If you divide the cost of the card by the pixel fill rate, the 4090 is a bargain.
If you’re looking for a pro card and have the cooling and power to support it (not to mention, the workflow needs that could benefit), that’s great. If you’re a gamer or enthusiast, the price is still high enough (not to mention the other changes you might need to make to your rig to support the card) that the actual delta between the potential and what you’ll actually do with the card means you should probably just stick with either a 3090 that is now half the price, or hold out for the other 4000 series cards if you must get a next-gen card.
There is a lot of confusion because some people assume turning on DLSS increases settings but it actually lowers quality. Sure DLSS is good enough that most don’t notice, but you can say that about lowering most settings slightly to improve performance.
Which just means better hardware beats DLSS at the same settings.
DLSS 3 still only looks fine on cherry picked screens: “It looks like DLSS 3's weakness lies in hidden geometry, where information is missing between two frames due to geometry overlapping another set of geometry while in motion. This can cause DLSS 3 to "shutter" and output ugly artifacts as it tries to fill in the void of missing detail.” https://www.tomshardware.com/news/dlss-3-early-review-rtx-40...
It doesn’t apply to every game or every resolution, but there are actual game scenarios where a 4090 makes sense. Price and heat and power and space not withstanding, it’s a meaningful upgrade.
Most games don't meet this bar, and I think gamers who expect to be able to set every slider to the max and go to town—yes, even on a $2K GPU—are mostly responsible.
Releasing an "extreme" preset that no consumer PC can really run yet is just a matter of leaving those higher-quality assets in + enabling those full resolution options for VFX. Allowing users to use ray tracing at full resolution(something which will kill any GPU on the market) requires just adding an extra config option - there's very little engineering time required. QC will need to do a full pass of these options, but in the grand scheme of things tested that's not a huge deal.
This is especially beneficial when deadlines slip and suddenly consumers have a lot better hardware than you where expecting.
I'm fine with nearly all gamers getting a better experience even if that means the tiny fraction of gamers who can and are willing to spend insane amounts of money on the best of all possible video cards are not able to take full advantage of their crazy hardware in most games.
I find it difficult to return to games such as Assassin's Creed II because of their muddy textures and low draw distances. These issues feel like something that could have been avoided with just a tad more forward thinking!
There are also games like Quantum Break which (at least at launch, not sure if it was ever fixed) included mandatory upsampling which the user couldn't disable. The reason given was that the game wasn't designed to be playable at native resolution, and no current hardware would be able to run it that way.
Extreme draw distances: large open-world games or those with very long, content-intensive environments need to resort to tricks to unload parts of the world that are not visible or quickly accessible. Extreme draw distance can mean keeping orderS of maginitude more objects resident, which could mean a lot more materials loaded, more draw calls, more VRAM usage, or more swap.
Higher resolution shadows: Hard shadows tend to look bad, soft shadows tend to perform bad, and worse with more lights. It takes a lot of deep GPU knowledge to do these in a visually convincing and high quality manner. The difference between "good enough" and "perfect" will easily cost you double digit fps at a minimum.
Full quality textures: As with the draw distance caveat, implementing LODs is rather work-intensive. Some people will tell you that you can automate it, and they're half-right. If you are looking for top-notch game quality, that absolutely does not cut it, but if you're not trying to go the extra mile it can be serviceable.
Games are super inconsistent in how far they push technology vs push the art, but there is rarely a "turn the dial to 11" knob ready to turn. The production requirements and technical limitations mix in unpredictable ways.
Other times, games push ridiculously far in certain directions that later become mainstream, and execute well enough that, after they are copied into other mainstream games, they feel deficient - not in spite of their success, but as a direct result of it!!
The point is that all of this merely requires more resources at runtime, not any additional work on behalf of the developer. So, by allowing limits higher than is practical on hardware at the time of release, the game can scale somewhat better on future hardware. What's the downside?
High-res textures are a different thing, since they actually have to be painted. Or upscaled, I suppose, but that's still code somebody has to write.
Ah, I want to clarify again—I was imagining the developers already had higher quality original textures, which they had downscaled for release. The textures in Assassin's Creed II, for instance, have the look of images that were saved down to a lower resolution from larger originals. But I could be wrong, or even if I'm not, it might be less common nowadays.
As you say, the goal is to include things that only require computational resources at runtime (even an order of magnitude more).
Many buying a Gaming PC will be buying a device that will do double duty and the price differential is best described by the GPU required for work vs play and if we want to truly compare it to console quality we ought to be comparing it to a budget GPU similar in performance to a PS5. For example a $370 nvidia gtx 3060 vs a $500 console.
https://www.logitech.com/en-us/products/keyboards/k400-plus-...
4090 has .28 GP/s/$ at base clock.
Considering this is coming out two years later, I'm not so inclined to call it a bargain.