Recent content by ericloewe

At least some current stuff seems to be available both from an old link and a new one. The old one can have fun things such as the BMC firmware link pointing to the system firmware page. And their own search function shows both types of links!

Since nobody's just edited the string to read "Dell", I take it the firmware image is signed with a key held by LSI and/or Dell?

And they have, for close to two decades now.

Shouldn't be too difficult, just flip a flag to unhide the bifurcation settings menu. But I'll believe it when I see it.

The datasheet for the Xeon D-15xx series (see here, section 5.2, page 84/82) states that both PCIe 3.0 root ports on the CPU (not to be confused with the PCIe 2.0 ports from the integrated PCH) can bifurcate down to x4/x4 and x4/x4/x4/x4. If the system firmware setup menu doesn't support this...

Haven't tested to that extent, but expect to not be bottlenecked when not using expanders.

Did you replace the cables as well, or switch them around?

That's really not the sort of thing that goes well with "tight budget", and your description does not make much sense. Why do you need 4-8 NVMe SSDs only to then cram everything into 2.5 GbE?

Decreased immunity to noise and significantly more complex transceivers. Since the former is handled by better coding and error correction and the latter will get easier over time as adoption grows, I would not be surprised to see a hybrid PCIe version using the new tech at a slower symbol rate...

I was going to say that no video out was probably for the best, but then you said the IPMI LAN had been deactivated. That is short-sighted and frustrating.

Pretty much. Although using two is conceptually possible, either for increased bandwidth or for multipath, that adds complexity that is often pointless. This is going to depend on the specific unit. The SAS2 stuff is getting old, but if it works, no reason to avoid it. As for bandwidth, the...

In this case, it's actually more along the lines of <=2133 to 2400.

As a reference, Xeon E5 v3 -> v4 brings a small (1-3%) improvement in IPC and support for faster DIMMs.

I'm looking for information around this matter, too. As far as I can tell, Dell used these mezzanine cards on the single-socket Rome/Milan servers (R6515 and R7515) and the R740xd2 and they sure look like OCP NIC 2.0 form factor cards. But... Is there a mechanical change that makes regular OCP...

Similar product from Asus that flew under my radar: IPMI EXPANSION CARD｜Motherboards｜ASUS Global They claim no compatibility outside of specific Asus motherboards, but that seems to be more a practical decision than the result of a different design.