dc7800 has quit [Read error: Connection reset by peer]
<deathmist>
I assume eGPU via USB4 isn't a thing still? I've seen the X1E series from late last month but it hints at seemingly still missing "USB4 router driver" side changes(?). at least the "Prepwork for non-PCIe NHI/TBT hosts" series seems to be in 7.2 already. works on Windows at least so far with the "top-left" port on Vivobook as the other kills watchdog
<penguin42[m]>
deathmist: Yeh apparently there are a few more chunks needed yet
dc7800 has joined #aarch64-laptops
dc7800_ has quit [Read error: Connection reset by peer]
dc7800_ has joined #aarch64-laptops
dc7800 has quit [Read error: Connection reset by peer]
bandini has quit [Quit: WeeChat 4.9.5]
dc7800 has joined #aarch64-laptops
dc7800_ has quit [Read error: Connection reset by peer]
dc7800_ has joined #aarch64-laptops
dc7800 has quit [Ping timeout: 480 seconds]
<notHorseface>
I didn't buy it yet although I am waiting for the people...
<notHorseface>
lol
dc7800 has joined #aarch64-laptops
<Mis012[m]>
deathmist: there's always the m.2 port though
dc7800_ has quit [Read error: Connection reset by peer]
<Mis012[m]>
and the risers for that are much cheaper :P
<deathmist>
yeah and then I lose SSD, nice
dc7800_ has joined #aarch64-laptops
dc7800 has quit [Remote host closed the connection]
<nergzd723>
does x1e support bifurcation? :P
<HdkR>
Bifurcate those 4x M.2 lanes down to 4x1 lanes
<valpackett>
(btw the funniest thing is that on the sc8180x / surface pro x, NVMe is attached to such a split/multi phy)
<valpackett>
dang I just realized how convenient the SPX is for the m.2-to-eGPU thing!! with its quick-access m.2 door on the back :D
<valpackett>
shame the CPU is light years behind lol
<JensGlathe[m]>
you want gpu power, you get pcie lanes
<valpackett>
if i'm not mining buttcoins or slop, i also want "some" cpu power.. some games can easily choke on cpu :D
<JensGlathe[m]>
alsa ucm2 update killed my sound again, becoming a ritual
<Mis012[m]>
I get the feeling that there is some misunderstanding about uapi stability somewhere in the alsa chain
<JensGlathe[m]>
need to upstream my changes for volterra I guess
<JensGlathe[m]>
there was something about detection script vs x13s
<valpackett>
IME it's just really aggravating that it's so all-or-nothing and there's no way to make subdevices optional / allow some to fail
<JensGlathe[m]>
thats bad yes
<valpackett>
like if a displayport was added to ucm but it's not there yet on the kernel level.. why the f should speakers fail
<Mis012[m]>
<valpackett> "if i'm not mining buttcoins or..." <- well you see, that's why you need a second pcie riser to plug the laptop into some Epyc server
<JensGlathe[m]>
yes, true
<Mis012[m]>
valpackett: that's a very good question that should probably be asked elsewhere :P
<HdkR>
Mis012[m]: Ah yes, PCIe switches instead. That's a solution as well.
<Mis012[m]>
I think a breakout board for AM5 chipset thingy would be quite popular, probably paired with a usb4 enclosure because nobody likes fun :P
<Mis012[m]>
also re: ssd obviously there should be more than one m.2 slot in these laptops to begin with
<Mis012[m]>
there's a lot of unused PCIE left
<HdkR>
It has the PCIe lanes for it, might as well
<Mis012[m]>
it has enough for x8 oculink :P
<HdkR>
Expose all 16 of the PCIe lanes, and put it in an mATX board
<Mis012[m]>
I think it was only x8?
<HdkR>
12x5.0 and 4x4.0 in X2E
<Mis012[m]>
ah, I was thinking x1e
<Mis012[m]>
but yeah I assumed you meant 16 in one slot
<Mis012[m]>
which would not be wise :P
<HdkR>
Yea no, that would of course break
<HdkR>
Gives you enough for an M.2 slot, an 8x slot, and then another 4x slot. Would compete well with IGX Thor.
<Mis012[m]>
I feel like with something like oculink you could still do laptop size board
<Mis012[m]>
two m.2 slots and oculink
<valpackett>
i wonder if some mega cursed thing could be possible to "merge" multiple separate links to the host.. like a driver that would talk over multiple pcie devices on the host + all those links going to some cursed hw that connects them all to an x16 slot
<HdkR>
Definitely cursed
<Mis012[m]>
I suppose maybe you could do something cursed if both devices have the same lane configs
<Mis012[m]>
like a cluster of 2 glymurs
<valpackett>
i mean like a "PCIe octopus" connecting to the wlan, wwan, and ssd slots to make all of those link to an egpu :D
<Mis012[m]>
well, with some cursed expensive AF ASIC you could probably do a lot
<Mis012[m]>
the host side would be sad in any case
<Mis012[m]>
Val Packett: but I mean... why
<Mis012[m]>
5.0 x8 is enough bw for any gaming needs?
<valpackett>
just for fun, as always
<valpackett>
for specific kinds of fun with hosts that don't have anything like an x8, i mean
<Mis012[m]>
if you have the money to tape out an ASIC I can think of some more fun things to make
<valpackett>
target market: youtubers that make "gaming on X shitty hardware" videos 🤣
<Mis012[m]>
not even close
<Mis012[m]>
valpackett: don't think the ROI is there ngl
<Mis012[m]>
how many would you need for the video revenue to pay for fabbing an ASIC 🤔
<HdkR>
If I was taping out an ASIC, I would definitely just glue more PCIe lanes on the SoC :D
<JensGlathe[m]>
Nvidia style, pcie to pcie bridges?
<Mis012[m]>
Mis012[m]: something tells me that the number is large enough that you can't assume independence and have to consider the entire youtube platform's ability to provide that many <del>products </del> viewers
<Mis012[m]>
HdkR: why not AXI though
<Mis012[m]>
and more AXI for the GPU :P
<Mis012[m]>
have to think big
<HdkR>
Might as well as get a Spark and glue a Radeon to it at that point.
<Mis012[m]>
you mean you'd get only half of a spark? xD
<Mis012[m]>
since you don't need the GPU clearly
<HdkR>
Yea, I don't need that blob GPU
<Mis012[m]>
is it nvlink between the two dies? :P
<Mis012[m]>
would need a pcie controller die unless you manage to steal AMD's HDL
<Mis012[m]>
to make a radeon die
<HdkR>
Yea, it's NVLink-C2C fabric
<Mis012[m]>
I guess they're gonna call it NVLink no matter if it bears any similarity to the other nvlink
<HdkR>
Would be fun to know what the difference is
<Mis012[m]>
I wonder if it's purpose made to do PCIE glue shit or if that's completely separate
<Mis012[m]>
but yeah I think to have proper fun you'd want to get Intel's Lunar Lake (well, next next next gen equivalent) tiles, specifically a nice GPU and some nice x86 cores, and add that to your sane aarch64 SoC :P
<Mis012[m]>
hw accelerated fex container go brrrr
<HdkR>
Just get that x86 coprocessor to make the JIT go fast.
<Mis012[m]>
lol
<HdkR>
The Asus Ascent QN10 is on Newegg
<HdkR>
But it says it has an Intel CPU
<HdkR>
Do I risk it?
<Mis012[m]>
do you promise to make a dts for it no matter which SoC it comes with /hides
<HdkR>
I'll ACPI boot it instead
<Mis012[m]>
HdkR: I still think managing to make userspace on x86 cores talk to the kernel on aarch64 cores with negligible performance penalty would be more fun
<Mis012[m]>
HdkR: booooring
<HdkR>
I bought it
<HdkR>
Let's see what arrives!
<Mis012[m]>
creative gambling
<Mis012[m]>
most people are so unimaginative with their execution on gambling
<Mis012[m]>
probably also not that hard to offload it if it ends up being intel
<Mis012[m]>
(you can read that two ways xD)
<HdkR>
Indeed
<HdkR>
I can throw it down a flight of stairs if an Intel board arrives
<HdkR>
Should make for a good time
<Mis012[m]>
don't think that falls under any of the definitions of the word "offload" but then again I'm not a native speaker
<HdkR>
Offload it from my hands that happen to be above a flight of stairs.
<Mis012[m]>
you do you
<HdkR>
It claims it's going to be here from Aug 19-21 shipped direct from China
<clover[m]>
this is useful on arch ports where dtbs are packaged in /usr/lib/modules and not directly in the EFI mount location. not sure how useful it would be on other distros, but if it is, get with me to get your distro supported
ginger has joined #aarch64-laptops
<robclark>
hmm, dtb not in ESP sounds kinda like an arch bug
<Mis012[m]>
usually they go to /usr/share/dtbs, not sure how much it matters if they are on the ESP when relevant parties refuse to codify how they should be laid out on the ESP anyway
<Mis012[m]>
I assume arch's boot selector configs don't reference the dtb location, otherwise they would care whether it's accessible to dumb efi firmwares that only support fat32
<robclark>
I mean.. if you needed the dtb to boot, having it in /usr seems like a bit of chicken/egg problem
<Mis012[m]>
well, not sure how arch deals with that, but I suspect they don't otherwise they would've noticed the issue themselves
<Mis012[m]>
/lib is probably symlinked to under /usr anyway
<clover[m]>
archinstall installs linux linux-dtbs and dtbsync alongside each other, the pacman hook takes care of the rest
<Mis012[m]>
I think pmos also copies stuff from /usr/share/dtbs fwiw, instead of making the kernel package install it in /boot
<Mis012[m]>
but that's technically still the distro making sure it ends up where it needs to
dc7800_ has quit []
dc7800 has joined #aarch64-laptops
* clover[m]
shrugs
<clover[m]>
dtbs in /usr/lib/modules/linux version/dtb is what the arch ports folks landed on for their design
<clover[m]>
apparently a lot of them don't even need linux-dtbs, something to do with uboot
anthony25 has joined #aarch64-laptops
<clover[m]>
so its an optional dependancy of linux package
<misyl>
well, hopefully the installer doesn't make people go thru and manually add the linux-dtbs package and tries to do something at least partially user friendly here
<Mis012[m]>
well, if you have u-boot present an up-to-date dtb from it's own stash sync'd from Linux tree then it's fine yeah
ginger has quit [Remote host closed the connection]
<clover[m]>
misyl: Well there has to be a balance of ease of use and control. Like I said not everyone is wanting to install linux-dtbs
<misyl>
i totally expect users who pacstrap themselves to be able to not install that, but the typical install path should install dtbs
<HdkR>
I'm fine with my ACPI platforms carrying the burden of other platform's DTB files
<clover[m]>
<Mis012[m]> "well, if you have u-boot present..." <- Can you explain how that works exactly?
min1123[m] has joined #aarch64-laptops
<min1123[m]>
U-Boot can embed DTBs, and will pass that in to the kernel as part of a boot argument, giving it the memory address of the DTB.
<min1123[m]>
Alternatively, u-boot can load a DTB into RAM at an address from a filesystem, and it can pass that as the DTB memory address to the kernel.
<min1123[m]>
I initially thought you were asking about the ACPI, and I was going to be all snarky about how MS and Intel wrote a giant standard, then didn't actually use the standard and just BS'd about, then Qualcomm came in and decided that it'd be great to make blank spaces instead of filling in the data, and have the Windows drivers fill the data in as Windows boots, thus ensuring extra lock-in.
<Mis012[m]>
the ACPI standard is not even that hard to follow
<Mis012[m]>
with very few exceptions you can invent any bullshit you please as long as you strap your vendor prefix onto it
<clover[m]>
I see. And how does it keep synced on what an up-to-date dtb is? Is it pegged to the latest linux release?
<min1123[m]>
It's in the Linux kernel source tree.
<Mis012[m]>
in principle if you pass the dtb via the boot firmware (which u-boot is supposed to be, if you own the hw and can yeet whatever was there, or if you are the OEM) it should ideally not be WIP
<Mis012[m]>
otherwise the user may choose to supply a more up to date one themselves
<min1123[m]>
Problem is, DTBs normally are paired with vendor kernel forks (frequently made with Android kernels), and then the drivers and the DTBs have to be redone to go upstream.
<Mis012[m]>
well, yeah, we're talking hypothetical upstream first platforms
<min1123[m]>
That would be fantastic. Have those companies do some lifting for a change.
<Mis012[m]>
I still wonder how Valve will handle that, since they very clearly don't want to have the bindings reviewed day one
<Mis012[m]>
maybe they will have a bootloader update at some point
<clover[m]>
<min1123[m]> "It's in the Linux kernel..." <- Ok so when linux updates it updates. Because its part of linux basically?
dc7800 has quit [Remote host closed the connection]
dc7800 has joined #aarch64-laptops
<HdkR>
Gah, I updated Ubuntu and it bricked my install.faaaaah
<rigo>
I've been tinkering with the C630 for a while now and now on 7.1 it seems to be running great!
<steev>
rigo: lucky. my c630 died (the display backlight turns on but it never gets to bios/lenovo screen)
<rigo>
dang, that unfortunate! is that a firmware thing or hardware?
<rigo>
on this new kernel I get the night light mode now. the display was too cool before
<steev>
i'm assuming its hardware, because my best friend's did the same thing (so he sent it to me, and i'm back to having 2)
<rigo>
that's strange it'd happen to both of them!
<Mis012[m]>
almost like it's some sort of issue inherent to the hardware that should be covered under warranty
<rigo>
The only things that aren't quite working now are the RTC and speakers. I got the LTE modem working. I guess the accelerometer and fingerprint sensor don't work either, but I'm not really loosing sleep over that
<rigo>
they are getting kinda old now
<penguin42[m]>
steev: Unless it's bad efi vars or similar? Does it have an emergency reflash type of thing ?
<steev>
Mis012[m]: they're like.. 8 years old... no warranty
<rigo>
you could probably script something to rewrite the efi partitions
<rigo>
there's a bunch of wierd programs there
<Mis012[m]>
penguin42[m]: there's EDL, generally not gonna do much for you if you don't own the device unless the programmers leak
<Mis012[m]>
and I don't think that they leaked for anything WoA?
<penguin42[m]>
Mis012I was more thinking a vendor flash restore off USB stick
<Mis012[m]>
if that was going to work then presumably it should be booting further than it is
<penguin42[m]>
shrug
<steev>
yeah, i think edl would be my only hope
<rigo>
I almost connected a uart to the debug connector, but I'm not quite sure what I'd even get there
<Mis012[m]>
logs
<rigo>
if linux would be you'd also get a console too, right?
<rigo>
but otherwise maybe only bios things
<gwolf>
rigo: I'm also saddened to say my C630 is not operative anymore — but I very much enjoyed it
<steev>
oh, iirc, you can get serial output from the c630
<gwolf>
Mine does boot and work, but no longer starts the wifi interface
<steev>
i can't remember what her handle is, but she wired one up and ran it out the headphones port
<gwolf>
and it's much tougher to work without network nowadays... Anyway, I got a X13s thinkpad, which seems to be much its successor
<steev>
maybe it's kcxt_ ?
<gwolf>
But if you want testers, I'll be happy to give your iso a go — maybe I'll even gain some function!
<rigo>
the iso has everything to start up the wifi, you just gotta start up the right services and use nmcli. maybe the wifi starts working again haha
<gwolf>
Right! I won't lose anything other than time in trying it 🙂
<gwolf>
I will download it later today and test where it gets me
<gwolf>
but the link you shared is for the project to build the script, right? Are you interested in distributing a binary version of it?
<gwolf>
(yes, I know, never trust binaries generated by who-knows-who on the Internet, but... for the use case... 😉)
<rigo>
I also put a release up so you don't have to build it
<gwolf>
👍
<rigo>
hope it works! keep me posted
<gwolf>
will do. Do you idle here on IRC? If so, I'll tell you here.
<rigo>
i'm off to bed soon, but i'll check in the morning
<gwolf>
Perfect. Hope to have a review for you by then!