The RTX 4000 generation is powerful, but the graphics cards are also significantly more bulky than their predecessors.
Since the release of the first GeForce RTX 4090 Founders Edition, it has been said repeatedly: the new Ada Lovelace generation from NVIDIA is not particularly compact. Just compare the "small" RTX 4070 to previous generation models and this is even more true for the majority of "third party" models, those directly designed by NVIDIA partners. For example, even on this mid-range card, most manufacturers employ a necessarily bulky three-fan cooling system. RTX 4070s are often around 30 centimeters in length, and it's even more noticeable with the higher models: RTX 4070 Ti, RTX 4080 and RTX 4090 often exceed 30 cm and in some cases flirt with 35 cm. We are not even talking about the thickness of these cards which systematically condemn three PCI Express slots... and sometimes four!
However, a very small model of GeForce RTX 4000 has just been released thanks to NVIDIA of course, but also to its partner PNY which distributes the said card which is only available on the American market for the moment. The NVIDIA RTX 4000 SFF Ada Generation as PNY calls it is a so-called "SFF" card. The acronym does not speak to you? Well, it stands for " small form factor ", which means "small size"... and this card is very well named. Announced by PNY, the dimensions of 16.7 centimeters long and 6.6 cm high have nothing to do with those of the RTX 4070 Verto model, one of the most compact so far (24.7 x 12 cm). Such a compactness logically allows to take less space in the case and therefore destines this card to small models, cases in SFF format of course! Note that micro-ATX and mini-ITX will also be perfect for such a card.
Unsurprisingly, to achieve such a compactness, PNY has been forced to make some sacrifices. First of all, we notice that the four display connectors are in MiniDisplayPort 1.4a format, to avoid taking up too much space on the I/O panel of the card. They are numerous (4) in order to allow the same multi-display management as on other GeForce RTX 4000. The GPU used here by NVIDIA and PNY is a bit higher than the RTX 4070 with 6,144 CUDA cores against 5,888. The number of transistors is identical (35.8 billion). On the other hand, the technical specifications differ when it comes to VRAM. Indeed, the NVIDIA RTX 4000 SFF Ada Generation has 20 GB (against 12 GB) but only in GDDR6 ECC (against GDDR6X) with an interface bus of 160-bit (against 192-bit), which limits DDR5 to a bandwidth of 320 GB / s (against 504 GB / s).
PNY announces a raw power of 19.2 TFLOPS in FP32 and highlights the very low power consumption of the card: at only 70 Watts, it does not need an external power connector to satisfy its needs: in fact, one can be satisfied with the power supply provided by the PCI Express and thus have no power connector. Unfortunately, not only is this card only available in North America for the moment, but it is also offered at an exorbitant price of almost $1,500 ($1,444.22 at ShopBLT). This pricing is not surprising since it is a card intended for professionals in a niche segment: half-height cards. Too bad for the potential amateurs.
