6 ms·
It’s no secret that higher pixel counts are harder to achieve since computational power required grows approximately quadratically while GPU and CPU power impro
by valval 3y ago
It’s no secret that higher pixel counts are harder to achieve since computational power required grows approximately quadratically while GPU and CPU power improvements are more linear.
- colordrops 3y ago> GPU and CPU power improvements are more linear Do they?
- elabajaba 3y agoWe also just didn't have any good ways of driving such high PPI displays until recently without having to resort to janky hacks like dual cable setups (and the lines down the middle that those tend to have). A lot of laptops still don't support HDMI 2.1, and DP 2.0/2.1 only just came out (and I don't know if any displays are using it yet). There's also the bullshit the HDMI forum pulled where they relabeled HDMI 2.0 as HDMI 2.1 (years after HDMI 2.1 was out and established as "the one that can do 4k120hz no chroma subsampling" (a lot of implementations were around 40Gbps instead of the full 48Gbps because it was cheaper, and they didn't need the full 48Gbps for 4k120hz 10bit)), despite them having hugely different specs. Displayport 2.0/2.1 is also a clusterfuck because they decided that it would max out at either 40Gbps or 80Gbps depending on the implementation (DP40 and DP80), and companies don't generally tell you what their implementation actually supports. Macbook Pros only got HDMI 2.1 this year, iirc base and airs still don't have it. Intel iGPUs only do HDMI 2.0 and DP 1.4, AMD's Zen4 laptops can do HDMI 2.1 and DP 2.1 (idk what bandwidth). For dGPUs, AMD's 6000 series is HDMI 2.1+DP 1.4 (but there are issues with DSC), while 7000 series is HDMI 2.1 and DP 2.1 (DP80). Intel Arc dGPUs are DP 2.0 (DP40) and HDMI 2.1. Nvidia's 30 and 40 series are HDMI 2.1 and only DP 1.4. edit: Also, people don't realize just how fucking huge an 8k framebuffer is, and you need multiple for a swapchain. Even for just basic 8bit colour you're looking at ~133MB per framebuffer.
- NavinF 3y ago> ~133MB per framebuffer That's not a lot for people in the market for 8K monitors. 24G VRAM in the form of used 3090 cards is way cheaper than high end monitors. But yeah lack of DP2.1 support on current-gen cards like the 4090 is a problem > you need multiple for a swapchain There should be just two with variable refresh rate
- elabajaba 3y agoOn an 8GB (eg. base Macbook) or 16GB laptop where you're sharing ram between the CPU and iGPU it's a ton. People want 8k for better PPI for text rendering, not for games, meaning they want to use it with their work laptop. > There should be just two with variable refresh rate There's still a lot of software out there that's using DXGI blit model instead of DXGI flip model, and a triple buffered vsync+windowed blit app means you're looking at 5? framebuffers in flight at once. (ignoring any additional intermediate buffers for each app, and potentially having multiple overlapping windows drawing onscreen at one time).
- 111111IIIIIII 3y agoGreat comment. I just bought a new Dell 6K display and I'm struggling to imagine a purpose for 8k, but it's incredible that we have the technology to drive that.
- nordsieck 3y ago> I just bought a new Dell 6K display and I'm struggling to imagine a purpose for 8k It's for large monitors. I run a 43" 4k tv as a monitor. I would absolutely benefit from a similar sized tv in 8k. And people who run closer to the 50" size would benefit even more. I think you're probably right about traditionally sized monitors, though.
- rbanffy 3y ago> janky hacks like dual cable setups I’m a bit surprised we don’t see point-to-point ethernet cables doing this. They are standardised, reliable, and carry a lot of data. My napkin math says an 8K display at 40 bits per pixel and 60 fps is about 18 Gbps, well into the range of fancy ethernet. And, if consumer electronics can drive down the costs of these interfaces, everyone wins.
- jiggawatts 3y agoActually... Thanks to Moore's Law, computer power has been growing exponentially since forever. Single-core performance has stagnated, sure, but total performance is still following the curve. Graphics is easy to parallelise, so throwing more cores at it is not a problem. In fact, this is precisely what GPUs do. In other words, the available "operations per pixel per second" have gone up over the years, despite the increasing resolution! In fact, we're at the point now that we can even do real time ray tracing, which was unthinkable at any typical resolution just a decade ago. Top-end RTX 4090 GPUs can even do ray tracing at 4K, which is just crazy.
- valval 3y agoMoore’s law is a thing yes, but in the real world, power consumption and heat dissipation issues seem to be the actual constraints rather than transistor counts. Vertical scaling is of course possible, but it’s not really useful for driving a higher resolution monitor in your living room because of bottlenecks being completely elsewhere.
- forgotpwd16 3y ago>total performance is still following the curve Nitpick: Moore's law is about number of transistors on microchips not performance.
- jiggawatts 3y agoYes, but for GPUs performance scales fairly close to 1:1 with increasing transistor counts. Newer GPUs generally have more of the same type of compute units tiled out. The original general-purpose GPU, the NVIDIA 8800 GT had 112 cores. The RTX 4090 has 9,728! Each core has pretty much remained the same: a 32-bit arithmetic/logic unit (ALU).
- prmoustache 3y ago>In fact, we're at the point now that we can even do real time ray tracing, which was unthinkable at any typical resolution just a decade ago So are you saying that we should go back to "just wait for the picture to render..." era just because we can have higher density? Top-End RTX 4090 cater for a niche market.
- mumumu 3y agoThe bottleneck is bandwidth and storage not computer power. Data from sensor and to display are usually (but not always) raw and uncompressed. To go beyond 8k 120fps we will need fiber, or better compression. We have good codecs to move data online. And specialized ASIC to do live encoding with good quality. The current compression defined by HDMI is very rudimentary.
- beebeepka 3y agoPower is definitely a major factor. Unless games do not exist in your world
- leptons 3y agoGames do not exist in my world. Playing games is a waste of silicon, and human potential.
- beebeepka 3y agoYou're not the person I replied to. Why the hell did you respond to this. Having a bad day or something? Quite ironic to complain about waste at the same time as making this pointless and aggressive comment
- leptons 3y ago>Power is definitely a major factor. Unless games do not exist in your world Maybe you don't realize you're on a worldwide forum where anyone can respond to anyone else for any reason? I'm perfectly entitled to respond to your comment as someone who does not game. The person you replied to isn't the only person in this entire thread. And suggesting that it's strange for me to reply to your comment is just strange in itself.
- oblio 3y agoHow does commenting on anonymous forums rank in your world?