Oh well. I think it is often a problem running into a help- and clueless hotline/email person that doesn't know shit and is scared of that strange L-Word. So they block it with a copy and paste "blahblah" lame answer. Then, a lot of stuff lies in the hands of the BIOS Vendors themselves (see agd5f's post here).
I personally use pcie_aspm=force and it seems to be ok. Still, we really need that secret password to reach the better informed hotline.
My experience: I called a few times hotlines when I really ran into a problem I couldn't seem to fix on my own but that person was normally 100x more clueless than I. One of the few exception were a now defunct internet provider (but really depends whom you reach in the callcenter) and iiyama (the monitor folks).
Especially with mainboard vendors it seems that they have one single support person that barely speaks English and is overloaded with support request from the whole world. Often they just send you pre-compiled blahblah from the FAQ not even having actually read your question. Sad thing that support is often so lousy.
Can't tell about Gigabyte since I didn't have one but I checked their webpages from time to time and afair they had this "Linux users pls download drivers from chipset manufacturer" and iirc. they provide a BIOS update mechanism that enables you (like ASUS) to put the BIOS image to a FAT32 medium (floppy, HDD, stick) and just load it from a small program "inside" the BIOS setup. No DOS or Windows required.
And when it comes to boards I guess a lot is just
a) chipset from AMD, intel or nv or whom ever
b) additional network chip (Realtek, via, something)
c) audio chip
d) USB3 from Renesas or NEC if not in chipset
e) firewire from via
f) SuperIO from e.g. ITE or Fintek
... whatever I forgot.
But all compiled mainly from "standard" components without much own input.
The thing is that this isn't MY problem, I'm not having any problem, it's THEIR faulty product. And I won't ask for any support agreement with a company that doesn't care at all when they are noticed their products are DEFECTIVE (look at it how you want, but a faulty or incomplete bios implementation makes a board defective). That's why my boss ordered me to buy a different branch.
They have two problems now, defective products and less people wanting to buy their products.
It sounds like you're being sensationalist. You registered just to post that you were changing from Gigabyte to whatever. The ASPM bug affects more boards than it doesn't. Gigabyte is not doing anything different than nearly every other manufacturer does.
I have no lower opinion of Gigabyte than I did before this e-mail. Nearly every board in my house (Main PC, Garage PC, 3 HTPCs) is a Gigabyte. They have never had any issues with Linux. You (and your boss) are making a mountain out of a mole hill.
the good point of gigabyte mainboard is the dual-bios they have on all mainboards .
the highest model of it with x58 , is very liked by those doing sli/crossfire with 3 cards or more @ 16x
this model also has radiators for chipsets ready to get in water cooling loop . that is a great good point and i never seen others doing that . http://www.gigabyte.com/MicroSite/19...-power-sp2.htm http://www.gigabyte.com/products/pro...px?pid=3434#ov
for the rest , like all others motherboards makers , they have all prices and quality . benchmarks based on chipset show clearly that all brands have close results when no overclocking . overclocking leads intel boards to be last ;']
Is there a reason why Linux can't provide signatures that allow installation?
Ya there is a couple of issues. First of all using a signed key could be seen as a form of DRM which has it's issuses with gpl code. Secondly the code verification would have to be done in a open manner which kind of defeats the purpose of signature protection.
I think I heard the UEFI signing is done by MS, but please don't quote me on that....
Not quite, MS has their own signature, every other entity would be responsible for their own signing. It is up to the UEFI vender to determine what signatures the UEFI will accept. Even if licensing issues where not a factor (but they are as well) it would still be up to those MB manufacturers to add acceptance to other signatures or to provide a toggle to disable the signature checking. Either way I do not put a lot of faith in motherboard manufacturers providing those options given their reluctance to even fix ASPM issues as well as provide any other data to kernel developers for their vender specific feature set.
Server mobos do need very short boot times. Shorter than Desktops actually
Originally Posted by deanjo
Unfortunately, server motherboards are not really where 2 second boots are needed.
I can see why you would think that though, but obviously are around the enterprise server market and not in the embedded server market. Servers are EXACTLY where short boot times are needed. When you are using servers for embedded applications (like VoIP, CallCenters, etc), you want short boots to help you keep five 9's when you have to do a reboot for whatever reason. The current IBM uEFI/IMM based servers take around 5 minutes to start booting the OS.
i do have a gigabyte mainboard right now but to read this about gigabyte makes me sad--
maybe i really buy supermicro or tyan next time.
i do have a Tyan mainboard in the past but the nvidia nfoce 3600 pro makes many trouble with the graphic driver and the bios do have tons of bugs like the 128gb first boot pardition bug if you have a pardition greater than 128gb it fails to start grub. also the bios do have a bug that force me to use ext3 or raiserfs it does not start with ext4...
but maybe the modern mainboards with amd chipset are much better than that.
I've got several servers running Tyan mainboard. Due to my loath for nvidia products, they are ALL AMD chipsets. I've never had a problem with one of those boards, brutally stable. I've got a couple with uptimes pushing a decade.