For reference, while this seems to fix the issue for the B50, I have a separate NIC that refuses to initiate even though everything is kaptan'd off (against shorting). Directly when it POSTs the activity led on the NIC just dies out. On a P320 it keeps working after post and also successfully sets up a link to the other device.
Every found an answer to that question? M75q gen2 continues to receive BIOS updates, I wonder if anyone eventually got it working with ECC.
Would you pleae be able to check the M75q1 schematics as well to see whether there's any safe 12V source for the same?
The M75q gen 1's 12V regulator is only rated for 1A continuous (OCP 5.1A) since it's only meant to power the fan and some other small things. It was never designed to support a PCIe riser slot or anything so it'll be very limited in what you could even use it for. I have the schematic but not the boardview so I can't really tell you where all you could best get access to the 12V rail.
No. It seems the x16 physical riser does only work for graphics cards on the arrow lake platform. I have the same issue with my P3 Tiny G2. The x4 riser does work.
You do have the Thunderbolt Card in the PCI Express riser? So the riser does work for you?
These values are with connected mouse, keyboard and display? I just got a M90q Gen 5 with an i5-14600T and 32GB RAM. With only the internal NIC connected to the network it idles at ~4.6W. (Debian Linux) With two NVMEs and an i350-T4 it's at ~6W. Essentially the same idle values as my old M920x.
Its a extra power source for the thunderbolt card or one of the B2B slots I believe. I purchased the cable when Lenovo had it on their parts sale, part number is 01YW371. Never got around to using it as I ended up getting a P340.
I get 3-4W with a 235, 48GB RAM and 2x 2TB 970 Evo. So Arrow Lake is capable of similar values. It's just the C package states which are flawed as soon as you add any PCI Express card to any riser (either x16 phys or x4 phys). Then you drop to C3 from C8 and its up to ~11-13W. Then add 3-4W for a PCie Express NIC and you're at 18W.
thanks for confirming this issue, had a feeling it was just for that gpu in there but it would have been nice to have other pci e cards work with it. Gonna have to sell £50 worth of riser cards now.
Well this was the issue for me, £480 with black friday discounts with 235T, 8gb DDR5 and 512gb nvme and thunderbolt 4 and realtek 2.5gb nic but the price has gone through the roof due to memory prices so its £740 now, so its not worth it.
re-edited my post, meant to say just sitting in windows desktop not doing much, ie few websites open and few apps running in background.
At least for (genuine!) Intel i350-T4 NICs I can say that both my M920x and m90q Gen 5 can hit C7 without problems. Not only in a synthetic test but in real circumstances. The M920x (32GB, 2x 1TB NVME) has several Incus containers spun up, an attached 2.5" drive, some NFS and CIFS shares and also my OPNSense libvirt VM. According to powertop it hits C7 ~80% of the time and stays below 11W on average(24h). I bought the m90q not only for the additional CPU power but also for propper IOMMU groups so I can switch the OPNSense network interfaces from virtual devices to passed through SR-IOV ones. Righ now even 200 MBit/s of traffic will bring up energy consumption to around 25W and I will finally get a Gigabit Internet conneciton soonish.
I have the same problem with my M90q Gen 5 as I described earlier in the thread. I opened a ticket with Lenovo, but they claim it's an electrical incompatibility with the riser... I guess if you want to use anything other than a GPU you need the smaller riser.
