Posts by zipityzi 15 publicly visible posts • joined 12 Mar 2024
Interestingly, even some of Arm's smallest &w weakest cores have SMT: see the Neoverse E1.
"The Arm Neoverse E1 CPU delivers best in class throughput efficiency. It incorporates a new simultaneous multithreading (SMT) microarchitecture design. With SMT, the processor can execute two threads concurrently resulting in better aggregate throughput performance."
Now, how effective and how useful is SMT on the E1? I can't say that I've ever seen a design with it.
I also didn't know this was possible: Arm allows some Neoverse configuration options, but nothing deep inside the microarchitecture.
However, Arm's interview The Next Platform by Timothy Prickett Morgan appears to aligns with the El Reg's reporting.
Q:Is there any cloud builder or hypervisor to date that doesn’t use a Neoverse core and that uses a customized core in any way? To my knowledge, there isn’t, although there might be something happening with Nvidia’s future “Vera” server CPU. I strongly suspect that the Nvidia slide that said “custom Arm core” meant to say Compute Subsystem.
Mohamed Awad, Arm: There is no hyperscaler or cloud builder today that is building an SoC based on a custom architecture. Their SoCs are all built on Neoverse."
That was quite surprisingly, as NVIDIA always calls it "custom".
My hypothesis: I would not be surprised if Arm and NVIDIA had a quiet sweetheart deal where Arm's Neoverse architects worked with NVIDIA's architects to design the Olympus core (for a fee above the typical ALA & TLA prices). I do not imagine it was cheap, as Arm would have to go above & beyond their typical ALA / TLA obligations to make this work.
Native apps will be ported more slowly the longer the market share is so weak. Qualcomm really should’ve made these much cheaper. People will gloss over flaws if the product didn’t cost this damn much.
I also hate how hidden and coy they are about compatibility. Did you know the entirety of Snapdragon X1 product line cannot run Windows Night Light on an external display?
I didn’t even know that feature required a compatibility matrix. It’s such an ancient, minor feature that has nothing to do with Arm the ISA, but Qualcomm never implemented it.
"Fresh" is relative, but Metronet is a US fiber ISP founded in 2005, so 20 years ago.
Metronet has—since 2005, mind you—implemented an exclusively 100% CGNAT network. The "holdover" technology is apparently the state of the art for some ISPs.
Anyways, Metronet were recently acquired by private equity firm KKR and T-Mobile.
What’s unusual to me: 20A as a testing ground was the financially prudent choice. Isn’t $500m of validation worth it to ensure 18A rolls out near-perfectly?
20A as a validation step was billed as critical. Intel getting cocky about 18A and trying to save costs sounds like a terrible 2010-era Intel mentality.
It is illogical, or at least inconsistent.
Cut costs in other places, not anything related to 18A’s delivery. The latest nodes are the wrong place to cut funding, especially in what is essentially a major validation test.
I read 1Password Release notes pretty religiously and 1Password never mentioned this CVE (nor any security vuln fix) in the 8.10.36 patch notes. Luckily, I'd already updated.
https://releases.1password.com/mac/8.10/#changelog
Critical security patches should always be mentioned in release notes: it gives users a heads up & pushes them update other devices. Not a good look as a security-focused company.
> Snapdragon maker accounted for 10 percent of Arm's total revenue for the fiscal year ended March 31, 2024, meaning that any victory against Qualcomm could be a Pyrrhic one.
IIRC, doesn’t Qualcomm use Arm codes in its WiFi chipsets and what-not? So Snapdragon SoCs alone can’t be all 10%.
Maybe Arm thinks Qualcomm is moving to totally abandon Arm and migrate to RISC-V? I know they’re making an Android watch SoC, but that’s a far cry to replace the likely thousands of individual Qualcomm dies that incorporate Arm cores.
Nah, a lot of macOS users in the past few years had never used macOS for any serious amount of time.
macOS jumped from 18% to 24% in six years across all U.S. desktops, a significant +33% growth:
Source: https://gs.statcounter.com/os-market-share/desktop/united-states-of-america/#quarterly-201801-202402
It’s a messy measurement rate, but it’s clear a lot of Windows users (see decline) have no problem with macOS or ChromeOS.
Very few PC consumers care about ancient Windows software. There’s a reason old software hasn’t been updated in decades: it probably has good enough perf on a potato emulator via four levels of virtualization.
The # of lines of x86 code that simply exist are irrelevant. It needs to be x86-exclusive code that doesn’t have an official native Arm port times the number of active users. And then subtract any with no enough alternatives that do have a popular Arm native port from other vendors. And subtract any that work well enough under Prism.
Major Windows software has already been ported to Arm native that you conveniently ignore: Chrome, VLC, Handbrake, MS Teams, GIMP, Paint.net, Da Vinci Resolve, the Affinity Suite,7-zip, NotePad++, Everything, MS PowerToys, etc.
Active Backup is one of the least reliable, most finicky backup suites I’ve used. Well, backing up was easy. Restoring was a nightmare: nonsense criteria restrictions (even on the same hardware), convoluted driver backup with incomplete drivers in their PE environment, and incomprehensible error codes.
I used it on DSM 7.1 & 7.2 for two weeks, stress testing its backup & restore process. I felt like they’d borrowed Microsoft’s 2010-era disk imaging platform with how incomplete it was.
Just read Synology’s own docs to see how finicky it is.
I’ve seen too many YouTube influencers peddling it, without making a serious comparison to veteran SME / SMB backup platforms.
I’d much rather point people to Veeam, Macrium, Nakivo, etc.
Active Backup for Business, I fear, is ”free” for a reason, like much else first-party apps in Synology’s Package Center: barely supported, looks pretty and useful on the surface, and full of annoying or fatal restrictions that other vendors fixed and improved decades ago.
I don’t mean to be overly harsh, but so much feels like abandonware that was developed to “fill a niche, poorly, but just check off the box.”
Maybe with this new push, Synology will take it more seriously. But I doubt it: ABB has shown little improvement in years.
Why did more Intel Foundry have more sales / revenue in 2022 than 2023?
I understand Intel used “modeling” to edit its older financial data to create the split between Intel Foundry vs Intel Product, so maybe 2022 had been previously underreported.
But 2022 & 2023 were under Gelsinger’s reign. Surely Intel Foundry revenue should at a minimum hold steady year over year, if this plan is working.
How did Intel Foundry sell less in 2023? Lower profits, sure. Lower revenue?
> However, "Using a different computer with a different amount of memory, a different processor or a different operating system makes the biggest difference of all, and these tests do not cover any of that."
That is an unexpected comment. Shouldn’t Speedometer 3.0 also include the JS execution time, which would depend on the CPU?
