Snapdragon X Elite in the wild is allegedly slower than iPhone 12 | Hacker News Reader

Snapdragon X Elite in the wild is allegedly slower than iPhone 12

tomshardware.com

20 pointsz_27 comments

This is certainly plausible for early hardware but I’d wait for more widespread confirmation before treating it as anything more than a fluke. It’s very easy to imagine that a new platform might have something like power management or driver issues which can be resolved without touching the hardware.

The post mentioned in article is full of details / photos:

https://www.reddit.com/r/GalaxyBook/comments/1dd7t0v/samsung...

I'm under the impression that apple silicons dominance largely comes from TSMC. And that Qualcomm won't be able to match that. Am I wrong here?

Snapdragon X Elite is manufactured using TSMC N4.

Source: https://www.anandtech.com/show/21445/qualcomm-snapdragon-x-a...

It’s not just TSMC - they also have a very good CPU team giving AMD heartburn even when you compare the chips on the same TSMC process, and they have the part which is hardest to match: they control the OS, compilers, libraries, and key apps. That allows them to ensure that hardware acceleration is actually used, tailor chips for the power and usage patterns their devices need, and avoid the conflicts of interest inherent to multi-vendor systems because they share responsibility for a great device and profits aren’t unevenly distributed. That’s not perfect and they’re certainly in a bad place if they misjudge but it’s worked so far.

I haven't looked at CPU architectural comparisons, but Apple must have a major benefit in owning the whole stack: from the lowest hardware level to the OS to many of the applications. And with the right focus (which they almost certainly have) the ability to optimize the system as a whole instead of each level taking care of just their own section.

If the x86 translation team can make their case that a specific CPU feature will be a significant enough performance benefit, chances are that a discussion within the company will have less friction that a discussion between, say, a Microsoft SW team and a Qualcomm or ARM CPU design team.

And while Microsoft may care about power consumption and power management, I can't image that they care as much as the Apple MacBook team. (My POS high-end Dell laptop is still a power management disaster. It would fail any internal go-to-market review at Apple.)

But there's also the simple fact that the Apple silicon team is very good, and they've been cranking out best in class silicon IPs for more than a decade. A lot of IPs on an SOC are developed incrementally, with smaller improvements from one generation to the next accumulating into major advancements over the years. Even if the competition builds an equally capable silicon team, they may not have the solid foundation to build on and it can take generations to match what the competition already had.

It's also true that for Apple, the silicon is not a product that's being sold on its own. I'm sure they care about silicon area (and thus cost), but they can once again put this cost in the larger context of the full laptop. For Qualcomm, increasing a cache by 50% is a much bigger hit on their gross margins than for Apple. If that cache increase will result in significantly better laptop characteristics, Apple may decide to go for it while Qualcomm may not.

For Dell, the customer is corporate Information Technology departments and accountants. Most of their end-user kit is minimum viable brick for business use: if it isn't in the box, it can't break. I find their engineering impressive.

Apple tells corporate IT to get out of the way of their customers, the users. Thus repairing their stuff is a nightmare.

(My laptop is an M1 MacBook Air. My servers are off-lease salvage Dell or Intel OEM hardware.)

If I close my laptop, put it in a backpack, and it comes out hot like a bun out of the oven, I don't consider that a minimum viable solution.

Apple's core "Silicon" team was the Centaur Technologies chip design group. At the time, they had brought to market a PowerPC CPU chip that was a breakthrough in performance per watt -- for 32-bit PowerPC. Which had already lost to Intel for mainstream applications.

At Apple, Centaur joined engineers who had helped to develop early ARM RISC processors.

While I know that the talent pool migrates, engineers have moved on, Apple has retained deep institutional knowledge of how to build systems that work really well. And developed teams worldwide to deliver manufacturing at scale.

By no means are they the only company to do so. Vast resources in competition. But please understand that Apple isn't new to full-stack hardware design.

I believe you’re thinking of PA Semi not Centaur (which designed x86 chips not PowerPC or ARM)

Oops. Thanks for the correction! It was PA Semi, indeed.

Review embargo lifts next week.

Based on the original reddit thread discussion benchmarks are running with the cpu peaking at 2.5ghz when x elite should be able to go up to 4ghz. Seems like there me be some firmware issue in this model. What's more interesting is that the reviewer was able to play resident evil village at 60-100fps on battery using the x64 version of the game from steam.

Like any PC laptop, the results are going to vary wildly based on the manufacturer's thermal solution. You can't buy based on the CPU part number. You have to wait for reviews of specific machines and buy the one that's actually good.

Sensationalist article based on a reddit post which the OP already updated stating it has to do with a driver or chipset issue as the CPU did not go over 2.52GHz.

Bit of a clickbait headline

This is based on a Reddit report of a CPU that seems like it’s not boosting…ofc the score will be low

Yeah a cpu clocking 2.5 isn’t going to give you the same score you see at designed 4+. Unless it was limited that way due to thermals this is likely not a big deal in long run

So it’s slow by default but if it boosts properly will it still retain the battery life improvements?

I think the implication is that there's just currently a bug.

Can I run linux on the iPhone 12? And not talking about UTM VMs.

You can not run Linux on IPhone 12. Neither you can run Linux on Snapdragon X laptops in any meaningful way, today. While there was some Linux demos before Computex, there was no Linux demos on Computex, and people presenting the laptops were declining to comment about Linux.

Is WSL2 available at launch day? If not that will be a deal breaker for me.

WSL2 already works on Windows on ARM, so I would assume so.

Incorrect.

https://www.tuxedocomputers.com/en/In-the-media-TUXEDO-proto...

"TUXEDO Computers showcased a prototype of its ARM notebook at the Computex trade fair in Taiwan. This prototype is part of the company’s ongoing project, internally dubbed “Drako,” aiming to port TUXEDO OS with KDE Plasma to the ARM platform. The development is currently in the alpha stage, with some drivers still pending. TUXEDO is optimistic about resolving these issues with the next two kernel versions. While a release date is not yet set, the possibility of an ARM notebook under Christmas trees in 2024 is tantalizingly close."

"Qualcomm has been working to ensure good Linux support for the Snapdragon X Elite high-end laptop SoC. That work is still ongoing but the initial bits are already in the mainline kernel while other optimizations and other features remain in the works. So far for the Snapdragon X Elite has been many vendors showing off Microsoft Windows based laptops but it looks like from at least TUXEDO Computers there will be a Linux-focused laptop using this ARM64 chip."

Article isn't saying the Snapdragon is worse than iPhone 12, just that real world performance does not seem to match prerelease benchmarks.

No but you can run Linux on the Apple m-series laptops using Asahi Linux.

This is a laptop chip, right?

When you rely on volunteers blindly reverse engineering Apple hardware you're never going to reach a level of support and stability that meets mission-critical requirements. It's a hobbyist solution.

https://www.youtube.com/watch?v=_TEBKEhhcqM

It's amusing because Linux started as a hobbyist solution. I think the results will speak for themselves. Asahi Linux is quite usable with some limitations. Corporate support for Linux doesn't guarantee usability based on past efforts.

Reply on news.ycombinator.com