Nvidia won, we all lost • Hacker News Board

Re: Nvidia won, we all lost

#341

Post by keyringlight » Sat, Jul 05, 2025, 9:34 AM UTC

mcdeltat wrote: post #251

Anyone else getting a bit disillusioned with the whole tech hardware improvements thing? Seems like every year we get less improvement for higher cost and the use cases become less useful. Like the whole industry is becoming a rent seeking exercise with diminishing returns. I used to follow hardware improvements and now largely don't because I realised I (and probably most of us) don't need it. It's staggering that w…

What stands out to me is that it's not just the hardware side, software production to make use of it to realize the benefits offered doesn't seem to be running smoothly either, at least for gaming. I'm not sure nvidia really cares too much though as there's no market pressure on them where it's a weakness for them, if consumer GPUs disappeared tomorrow they'd be fine.

A few months ago Jensen Huang said he sees quantum computing as the next big thing he wants nvidia to be a part of over the next 10-15 years (which seems like a similar timeline as GPU compute), so I don't think consumer GPUs are a priority for anyone. Gaming used to be the main objective with byproducts for professional usage, for the past few years that's reversed where gaming piggybacks on common aspects to compute.

Re: Nvidia won, we all lost

#342

Post by PoshBreeze » Sat, Jul 05, 2025, 9:42 AM UTC

esseph wrote: post #334

PoshBreeze wrote: post #326

Earlier quoted context omitted.

I will be honest, I just gave up. I couldn't get consistent performance on HellDivers 2. Many of the things you have mentioned I've tried and found they don't make much of a difference or made things worse. I did get it running nice for about a day and then an update was pushed and it ran like rubbish again. The game runs smoothly when initially running the map and then massive dip in frames for several seconds. This…

I used Debian for about 15 years. Honestly? Fedora is really the premier Linux distro these days. It's where the most the development is happening, by far. All of my hardware, some old, some brand new (AMD card), worked flawlessly out of the box. There was a point when you couldn't get me to use an rpm-based distro if my life depended on it. That time is long gone.

I don't want to use Fedora. Other than I've found it unreliable I switched to Debian because I was fed up of all the Window-isms/Corporate stuff in the distro that was enabled by default that I was trying to get away from.

It the same reason I don't want to use Bazzite. It misses the point of using a Linux/Unix system altogether.

I also learned a long time ago Distro Hopping doesn't actually fix your issues. You just end up either with the same issues or different ones. If I switched from Debian to Fedora, I suspect I would have many of the same issues.

e.g. If a issue is in the Linux kernel itself such as HDMI Audio on AMD cards having random noise, I fail to see how changing from one distro to another would help. Fedora might have a custom patch to fix this, however I could also take this patch and make my own kernel image (which I've done in the past btw).

The reality is that most people doing development for various project / packages that make the Linux desktop don't have the setup I have and some of the peculiarities I am running into. If I had a more standard setup, I wouldn't have an issue.

Moreover, I would be using FreeBSD/OpenBSD or some other more traditional Unix system and ditch Linux if I didn't require some Linux specific applications. I am considering moving to something like Artix / Devuan in the future if I did decide to switch.

Re: Nvidia won, we all lost

#343

Post by tankenmate » Sat, Jul 05, 2025, 9:45 AM UTC

jezze wrote: post #320

datagram wrote: post #309

Earlier quoted context omitted.

AMD cards are fine from a raw performance perspective, but Nvidia has built themselves a moat of software/hardware features like ray-tracing, video encoding, CUDA, DLSS, etc where AMD's equivalents have simply not been as good. With their current generation of cards AMD has caught up on all of those things except CUDA, and Intel is in a similar spot now that they've had time to improve their drivers, so it's pretty e…

I have no experience of using it so I might be wrong but AMD has ROCm which has something called HIP that should be comparable to CUDA. I think it also has a way to automatically translate CUDA calls into HIP as well so it should work without the need to modify your code.

`I think it also has a way to automatically translate CUDA calls`

I suspect the thing you're referring to is ZLUDA[0], it allows you to run CUDA code on a range of non NVidia hardware (for some value of "run").

[0] https://github.com/vosen/ZLUDA

Re: Nvidia won, we all lost

#344

Post by ohdeargodno » Sat, Jul 05, 2025, 9:45 AM UTC

kbolino wrote: post #69

ohdeargodno wrote: post #44

Earlier quoted context omitted.

It's not that it's difficult to turn off TAA: it's that so many modern techniques do not work without temporal accumulation and anti-aliasing. Ray tracing? Temporal accumulation and denoising. Irradiance cache? Temporal accumulation and denoising. most modern light rendering techniques cannot be done in time in a single frame. Add to that the fact that deferred or hybrid rendering makes implementing MSAA be anywhere…

A lot of this is going to come down to taste so de gustibus and all that, but this feels like building on a foundation of sand. If the artifacts can be removed (or at least mitigated), then by all means let's keep going with cool new stuff as long as it doesn't detract from other aspects of a game. But if they can't be fixed, then either these techniques ought to be relegated to special uses (like cutscenes or the ba…

So, there is a way to make it so that TAA and various temporal techniques look basically flawless. They need a _lot_ of information and pixels.

You need a 4k rendering resolution, at least. Modern effects look stunning at that res.

Unfortunately, nothing runs well at 4k with all the effects on.

Re: Nvidia won, we all lost

#345

Post by duckmysick » Sat, Jul 05, 2025, 9:48 AM UTC

__turbobrew__ wrote: post #49

> With over 90% of the PC market running on NVIDIA tech, they’re the clear winner of the GPU race. The losers are every single one of us. I have been rocking AMD GPU ever since the drivers were upstreamed into the linux kernel. No regrets. I have also realized that there is a lot out there in the world besides video games, and getting all in a huff about it isn’t worth my time or energy. But consumer gotta consoooooo…

> I have also realized that there is a lot out there in the world besides video games, and getting all in a huff about it isn’t worth my time or energy.

I'd really love to try AMD as a daily driver. For me CUDA is the showstopper. There's really nothing comparable in the AMD camp.

Re: Nvidia won, we all lost

#346

Post by tankenmate » Sat, Jul 05, 2025, 9:51 AM UTC

senko wrote: post #321

rewgs wrote: post #297

Earlier quoted context omitted.

> I use Linux and have learned not to touch AMD GPUs (and to a lesser extent CPUs due to chipset quality/support) a long time ago. Even if they are better now, (I feel) Intel integrated (if no special GPU perf needed) or NVidia are less risky choices. Err, what? While you're right about Intel integrated GPUs being a safe choice, AMD has long since been the GPU of choice for Linux -- it just works. Whereas Nvidia on L…

Had major problems with xinerama, suspend/resume, vsync, probably a bunch of other stuff. That said, I've been avoiding AMD in general for so long the ecosystem might have really improved in the meantime, as there was no incentive for me to try and switch. Recently I've been dabbling in AI where AMD GPUs (well, sw ecosystem, really) are lagging behind. Just wasn't worth the hassle. NVidia hw, once I set it up (which…

I run llama.cpp using Vulkan and AMD CPUs, no need to install any drivers (or management software for that matter, nor any need to taint the kernel meaning if I have an issue it's easy to get support). For example the other day when a Mesa update had an issue I had a fix in less than 36 hours (without any support contract or fees) and `apt-mark hold` did a perfect job until there was a fix. Performance for me is within a couple of % points, and with under-volting I get better joules per token.

Re: Nvidia won, we all lost

#347

Post by duckmysick » Sat, Jul 05, 2025, 9:53 AM UTC

theshackleford wrote: post #285

globalnode wrote: post #136

Earlier quoted context omitted.

good move, thats why i treat my windows install as a dumb game box, they can steal whatever data they want from that i dont care. i do my real work on linux, as far away from windows as i can possibly get.

Same way I treat my windows machine, but also the reason I wont be swapping it to linux any time soon. I use different operating systems for different purposes for a reason. It's great for fompartmentalization. When I am in front of windows, I know I can permit myself to relax, breath easy and let off some steam. When I am not, I know I am there to learn/earn a living/produce something etc. Most probably do not need…

What works for me is having different Activities/Workspaces in KDE - they have different wallpapers, pinned programs in the taskbar, the programs themselves launch only in a specific Activity. I hear others also use completely different user accounts.

Re: Nvidia won, we all lost

#348

Post by tankenmate » Sat, Jul 05, 2025, 9:54 AM UTC

trynumber9 wrote: post #168

kyrra wrote: post #129

Earlier quoted context omitted.

Frame per watt they aren't as good. But they are still decent.

They seem to be close? The RX 9070 is the 2nd most efficient graphics card this generation according to TechPowerUp and they also do well when limited to 60Hz, implying their joules per frame isn't bad either. Efficiency: https://tpucdn.com/review/gigabyte-geforce-rtx-5050-gaming-o... Vsync power draw: https://tpucdn.com/review/gigabyte-geforce-rtx-5050-gaming-o... The variance within Nvidia's line-up is much larger…

I run 9070s (non XT) and in combination with under-volting it is very efficient in both joules per frame and joules per token. And in terms of purchase price it was a steal compared to similar class of NVidia cards.

Re: Nvidia won, we all lost

#349

Post by kevincox » Sat, Jul 05, 2025, 10:14 AM UTC

klipklop wrote: post #295

__turbobrew__ wrote: post #49

> With over 90% of the PC market running on NVIDIA tech, they’re the clear winner of the GPU race. The losers are every single one of us. I have been rocking AMD GPU ever since the drivers were upstreamed into the linux kernel. No regrets. I have also realized that there is a lot out there in the world besides video games, and getting all in a huff about it isn’t worth my time or energy. But consumer gotta consoooooo…

You are certainly right that this group has little spending self-control. There is no limit just about to how abusive companies like Hasbro, Nvidia and Nintendo can be and still rake in record sales. They will complain endlessly about the price of a RTX 5090 and still rush out to buy it. I know people that own these high end cards as a flex, but their lives are too busy to actually play games.

I'm not saying that these companies aren't charging "fair" prices (whatever that means) but for many hardcore gamers their spending per hour is tiny compared to other forms of entertainment. They may buyba $100 game and play to for over 100 hours. Maybe add another $1/hour for the console. Compared to someone who frequents the cinema goes to the pub or does many other common hobbies and it can be hard to say that games are getting screwed.

Now it is hard to draw a straight comparison. Gamers may spend a lot more time playing so $/h isn't a perfect metric. And some will frequently buy new games or worse things like microtransactions which quickly skyrocket the cost. But overall it doesn't seem like the most expensive hobby, especially if you are trying to spend less.

Re: Nvidia won, we all lost

#350

Post by solatic » Sat, Jul 05, 2025, 10:24 AM UTC

strictnein wrote: post #105

This really makes no sense: > This in turn sparked rumors about NVIDIA purposefully keeping stock low to make it look like the cards are in high demand to drive prices. And sure enough, on secondary markets, the cards go way above MSRP Nvidia doesn't earn more money when cards are sold above MSRP, but they get almost all the hate for it. Why would they set themselves up for that? Scalpers are a retail wide problem. A…

Scalpers are only a retail-wide problem if (a) factories could produce more, but they calculated demand wrong, or (b) factories can't produce more, they calculated demand wrong, and under-priced MSRP relative to what the market is actually willing to pay, thus letting scalpers capture more of the profits.

Either way, scalping is not a problem that persists for multiple years unless it's intentional corporate strategy. Either factories ramp up production capacity to ensure there is enough supply for launch, or MSRP rises much faster than inflation. Getting demand planning wrong year after year after year smells like incompetence leaving money on the table.

The argument that scalping is better for NVDA is coming from the fact that consumer GPUs no longer make a meaningful difference to the bottom line. Factory capacity is better reserved for even more profitable data center GPUs. The consumer GPU market exists not to increase NVDA profits directly, but as a marketing / "halo" effect that promotes decision makers sticking with NVDA data center chips. That results in a completely different strategy where out-of-stock is a feature, not a bug, and where product reputation is more important than actual product performance, hence the coercion on review media.