FW13 Pro (Ultra 5 325): 3-minute firmware phase on every boot

New Framework Laptop 13 Pro DIY, assembled without issue. Every single boot spends 3 min 05 as a black screen before the Framework logo, then reaches the login screen in 15 seconds. I’ve spent a while narrowing this down and would like to know whether other 13 Pro owners see the same.

System: Core Ultra 5 325, Framework 32 GB LPCAMM2 (Micron MTD16C20325N4FN023F1), WD_BLACK SN7100 1 TB, BIOS 3.02, EC sakura-3.0.2-cf48815, NixOS 26.05, kernel 7.2.5.

The delay is entirely pre-OS. systemd-analyze:

Startup finished in 3min 1.902s (firmware) + 5.335s (loader) + 507ms (kernel)
+ 2.364s (initrd) + 3.486s (userspace) = 3min 13.597s

ACPI FPDT gives firmware_start_ns = 543013218, so the firmware starts executing normally and burns the 181 s internally. Nothing from the distribution can be involved - this is all before the bootloader is even loaded.

It is not memory training. The “Memory Training in Progress” screen shows up exactly where the guides say it should: first boot after assembly, first boot after reseating the LPCAMM2. It lasts about 30 seconds - and is then followed by the very same 3-minute black screen every other boot shows. Training is a short, visible, separate phase; the 181 s start once memory is already initialised.

The panel is also alive well before the logo: for roughly the first minute of the black screen the backlight is clearly on, then it switches off entirely for the remaining two minutes, then the logo appears. Display up, display down, display up again, all within the 181 s. The LED diagnostic fires at 1 min 30, just after that transition. Looks more like a platform reset mid-POST than a long continuous operation, but I can’t prove that.

Other observations:

  • Warm reboot is as slow as a cold boot. Repeated cold boots give the same timing within a few seconds.
  • Holding F2 from power-on never enters setup; it only registers once the logo is up. Setup is reachable only via the bootloader’s firmware entry.
  • LED diagnostic sequence: white, 12 green, orange, then 10011001 (blue=1, green=0) = 0x99. All twelve checks pass, including S0 and the memory check.

Ruled out, with no change in timing: all expansion cards removed; battery only vs. AC; USB4 device measured boot disabled; Quick Boot already enabled, Network Stack disabled, boot Timeout 0, Boot performance mode Turbo; LPCAMM2 removed and refitted with the interposer inspected (clean, seated on its pegs, contact marks even) and progressive tightening in the 1-3 order; mainboard reset procedure (guide 737) performed at support’s request; BIOS is 3.02, which is the only release ever published, and fwupd confirms System Firmware 0.0.3.2 as current with nothing newer on LVFS.

I have no idea what actually consumes those 181 s. The EC console ring buffer only retains about twenty seconds, so the firmware phase is overwritten before I can read it from userspace.

If you have a 13 Pro - ideally an Ultra 5 325 - could you post this?

systemd-analyze

Valerii_Myronov has since reported 6.204s of firmware on an Ultra X7 358H with the same BIOS, EC and memory module, so the firmware can clearly boot this hardware in six seconds. The CPU tier is the last uncontrolled variable between us, which is why an Ultra 5 325 data point would be the most useful thing right now.

If someone knows a way to capture the EC console or POST codes across a full boot, or what 0x99 maps to in Insyde’s table on this platform, I’m all ears.

A ticket is open with support; I’ll update this thread with whatever comes back.


Edit 19/09/2026: removed the memory speed observation. I originally flagged 6800 MT/s as a possible anomaly, based on the 7467 MT/s figure in Intel’s specs for the Ultra 5 325. Valerii_Myronov pointed out that the board listing gives 6800 MT/s for this CPU and notes the 7467 there was a listing error - so 6800 appears to be the expected speed and I’ve dropped that line. Also added the mainboard reset, and his 6.204s firmware timing on an X7 358H as a comparison point.

I have Ultra X7 358H but the same memory module. Also the same BIOS with the same ECversion. These are my numbers.

Boot timings:

Startup finished in 6.204s (firmware) + 7.839s (loader) + 552ms (kernel) + 1.660s (initrd) + 10.463s (userspace) = 26.720s
graphical.target reached after 10.415s in userspace.

Memory:

Part Number: MTD16C20325N4FN023F1 YF
Configured Memory Speed: 7467 MT/s

But based on the board listing page Ultra 5 325 is configured to 6800 MT/s, so your readings are correct for the board:

LPCAMM2 LPDDR5X compatible up to 6800MT/s (previously listed as 7467MT/s in error).

So it looks like memory is reporting expected values for the board.

Thanks, that settles the memory question - 6800 is the documented figure for the Ultra 5 325 and the 7467 on the board listing was an error. Dropping that line.

The interesting bit is your 6.204s firmware phase against my 3min 1.902s. Same BIOS, same EC, same module, same board. So 3.02 clearly can boot this hardware in six seconds and I’m not looking at a general firmware bug.

One variable left between us though: you’re on the X7 358H, I’m on the Ultra 5 325.

Any Ultra 5 325 owner around who could post their systemd-analyze? That one number tells me whether my unit is faulty or whether it’s something about this SKU. It’s the only test I can’t run myself.

Also, support had me run the mainboard reset procedure (guide 737). No change at all.

Hi, my Pro with an Ultra 5 and 16GB Ram arrived yesterday.

Here are my timings:

Startup finished in 5.996s (firmware) + 18.300s (loader) + 501ms (kernel) + 2.431s (initrd) + 4.080s (userspace) = 31.309s 
graphical.target reached after 4.024s in userspace.

Hope it helps.

The led flashing is controlled by the BIOS and not the EC.
It is worth contacting FW support via their web page and ask for a list of led flash codes.
AMD refuse to release them, but hopefully Intel will release them.

Intel is also better than AMD, as Intel can display messages on the screen during memory training and AMD cannot.

For the F2 into BIOS, i find i have to keep pressing and releasing the key quickly during boot to get into the BIOS on my FW16 AMD. Maybe that is needed on your FW13 intel?
After a while, i get a feeling for when in the boot cycle i need to press it, so i only need to press it while it is counting RAM on my FW16 AMD. I don’t know when you need to press it for Intel.
Intel CPUs have a different boot sequence than AMD.

Some other threads on here have seen fixes by setting the date/time in the bios.

This is tangential to the topic at hand, but: the LEDs are controlled only by the EC. The LEDs are not connected to the SoC.

Hi @DHowett

I would be interested to find out how the EC does the fault led flash codes.
The reason I am asking, is because in the EC source code, I wished to make the EC also print a message to the EC console, detailing what the led flash code was. It was only when looking for that through the EC source code for where to put the extra log message, that I found that the EC was not involved.
Can you point me to the EC source code line that does it?