<lynxy>
is anybody currently maintaining a kernel fork for the x1e devices- specifically the dell xps 9345? where are we on resolving the system crashes, and on usb4 support? i'm seeing some patches in the kernel mailing lists, but i'm struggling to discern exactly how close we are to a functional system.
<lynxy>
more recent kernels (7.1+) seem to result in a rather high power draw at idle (11w+)
dc7800_ has quit [Read error: Connection reset by peer]
dc7800 has joined #aarch64-laptops
<valpackett>
something's really wrong in your setup if you idle at 11w+, my dell 7455 idles at like 3.5w at min brightness
<penguin42[m]>
lynxysystem crashes I think was sorted - those I think were mostly a cache oddity
<valpackett>
newest releases, lots of fixes, most devices, everything
<valpackett>
lynxy: ^^^
<lynxy>
you're always so helpful, val- thanks. i'll take a look through your sources and jglathe's and see what's up. I might also perform a fresh install at some point- it might just be that my system is in a funky state
<lynxy>
are you running el2? I never got around to trying
<lynxy>
re: random hangs- it does seem that the cpuidle state removal stops the hangs, but I'd rather not have to. I swear, and I don't know if this is something anybody else experiences, that the hangs *must* be related to the networking stack. bluetooth or wifi. I *only* get them in crowded spaces, never at home.
<lynxy>
I've also never managed to get audio working, even following the guides? I only ever see a dummy output device
<valpackett>
yea that's why the linked patch (jokingly called "Don't compute zero value assets") is better than the cpuidle one xD
<valpackett>
for me, the hangs were rather associated with idle-ish low load or being afk
<valpackett>
yes i run el2, i even used nested virtualization :D
<lynxy>
is it a pita to get setup?
<lynxy>
what benefits does it provide?
<valpackett>
EL2 provides KVM virtualization, if you don't need to run VMs you don't really need to bother
<lynxy>
ahh I see- a cool-to-have, but I'll take a working system first
<lynxy>
my old laptop is on the brink of death and I'd love to finally be able to use this daily
<valpackett>
i've been using this dell daily since i got it many months ago, but evidently I have a very high tolerance for barely working stuff since the random hangs were fine by me lol
<lynxy>
they're somewhat frustrating in the middle of a fast-paced lecture haha
<lynxy>
a complete crash and restart will set you back a bit
<valpackett>
right.. I even ran OBS recording for some live events with the risk of hangs and luckily never got one in the moment haha
<lynxy>
I was working technical at a con and my machine crashed ten minutes into a panel- at least the important stuff was running remote in a screen session- but it was stressful enough without that
<clover[m]>
Its going to have a linux partition for sure. Another platform to test my firmware extraction zig program
tobhe has joined #aarch64-laptops
tobhe_ has quit [Ping timeout: 480 seconds]
dc7800_ has joined #aarch64-laptops
dc7800 has quit [Read error: Connection reset by peer]
dc7800_ has quit [Read error: Connection reset by peer]
dc7800 has joined #aarch64-laptops
hexdump0815 has joined #aarch64-laptops
hexdump01 has quit [Ping timeout: 480 seconds]
* travmurav[m]
somehow thought the perf difference should've been much bigger betwen 4/4 and homogeneous-ish 8
<travmurav[m]>
(just assuming it's x1p-4x)
dc7800_ has joined #aarch64-laptops
dc7800 has quit [Read error: Connection reset by peer]
<mmediouni[m]>
travmurav: is it a high boost X1 bin
<mmediouni[m]>
Because for the 3.4GHz max clocks ones they're really not that much faster on single core than the Cortex-X1C
<travmurav[m]>
oh interesting
<travmurav[m]>
but seems like 1 core is +25% but +30% multi is weird to me
<mmediouni[m]>
oryon-1 on the X Elite never clocked all that well, but that was fixed by the 8 Elite implementation that ended up shipping less than 6 months later
<travmurav[m]>
since 8c3 only has 4 x1 cores
<travmurav[m]>
and I'd not expect whatever half bw they have for the other cluster make it effectively comparable to a78 :/
<mmediouni[m]>
The 8cx Gen 3 was a quite nice part but Qualcomm just didn't give it the support it needed to succeed at the time unfortunately
<mmediouni[m]>
It has working USB4 in hardware and all too...
<travmurav[m]>
I mean yes, I love 8c3 (more than anything newer too since it's a low end enough soc to be fully passive)
<mmediouni[m]>
Iirc Snapdragon X1(P)-42-100 is also passive-capable
<mmediouni[m]>
And ships in such a configuration on the 12 inch surface pros
<travmurav[m]>
but this is what baffles me - only +30% ish benchmark perf between a fully passively laptop with only 4 large cores and 8 "brand new" presumably nuvia server cores in a presumably actively cooled chassis...
<mmediouni[m]>
The higher end X1P/E had 4GHz boost so they won by significantly bigger margins
<mmediouni[m]>
They launched the X1P-46-100 much later which is the 42-100 config but with boost clocks unlocked
mbuhl has quit [Remote host closed the connection]
<travmurav[m]>
so you say it runs at such a low freq that MT doesn't beat an average between x1 and a78 by more than 30%? xD
<travmurav[m]>
oh well
<mmediouni[m]>
Btw the Snapdragon X2 surfaces have kinda something similar. They're at 4.0GHz max and aren't really that much faster than the X1 devices at all
<mmediouni[m]>
While the Asus X2 one ships with the high clock version
<travmurav[m]>
but i think x2 is back to heterogenous design so it's also a bit eh
<HdkR>
travmurav[m]: When I benched X1E versus SC8280xp, there were cases that sc8280xp won in purely CPU bench situations :)
<travmurav[m]>
lol
<mmediouni[m]>
An interesting thing about the 8cx Gen 3 is that it had all cores on the same cluster with relatively low core to core atomic latencies between them
<mmediouni[m]>
while.... that's not the case on X1 where core to core sync across clusters is much more expensive
<travmurav[m]>
uh I just looked at the brief for x1-26 for fun
<valpackett>
did that cursed one make it into any of the retail laptops?
<mmediouni[m]>
Yes
<mmediouni[m]>
X1-26-101 ships in Vivobook S14
<travmurav[m]>
the question you should be asking is "is there a way to hack the firmware to unlock the cores and clocks" /hides
<mmediouni[m]>
Specifically the S5408QA variant
<mmediouni[m]>
> US MIL-STD 810H military-grade standard
<mmediouni[m]>
Oh no
mbuhl has joined #aarch64-laptops
<mmediouni[m]>
I wish that QCOM would have been able to ship the phone SoCs on laptops on a 1 year cadence. But alas
<travmurav[m]>
yeah 8 elite on low end laptops could've been so good :(
<travmurav[m]>
and especially if they'd eventually add say 2nd combphy/usb/dp to make internals not insane
<mmediouni[m]>
But reality on SoCs shared with phones is that Snapdragon C thing
<mmediouni[m]>
which is very unfortunate
<valpackett>
lol, i guess with how much you have to pay for ram and flash, budget oems could only afford kodiak with what's left
<mmediouni[m]>
They outright ressurected Kodiak WP back from the dead for this
<mmediouni[m]>
The 7c+ Gen 3 was a failure of an SKU that didn't sell
<travmurav[m]>
ig they hoped for cros and that didn't work out
<valpackett>
it's a perfectly cromulent soc, i have it in my phone :D
<valpackett>
there is some old smasnug WP laptop with it heh
<mmediouni[m]>
That was the one machine to ship with it
<mmediouni[m]>
Shipped in plenty of mainstream phones as the 778G too
<mmediouni[m]>
After the 845 they stopped entertaining shipping high end phone chips on PC so what's left is the low end
dc7800_ has quit [Remote host closed the connection]
dc7800 has joined #aarch64-laptops
dc7800_ has joined #aarch64-laptops
dc7800 has quit [Read error: Connection reset by peer]
<Mis012[m]>
<penguin42[m]> "nod; I had run Zoom's x86 app..." <- isn't it electron
<Mis012[m]>
or can run it in browser
<Mis012[m]>
<travmurav[m]> "the question you should be..." <- even if only top SKU wasn't locked down PoS I'd be happy...
<travmurav[m]>
so x1e-de? xD
dc7800 has joined #aarch64-laptops
dc7800_ has quit [Read error: Connection reset by peer]
<Mis012[m]>
well, also would want to be able to buy one but sure
dc7800_ has joined #aarch64-laptops
icecream95 has joined #aarch64-laptops
dc7800 has quit [Ping timeout: 480 seconds]
<icecream95>
I'm confused by the TFLOPS for the X1-26-* GPU; from my calculations it should be 1.9 TFLOPS, not 1.7 as listed. Maybe this is achieved performance rather than theoretical?
<steev>
under promise, over deliver?
<icecream95>
Also X1-26-101 is listed at 0.63 GHz, but the closest OPP in hamoa.dtsi mentions Fmax of 687MHz... which would be 2.1 TFLOPS. But of course the firmware could lie
<icecream95>
purwa.dtsi also mentions SKUs which only go up to 666Mhz, perfect for running demanding titles such as Quake
<JensGlathe[m]>
<travmurav[m]> "so x1e-de? xD" <- Try to get one. Since you own that one, you need to write your own firmware 😉
<JensGlathe[m]>
Interesting bit is the x1-85 gpu on that binning(?). x1-85 is fast, noticing the difference every time I sit at angrybox
* travmurav[m]
wonders if double cores half the clock is better or worse than full clock half the cores, both with power consumption and performance
dc7800_ has quit [Read error: Connection reset by peer]
dc7800 has joined #aarch64-laptops
<icecream95>
I assume it's better for both. At half clock speed, per-clock memory latency should be a lot lower. I wonder if idle or low-load power consumption could be worse however?
<notHorseface>
the soviets are invading australia.
<notHorseface>
wrong channel.
<notHorseface>
brb
dc7800_ has joined #aarch64-laptops
dc7800 has quit [Read error: Connection reset by peer]
antoinelassagne has quit [Read error: Connection reset by peer]
antoinelassagne has joined #aarch64-laptops
dc7800_ has quit [Ping timeout: 480 seconds]
dc7800 has joined #aarch64-laptops
notHorseface has quit [Remote host closed the connection]
notHorseface has joined #aarch64-laptops
lynxy has quit [Remote host closed the connection]
lynxy has joined #aarch64-laptops
Acelogic has joined #aarch64-laptops
Acelogic_ has quit [Ping timeout: 480 seconds]
icecream95 has quit [Ping timeout: 480 seconds]
dc7800_ has joined #aarch64-laptops
dc7800 has quit [Read error: Connection reset by peer]
vanveldt has quit [Ping timeout: 480 seconds]
lynxy has quit [Remote host closed the connection]
lynxy has joined #aarch64-laptops
<penguin42[m]>
Mis012Last time I did it (few years ago), some calls you had to use the app for - I think ones with extra passwords
dc7800_ has quit [Read error: Connection reset by peer]
<programmerin-wonderland[m]>
I'm also wondering if the same is true for cdsp and adsp firmware? Is it the end of needing vendor signed firmware?
<JensGlathe[m]>
That would be too easy, no
<programmerin-wonderland[m]>
JensGlathe[m]: Sooooo GPU won't need vendor signed firmware but audio and NPU still will?
<JensGlathe[m]>
I have no idea
<programmerin-wonderland[m]>
weirdtreething: I know you got Linux working on your X2EE Asus zenbook a16, what did you need to extract firmware wise for GPU and audio if anything
<JensGlathe[m]>
But as things evolve, security first dies hard
<HdkR>
programmerin-wonderland[m]: I just used the GPU firmware from linux-firmware.
<weirdtreething>
programmerin-wonderland[m]: cdsp, adsp, soccp are from the vendor and iirc gpu was generic
<programmerin-wonderland[m]>
Good to know.. at least the GPU firmware is generic I guess
<weirdtreething>
programmerin-wonderland[m]: `msm_dpu ae01000.display-controller: Zap shader not enabled - using SECVID_TRUST_CNTL instead`