Motherboards With Broken ASPM On Linux

[Chewi]

I think it should include servers. Electricity is a significant cost in running a server and not all servers push all their hardware to the maximum all the time. My own desktop is also my personal server that stays on all the time. This bug certainly matters to me.

[Michael]

Does it matter? It's an autogenerated list.

Right, it's a matter of what the vendor exposes. Some vendors expose their server model as the motherboard product string.

[enihcam]

http://forum.lenovo.com/t5/Linux-Dis...ng/td-p/560387

Lenovo users, pls second this post.

[fewt]

Does it matter? It's an autogenerated list.

Yes, it matters. The list should include devices that matter - portable devices.

[DeepDayze]

I think it should include servers. Electricity is a significant cost in running a server and not all servers push all their hardware to the maximum all the time. My own desktop is also my personal server that stays on all the time. This bug certainly matters to me.

many server boards don't fully implement power-saving so something like this definitely should apply to server-class motherboards as well. Properly implemented powersaving/ASPM in server motherboards should indeed help reduce power usage in a data center filled with servers.

[DeepDayze]

It's not the chipset that has the quirk, it's the BIOS that has the quirk...

I think it would have to be thoroughly tested on each specific motherboard before they go overriding what the BIOS is trying to tell the OS and enabling it anyway for everybody. That's probably the reason why the mobo manuf. didn't bother enabling it in the BIOS to begin with.... The manuf. may have never thoroughly tested it.... Or perhaps they just felt that setting it in the BIOS was redundant if the drivers enforced it.

If someone knows what parameters to change I could maybe hack the FADT myself to enable ASPM on my Abit mobo.

[mdias]

Yes, it matters. The list should include devices that matter - portable devices.

FFS, go cry to another corner! If you feel that only portable devices should be included, do us all a favour and give us the link to your filtered list.

[ChrisIrwin]

Yes, it matters. The list should include devices that matter - portable devices.

It does, but it only reports whatever string is in place of the board name. While it would be handy to see "Lenovo T510" on the list, it is simple to check what string identifies your motherboard:

$ cat /sys/class/dmi/id/{product_version,board_name,product_name}
ThinkPad T510
4313CTO
4313CTO

I'm unsure if this list is board_name or product_name. They are the same on my system, so I'm not sure it matters.

[Michael]

I'm unsure if this list is board_name or product_name. They are the same on my system, so I'm not sure it matters.

It depends upon what's detected... I have many different fallbacks in place as some vendors report useful information to board_name while nothing (or crap) to product_name, while others do the opposite, etc. So from looking at thousands of reports over the years and from continued monitoring of what I find with new motherboards / BIOS, it's refined to try to be universal and to always provide as useful information as possible.

[Ex-Cyber]

Looking at the Smolt statistics is always depressing. Even the mighty HP has lost the market share war to the up-and-coming "System manufacturer", thanks to the overwhelmingly popular "System Product Name System Version"