The RTX 4090 easily outperforms prior generation models in terms of technology and performance. With considerable upgrades to all of the core hardware, significantly faster clock rates, and cutting-edge technology like DLSS 3, a new standard has been set, albeit at a very expensive cost.
For the majority of 2022, the Nvidia GeForce RTX 4090 hype train has been growing. CEO Jensen Huang unveiled important details at GTC 2022 after more than a year of exorbitant GPU prices and shortages, with a price that would no doubt cause many to sob in sorrow. for Nvidia’s Ada Lovelace architecture’s top product? Actually, it only amounts to $100 more than the RTX 3090 at launch, and if the card can deliver performance that even comes close to what Nvidia claims—between 2 and 4 times that of an RTX 3090 Ti—there will undoubtedly be buyers prepared to spend that price. At least at 1440p and 4K, the RTX 4090 is currently seated atop the GPU benchmark hierarchy throne.
That said, it can be debatable if the RTX 4090 is a decent deal or not. It ranks dead last out of 68 GPUs from the last ten years when only considering the FPS generated by different GPUs per dollar invested. Except that the 4090 is most definitely not a card made to perform well at 1080p, and our regular ranking is based on 1080p ultra performance. In fact, even while gaming at 1440p extreme, CPU bottlenecks are still a worry. See what we mean about value being subjective? Take a look at 4K performance and take into account ray tracing, and you might argue it’s arguably one of the greatest deals.
Again, owning an RTX 4090 card will cost you dearly; the RTX 4090 Founders Edition basic model sells for $1,599, and partner cards can raise the price. But if you want the best and have the money, this is the card you should buy right away. We doubt anything will outperform it in this generation, barring the release of an RTX 4090 Ti in the future.
Here is a list of the top extreme performance graphics cards, including the newest models from Intel, AMD, and Nvidia. However, it’s still intriguing to see how Intel’s Arc A770 compares on paper even though it competes on a totally different playing field.
If you want to understand about all the new innovations and adjustments introduced with the RTX 40-series, we’ll just point you to our Nvidia Ada Lovelace Architectural deep dive. You can learn much of what you need to know from the specs table above. The number of transistors has almost tripled compared to Ampere; the RTX 4090 has 52% more cores than the RTX 3090 Ti; the GPU clock speeds are 35% faster; and the GDDR6X memory? The only significant change is that there is now 12 times as much L2 cache, which prevents the GPU from frequently needing to seek data from memory.
According to the numbers, the RTX 4090 has slightly more than double the computational performance of the RTX 3090 Ti, and there are workloads where you’ll see benefits of that magnitude. Other adjustments made internally, though, have the potential to increase the margin even more.
Ray tracing is once more given a lot of attention, and three new technologies—Shader Execution Reordering (SER), Opacity Micro-Maps (OMM), and Displaced Micro-Meshes (DMM)—all promise to be better than what is already available. Existing games and engines won’t gain from them because they also mandate devs utilise them.
Workloads in AI and deep learning also stand to see significant generational improvements. Ada comes with support for the FP8 number format and the FP8 Transformer Engine from Hopper H100. For algorithms that can use FP8 instead of FP16, this translates to twice as much compute per Tensor core and up to four times the 3090 Ti’s computational power.
DLSS 3 is one algorithm that can make use of the new Tensor cores and an enhanced Optical Flow Accelerator (OFA). In fact, previous RTX GPUs won’t benefit because DLSS 3 requires an RTX 40-series graphics card. What is DLSS 3 used for? It creates an additional in-between frame to fill the space between the currently rendered and preceding frames. In some instances, it can perform nearly twice as well as DLSS 2. Later in this review, we will examine DLSS 3 in further detail.
Time is money, and tripling or quadrupling throughput will undoubtedly save time, so you can easily justify the cost of the RTX 4090 from a professional standpoint, especially for anyone interested in deep learning. There are several features that content creators will like, and upgrading from a 3090 or 3090 Ti to a 4090 is quick and simple. We’ll also examine ProViz performance.
What about players, though? Nvidia isn’t making a big deal about how the RTX 4090 is made for pros, unlike the RTX 3090 and 3090 Ti. Yes, it will be perfect for such users, but it also belongs to the GeForce family, and Nvidia isn’t holding back on its comparisons and claims regarding gaming performance. Perhaps the past two years of cryptocurrency mining are to blame, but at least players won’t have to compete with miners for cards this time around since GPU mining is no longer profitable.