Aapje

You are still thinking in 3D, not 4D. DLSS 1 and FSR 1 are not temporal and do what you say. They take a single frame and upscale that to a higher resolution. But ever since DLSS 2 and FSR 2, the upscalers use previous frames as well. If you run DLSS Quality in 4K, then the real render resolution is 1440p. But temporal upscaling uses multiple frames. So if it uses 2 frames, the actual number of pixels used by the upscaling is 1440p x 2. If the upscaler uses 5 frames, then the number of pixels is 1440p x 5, which is more than 4k. But it's actually more complicated, because with temporal upscaling the same pixel in the source material is sampled multiple times, so the actual number of samples is even higher. So when you use DLSS/FSR to generate a 4K image, the algorithm will create more than 4K samples internally and then downsamples that to 4K. This is key to making DLSS/FSR look as good as it does, rather than merely looking like a scaled up image. PS. Key to temporal upscaling is jitter, where they shift the rendered image for each frame a little bit. This means that they can notice the effect of small features that are not visible on screen, by how it changes the rendered image when that feature moves to a different pixel. For example, let's say that there is a light in the distance that is so small that it is smaller than one pixel. In frame A, the light falls entirely within pixel 1 (at the right side of the pixel), so that pixel overall brightness is increased due to the light. But in frame B, where the frame is moved a bit to the left, that light falls partially or entirely in pixel 2. The result is then that pixel 1 is less bright and pixel 2 is more bright. Then the DLSS/FSR algorithm can conclude based on both frames that there must be small bright thing at the edge region of these two pixels. And when it upscales and thus has way more pixels to create, so those two pixels turn into 8 pixels, it can show a bright pixel and thus show a light that was never visible in each individual frame. Note that this is a bit similar to how astronomers detect planets, where they notice that the brightness of a star changes over time when a planet moves in front of it. So even though they cannot see the planet itself, they can still conclude that it is there. In principle, astronomers could use a DLSS-like algorithm to create video of the planet moving in front of the star.

This is quite unacceptable for someone with your username. We are all waiting for the PJ Pants reveal. But I guess you don't want to just show that off for free, and will monetize it on your OnlyPants account.

And Crypto & AI demand, and increases in die prices.

This is again, just a matter of expectations. Fact is that a midrange Geforce GTX 970 from 2014, is almost three times slower than a 2023 4060 Ti, that is a little cheaper in inflation adjusted dollars. So objectively speaking, you have a far faster card. The issue with focusing on subjective terms like like mid-range, is that you can get very weird results. For example: - Scenario 1: You buy card X and this is a midrange card, because there are 2 products above it and 2 products below - Scenario 2: You buy card X and this is a lower-midrange card, because Nvidia decided to released two extra high end products According to your way of looking at things, you are worse off in scenario 2, but you actually got the exact same product, with the exact same capabilities, for the exact same price. If scenario 2 happens because the average gamer got way richer, and Nvidia thus sees more reason to cater to richer gamers, then your conclusion is that gamers are worse off, even though the reality is that gamers got richer on average, and they chose to spend more money on GPUs. It seems backwards to see poverty as good and wealth as bad, and to see a voluntary choice to get a more premium product as a bad thing. I still have one lying around as a backup-card. Absolutely tiny heatsinks. It's pretty obvious that we are now getting a lot more product for our money, as even the lowest of low-end card have way bigger heatsinks.

Performance per dollar today is better than it has ever been, so this is more a matter of expectations, than of reality. The 5090 is an entry level product in the same way that a Ferrari is...not at all. You can run DCS on a 4060, on a flat screen at 1080p or 1440p. Back in the day, progress went faster, and it didn't take all that long for top-tier performance to get down to very low prices. This is no longer the case. But that doesn't mean that you can't have tons of fun, but you can't expect to get 3090 performance for 5060 prices this generation. So the choice is either paying more, or accepting that you are further away from the top tier.

Brunner doesn't really need their own section. Very few people have it and threads involving Brunner are thus quite rare.

All the people for whom money is no (major) concern obviously bought a 4090, so for them, it doesn't matter if the 5090 is not better value, as long as it is faster. You also have 4090 buyers who do care about value and some of them will skip it.

Downsampling is part of supersampling, so this is a bit of a nonsensical statement. Note that to understand DLSS, you have to think in 4D, since the DLSS doesn't look at a single frame, but supersamples across frames.

Which is even more reason why we need Meta to implement these codec improvements. Anyway, it seems like they removed the spam flag, so my post should now be visible to all.

It's a bad generation anyway, so most people should skip it.

Yes, the idea is that they would add these variants, so there would be 2 extra options in the dropdown. I personally don't use the ethernet method, but I've heard of a bunch of simmers who do use it.

Oh, my request got marked as spam, so you might not be able to see it. I have appealed and hopefully they will fix it. Meanwhile, this is the text of my request: The Plus variants of these codecs are optimized to require less bandwidth for the same quality. They are compatible with most decoding hardware/software, since they conform to the H.264/AVC and H.265/HEVC standards, respectively. They were developed by a camera surveillance company, which is a fairly similar use case to the needs for Link: sending high-quality video in real time over a connection with limited bit rates, using relatively low-performance hardware (compared to a GPU). The H.264+ variant is already supported by Virtual Desktop, so I see no reason why it shouldn't work for wired or air Link. In fact, it works so well that there are wired PCVR users who connect to their headset using a ethernet cable (with an USB-adapter), just so they can use Virtual Desktop over a wired connection. So it seems proven that it would be a benefit if implemented for Link. Whether H.265+ works well for the Quest is unproven, but it might provide a big benefit for low bandwidth air Link connections (H.265 works better for air link due to the bitrate limitation, but with a high-speed connection, H.264 works better). Even if these are only supported through the Debug Tool, this would provide a nice benefit for users & developers who want to optimize their connection. Of course, it would be even better if Link auto-selects the best codec and bitrate, so less tech-savvy users would benefit as well. https://www.securitymagazine.com/ext/resources/whitepapers/Hikvision-H264-Encoding-Technology.pdf https://www.hikvision.com/content/dam/hikvision/en/brochures-download/product-brochures/White-Paper_H.265-Plus-Brochure.pdf

There are a ton of problems with your example: - Even with 4x Frame gen, you will get at most 60 FPS not 120. And in reality it tends to be lower - Frame gen merely makes things more smooth on screen. The latency will actually become worse than with 15 FPS native. - Since the latency of 15 FPS is unplayable, the game is still unplayable with frame gen in your scenario A better analogy in your case is that you don't like the mediocre looks of the ladies at your local bar, so you go online to hire a cheap escort from a dodgy website that shows you that you can have Rihanna come over. But then the person who actually shows up is a big sweaty dude who steals your money.

I made a new topic for the video encoding discussion, with way for you to help to make it happen (perhaps):

In the DLSS 4 topic, there was a very interesting discussion about H.264+: Basically, it has been known for a while that Virtual Desktop uses a better video encoder, and that you get optimal quality when using a wired ethernet cable + USB-adapter to connect to the Quest. That way, you have the benefits of a fast wired connection, but also the better video encoder. However, it is a bit ridiculous that this is needed, when there is no reason why this can't work over USB directly. I wrote a feature request on the Meta forum to support H.264+ and H.265+ : https://communityforums.atmeta.com/t5/PCVR-Link-Development/Feature-request-H-264-and-H-265-support/m-p/1291237#M26324 As noted above, this should be a big benefit to us, basically giving us a free visual boost, whether we use AVC (H.264) or AV1 (H.265). It probably helps if some of you would give an upvote ('kudos') and leave a statement in support. However, better not mention flight simming/DCS too much, since then it looks like only our small niche cares and we are brigading. There is some truth to this, but they don't have to know that, and it is for a good cause, since all PCVR users could benefit from this, not just (DCS) simmers.

I was hoping that the extra bandwidth would do more. However, it was always unlikely to be such a big gain, because the 30-series was made on a very poor (but cheap) node, and the 40-series benefited a lot in raw performance and efficiency from going back to TSMC. In contrast, the 50-series does not even benefit from a single node improvement.

You've already been conned. The entire reason why they put RT-hardware on these cards, at the expense of either adding extra rasterization cores or reducing the size of the die and thus price, is because those RT-cores are useful for the cards aimed at business. So asking for better RT just means that that you buy into the BS that Nvidia has been peddling. In fact, what they market as raytracing is actually rasterization with a sprinkle of raytracing. That's why RT in games is generally so underwhelming.

According to the reviewers themselves, it is more the other way around, where popular products increase their views and bad products mean low viewership.

The problem with an FFB subsection is that it will become a mess if you have manufacturers make FFB devices, but who also have their own dedicated section. That will happen once Winwing and Virpil bring out theirs.

Apparently, people in some European countries had no chance at all at getting an FE, since scalpers figured out that they could start ordering 40 minutes before launch:

You can change the boundary sensitivity in the headset to reduce how soon the blue circle appears and how soon you go to passthrough at the edge of your boundary.

You should really make or use a different thread instead of this on, since it is offtopic. Anyway, this is how you speed test:

He found an American that would receive them, and then sent them on to Georgia. As for the other comment that only people in Russia are at risk, we've had assassinations in London, Berlin and Vienna, so if they care enough, you are not safe in the West.

You need to keep in mind that the AMD-chips with more than 8 cores have two compute-chiplets, with a high latency between them. Because of the high latency, you generally want to run the game on just one die. This is even more true when dealing with 3D-cache, which is only on one chiplet of the two. You then want to run the game only on the chiplet with X3D. The 9800X3D has one 8-core chiplet with X3D. The 7900X3D has a 6-core chiplet with X3D, plus another 6 core chiplet that is generally not much use in gaming. So from a gaming perspective, you can argue that it actually has fewer cores. Furthermore, the 9800X3D can boost higher due to them moving the X3D-cache below the compute, which means that it suffers less from the heat, and the CPU can thus clock higher. So my answer is the 9800X3D, if it's just for gaming.

Indie games are probably not any harder to produce, since devs can now use UE and Unity, to keep up with the increased demands. A single guy made VTOL VR, a solid arcade flight sim. And there is plenty of passion, because there are a lot of people who make games with very little chance and expectation to earn a good income out of it. Even one of the candidates of game of the year, Balatro, was made by a person who only made it for himself and had no expectations that others would like it.