Hacker News new | ask | show | jobs
by jayd16 5 days ago
>[..] take full use of the PCIe bandwidth of the GPU or the bandwidth of its GDDR memory.

I'm honestly a little confused by what you mean here. Why would we want to maximize those things? Games are about consistent output under the frame deadline, not full saturation of the hardware.

Why would anyone try to saturate a 5090 with their game? The addressable market is tiny and you'd have to hope their full spec runs as well as or better than your test rig or they'll still not hit framerate.
1 comments
simonbw 5 days ago
You could do some sort of adaptive quality where you spend time incrementally improving fidelity until your frame budget is up. In practice I think that might be trickier than it sounds, but I feel like theoretically there's something there that could get you the best graphics your rig can handle without dropping frames. I've been considering doing something like this when I've been building a game/engine lately.
link
Rohansi 5 days ago
There's only so high you can go because the game assets have a maximum quality. Maybe you'll be able to max out the 5090 but what about the next flagship GPU?

You're also likely not going to maximize all of bandwidth, compute, etc. because one of them will likely be your bottleneck. And it might be different depending on the GPU, too.

link
rustystump 4 days ago
Most games are strictly scaled on resolution due to how deferred pipelines run. This is exactly the slider to max or not max everything on a gpu for games. The more pixels the more memory and the more compute.
link
Rohansi 4 days ago
If you're rendering at native resolution, which many PC gamers do, going higher isn't significantly better because it just helps with antialiasing via supersampling. There's no point rendering so much more pixels just because you can, that's just a waste of electricity.
link
jayd16 4 days ago
The more pixels, the more compute of fragments but not necessarily more memory. A fragment might hit the same texel as an adjacent fragment.

Certainly not more from main memory, and maybe not more from the vram either depending on how the pipeline goes.

It's not a linear slider.

link