Mesa's Classic Drivers Have Been Retired - Affecting ATI R100/R200 & More

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts
  • MrCooper
    Senior Member
    • Aug 2008
    • 624

    #81
    Originally posted by syrjala View Post

    I have quite a lot of i915c patches lying around. The real problem is getting them reviewed.
    If you step up to be the i915c maintainer (on the Amber branch now), you can merge patches even if you can't get them reviewed.

    Comment

    • Jabberwocky
      Senior Member
      • Aug 2011
      • 1192

      #82
      Originally posted by oiaohm View Post

      That dell poweredge R300 would have to be fairly old if it came with Ati R200 or R100. Ati R200 cards stopped being fitted by Dell in 2003 replaced by Ati R300 cards. They also did a new version of the Poweredge R300 motherboard changing to polymer electrolytic capacitors few months after dell stopped fitting Ati 200 cards. Yes the prior capacitors really only have a 20 year life span. The polymer electrolytic capacitors are 30-40 years.

      2003-2005 we see vendors change over to the polymer electrolytic capacitors. So there is a technical line in the sand here.

      Yes the Ati R100 and R200 used the same 20 year life span capacitors as the motherboards from that time frame. Same capacitor problem exist on dropped Nvidia gpus.

      Basically its not surprising that those parts have been working up until this point because we had not crossed the end of life points. Problem is we getting close to the hardware end of life points if people don't rework the boards replacing capacitors and so on there is going to be a increasing failure rate.

      Yes no R100/R200 drivers will not stop you from using simpledrm with software rendered opengl and I am not sure if this will be a better or worse expensive than using the R100/R200 drivers..
      Interesting. I just checked (booted up the machine). The chip uses a PCI bus and is called ATI ES1000, from what I can tell it's an R100 based Rage 6. The system currently has Linux 3.13 installed and the driver in use is "radeon".

      According to techpowerup the ATI ES1000 was launched October 18th, 2007. https://www.techpowerup.com/gpu-specs/es1000.c2102

      It seems I made a mistake with the name it's a poweredge R200 not R300. The user manual has the date Auguest 2007 on it. Dell claims sales typically goes on for 5 years after release so that means that the device was sold until late 2012. I suspect it was probably closer to 2010 though. I bought it around 2011 from a private company.

      It was fun going down the rabbithole, but now I need to get back to work lol.

      Comment

      • guara
        Junior Member
        • Feb 2021
        • 31

        #83
        Originally posted by oiaohm View Post

        [...] the prior capacitors really only have a 20 year life span. The polymer electrolytic capacitors are 30-40 years.

        2003-2005 we see vendors change over to the polymer electrolytic capacitors. So there is a technical line in the sand here.

        [...]Same capacitor problem exist on dropped Nvidia gpus.
        The old electrolytic caps failed way earlier, especially if one relied on questionable power supplies.
        My old Pentium 4 631 + Geforce 6200 AGP machine had to be recapped after roughly 10 years of being subjected to a low quality undersized PSU.

        Comment

        • oiaohm
          Senior Member
          • Mar 2017
          • 8268

          #84
          Originally posted by Jabberwocky View Post
          Interesting. I just checked (booted up the machine). The chip uses a PCI bus and is called ATI ES1000, from what I can tell it's an R100 based Rage 6. The system currently has Linux 3.13 installed and the driver in use is "radeon".
          Horrible part ES1000 is not a R100 that why I missed it. Darn custom make as cheap as possible chip. ES1000 only support opengl 1.0 and Direct X 6. Everything else in the Ragc 6 group including R100 and R200 supports direct X 7 and opengl 1.3 with hardware support.

          Yes ES1000 is 8 million transistors and all the other Rage 6 technology are 30 million+ transistors. ES1000 is a cut back bit of work. I am really not sure if dropping a bit of hardware like this back to simpledrm and software rendering will be a improvement or not. Serous-ally this could be a improvement due to how little GPU the ES1000 really is.

          Sorry I had to add this. ES1000 8million transistor count matches that of a Rage 4 8 million transistor count that is also direct x6 but is opengl 1.1 not 1.0. So yes its less functional than a Rage 4 card yet its mean to be a Rage6.
          Last edited by oiaohm; 06 December 2021, 11:02 AM.

          Comment

          • billyswong
            Senior Member
            • Aug 2020
            • 694

            #85
            Originally posted by Jabberwocky View Post

            Interesting. I just checked (booted up the machine). The chip uses a PCI bus and is called ATI ES1000, from what I can tell it's an R100 based Rage 6. The system currently has Linux 3.13 installed and the driver in use is "radeon".

            According to techpowerup the ATI ES1000 was launched October 18th, 2007. https://www.techpowerup.com/gpu-specs/es1000.c2102

            It seems I made a mistake with the name it's a poweredge R200 not R300. The user manual has the date Auguest 2007 on it. Dell claims sales typically goes on for 5 years after release so that means that the device was sold until late 2012. I suspect it was probably closer to 2010 though. I bought it around 2011 from a private company.

            It was fun going down the rabbithole, but now I need to get back to work lol.
            Windows Aero Effect was introduced as an extension of DirectX 9 in Windows Vista, "released to manufacturing" in November 2006. Then this iGPU first released in August 2007 only supported DirectX 6. Wow, I guess someone was recycling old chips for bare-bones display need there.

            Comment

            • oiaohm
              Senior Member
              • Mar 2017
              • 8268

              #86
              Originally posted by guara View Post
              The old electrolytic caps failed way earlier, especially if one relied on questionable power supplies.
              My old Pentium 4 631 + Geforce 6200 AGP machine had to be recapped after roughly 10 years of being subjected to a low quality undersized PSU.
              Problem is in those older system there will be parts people did not recap as well. Like that Geforce 6200 AGP in 2005 it was still being made with the older generations of electrolytic caps. Yes low quality undersized PSU could half the projected lifespan under good conditions on the motherboard and most doing recapping 10 years ago left the GPU and other problem parts alone because those has been shielded by the motherboard caps. Remember those caps that were shielded by the motherboard caps so did not show symptoms 10 years ago will start showing symptoms now. Yes if a person did a part recap the caps they left behind on the motherboard is going to come back and get them now.

              Basically this is the second round of o darn my electrolytic caps have decide to bulge and cease to operate correctly except now the hardware is worth even less. Yes the 10 year old bits of hardware getting this issue they were still worth enough to fix but this time around they are now 20 year old parts.

              Comment

              • bridgman
                AMD Linux
                • Oct 2007
                • 13183

                #87
                Originally posted by oiaohm View Post
                Horrible part ES1000 is not a R100 that why I missed it. Darn custom make as cheap as possible chip. ES1000 only support opengl 1.0 and Direct X 6. Everything else in the Ragc 6 group including R100 and R200 supports direct X 7 and opengl 1.3 with hardware support.

                Yes ES1000 is 8 million transistors and all the other Rage 6 technology are 30 million+ transistors. ES1000 is a cut back bit of work. I am really not sure if dropping a bit of hardware like this back to simpledrm and software rendering will be a improvement or not. Serous-ally this could be a improvement due to how little GPU the ES1000 really is.

                Sorry I had to add this. ES1000 8million transistor count matches that of a Rage 4 8 million transistor count that is also direct x6 but is opengl 1.1 not 1.0. So yes its less functional than a Rage 4 card yet its mean to be a Rage6.
                Sounds like an RV50 aka "Peanut" - half of the pixel processing of an R100, and possibly no vertex shader block. I'm not sure if ES1000 was the same chip or same/similar design moved to a newer process... I think it was the latter but not sure.

                Anyways, they were a common server console GPU for a long time and used R100 driver code (probably with SW vertex processing).

                EDIT - RN50, not RV50. It was also called RV100 if I remember correctly.

                There was also an RV200 but I don't remember the specifics.
                Last edited by bridgman; 06 December 2021, 04:12 PM.
                Test signature

                Comment

                • oiaohm
                  Senior Member
                  • Mar 2017
                  • 8268

                  #88
                  Originally posted by bridgman View Post
                  Sounds like an RV50 aka "Peanut" - half of the pixel processing of an R100, and possibly no vertex shader block. I'm not sure if ES1000 was the same chip or same/similar design moved to a newer process... I think it was the latter but not sure.

                  Anyways, they were a common server console GPU for a long time and used R100 driver code (probably with SW vertex processing).
                  The documentation of the ES1000 says no pixel shaders. Full R100 did not have hardware vertex processing so software vertex processing was part of R100s.

                  ES1000 is very weak. I am basically not sure if the R100 driver running ES1000 is really doing anything more beneficial than using simpledrm vesa/EFI framebuffer and software rendering with the ES1000. Of course ES1000 might be a reason for someone to make a new R100 driver like what has happened with matrox that been used in the same area if there is a benefit. Yes matrox g200 driver update in the year 2020 mostly come about due to server console usage as well.

                  bridgman RV50 Peanut might be less cut down than the ES1000. So the ES1000 a very weak GPU even that its made in 2007.

                  I don't have a ES1000 to play with to check how it performs with simpledrm and software rendering vs r200 classic. Yes ES1000 is about the only part I can see that uses the R100 driver that not likely to fall in the next 3 years due to the parts used in construction. I am not sure that the ES1000 has enough functionality to justify not using a generic driver like simpledrm.

                  The reality here hardware does have a functional life. At point of driver desperation it is a good time to have a close look at the remaining hardware and see if the old driver is doing anything beneficial with the hardware it supports.





                  Comment

                  • agd5f
                    AMD Graphics Driver Developer
                    • Dec 2007
                    • 3939

                    #89
                    ES1000/RN50 were chips for server consoles. The 3D hardware was not validated on them, so it was never enabled in the driver so you aren't losing anything here.
                    Last edited by agd5f; 06 December 2021, 04:17 PM. Reason: typo

                    Comment

                    • bridgman
                      AMD Linux
                      • Oct 2007
                      • 13183

                      #90
                      Originally posted by agd5f View Post
                      The 3D hardware was not validated on them
                      Ahh, that solves the discrepancy between my recollection of what the hardware did and other people's recollection of what the product did
                      Test signature

                      Comment

                      Working...
                      X