range of 21-22 TFLOPs in FP32, depending on the frequency of work that NVIDIA uses in this new graphics card. Except for surprise, this model should be available between December of this year and January of next year.
With the launch of these types of products, NVIDIA better covers the gaps between different graphics solutions, and offers solutions better adapted to the particular needs of each user profile. It is also important for budget reasons, since with a more varied catalog a potential customer is more likely to find a graphics solution that suits his needs in terms of power, and also the money he can invest in it.
