Qualcomm Snapdragon X Dev Kit for Windows

geerlingguy commented 1 month ago

Basic information

Board URL (official): https://www.arrow.com/en/products/c8380-12c-mp-32g/thundercomm
Board purchased from: Arrow
Board purchase date: 2024-07-16 (Arrived 2024-09-25)
Board specs (as tested): X Elite - 12 core, 32GB RAM, 512 GB NVMe SSD
Board price (as tested): $855.42

Linux/system information

# output of `screenfetch`
PASTE_HERE

# output of `uname -a`
PASTE_HERE

Benchmark results

CPU

Geekbench 6: (TODO single / TODO multi - PASTE_URL)
TODO Gflops (geerlingguy/top500-benchmark HPL result)

Power

Idle power draw (at wall): TODO W
Maximum simulated power draw (stress-ng --matrix 0): TODO W
During Geekbench multicore benchmark: TODO W
During top500 HPL benchmark: TODO W

Disk

MANUFACTURER_AND_MODEL_OF_DISK_HERE

Benchmark	Result
iozone 4K random read	TODO MB/s
iozone 4K random write	TODO MB/s
iozone 1M random read	TODO MB/s
iozone 1M random write	TODO MB/s
iozone 1M sequential read	TODO MB/s
iozone 1M sequential write	TODO MB/s

wget https://raw.githubusercontent.com/geerlingguy/pi-cluster/master/benchmarks/disk-benchmark.sh
chmod +x disk-benchmark.sh
sudo MOUNT_PATH=/ TEST_SIZE=1g ./disk-benchmark.sh

Run benchmark on any attached storage device (e.g. eMMC, microSD, NVMe, SATA) and add results under an additional heading.

Also consider running PiBenchmarks.com script.

Network

iperf3 results:

iperf3 -c $SERVER_IP: TODO Mbps
iperf3 -c $SERVER_IP --reverse: TODO Mbps
iperf3 -c $SERVER_IP --bidir: TODO Mbps up, TODO Mbps down

(Be sure to test all interfaces, noting any that are non-functional.)

GPU

glmark2-es2 / glmark2-es2-wayland results:

1. Install glmark2-es2 with `sudo apt install -y glmark2-es2`
2. Run `glmark2-es2`
3. Replace this block of text with the results.

Note: This benchmark requires an active display on the device. Not all devices may be able to run glmark2-es2, so in that case, make a note and move on!

TODO: See this issue for discussion about a full suite of standardized GPU benchmarks.

Memory

tinymembench results:

Click to expand memory benchmark result

``` # Run the two commands below, then replace this code block with the full result. git clone https://github.com/rojaster/tinymembench.git && cd tinymembench && make ./tinymembench ```

`sbc-bench` results

Run sbc-bench and paste a link to the results here:

wget https://raw.githubusercontent.com/ThomasKaiser/sbc-bench/master/sbc-bench.sh
sudo /bin/bash ./sbc-bench.sh -r

Phoronix Test Suite

Results from pi-general-benchmark.sh:

pts/encode-mp3: TODO sec
pts/x264 4K: TODO fps
pts/x264 1080p: TODO fps
pts/phpbench: TODO
pts/build-linux-kernel (defconfig): TODO sec

Provisional Tests under Windows

Idle power draw (Windows 11 Home, no apps running) - 11 W
Sleep power draw - 4.4 W
Power off power draw - 0-0.9 W
Cinebench 2024 (Arm Windows) - 131 single (at 28 W) / 1227 multi (at 99 W)
Geekbench 6.3.0 (Arm Windows) - 3020 single (at 28 W) / 15969 multi (at 80 W) (result)
Heaven (https://benchmark.unigine.com/heaven) - TODO (at 40 W)

Civil commented 2 weeks ago

Ok...

I went the other route - I've looked at YOCTO project and how they did and fetched the firmwares from the WOA repo with simple script:

#!/bin/bash
DEVICE="8380_RUN"
VER="200.0.4.0"
DESTINATION="/lib/firmware/qcom/x1e80100/Thundercomm/DEVKIT/"
mkdir -p "${DESTINATION}"
for f in qcdx8380.cab qcsubsys_ext_adsp8380.cab qcsubsys_ext_cdsp8380.cab qcwlanmsl8380.cab; do
    wget "https://github.com/WOA-Project/Qualcomm-Reference-Drivers/raw/refs/heads/master/${DEVICE}/${VER}/${f}"
    cabextract "${f}"
done

cp -a *.mbn "${DESTINATION}"
cp -a *.elf "${DESTINATION}"
cp -a *.jsn "${DESTINATION}"

# https://github.com/jglathe/wdk2023_fw_fetch/blob/jg/x1e001dev/bin/fetch_x1e001de-devkit_fw.sh#L78
mv "${DESTINATION}"/qcadsp8380.mbn{,.disabled}

jglathe commented 2 weeks ago

Could also be that you might need a modules file for proper booting. The one I have on the devkit is this one:

pwm_bl
tcsrcc_x1e80100
phy_qcom_qmp_pcie
pcie_qcom
phy_qcom
qmp_pcie
phy_qcom_qmp_combo
qrtr
drm-dp-aux-bus
phy_qcom_edp
gpio_sbu_mux
i2c_hid_of
i2c_qcom_geni
evdev
pmic_glink_altmode
leds_qcom_lpg
qcom_q6v5_pas
panel-edp
gpucc_sc8280xp
dispcc_sc8280xp
msm
nvme
usb_storage
uas

This one is from the multiboot image, unchanged. It loads some for sc8280xp, too, could be improved but works for now. The Ubuntu 24.10 X1E concept has no modules :man_shrugging: so there may be other ways.

jglathe commented 2 weeks ago

In the end: Whatever works. Nice one.

jglathe commented 2 weeks ago

The image for booting the Dev Kit blindly into Linux is ready enough. It is available for download here. Write it to an USB type-c stick or SSD. The type-A connectors don't work yet.

Precautions you should take:

Back up your internal SSD. We need it for UEFI access, so removing it is not an option (yet).
Decrypt the local Windows drive. We will disable Secure Boot, and this would trigger BitLocker.

To get it to boot from the stick:

Boot to the UEFI menu. I found that most documented ways don't work, but entering shutdown -s -t 0 /fw in an elevated CMD does.
Disable Secure Boot in the UEFI menu.
Enable "Boot from USB first" in the UEFI menu.
Insert the prepared boot stick / SSD. Notice that orientation is "logo down", the UEFI doesn't seem to recognize the device in a swapped configuration.
Power on / reboot.

The box should boot to grub, and boot up - without a display. The working network connector is Ethernet, Wifi not yet. It should register with host name angrybox, please monitor your dhcp server to find out which ip address. When you can ping it, ssh into it:

Screenshot from 2024-10-27 07-39-36

Access credentials:

user: devkit
pass: Canceled!

The image comes without the device-specific firmware, it needs to be fetched from the Windows partition. To do this, you can execute:

sudo chmod -x /etc/initramfs-tools/hooks/x1e-firmwares 
sudo chmod +x /etc/initramfs-tools/hooks/x1e001de-firmware 
sudo fetch_x1e001de-devkit_fw.sh

This changes the hook for initramfs, and fetches the firmware files from Windows. Afterwards it updates the initramfs. After rebooting it should show a lot of firmware loaded. The effect for now is - none, but its good to know the gpu firmware is operational (and remoteprocs, too) for further experiments.

Screenshot from 2024-10-27 08-53-52

Happy hacking!

Civil commented 2 weeks ago

Back up your internal SSD. We need it for UEFI access, so removing it is not an option (yet).

I've actually booted from the local SSD successfully.

I've backed up original partitions (I did that a while ago and actually store the original SSD on a shelf, just in case), then removed 13th, 14th and 15th ones (all after EFI Boot).

On boot partition, I've replaced MS bootloader with grub (removed everything, and put bootaa64.efi and grub.cfg in Boot folder)

Btw, uuid in it was wrong, it should've been 9289de4f-484c-4477-8f5f-a9264ab69033 as in main grub file.

And after a while (it takes A LOT of time) I have a working system.

I've used my method to fetch fimware from github, just putted it into folder where your scripts are expecting to see it.

geerlingguy commented 2 weeks ago

@jglathe - haha love the default password :D

Civil commented 2 weeks ago

Update on getting display to run: Not yet. A bit frustrating, but there are parts missing, need to drill into the i2c buses to get retimers and repeaters. I'm afraid some of these are required for the dp altmode switchery. What a mess. I will try to get the dedicated ssh-enabled image up before doing that, hopefully this afternoon.

That is actually for sure required. If you look closely near usb-c ports you can see Parade 8830 retimers, which needs to be configured via i2c and devkit devicetree completely missing all retimers configuration and relevant i2c configuration blocks.

As far as I understand: https://github.com/jglathe/linux_ms_dev_kit/blob/jg/ubuntu-x1e80100-v6.12.rc/arch/arm64/boot/dts/qcom/x1e80100-crd.dts#L1062-L1171

At least that is missing and probably should be copied from crd...

jglathe commented 2 weeks ago

@Civil yeah I know :grin: too bad it doesn't come up with it, though.

https://github.com/jglathe/linux_ms_dev_kit/blob/jg/ubuntu-x1e80100-v6.12.rc/arch/arm64/boot/dts/qcom/x1e001de-devkit.dts#L961-L1181

Civil commented 2 weeks ago

Ok, I haven't seen that you've already added a bunch of those :)

jglathe commented 2 weeks ago

Actually, didn‘t push since it doesn’t work yet. Need to activate in parts.

Civil commented 2 weeks ago

Also I'm not sure how its for you, but for me devkit doesn't always boot (not sure why though, can't get any logs, but symptoms are same - LAN goes down, then goes up and then goes down, LEDs are as usual).

And second problem - if I connect an external monitor over usb-c, it back-feeds power over usb-c to board and it is enough to keep it running once its booted (or if its stuck)... so I kinda forced in those cases to unplug monitor and plug it back after I turn on power.

jglathe commented 2 weeks ago

There is some oddity, yes. Haven't had it from the USB stick, but yet "hard" to debug. The Dev Kit has a usb2 serial debug connector, though - so maybe that's the way to go to find out. Regarding type-c power draw from display, yes. No capacitors, but getting power over the type-c connectors. "Odd", but makes some sense for a dev kit.

Civil commented 1 week ago

Btw, I haven't tried that, but in the very first patch series there was a DP Alt Mode support added here: https://lore.kernel.org/all/20241025123551.3528206-4-quic_sibis@quicinc.com/

AFAIR it was somehow lost in V3.

Civil commented 1 week ago

I probably forgot to copy some of the thing over, but I have a slight improvement - now usb-c can power my portable screen (it was constantly losing power before)

jglathe commented 1 week ago

patch v1 again... odd that later versions have reduced functionality. Anyway, I did a cursory check, and at least the PS8830 retimers are on the same i2c buses. Newer versions have other enablement wiring (gpio). The patches hint at other patches needed, haven't had the time to investigate. Will do another debug/test round at the weekend probably. A good source for further research may be this: https://git.codelinaro.org/abel.vesa/linux/-/commits/x1e-next-20240930

Civil commented 1 week ago

Yeah, I've tried applying the patches from the link but without any results.

P.S. I've also needed to increase rootwait to 40 seconds to make boot process more predictable.

Civil commented 1 week ago

Also, reading the V1 patchset, it mentions that dispalys were tested if kernel built from the x1e-next branch... Maybe it actually had more changes than what I've applied...

jglathe commented 1 week ago

Try looking at the branch I pointed to, there is stuff for enabling DP in aux-dp. But from what I've read you need more to get it stable.

jglathe commented 1 week ago

On 11/7/24 09:09, Cytochrome P450 2D6 wrote:

Thanks for the image you provided. I have successfully booted Linux and loaded the firmware. But when I execute some commands, it freezes, even

|iperf3 -c x.x.x.x| or |cat /sys/devices/system/cpu/cpu*/cpufreq/cpuinfo_cur_freq| will cause the machine to shut down. Does your machine have such random halt?

Not observed, but read about these on #arch64-laptops. Should be resolved with newer 6.12rc5 or 6.12rc6 kernel. I have seen odd things with inxi when GPU firmware was not loaded.

Message ID: @.***

CYP4502D6 commented 1 week ago

On 11/7/24 09:09, Cytochrome P450 2D6 wrote: Thanks for the image you provided. I have successfully booted Linux and loaded the firmware. But when I execute some commands, it freezes, even |iperf3 -c x.x.x.x| or |cat /sys/devices/system/cpu/cpu*/cpufreq/cpuinfo_cur_freq| will cause the machine to shut down. Does your machine have such random halt? Not observed, but read about these on #arch64-laptops. Should be resolved with newer 6.12rc5 or 6.12rc6 kernel. I have seen odd things with inxi when GPU firmware was not loaded. … Message ID: @.***

Thanks, I will try to compile a new kernel to see if this solves the problem.

Civil commented 1 week ago

My DevKit is actually stable when it boots :)

But I've installed Linux on PCIe SSD (swapped stock SSD for a brand new one and copied most of the partitions there, then removed all windows related stuff and moved grub's efi file to the place where UEFI expects to find Windows bootloader) and I've immediately imported firmwares into the initrd, so I don't know if your instability caused by their absence or not.

CYP4502D6 commented 6 days ago

On 11/7/24 09:09, Cytochrome P450 2D6 wrote: Thanks for the image you provided. I have successfully booted Linux and loaded the firmware. But when I execute some commands, it freezes, even |iperf3 -c x.x.x.x| or |cat /sys/devices/system/cpu/cpu*/cpufreq/cpuinfo_cur_freq| will cause the machine to shut down. Does your machine have such random halt? Not observed, but read about these on #arch64-laptops. Should be resolved with newer 6.12rc5 or 6.12rc6 kernel. I have seen odd things with inxi when GPU firmware was not loaded. … Message ID: @.***

I observed again that the machine did not shut down, but reported an error [72387.229430] energy_model: Accessing cpu4 policy failed and the speed was abnormally slow when compiling a new kernel. I was using another machine (Kunpeng 920) to compile the kernel and install it into the system, but it crashed again when unpacking tar.

jglathe commented 6 days ago

You could try the 6.12-rc5-x1e package I built. Just don't use the dtb that is in there for the dev kit, this at least doesn't work here yet - use the current one. Also, what about the SSD you're using. Some stability issues may come from sub-par USB sticks. What I've seen was some odd behaviour with the rtl8125 ethernet port (with 2.5G switch on it). Changed to 1GB connection (by using a 1GB switch) and this mostly vanished. It's early days for us, still.

CYP4502D6 commented 6 days ago

You could try the 6.12-rc5-x1e package I built. Just don't use the dtb that is in there for the dev kit, this at least doesn't work here yet - use the current one. Also, what about the SSD you're using. Some stability issues may come from sub-par USB sticks. What I've seen was some odd behaviour with the rtl8125 ethernet port (with 2.5G switch on it). Changed to 1GB connection (by using a 1GB switch) and this mostly vanished. It's early days for us, still.

I used the 6.12.0rc6 kernel. There were no errors during installation, but it failed to boot and then rebooted. In addition, my system was loaded on the newly installed internal Nvme SSD. I mounted the original hard drive to /mnt using a hard drive box and successfully extracted the firmware in Windows.

CYP4502D6 commented 6 days ago

Thank you for building the image, I will try it. In addition, when using the old kernel, running the command cat /sys/devices/system/cpu/cpu4/cpufreq/cpuinfo_cur_freq can stably reproduce the energy_model: Accessing cpu4 policy failed error code. Could this be a hardware stability issue?

jglathe commented 6 days ago

Not too likely. I recall seeing some talk about cpufreq and reboots when trying that command. Would assume it is fixed in newer rcs.

Civil commented 6 days ago

I won't be able to try it until this evening at least or even until tomorrow, but overall I would not be counting on stable cpufreq/powersaving yet and one of first things I did was to disable sleep and other powersaving features.

can you check if devkit is stable under load in windows? If it is, then it is for sure Linux issue, question is what exactly causes problems

CYP4502D6 commented 6 days ago

So it seems to be a problem under old linux kernel. At least I did not encounter any problem when running Cinebench and CPUz under Windows.

SpieringsAE commented 6 days ago

the crashing when accessing cpufreq is talked about here, I believe it is some patch that causes it, it was later talked about in the aarch64-laptops irc that reverting a certain patch stopped the crashing https://lore.kernel.org/all/ZypOY-NCDN9fdMAR@hovoldconsulting.com/

From what I remember it is a firmware issue

Civil commented 6 days ago

There were some news with more patches around cpufreq on sdx: https://www.phoronix.com/news/Snapdragon-X1-Elite-CPUFreq-V7

Or direct link to patch series: https://lore.kernel.org/lkml/20241030130840.2890904-1-quic_sibis@quicinc.com/

But I haven't tried it myself. I want to get video out working first :)

P.S. I also will try to replace onboard NIC with Intel 225 because I had enough problems with realtek under linux in my life :)

Civil commented 3 days ago

https://git.codelinaro.org/abel.vesa/linux/-/commits/x1e-next-20241101 - they have a devicetree there for the devkit. I've tried to compile it and boot, however with no luck (with defconfig I don't get network and I don't get display output, so I don't know whats wrong with it). I'll try to have a better look at it tomorrow.

Civil commented 1 day ago

defconfig for devkit there is probably missing something like NIC support, if I use config from the kernel - kernel boots, but still no display output.

talynone commented 15 hours ago

Microsoft just released the Windows ARM ISO for download https://www.microsoft.com/en-us/software-download/windows11arm64

geerlingguy commented 14 hours ago

@talynone - Indeed! Now who wants to be the daring person to nuke their Dev Kit and see if they can get the ISO installed? :D

geerlingguy / sbc-reviews