Announcement

Collapse
No announcement yet.

AMD To Support Coreboot On All Future CPUs

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • #46
    On all of my hw since 2006 POST has been the bottleneck on boot. On the current system it takes 5-6s, when the rest of the boot is 3-4s.

    Comment


    • #47
      Originally posted by Veerappan View Post
      Then your work laptop is special. My desktop at home gets from login to usable desktop (Win7 64-bit) in 30 seconds (with my few auto-loaded programs fully ready). Is your machine force-launching a re-index or virus scan every startup? Are network mounts timing out and confusing explorer while they wait? 8 minutes to a usable machine is not normal.
      Nope, no timeouts. Virus Scan has something to do with it. A lot of corporate crap is set to load on startup. I know Win 7 is faster than that, but Vin!cent's example as corporate folks waiting minutes to use their computers. It's not due to a POST sequence, it's due to crap like this.

      I can definitely say that I wouldn't mind having a partition manager, drive imager, and maybe some form of file manager/virus scanner that could be launched from outside of an OS. Even if I had to load the virus definitions from a thumb drive, it could be very useful when attempting to rescue a system which has trashed itself (or a family member infected the machine and it falls to me to clean it up).
      All of those functions can be done from a Pen Drive or PXE boot. I'm not debating their utility; they do not belong in the same process as bringing up hardware.

      The whole point to LinuxBIOS and Coreboot was "Hey, BIOS is initializing hardware, and then the kernel just does it a second time, that's a waste of time." It was not to cram an entire operating system onto a flash chip.

      Your computer may post in 2 seconds, but mine takes about 15 before I hit the boot loader (almost half of which is due to the hardware-RAID card). Considering that I often have to sit in front of the computer to make sure that the correct OS is selected for boot, that's 15 seconds that I could be heading to the coffee maker to get a pot of coffee started. Yeah, it's only once or twice a day, but it's still annoying.
      You're not a typical user. The discussion was that AMD's announcement increased the possibility Coreboot was going to be on mainstream, commodity hardware. Hardware RAID cards are not mainstream, commodity hardware and are not in use by said audience. The mainstream, commodity market that has HPs, Dells, etc. already have a POST done in 2-3 seconds. Coreboot's advantages will not be seen by that market.

      Comment


      • #48
        Originally posted by locovaca View Post
        The whole point to LinuxBIOS and Coreboot was "Hey, BIOS is initializing hardware, and then the kernel just does it a second time, that's a waste of time." It was not to cram an entire operating system onto a flash chip.
        Yes, but once you limit the core firmware to just the bare minimum needed to bring up something else, it implies that the user (or OEM) has a choice of what the "something else" should be. That's why it's not called LinuxBIOS anymore. I agree that most users will not be interested in trying to "cram an entire operating system onto a flash chip".

        You're not a typical user. The discussion was that AMD's announcement increased the possibility Coreboot was going to be on mainstream, commodity hardware. Hardware RAID cards are not mainstream, commodity hardware and are not in use by said audience. The mainstream, commodity market that has HPs, Dells, etc. already have a POST done in 2-3 seconds. Coreboot's advantages will not be seen by that market.
        The typical user will never know the difference between legacy BIOS and UEFI, or the difference between NTLDR and WINLOAD. The typical user shouldn't notice these things, which is the main reason that so much effort goes into making them work on mainstream, commodity hardware. Isn't coreboot in the same boat?

        Comment


        • #49
          Originally posted by Ex-Cyber View Post
          The typical user will never know the difference between legacy BIOS and UEFI, or the difference between NTLDR and WINLOAD. The typical user shouldn't notice these things, which is the main reason that so much effort goes into making them work on mainstream, commodity hardware. Isn't coreboot in the same boat?
          It is, except that Coreboot has never made itself a serious alternative. As I said before, even if they magically get things working for a large number of boards, they still have to get some sort of support system in place. Right now it's more or less a volunteer to keep a board running with each new version of code. Motherboard companies have not been very helpful in release their source code to the public; look at the EEE PC and Splashtop. It was pulling teeth to get source code released for these. Why would a motherboard manufacturer suddenly want to release their proprietary BIOS-replacement code to the world?

          Either Coreboot needs to offer a paid support, closed source version for motherboard manufacturers, or else they have to be prepared implement all functionality and support it 100% through the open source movement. Motherboard manufacturers have no interest in Open Source; the "pay for a Phoenix/AMI/whatever BIOS" has worked for decades. There is no business incentive for a switch.

          Comment


          • #50
            Originally posted by locovaca View Post
            It is, except that Coreboot has never made itself a serious alternative. As I said before, even if they magically get things working for a large number of boards, they still have to get some sort of support system in place.
            That's not part of the project as such, though. It'll happen if/when more people and companies start taking an interest in coreboot. Red Hat wasn't founded by Linus.

            Motherboard companies have not been very helpful in release their source code to the public; look at the EEE PC and Splashtop. It was pulling teeth to get source code released for these.
            That's happened with a bunch of companies, usually because they're clueless about the requirements of copyleft licenses. I don't see how motherboard vendors are special here.

            Why would a motherboard manufacturer suddenly want to release their proprietary BIOS-replacement code to the world?
            Why wouldn't the proprietary parts be in the payload? I doubt board vendors are differentiating their BIOSes with secret chipset configuration registers. libpayload is permissively licensed, so there shouldn't be a problem with shipping proprietary payloads. Besides, a lot of what motherboard companies are differentiating lately isn't really in the BIOS at all, but rather in Windows apps to configure settings and monitor performance.

            Motherboard manufacturers have no interest in Open Source; the "pay for a Phoenix/AMI/whatever BIOS" has worked for decades. There is no business incentive for a switch.
            I think it's not so far-fetched that motherboard vendors might like to stop paying Phoenix/AMI/whoever and be able to offer a royalty-free customizable firmware to OEM customers. I'm not in the business, though, so I guess that's just speculation.

            Comment


            • #51
              Originally posted by Ex-Cyber View Post
              That's not part of the project as such, though. It'll happen if/when more people and companies start taking an interest in coreboot. Red Hat wasn't founded by Linus.
              It's a chicken and egg problem. They won't provide paid support unless they get enough companies, and a lot of companies won't look at open source products without some sort of proprietary license and paid support.

              Our company is currently switching from WebLogic's Java App Server to tcServer. Why tcServer? Because it's a commercially supported version of Tomcat. If crap blows up, we have someone to ring up and yell at to fix. For a lot of organizations this is an important safety net.

              That's happened with a bunch of companies, usually because they're clueless about the requirements of copyleft licenses. I don't see how motherboard vendors are special here.
              In this case it's software that interacts with otherwise the same hardware. If Asus comes up with a spiffy way of overclocking a board, do you think they really want Gigabyte to get their hands on that?

              Why wouldn't the proprietary parts be in the payload? I doubt board vendors are differentiating their BIOSes with secret chipset configuration registers. libpayload is permissively licensed, so there shouldn't be a problem with shipping proprietary payloads. Besides, a lot of what motherboard companies are differentiating lately isn't really in the BIOS at all, but rather in Windows apps to configure settings and monitor performance.
              Well don't take my word for it:

              http://www.gigabyte.us/products/prod...px?pid=3856#ov

              Navigating through the BIOS to change system settings can be a daunting task for users not familiar with control “F”functions and mouse-less navigation. While some EFI BIOS try to address this with a mouse friendly environment, many implementations still lack a certain ease-of-use necessary for most people. With GIGABYTE Touch BIOS™, GIGABYTE engineers have completely re-imagined how users can interact with their BIOS, allowing for a more intuitive user experience. In fact, with a touch screen monitor, GIGABYTE Touch BIOS™ is as easy to use as most apps on your iPhone.
              http://www.asus.com/Motherboards/AMD...ir_IV_Extreme/

              Two BIOS ROM, Two BIOS settings, Twice the overclocking flexibility
              Flashback act as a SaveGame function for BIOS management. Go back to any other BIOS version you like!
              http://www.abit.com.tw/page/en/mothe...&fMTYPE=LGA775

              abit SoftMenu™
              The original jumperless motherboard design allows for CPU setting changes completely through the BIOS. For GigaOverclocking! Boost your PC's Performance by up to 50%. Convenient and easy-to-use fine tuning from within a self-explanatory BIOS menu.
              Many MB manufacturers pimp their "exclusive" BIOS features. Whether or not they are truly exclusive is another point, but if you told your president that you could switch to this fancy new system but you'd have to give away all the "exclusive" code that you wrote, who would do that? Coreboot has to re-license, even if it's seemingly stupid.

              Comment


              • #52
                Originally posted by locovaca View Post
                Many MB manufacturers pimp their "exclusive" BIOS features.
                I never said otherwise. My point is that the "exclusive" part tends to be either built into the hardware configuration (for which the main expense is hardware, and for which the support code would be small and highly board-specific) or based on UI enhancements (which would go into the payload).

                Whether or not they are truly exclusive is another point, but if you told your president that you could switch to this fancy new system but you'd have to give away all the "exclusive" code that you wrote, who would do that?
                On what basis would the "exclusive" code would have to be given away?

                Comment


                • #53
                  Originally posted by locovaca View Post
                  It is, except that Coreboot has never made itself a serious alternative. As I said before, even if they magically get things working for a large number of boards, they still have to get some sort of support system in place.
                  Huh? You mean like the support system put in place by Aware/phoenix/whatever? Oh wait... they have NO SUPPORT SYSTEM AT ALL, and not only that, their code is BLOB JUNK.

                  Comment


                  • #54
                    Originally posted by droidhacker View Post
                    Huh? You mean like the support system put in place by Aware/phoenix/whatever? Oh wait... they have NO SUPPORT SYSTEM AT ALL, and not only that, their code is BLOB JUNK.
                    I thought he meant "support system" in a business sense for OEMs, i.e. companies that OEMs can outsource their BIOS development/QA to and/or harass/sue/etc. if something goes wrong. This sort of thing isn't totally nonexistent for coreboot (see e.g. Sage Electronic Engineering), but the market does appear to be much more developed for PhoenixBIOS and the like.

                    Comment


                    • #55
                      Most of the features of a modern BIOS are useless* anyways.

                      I for one will not miss 16-bit Kruft mode VGA text based, Ill thought out buggy BIOS's with NO hope of anyone EVER updating again.

                      It would be nice to be able to adjust usually 'hidden' subsystems (ACPI, System Management Mode, ect...).

                      Better fan PWM control, suspend states that work better, fixes and workarounds for buggy hardware/firmware.

                      Comment


                      • #56
                        Coreboot is open-source. You can change whatever you want and there are already quite some motherboards with Coreboot support.

                        Comment


                        • #57
                          I think Coreboot will come to prominence as people/OEM's realize what uEFI really means.

                          Total abandonment of 'legacy' hardware AND software!

                          Think about it:

                          If you have a server, odds are, you are using one of the below:

                          Gigabit NIC, SAS RAID/HBA, Fiberchannel Controller

                          -Can't PXE boot without the NIC being supported in the new EFI firmware!
                          -Can't Boot OS from Raid Card without the controller's BIOS extensions!
                          -Can't do.... several of the above without FC Controller's BIOS extensions either!

                          I've realized uEFI will just make the same (or worse) mistakes than the original real-mode x86 BIOS,
                          vendors will never make the effort to improve it - other than to add marketing chrome or shiny bling
                          (which is generally consumer grade garbage-ware).

                          Comment


                          • #58
                            Originally posted by locovaca View Post
                            It takes 8 minutes from password entry to a workable desktop.
                            But you have to admit that this is a nice feature to have, which is why RHEL has copied it for 6.x. 10 minutes from power button to a functioning desktop gives you an extra break in the morning, right when you need it most.

                            Comment


                            • #59
                              Originally posted by locovaca View Post
                              It's a chicken and egg problem.
                              No, it's a catch-22.

                              Considder evolution, where you draw the line between species based on DNA. At some point one declares "From this DNA on, it's a chicken.". Given that a chicken grows from an egg; the egg was there first.

                              Comment


                              • #60
                                Originally posted by V!NCENT View Post
                                Given that a chicken grows from an egg; the egg was there first.
                                That assumes that the first chicken hatched from an egg. If it in fact was the result of a mutation of an animal which didn't hatch from an egg, then it could be said that the chicken came first. I find that unlikely though; for that to have happened, the chicken's ancestor would have to had not hatched from an egg. I suspect that is not the case, but I don't know for sure. :/

                                I agree that it's a catch 22.
                                Last edited by Nobu; 12-24-2011, 06:28 PM.

                                Comment

                                Working...
                                X