Reading anti-tech innovation comments from sinophobic losers is my new Roman Empire.
Education should be an issue for you though. 38% of electricity is generated from green energy in China. The carbon intensity of electricity generation in China was 560g of CO2/kWh in 2024, which was at the same level as the U.S. in 2010. Where were you back then criticizing the environmental impact of iPhones? Or is your greenwashing bs reserved only for China?
You are wasting your time, these anti-tech innovation sour losers don't want to see the reality, they prefer to believe lies to feel safer.
It's truly sad to see individuals blinded by hate against China becoming anti-tech advocates. Living in fear and spreading hate will not bring anything positive to your life.
This site is about learning and sharing of tech news and information. You're right that some people are unreachable, but the effort isn't wasted on many of us.
"Breakthroughs" are magnitudes easier to achieve when you're starting from garbage and have the competition to copy from. The only western tech company that gets praise for copying the competition is AMD. Everyone else has to develop their own breakthroughs to get praise, and most of the time they don't get it then.
Spoken like a disgruntled Intel shareholder!
Spoken like someone who thinks they are clever and snarky, but as usual missed the point the poster was making.
If it were as you say, they wouldn't have been way out in front of everyone else in their use of chiplets and 3D cache.
Yeah, now that they had early partners, like AMD, who helped them refine it. Even then, it's not like you just click a checkbox in your chip layout tool.
That's backwards logic: "Intel isn't doing it, so it must not be smart or worthwhile." I think the real answer is that Intel is still trying to master chiplets, as we see from Arrow Lake's underwhelming performance. They'd better figure out how to do that right, before adding any more complexity.
AMD's GPU situation mirrors where their CPU situation was, about 5 years ago. They had fallen behind their competition and were steadily catching up. The big difference is that Nvidia never took its foot off the gas, like Intel did (i.e. due to its 10 nm problems).
GPU Direct is somewhat of a gimmick, in that it only works in very specific hardware setups. For its part, AMD had SSG back in 2016.
No, pretty much just DLSS. And that was pretty bad, until version 2.0.
That's not what I see, but maybe we run in different crowds.
Which is something they accidentally discovered was meaningful for client parts (bit_user already explained the co-design aspect). I don't think this was anywhere near as innovative as their chiplet strategy. They've shared enterprise and client chiplets which allows them to gain benefits from each side of the equation (3D V-cache and AVX512 may not have ever graced client without this). AMD also streamlined their CPU design much sooner than Intel did which has allowed them to do revisions faster.
I can't think of a single thing you've listed that people said one should buy AMD over nvidia to get. AMD has certainly gotten praise for the more open nature of things like VRR, upscaling and frame generation but none of those have been better than nvidia's implementation.
I don't know about that, but it certainly wasn't as fundamental to their successful CPU business strategy. What it did accomplish was letting them put a feather in their cap and winning them a position of prestige, among gamers.
Who can say, since even the Zen 4 APUs added support for it? Maybe they only did that because their desktop processors supported it, but maybe they'd have done it anyway, since the hard part is the initial implementation it and it doesn't really cost that much to leave it in the mobile version of the core.
That was effectively an unintended side effect of their chiplet strategy. Its primary purpose hasn't been anywhere near the success.
I think doing it with SMT may have also made this trickier than it otherwise could have been.
I disagree. As Intel has said, their E-cores are faster than if you pair the thread with another one, on the same P-core. It should therefore be no more difficult to schedule than SMT and even provide a middle step between running a thread exclusively on a P-core vs. having it share a P-core with another. In fact, if Intel had done more to help latency-sensitive apps effectively use SMT, then it should've carried over nicely to hybrid.
My train of thought is more about addressing the software side in a more simple fashion rather than having to optimize scheduling for three potential thread locations.
E-cores were never going to stay small with the designs Intel was using in mobile and the enterprise application.
It's not well-optimized for two. The existing core-affinity APIs are not a good solution and I'm therefore not sure how many apps even use them.
I know the rise of the E-core has turned more into the rise of the Medium-core, but PPA would be better with more Gracemont/Crestmont-style cores than Skymont. I think a lot of the reason why Skymont improved quite so much was to reduce the scheduling hazard of a latency-sensitive thread landing on one. Also, since AMD hasn't gone to a true E-core model, there wasn't really pressure on Intel to stick with that.
E-cores seem to be used a lot more for more standard operations with ARL whether that's due to necessity or bad software I can't say. It is pretty wild to see overclocking E-cores improving game performance for some titles. I don't recall hearing of that being the case with ADL/RPL though that may also have been due to less headroom.
Just have to point out that it did borrow from MIPS, at least on the system level. When they first added support to it in the Linux kernel, they copied vast amounts of code straight out of the MIPS tree, to the point where the kernel maintainers wondered why they didn't just add it as another flavor of MIPS.
Borrow from? That sure is biased, they were all in on MIPS, they built MIPS chips like AMD builds x86, or Apple builds ARM, and they made good headway in that area, but MIPS kind of pulled the rug from under them. So they pivoted away.
It's not biased, it's a fact. They lifted a lot of the system architecture stuff from MIPS.
Facebook
Twitter
Instagram