<davidlt[m]>
As expected Meteor Lake is at Computex
<davidlt[m]>
I just hope 16K monitor doesn't become a reality :)
jcajka has joined #fedora-riscv
<cwt[m]>
I'm upgrading to F38 vis chroot on my Arch Linux.
<cwt[m]>
s/vis/via/
<cwt[m]>
with bind mount /boot and /lib/modules, I hope I should be able to run dracut to create a working initramfs image for F38.
<Eighth_Doctor>
<davidlt[m]> "I just hope 16K monitor doesn'..." <- it'll happen... I'm pretty sure it already exists at an industrial level, it'll take a few years before it becomes available in the prosumer/consumer space
<davidlt[m]>
Conan Kudo: there is one at Computex, but not a final panel, I think.
<Eighth_Doctor>
doesn't surprise me
<davidlt[m]>
I just cannot imagine driving 16K as any meaningful FPS.
<Eighth_Doctor>
well, the memory requirements mean that GPU architectures will need to change
<davidlt[m]>
Not gonna happen :) I mean adding more memory to GPU is not what Nvidia wants to do :)
<Eighth_Doctor>
indeed
<davidlt[m]>
That would make then useful for other workloads (AI, HPC, etc.)
<Eighth_Doctor>
even ignoring that, it's not cost-effective to bolt on 64GB+ of VRAM on a GPU
<davidlt[m]>
There is already agreements that forbid you using gaming GPUs for scientific workloads IIRC.
<Eighth_Doctor>
ReBAR already gives us a bit of an out here
<Eighth_Doctor>
and we may see more on that path
<davidlt[m]>
GDDR6 chips aren't that expensive IIRC.
<cwt[m]>
<cwt[m]> "with bind mount /boot and /lib/..." <- Ha, selinux blocked me from booting 🤣
<davidlt[m]>
There will be new images today.
<Eighth_Doctor>
<davidlt[m]> "GDDR6 chips aren't that expensiv..." <- GDDR6X, however, is not yet standardized by JEDEC, and development on GDDR7 and HBM4 are both ongoing
<davidlt[m]>
Yeah, but I doubt we will see HBM on consumer hardware again.
<Eighth_Doctor>
depending on how things go, I'm not so sure
<Eighth_Doctor>
ReBAR changes things significantly, as does the need for consumer access to AI/ML capabilities
<davidlt[m]>
I get impression ReBAR is a mixed bag for existing game catalog.
<davidlt[m]>
And you also get different results on AMD and Intel.
<davidlt[m]>
I guess (but don't know, didn't check too) that Intel Arc was designed with ReBAR in mind.
<Eighth_Doctor>
It was
<Eighth_Doctor>
but yeah, ReBAR is a bit weird for games, but it'll be needed to allow for VRAM overflow to system memory
<davidlt[m]>
There aren't a lot of bandwidth on the system memory. Traditionally.
<davidlt[m]>
Apple uses LPDDR5 with tons of channels.
<davidlt[m]>
Nvidia in their Keynote at Computex mentioned that their latest greatest HW thing is also LPDDR5.
<Eighth_Doctor>
yeah
<davidlt[m]>
From what I have seen the current AMD APU iGPUs are starving for bandwidth. Performance goes quite up with faster DDR5.
<davidlt[m]>
And finally AMD seems be going for "Mega APU" to go against Apple M chips.
<davidlt[m]>
I just hope they will go beyond 2 channels to feed that massive iGPU.
<davidlt[m]>
Also LPDDR5 has way higher capacities.
<davidlt[m]>
Sasmsung sells 144Gb (18GB) per chip.
<cwt[m]>
<cwt[m]> "Ha, selinux blocked me from..." <- relabeled and booted. my VF2 is on F38 now.
<rwmjones>
cwt[m]: I missed all that .. are you running it with an F38 kernel?
<cwt[m]>
rwmjones: I run it with my kernel that built from StarFive source.
<rwmjones>
ah ok, sensible
<cwt[m]>
I found something weird, my mosh session with byobu got killed while I ran git clone. However, when I mosh back there is no error message in dmesg.
<cwt[m]>
It happened 3 times already, and it never happened on Arch.
<davidlt[m]>
Check SELinux and Audit logs (or disable them).
<davidlt[m]>
Basically the as previous, just more images (just in case).
<davidlt[m]>
The LEDs are controlled in a different way now.
<davidlt[m]>
Otherwise test, test, test :)
<davidlt[m]>
The next one will follow soonish with 6.3.X kernel.
<davidlt[m]>
Which exactly I don't know, that will depend on when 6.3 lands in Fedora (there were issues reported, especially with XFS IIRC).
<davidlt[m]>
Also I keep monitoring the queue for 6.3 kernel patch backports.
<nirik>
davidlt: any reason some of the files under alt/risc-v/disk_images/Fedora-Developer-38-20230519.n.0.SiFive.Unmatched.and.QEMU are mode 640? that messes up miroring...
<davidlt[m]>
nirik: no reason, change it to what you need.
<davidlt[m]>
It seems there is just one file like that.