Oof, the 7900 XTX only barely edges out the 4080 in raster performance by 2% at 1440p and 4% at 4K according to TPU. I was expecting a larger margin honestly.
And don’t even get me started on the XT. 15% less performance for 10% less price? And this is at the high end where extra performance is supposed to be at a premium. Even at $799 the XT would have been meh. Should have been $699.
@dave
Also, is it me, or do we live in an actual gaming clown show?
If ai understand it, almost all games are still on the level of last gen consoles, that can be matched by even Steam Deck at this point, so when it comes to gaming, these cards shouldn't be needed for another 6 years.
@caekislove @LukeAlmighty A quick google search says they’ve sold “over a million” units. While you can call that a success for a mobile PC and argue that devs will make sure they have a Deck-compatible configuration, if we’re talking consoles that’s peanuts. The Wii U was considered a commercial failure and that still sold 13 million units.
@caekislove @LukeAlmighty The switch is probably the closest console “competitor” and that has sold almost 115 million.
@caekislove @PapaPole @dave
I think it's good.
While I am not a big fan of "e-wase" argument, it is good to take things slow for a few years. It's not like new games are worth the investment anyway, and less demanding graphics might allow some more creative projects to grab the attention.
All the best games ran on total dog shit software so going from 128KB of RAM on the SNES, which had Mario Kart and Donkey Kong Country, games worth playing today, to 16GB of RAM on the PS5 is underwhelming. I’m not sure we need a 125,000-fold increase in RAM to render the blue hair on the mulatto transbian’s head but that’s just me.
@PhenomX6 @PapaPole @caekislove @dave @NEETzsche
This video is literally called:
Someone improved my code by 40,832,277,770%
I think that should speak for itself about "compiler knows the best"
https://m.youtube.com/watch?v=c33AZBnRHks
@LukeAlmighty @PhenomX6 @PapaPole @caekislove @dave @NEETzsche The “compiler knows the best” is more about avoiding micro optimizations like the Duff’s device. Compilers are actually pretty good at optimizing code at the dozen-of-lines level. They’ll do loop unrolling, vectorization, and even deduce closed-form versions of given code. The reason inline assembly is frowned upon is that it makes the optimizer retarded; it won’t try to figure out what it does, so it acts like a fence in the code where optimizations on one side can’t affect optimizations on the other. That said, the optimizer optimizes what’s there. It can’t fix bad design, and it isn’t omnipotent to what you want. If you really need that sonic speed you should first make sure what you’re doing is “conceptually optimized” and then hand write assembly at the function level after you’ve started benchmarking for bottlenecks and taken into consideration the branch predictor and CPU cache.