It's got about the same specs as my main webdev laptop -- which is a Thinkpad X200t -- with a modern GPU.

It's got a mechanical keyboard and a trackball!

[0] https://source.mnt.re/reform/mnt-reform-layerscape-ls1028a-s...

Same how you can't go to TSMC and be like "hey, you guys are making millions of M1 chips for Apple, ca you sell us some of those pls?"

Or how did you imagine such a scenario would play out for Framework?

Is there any movement for a smaller standardized connector? Even small phones could benefit.

Erm... "only"?! I've never seen anything faster than 1GB ethernet in a laptop before, or even in a consumer desktop/tower.

Am I missing something?

Going forward, both AMD and Intel have built 2.5 gbe controllers into their cpus and chipsets, so if partners want to use it, it's there. As such you will see it start to be rolled out even in consumer products over the next couple years. The initial rollout was hampered by a few "light errata" with intel's controller implementation (basically it didn't work at 2.5 gbit speeds at all, oops) but we're a couple gens into it now and it's going forward. AMD has one on their chips as well (if not faster, I want to say AMD has 10gbe in their dies for Ryzen Embedded usage).

(expecting it in a 2020-era product, however, is perhaps not reasonable, like I said this is something that has changed mostly in the last year or so. Although granted a Thinkpad is not a consumer product so maybe it should have been ahead of the curve. But I don't think multigig had penetrated even professional spaces that well in 2020, except perhaps 10gbe in certain high-end segments.)

The gotcha with multi-gig has always been switches, very few cheap switches existed and the ones that did were mostly SFP+ and not 10GBase-T (i.e. you either need new wiring anyway, or you pay $50 a set for SFP+ adapters). This is somewhat better with 2.5gbe, there are now some 8-port switches under $200, but it's still a long way from 1gbe where you can get a nice 8-port for like $25. Hopefully it is a matter of volume and prices come down over time.

I'm a little miffed that 10gbe got passed over - 2.5gbe seems like a half-measure, and the wiring argument isn't very compelling to me. If they implemented 10gbe you could always allow fallback to 2.5gbe if that's what the signal quality can support (many <10 meter Cat5e runs are good enough for full 10gbe anyway) and it's a much more significant step than 1->2.5gbit. And the economy of scale here is significant, if consumer stuff goes 2.5 then that's what gets made and 10gbit remains expensive.

Anyway currently I am eyeballing the Netgear MS510TXM - this does pretty much what I want in terms of giving me a few SFP ports for fiber links (or server usage - SFP does have lower latency), a few 10gbase-T links for consumer PCs with Aquantia 10gbit nics, and a few 2.5 gbit ports for consumer junk. At $500 it would be nice if it was a 12-port, but it checks most of the boxes at least.

Still, as much as I'd like it myself, I can't imagine there is any push at all from consumers, and very little from prosumers. I guess it's a bit of a chicken and egg thing.

It's not every day you need multi-gig, but it is very nice when you do end up doing something big that hammers your network hard. You do need to be serving from a SSD or a NAS, neither of which are particularly unusual anymore, but I guess for most consumers they don't really work with "their media" at all anymore, everything has transitioned to the cloud for average consumer. Everything is streamed, and if you want to back up a steam game or something, you just delete it and re-download it if you want it later.

Bet that kinda sucks for people in Canada or australia on super limited data caps though, it's kind of a weird US- and euro-centric perspective.

As a prosumer, it's a bit of a mentality change, but for things like photo or video editing, you don't need to have tons and tons of fast storage on your local machine, you can push it off to to your NAS and just manage it like the rest of your data, which is a huge convenience thing (in terms of backups, and in terms of capacity). 4-drive or 8-drive NAS can push a ton of data even if random isn't quite as fast as a SSD (and that's why files get cached in RAM anyway).

What?!? I live in Canada and have not had a data cap in over 10 years. They might exist for some ISPs but they are generally high enough that I've never heard anyone complain or get charged.

(SFP has a limit of 7 meters for copper cables, I think it can go higher for active cables, but most people use fiber for longer runs)

If you just wanted to plug into a switch that was like 5 feet away, SFP copper cabling would actually reduce power a fair bit. Not really sure where fiber falls, but as far as copper, the problem is specifically the 10GBase-T leg, and if you include a 10GBase-T leg then you pay that power penalty regardless of whether it's via a native NIC or whether it's a SFP NIC with a module.

I do totally agree it would be a nifty idea though. It actually would bring down power for short connections, and a lot of laptop ethernet connections are "plug me into my desk switch" type use-cases anyway. Meanwhile, if you do want a longer connection, it doesn't limit your options because you can use a module.

But I suppose a lot of that has been subsumed into docks anyway. Thunderbolt 3 is absolutely fine for that hop to your dock. Related discussion but as a weirdo, I would actually rather have a more robust connector with longer distance support than Thunderbolt natively allows - if I could plug a cable and get 10 meters or so, that would solve some use-cases for me. Currently the only option is Optical Thunderbolt cables and everything I've heard is that they are absolutely flaky as hell and usually die after a year or so. I would love if I could use QSFP copper or something.

(my use-case is I want to have an external PCIe enclosure with the adapter for my Vive Wireless so the PC doesn't have to be in the room, but that would involve at least a 10 meter run for me realistically. Being able to run HDMI 2.1 over QSFP would also be neat because I could have my desktop run a 4K120 TV properly without having to have a dedicated HTPC there...)

Anyway back to SFP in laptops - you're right that it's pretty bulky. Even the "box" is big but it's also a very long connector physically, and the tab that protrudes even farther. The other thing that may not be immediately apparent is that it's not rated for all that many plug/unplug cycles. My QSFP cables have a "10 cycles" tag on them. Obviously they don't instantly die after 10 cycles, but still it's not something you are going to plug and plug every day for years. I don't remember if my SFP cables have one but I don't think it's significantly different. The SFP series are not a "consumer" connector, they are for servers that get racked once in a datacenter and then sit for 5 years, maybe have the cables re-organized once or twice.

It might well be something that you could fix with a "micro SFP" connector though. Slightly smaller box, much shorter, try to aim for higher cycle life. I think you would still retain a lot of the electrical benefits of SFP.

It's not very relevant for laptops, but SFP adapters also have an order of magnitude lower latency than base-T adapters. For things like your fileserver, it's preferable to use SFP if possible, that's why I mentioned I liked the Netgear MS510TXM above, 2 SFP links gives you one for your fileserver and one for a fiber link (to another switch perhaps), or you could do two SFP links and bond. There's some nice flexibility there.

I own a GPD Pocket 3 which has it: https://gpd.hk/gpdpocket3

And 10GbE is basically limited to the MSI Katana: https://geizhals.de/?cat=nb&xf=31_LAN10g

> - Came with MediaTek 7961 WiFi, which while supported by mainline Linux, the performance is trash (~30MBit on 802.11ac).

It works perfectly for me on Fedora 35 (kernel 5.15), with consistent speeds around 800-900Mbps. The AP is an ASUS RT-AX86U.

Also somewhat surprisingly, while Intel AX200/201 is faster as it supports 160MHz (I had it before), the MT7961 has been more stable for me, both speed-wise and bug-wise, so far anyway. But YMMV.

> Luckily it's only M.2 so you can replace it with something like the AX201

I believe X13 Gen 2 has soldered WiFi, not M.2.

According to the hardware maintenance manual: "Wireless LAN card (for selected AMD models)" [1]

I can confirm from the internal photos I have of my X13 AMD Gen 2 (I swapped the stock SSD for a SK Hynix P31 Gold 1TB) that the WiFi is indeed M.2 socketed.

[1] https://download.lenovo.com/pccbbs/mobiles_pdf/t14s_gen2_x13...

I was originally looking to buy one of the X13 Gen 2 Intel models that have it soldered, I must've thought it applied to AMD as well.

I guess I'll try out Intel AX210 then, to see which one actually works better for me.

> The latest MacBook Pro does not come with an Ethernet port.

Thank you for catching all this. Fixed.

> - 1600p matte display looks amazing

I agree, matte display is great. And as I said, I calibrated mine.

> - proper support for S3, very low S3 power consumption

Strange, I've seen reports on the Lenovo support forums that Gen 2 laptops are also susceptible to the battery drain in S3. Was it always like this? Or was it fixed with some firmware update?

I am not sure. The first thing I did when I received my X13 AMD Gen 2 (20XH) was update the firmware to the latest available version at the time (1.14) and I haven't had any issues.

I thought we moved past whatever technical limitation it was that required matte surfaces on flat displays? It's like having Vaseline smeared on the screen, diffusing all the colour - why would you actively want that?

The issue is, like you said, the smearing and contrast, which was a bigger issue with newer, high DPI displays.

That being said, I do believe there are solutions - I have used 4k Matt displays that don't have the aforementioned issue, I think most glossy displays are there for there as a form over function thing.

Glossy displays are still annoyingly reflective and good antiglare coatings are exceedingly easy to damage

Why some manufacturers would want to use glossy on any non-touch device is beyond me.

Colour vibrancy and sharpness. A matte screen is literally a defuser between you and the actual content.

And displays arent high-resolution enough for it to be a big deal anyway.

For my digital-artist friends, they do prefer the over-bright glossy displays since it's actually a critical need for their process.

https://www.notebookcheck.net/The-Best-Notebooks-with-the-Be...

I mean it is great and all top places are pretty similar, but "nothing comes close" is just completely wrong.

Utter garbage in my opinion.

I already said why: reflections and glare make for a terrible work experience. That's why matte is the default preference for most people.

> Like if you wanted it blurry you could have done that in software.

The "grains" (for lack of a better word) in a matte coating are truly microscopic. They don't make anything blurry. I get the impression that either you're just trolling or you've never actually seen a matte display.

yep, those mediatek cards aren't just slow, they's complete trash (see my other comment: https://news.ycombinator.com/item?id=30248380). i swapped it out for an Intel AX210 on my P14s Gen2 (Ryzen 7 Pro).

on linux you need Kernel 5.15.13+ for AX210 since the bluetooth side (connected over M2's usb bus) fails to come back up after waking from sleep or soft reboots. i dont think that patch was in 5.16.0 when i tried.

https://www.spinics.net/lists/linux-bluetooth/msg96194.html

Do you still have to hack the firmware to bypass thee wifi whitelist? Whitelisting hardware is an engineering offense punishable by death in my book.

Can you measure it for those of us, who are interested in finding exact numbers? You can use script - https://github.com/savchenko/powerdrain

For example on Fedora35-Linux 5.16(T14s AMD Gen1), I get following numbers:

Feb 07 12:13:22 weathertop systemd-sleep[5438]: [INFO] Discharge rate: 0.94741 W/h.

So basically a fully charged laptop will discharge itself when left in sleep for 2 days or so.

Do you have a script that doesn't rely on systemd? I run a distro without systemd (Artix) so I am unable to use the script you linked.

From my casual observations S3 seems to use around 8% per 24h.

I learned to very carefully examine and in most cases wait for community feedback on BIOS updates after an HP update (that could not be rolled back) introduced a massive touchscreen delay on their 608 G1 tablet rendering mine effectively useless.

Previously I've read that this whitelisting was due to the fact FCC would certify the laptop with the particular adapter as a whole, and replacing the wifi/wwan adapter, or adding WWAN where there wasn't one, would make the whole device "uncertified".

the mediatek card was garbage; not just slow (ping times were 3x normal), but caused my machine to insta-reboot under git clone of a large repo in Windows 10, and simply dropped the wifi link in Manjaro during same op, so it's unlikely to be driver related.

kinda bizarre that they don't offer this card as an option on AMD builds and force mediatek upon you. the AX210 is a great card which supports monitor mode & packet injection.