The Nvidia GeForce RTX 4080 is the follow-up to the RTX 4090, which was released last month. It is now one of the best graphics cards and is at the top of our GPU benchmark hierarchy. Obviously, the melted 16-pin connectors have dulled the shine a little bit. The bright side: The fact that the RTX 4080 consumes less power suggests that it is less likely to channel sufficient power to melt the plastic connector… maybe. The blemishes: It inherits the RTX 3080 Ti launch price, which we also considered to be excessive, and is still priced out of reach for the majority of gamers at $1,199. It also represents a significant increase in generational pricing.
The only real question at this point is how performance varies as a result of having fewer GPU shaders, memory, cache, and a memory interface that is smaller, among other things. The majority of Nvidia’s Ada Lovelace architecture’s features are already known to us. Let’s quickly take a look at some of the best GPUs from AMD and Nvidia.
The RTX 4080 and the more powerful RTX 4090 are separated by a significant gap. The majority of an AD103 GPU—76 of the potential 80 Streaming Multiprocessors (SMs)—is available, but compared to the RTX 4090, there are still 40% fewer GPU shaders and other functional units. The rated TBP decreases by 29%, memory channels, bandwidth, and VRAM are reduced by 33%, and clock speeds remain the same. Concerningly, the theoretical compute performance suggests that the RTX 4090 could be up to 70% faster on paper.
Since the RTX 4080 costs $1,199, it seems like anyone even considering it should probably just save the $400 for the RTX 4090 and go broke or melt. However, the fact that the RTX 4090 has been sold out since its launch indicates that it may actually be a $900 upsell, which is significantly more significant.
When we take into account AMD’s forthcoming Radeon RX 7900 XTX/XT cards, the price becomes even more of a concern. We now have all of the relevant information about the first cards that use AMD’s RDNA 3 GPU architecture. Even though prices are still high, comparisons of the specs show that AMD might be able to beat the RTX 4080 for at least $200–$300 less. This means that you should at least wait until next month to see what the red team has to offer, unless you absolutely refuse to purchase an AMD graphics card.
However, AMD is unlikely to match Nvidia’s extras in the near future. For instance, the AI and deep learning power of the RTX 4080 far outperforms AMD’s plans. AMD’s FP16 and INT8 throughput will be less than a third of that of the RTX 4080 if our calculations are correct.
As a result of the improved Optical Flow Accelerator (OFA), Nvidia also provides DLSS 3. The technology is currently supported by ten games: Bright Recall: Destroy all humans, infinite! 2: Reformed, F.I.S.T. : Microsoft Flight Simulator, A Plague Tale, Forged in Shadow Torch, F1 22, Justice, Loopmancer, Marvel’s Spider-Man Remastered, and Super People and the Requiem. In less than a month, that’s roughly half as many DLSS 3 games as AMD’s FSR2 technology. Obviously, DLSS 3 requires an RTX 40-series GPU, whereas FSR2 is compatible with virtually all hardware.
Professional users, or at least their employers, tend to also favor Nvidia GPUs. Therefore, despite the fact that true workstations will likely select the RTX 6000 48GB card over a GeForce RTX 40-series card, it is definitely possible to acquire one or more RTX 4080 cards for use in AI and deep learning. Content creators may also find something they like, but if you’re willing to pay for a 4080, a 4090 may not be too much more expensive.
GPU mining continues to be unprofitable, which may or may not be good news for those on either side of the aisle. Although cryptocurrency mining won’t be able to cover the cost of a new graphics card, gamers should have access to more GPUs. Let’s take a closer look at what the brand-new RTX 4080 from Nvidia has to offer.
