3 ms·
I think much as with Windows, x86 is backstopped by gaming. It's going to be hard to kill because of that. Gaming is hardware-intensive, highly locked into cl
by hakfoo 2y ago
I think much as with Windows, x86 is backstopped by gaming. It's going to be hard to kill because of that.
Gaming is hardware-intensive, highly locked into closed source ecosystems, and users are not going to be amenable to alternatives.
You can't say "we recompiled the top 10 games for ARM" and have it be enough. You need to be able to offer something that can deliver high end native-level performance on pretty much 99% of the Steam catalog. That means either herding a lot of cats to get code recompiled for ARM, or developing breakthrough emulator performance.
Apple pulled off the transition because they were targeting a different audience. With productivity software, you could port a few key programs, with cooperative developers, and cover a lot of the big markets. They also dodged the raw performance question because they could target messages like "fanless", "lighter", and "longer-running" for M1, competing with existing mid-performance laptops, rather than a direct challenge to a fire-breathing gaming desktop.
- nothercastle 2y agoIt might take a while say 5-10 years but once 90% of games are ARM compliant It will be a non issue. Also ARM performance can be so much better than x86 and gamers love raw benchmarks and high fps scores I would bet they will switch to arm for 10-15fps.
- Wytwwww 2y ago> Also ARM performance can be so much better than x86 Why would that be the case? Also ARM/Qualcomm haven't been taking desktop seriously at all and there is no indications that that might change anytime soon. ARM doesn't have any reference designs between Neoverse and the low-power/mobile ones. Even Apple, if we look at raw performance the M2 Ultra in the Studio (for $4000) get destroyed by the $350 14700K. Their GPUs are nothing special either besides the shared memory which is useful for local AI.
- hakfoo 2y agoClaims about ARM performance tend to be qualified by "for the wattage". Yeah, M2 is impressive at a 10 or 20 watt envelope, but once you're looking at 45 and 65w parts, it's no longer a washout for Ryzen and Core. If you told the Qualcomm gurus "here's a 200W TDP budget, go to town", would they be able to make the most of it? I suspect when you're making a wattage-no-object desktop part, it opens a whole new candy shop of process and design choices, so you might have to start from a relatively blank slate to use them effectively. I wonder how it would compete with taking a modern x86 CPU and replacing the front-end decoder and microcode with ARM ones-- the other 90% already being optimized for a higher-wattage budget.