Announcement

Collapse
No announcement yet.

It's Still Going To Be Tough Getting The OpenChrome VIA KMS Driver In The Linux Kernel

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

  • #11
    This is not helped by the fact that VIA vastly overstated its power consumption targets and its chips are prone to burn out. As the unichrome developer, you know, the guy who came up with structured display driver development now called "modesetting", i have collected a ton of VIA hw over the years (later 2000s), and theres quite a few which just burned out. I am not sure i wish to go through my vast collection and find out which of them are permanently bricked.

    I also do not understand how one cannot go straight for atomic modesetting. Just like RandR1.2, KMS was a poor mans version of some hints of structured display driver development, and always had been.

    Comment


    • #12
      Originally posted by libv View Post
      This is not helped by the fact that VIA vastly overstated its power consumption targets and its chips are prone to burn out.
      Also the x86 VIA chipsets for Intel CPUs are full of quirks and have crappy (unstable) AGP or PCIe ports in many boards.

      Comment


      • #13
        Originally posted by libv View Post
        I also do not understand how one cannot go straight for atomic modesetting. Just like RandR1.2, KMS was a poor mans version of some hints of structured display driver development, and always had been.
        Could you please elaborate about this statement? It could be quite interesting to know. What solution/framework/subsystem would you do instead of KMS? Some graph would enlight us too.

        Comment


        • #14
          Originally posted by timofonic View Post
          Could you please elaborate about this statement? It could be quite interesting to know. What solution/framework/subsystem would you do instead of KMS? Some graph would enlight us too.
          Current KMS is getting to be pretty decent, but it wasn't always so.

          I was one of 3 guys interested in display driver development back in the 2003-2006 timeframe (the other two were egbert eich and thomas winnischhofer). The via driver that i was reworking at the time taught me that most or all necessary info for full display driver support without VBE/int10 was possible, and i was seeing more and more structure in the mess of register tables (read https://libv.livejournal.com/14035.html to see how my insights were accepted). Unlike most I actually did rewrite xf86 display drivers at the time, and touched the xf86 modesetting infrastructure in the xserver.

          My view for display drivers back in 2005 was:
          * structure: what the unichrome driver showed me is that encoders were separate hw blocks, making it logical to turn them into modular sw blocks. In a private discussion egbert eich said "but tv encoders are not ramdacs and ramdacs are not tv encoders", so i started naming them "output devices", which also stuck, but somehow this was morphed into a combination of encoder/connector/monitor mashup in RandR1.2. My subsequent RadeonHD code went for proper encoder(called output)/connector/monitor separation on a driver level (but not on an infrastructure level).
          * helper based: the main lesson learned when xfree86 went from 3.6 to 4.0. Drivers were high level hooks, with tons of helper functions available to do some of the infrastructural heavy lifting. This fundamental insight was completely overlooked with RandR 1.2, and which only started to re-appear in the atomic KMS world.
          * properties/attribute based: there's an enormous amount of complexity in the different display blocks, and structure only goes so far. This was added to RandR1.2 a month or two later in RandR1.2 development, when keithp actually did look at TV encoders more closely. Properties only became more central in the atomic KMS world.

          The original KMS was a header file blatantly created from the RandR1.2 header file by Jacob Bornecrantzer, pretty much copy-paste with some names kernelized. KMS only gained traction when the politics surrounding a free ATI driver had AMD force ATI to stop pushing a competing Xorg driver, and so the "players" ran to KMS.

          Atomic KMS was envisioned to be property based, with atomic transactions, by Ville Syrjala, in the first year of him working at intel. I worked with Ville while at Nokia, and when my whole team ran off to intel, i was not allowed to join for political reasons (and because i was not living in Finland). It is only this new generation of actual driver developers (read: not the powerplayers of the xfree86-xorg fork, but people who actually love poking at registers) that drove the transition to atomic and to make drivers extensively helper based. We could've had something close to modern atomic modesetting, initially still in userspace (which, when structured right, which i would've of course done, could've been ported to the kernel easily), somewhere in 2007-2008. This would've been 6-8 years ahead of when atomic actually hit mainline, depending on how you count it.

          Back in 2007, Egbert realized that i was not only barking up the right tree, but actually spot on. He got me into SuSE linux, where luck had it that a former SuSE manager at AMD needed some input on how to tame ATI, and SuSE was about to hire the then most talented and experienced display driver developer around. 6 weeks, 15kloc later, and ATI had no way back but to accept an open source strategy for radeon. And then the shit really started flying.

          As for openchrome, that was initially a fork from unichrome, as i would not accept VBE based display code back in, while i was in the middle of distilling out structure in the display driver. This fork was in march 2005. Now openchrome is all about getting display to work by employing logic and structure. So 13ys on, Kevin Brace is actually constantly stating that it was wrong to fork openchrome away, but his talk at XDC could fool anyone who does not know the actual history.

          Comment


          • #15
            Originally posted by oliw View Post
            The last bit of hardware for these was released 9 years ago, only supports DX10
            As far as I know, Chrome 640/645 (C-640/645) from the VX11 chipset (release date: August 2012) had support for DX11. And the latest GPU model is Chrome 860 (C-860) from the ZX-D SoC, released in December 2017.

            pre-DirectX 9 era

            Savage 3D, Savage 4 LT, Savage 4 GT, Savage 4 Pro-M, Savage 4 Pro, Savage/MX, Savage/IX, Savage 4 Xtreme, Savage 2000, Savage XP/AlphaChrome (only prototype, not released)
            Family: next generation after ViRGE
            Type: dGPU (AGP 2X/PCI for Savage 3D; AGP 2X for Savage 4 LT; AGP 4X for Savage 4 GT and later; Savage 4 Pro also supports PCI; Savage/MX & Savage/IX variants - notebook chips)
            NVIDIA equivalent: RIVA TNT-GeForce 256 series
            First version: Savage 3D - Jun 1998
            Latest version: Savage 2000 - Nov 1999; Savage XP/AlphaChrome (only prototype, not released) ~ 2002
            Key features: DirectX 7 (without T&L), hardware MPEG-2 decoding, trilinear-filtering, S3 Texture Compression (S3TC)
            Notes:
            - S3TL (hardware T&L): https://en.wikipedia.org/wiki/S3_Savage#Savage_2000

            SuperSavage MX, SuperSavage IX
            Family: Savage/MX & Savage/IX derivative
            Type: dGPU (AGP 4X, notebook chips)
            First and latest version: SuperSavage MX & SuperSavage IX - Mar 2001

            ProSavage 4, ProSavage 8, ProSavage 8 DDR
            Family: hybrid Savage 4/Savage 2000 (Savage 4 3D core and Savage 2000 2D functionality)
            Type: iGPU (embedded in the chipset)
            First version: ProSavage in the PM133 chipset (northbridge) - https://www.viatech.com/en/silicon/l...hipsets/pm133/ - Sep 2000
            Latest version: ProSavage 8 DDR in the P4M266 chipset (northbridge) for Intel CPU - https://www.viatech.com/en/silicon/l...ipsets/p4m266/ - Nov 2001; ProSavage 8 DDR in the KM266 chipset (northbridge) for AMD CPUs - https://www.viatech.com/en/silicon/l...hipsets/km266/ - Jan 2002

            Unichrome, Unichrome Pro, Unichrome Pro A, Unichrome Pro B, Unichrome Pro II
            Family: updated ProSavage
            Type: iGPU (embedded in the chipset)
            Intel equivalent: Intel Extreme Graphics/Intel Extreme Graphics II
            First version: Unichrome (originally named CastleRock) in the CLE266 chipset (northbridge) - https://www.viatech.com/en/silicon/l...ipsets/cle266/ - Jun 2002; Unichrome (this was the first use of the name Unichrome) in the KM400/KM400A chipset (northbridge) - https://www.viatech.com/en/silicon/l...ipsets/km400a/ - Apr 2003
            Latest version: Unichrome Pro II (C-Pro II) in the CX700M/CX700M2 chipset (single-chip solution) - https://www.viatech.com/en/support/eol/cx700m/ - Sep 2006; Unichrome Pro II (C-Pro II) in the CN800 chipset (single-chip solution) - https://www.viatech.com/en/silicon/l...hipsets/cn800/ - Nov 2006
            Key features: DirectX 7 (without T&L), hardware MPEG-2 decoding, X.Org 7.4, XGL/AIGLX, ACPI Suspend to Ram and Suspend to Disk, HW Video overlay by X extension XV, XAA 2D HW acceleration, EXA 2D HW acceleration, 3D DRI, OpenGL, thru AGP path, XRandR 1.2

            DirectX 9 era

            DeltaChrome S4 Pro, DeltaChrome S8, DeltaChrome S8 Nitro, DeltaChrome S8 ULP, GammaChrome S18 CE, GammaChrome S18 PRO
            Family: next generation after Savage XP/AlphaChrome
            Type: dGPU (AGP 8X for DeltaChrome; PCIe for GammaChrome and later)
            NVIDIA equivalent: GeForce FX (5xxx) series
            First version: DeltaChrome S4 Pro - mid 2004
            Latest version: GammaChrome S18 PRO - Mar 2005
            Key features: DirectX 9.0, Shader Model 2.0, Chromotion Video Engine 2.0, hardware MPEG-2/MPEG-4/WMV9 decoding, Duo-Rotate

            Chrome S25, Chrome S27, S3 Graphics 2300E
            Family: next generation after GammaChrome (S18 CE/S18 PRO)
            Type: dGPU (PCIe)
            NVIDIA equivalent: GeForce 6 (6xxx) series
            First and latest version: Chrome S25 & Chrome S27 - Nov 2005
            Key features: DirectX 9.0c, Shader Model 2.0+, OpenGL 1.5, AcceleRAM, MultiChrome, Chromotion Video Engine 3.0, hardware MPEG-2/MPEG-4/WMV9-HD decoding, Duo-Rotate, PowerWise Technology

            Chrome9, Chrome9 HC, Chrome9 HC3, Chrome9 HCM, Chrome9 HD
            Family: DeltaChrome derivative
            Type: iGPU (embedded in the chipset)
            Intel equivalent: Intel GMA
            First version: Chrome9 (C-9) in the K8M890 chipset (northbridge) - https://www.viatech.com/en/silicon/l...ipsets/k8m890/ - Oct 2005
            Latest version: Chrome9 HD (C-9 HD) in the VX900/VX900M/VX900H chipset (single-chip solution) - https://www.viatech.com/en/silicon/chipsets/vx900/ - Mar 2010
            Key features: DirectX 9.0, Shader Model 2.0, Chromotion Video Engine 2.0, hardware MPEG-2/VC-1/WMV-H/DH.264 decoding, X.Org 7.6, ACPI suspend to RAM and suspend to disk, HW Video overlay by X extension XV, UXA 2D HW acceleration, XAA 2D HW acceleration, EXA 2D HW acceleration, AGP path, XRandR 1.2/1.3
            Used in: VIA VB9001 3.5″ SBC, VIA EPIA-M910 Mini-ITX Board, VIA VB7009 Mini-ITX Board, VIA COMe-8X92 COM Express Module, VIA COMe-8X90 COM Express Module, VIA ETX-8X90 ETX Module, VIA ETX-8X90-10GR ETX Module
            Still supported by the manufacturer: yes (only Chrome9 HD)

            DirectX 10 era

            Chrome 430 GT, Chrome 430 ULP, Chrome 435 ULP, Chrome 440 ULP, Chrome 440 GTX, Chrome 4300E, Chrome 530 GT, Chrome 540 GTX, Chrome 5400E x2, Chrome eH1
            Family: next generation after Chrome S27
            Type: dGPU (PCIe)
            NVIDIA equivalent: GeForce 8 (8xxx) series-GeForce 9 (9xxx) series
            First version: Chrome 430 GT - Feb 2008/Mar 2008 (US market)
            Latest version: Chrome 540 GTX - Feb 2009; Chrome 5400E x2 - Feb 2010; Chrome eH1 - Apr 2010
            Key features: DirectX 10.1, Unified Shader Model/Shader Model 4.1, OpenGL 2.1/3.1, OpenGL ES 2.0, OpenVG 1.1, PCIe 2.0, AcceleRAM, MultiChrome, ChromotionHD 2.0 Programmable Video Engine, hardware MPEG2/MPEG4/VC1/WMVHD/AVS/H.264 decoding, features dual stream HD video playback and PiP (Picture in Picture), Duo-Rotate, PowerWise Technology, VGA/DVI/HDMI/DisplayPort output

            Chrome 520
            Family: Chrome 500 series (530 GT/540 GTX) derivative
            Type: iGPU (embedded in the chipset)
            Intel equivalent: Intel HD Graphics (Westmere-Sandy Bridge)
            First and latest version: Chrome 520 (C-520) in the VIA VN1000 chipset (northbridge) - https://www.viatech.com/en/silicon/l...ipsets/vn1000/ - Dec 2010

            DirectX 11 era

            Chrome 640, Chrome 645
            Family: next generation after Chrome 500 series
            Type: iGPU (embedded in the chipset)
            Intel equivalent: Intel HD Graphics (Ivy Bridge+, Bay Trail+)
            First and latest version: Chrome 640 (C-640) & Chrome 645 (C-645) in the VX11/VX11H chipset (single-chip solution) for VIA CPUs - https://www.viatech.com/en/silicon/chipsets/vx11/ - Aug 2012; Chrome 640 (C-640) & Chrome 645 (C-645) in the VX11H/VX11PH chipset (single-chip solution) for Zhaoxin CPUs - http://en.zhaoxin.com/Upload/201707061728520253.pdf - Q1 2014
            Key features: DirectX 11, Unified Shader Model/Shader Model 5.0, OpenGL 3.2, OpenGL ES 2.0, OpenCL 1.1, Chromotion Video Engine 5.0, hardware MPEG-2/WMV9/VC-1/H.264 decoding, LVDS/VGA/DVI/HDMI/DisplayPort output, X.Org X11R7.x with H/W 2D acceleration through S3G customized RXA architecture, SAMM / MAMM / Xinerama with multiple display, DVI dual-link up to 2560x1600 resolution, 90/180/270 degree display rotation, H/W accelerated direct-rendering OpenGL 3.2 API, H/W accelerated indirect-rendering OpenGL 3.2 API, Composite Desktop with Compiz, full featured RandR 1.2 function, Kernel Mode Setting with standalone module, full H.264, VC-1, WMV9 and MPEG-2 VLD bitstream H/W decoding through VDPAU or VA API driver
            Used in: VIA EPIA-E900 Pico-ITXe Board, VIA EPIA-P910 Pico-ITX Board, VIA EPIA-M920 Mini-ITX Board, VIA COMe-9X90 COM Express Module
            Still supported by the manufacturer: yes

            ChromE 1000
            Family: probably Chrome 600 series (640/645) derivative
            First and latest version: ChromE 1000 (C-1000) - Jan 2016
            Notes:
            - http://ranker.sisoftware.net/show_ru...dbbc8f5cd&l=en

            Chrome 320
            Family: Chrome 600 series (640/645) derivative
            Type: iGPU (embedded in the chipset)
            First and latest version: Chrome 320 (C-320) in the ZX-100S chipset (single-chip solution) - http://en.zhaoxin.com/Upload/201707061730130274.pdf - Q2 2016/Aug 2016 (mass production)
            Still supported by the manufacturer: yes
            Notes:
            - ZX-100S chipset has a "256 stream processors iGPU powered by S3 Graphics IP technology and ready for VR"
            - https://www.forum-3dcenter.org/vbull...&postcount=103

            Chrome 860
            Family: Chrome 600 series (640/645) derivative
            Type: iGPU (embedded in the SoC)
            First and latest version: Chrome 860 (C-860) in the ZX-D/KX-5000/KH-20000 (codename: "WuDaoKou") SoC - Dec 2017
            Still supported by the manufacturer: yes
            Notes:
            - https://www.forum-3dcenter.org/vbull...&postcount=103
            - https://en.wikichip.org/wiki/zhaoxin...aokou#Graphics

            Summary

            What we should take care of is:
            - Chrome9 HD (C-9 HD) iGPU from the VX900/VX900M/VX900H chipset - Vendor Id: 1106 ("VIA Technologies, Inc."), Device Id: 7122 ("VX900 Graphics [Chrome9 HD]")
            - Chrome 640 (C-640) iGPU & Chrome 645 (C-645) iGPU from the VX11/VX11H/VX11PH chipset - Vendor Id: 1106 ("VIA Technologies, Inc."), Device Id: 3a01 ("VX11 Graphics [Chrome 645/640]")
            - Chrome 320 (C-320) iGPU from the ZX-100S chipset - Vendor Id: 1d17 ("Zhaoxin"), Device Id: 3a02 ("ZX-100 C-320 GPU")
            - Chrome 860 (C-860) iGPU from the ZX-D/KX-5000/KH-20000 (codename "WuDaoKou") SoC - Vendor Id: 1d17 ("Zhaoxin"), Device Id: 3a03 ("ZX-D C-860 GPU")
            - Chrome iGPU from the ZX-E/KX-6000/KH-30000 (codename "LuJiaZui") SoC
            - Chrome iGPU from the ZX-F/KX-7000/KH-40000 SoC

            All of these GPUs have support for Windows 7/Windows 10.

            Enjoy the videos and music you love, upload original content, and share it all with friends, family, and the world on YouTube.



            Unfortunately, none of them is supported by the OpenChrome driver. However, Linux drivers exist ("via_chrome9_dri.so"/"via_drv.so", "s3g_dri.so"/"s3g_drv.so").

            BTW, there are also "Elite" GPUs included in the ARM SoCs, e.g. ZX-1000, ZX-2000, etc.

            Comment


            • #16
              Originally posted by the_scx View Post
              As far as I know, Chrome 640/645 (C-640/645) from the VX11 chipset (release date: August 2012) had support for DX11. And the latest GPU model is Chrome 860 (C-860) from the ZX-D SoC, released in December 2017.

              ...

              Unfortunately, none of them is supported by the OpenChrome driver. However, Linux drivers exist ("via_chrome9_dri.so"/"via_drv.so", "s3g_dri.so"/"s3g_drv.so").

              BTW, there are also "Elite" GPUs included in the ARM SoCs, e.g. ZX-1000, ZX-2000, etc.
              Quite a lot of what you list is vapourware and was never released.

              After the savages, the chrome dGPUs were few and far between, and mostly extremely rare early samples for reviewers or engineering samples. I have seen a gammachrome on ebay once in the previous decade, and i got a qimonda? Chrome S27 as those actually were sold. I also have a Chrome 435 as part of the VIA VB8003, which sits next to a VX800 northbridge.

              As for iGPUs... VX900 saw some sales in thin-clients. VN1000 never shipped. And while rare, some VX11 EPIA Pico boards can be found still today.

              Everything more recent is simply vapourware atm.

              Btw, I know this, since i have been collecting unichrome family devices and motherboards since 2003. I have between 40 and 50 different devices in this room, most of which i have not booted in years. They mostly are very interesting hardware, with very interesting display options. I also have a dozen or so display daughterboards. I lack a VX11H as those pico-itx boards are still falsely deemed as valuable by ebay sellers, i do not want to overspend on VIA electrowaste.

              Fun fact, the K8M890 (Chrome9) was the first Northbridge to have fully native fully free display support in coreboot: https://libv.livejournal.com/19432.html

              Comment


              • #17
                Originally posted by libv View Post
                Quite a lot of what you list is vapourware and was never released.
                I wouldn't say that. Some of these chips were probably not publicly available (e.g. Chrome 520 from the VN1000 chipset), while others were extremely difficult to buy (Chrome 5400E from the Chrome eH1 card or - in general - embedded variants of Chrome, like 2300E or 4300E), but most of them, despite being unpopular, could be obtained. For example, I remember that the Chrome cards from 400/500 series (e.g. Chrome 540 GTX) were available on e-commerce platforms, like eBay (North America) or Allegro (Central Europe). A lot of products could be ordered directly from the VIA Store. Currently, solutions based on VX900 (Chrome9 HD) and VX11 (Chrome 640/645) should be relatively easy to purchase.

                Originally posted by libv View Post
                I also have a Chrome 435 as part of the VIA VB8003, which sits next to a VX800 northbridge.
                I remember that board. VIA called this "Trinity Platform", because it combined VIA CPU (Nano), VIA MSP (VIA VX800 chipset with Chrome 9 series iGPU) and VIA/S3G GPU (Chrome 400 series dGPU).
                I regret that something like this has not appeared earlier. VIA for a long time lacked motherboards with a relatively modern GPU. There were boards with Unichrome graphics, even with Chrome9 series iGPU, but for a long time there were no motherboards with a built-in Chrome 400/500/600 series GPU.
                VIA really wanted to return to the desktop (or even gaming) market and it was obvious. That's why the integrated/embedded version of Chrome 400 (Trinity platform), Chrome 500 (VN1000 - unfortunately, samples only) and finally the Chrome 600 (VX11/VX11H) came along.
                https://www.tomshardware.com/reviews...1000,2779.html
                Unfortunately, at some point they turned out that it is too late to compete with Intel, AMD and NVIDIA on this market. Around 2012, they practically suspended the development of the desktop platform, including the development of graphics chips and Linux drivers for them.

                Fun fact: around 2009 year, NVIDIA had plans to create an ION2 platform compatible with VIA processors. It was a really great opportunity for VIA to conquer the desktop market (netbooks, nettops, barebones, etc.), because their Nano processor was generally better than Intel Atom. Unfortunately, nothing like that happened.
                https://www.tomshardware.co.uk/nvidi...ews-30361.html
                https://www.engadget.com/2009/02/20/...sors-this-year

                Originally posted by libv View Post
                Everything more recent is simply vapourware atm.
                Not exactly. VIA gave up efforts in this regard, it's a fact. But since 2014, Zhaoxin has been working hard to provide Chinese domestic x86 CPU with all the necessary elements, including BIOS, chipset and GPU. And they chose VIA solutions as a base. The ChromE 1000 (C-1000) is probably just a prototype, but the Chrome 320 (C-320) and Chrome 860 (C-860) had to be produced in thousands of pieces, as they are used in ZX-100 chipset and ZX-D SoC.
                "In 2018, Zhaoxin aims to ship about 100,000 to 200,000 computer chips" (Ye Jun, chairman and chief executive of Zhaoxin)

                Please note the progress that Zhaoxin has made over these four years:
                - They replaced AMI (American Megatrends) BIOS with ZD-Tech and BYOSOFT solutions.
                - They added support for Chinese Hash Algorithms SM3 and SM4 in the PadLock ACE (Advanced Cryptography Engine)
                - They provided new and complete SoC design.
                - Northbridge has been moved on die.
                - FSB has been abandoned and replaced by a new point-to-point high-speed interconnect crossbar. New SoC has "8 integrated x86 cores consisting of two clusters of four cores each connected over a new point-to-point crossbar, improving the internal bandwidth and latency considerably" (WikiChip).
                - They integrated the memory controller into SoC.
                - DDR4 replaced DDR3.
                - They integrated the graphics processor into SoC. It is Chrome 860 (C-860) iGPU, derived from Chrome 600 series.
                - PCIe 3.0 replaced PCIe 2.0.
                - Added support for USB 3.1 Gen2 (Type-C) ports.
                - The core has been improved: better out-of-order execution, optimized branch prediction, reduced pipeline, re-balanced execution engines, etc.
                - They decided to produce chips in HLMC (Shanghai Huali Microelectronics Corporation) and SMIC (Semiconductor Manufacturing International Corporation), instead of just TSMC.
                - They released new drivers supporting new operating systems: Window 10 government-version provided by C&M Information Tech (a joint venture between Microsoft and China Electronics Technology Group.), Ubuntu Kylin, NeoKylin, NFS China, iSoft Server OS, etc.

                I am very impressed with the progress they have made in such a short time. In my opinion, over the past four years they have achieved more (for desktop use) than the Russians with their Elbrus from the beginning (1999 - plans, 2008 - production).
                It is known that Zhaoxin focused primarily on the SoC core construction and probably didn't make a revolution in the GPU. However, sooner or later they will have to significantly improve their graphics chip. This means the development of Chrome iGPUs. They have no other choice. They can't just use Intel (U)HD Graphics/Iris (Plus/Pro) Graphics, AMD Radeon/Radeon Pro or NVIDIA GeForce/Quadro. ARM graphics (like Qualcomm Adreno, Imagination Technologies PowerVR, Vivante Vega, Broadcom VideoCore or even ARM Mali) also aren't an acceptable solution for them. Theoretically, they could buy SiS (Silicon Integrated Systems) and use their IP to build a new GPU, but both SiS Xabre 600 and XGI Volari 8300/Duo V8 Ultra are far behind the VIA Chrome 600 series. I really hope that in the next 5-10 years a modern GPU based on S3G IP will be created.
                The Chinese government already put 250 million USD in the joint venture between Zhaoxin and VIA in 2014. And this year, the Chinese government is said to have launched a fund of over 300 billion yuan (47 billion USD) to support its own semiconductor industry.
                Today Zhaoxin hardware is available only in China, but tomorrow, who knows? They are already talking about offering their products abroad in the future. And ten years ago, nobody expected that Huawei, ZTE, Xiaomi, Lenovo, and other Chinese brands will be so popular on the worldwide smartphone market. The next decade can be very exciting. Just saying.
                Last edited by the_scx; 30 May 2018, 01:25 PM.

                Comment


                • #18
                  I'm surprised that there's so much passion towards the crappy VIA hardware. How about implementing drivers for Mali400? Pretty common in chinese low end tablets and SBCs. So.. real hardware to work with.

                  Comment


                  • #19
                    Originally posted by caligula View Post
                    I'm surprised that there's so much passion towards the crappy VIA hardware. How about implementing drivers for Mali400? Pretty common in chinese low end tablets and SBCs. So.. real hardware to work with.
                    Guess who started that. It's the same person who lost a lot of revenue and job opportunities because it, and who got some crap for it too, and who cannot be bothered any more to waste his own time for ungrateful whiny w....... to reap those nice rewards.

                    Comment


                    • #20
                      Originally posted by the_scx View Post
                      I wouldn't say that. Some of these chips were probably not publicly available (e.g. Chrome 520 from the VN1000 chipset), while others were extremely difficult to buy (Chrome 5400E from the Chrome eH1 card or - in general - embedded variants of Chrome, like 2300E or 4300E), but most of them, despite being unpopular, could be obtained. For example, I remember that the Chrome cards from 400/500 series (e.g. Chrome 540 GTX) were available on e-commerce platforms, like eBay (North America) or Allegro (Central Europe). A lot of products could be ordered directly from the VIA Store. Currently, solutions based on VX900 (Chrome9 HD) and VX11 (Chrome 640/645) should be relatively easy to purchase.
                      Those later chipsets are still too expensive for me to waste money on. And while i have quite some rare "vintage" graphics cards, these S3/Chrome cards are very very rare indeed and the window for acquiring them was very very small.

                      Originally posted by the_scx View Post
                      Unfortunately, at some point they turned out that it is too late to compete with Intel, AMD and NVIDIA on this market. Around 2012, they practically suspended the development of the desktop platform, including the development of graphics chips and Linux drivers for them.
                      Typical bad VIA execution. They had found a nice niche market with mini-itx and the epia-M. And then they dropped the ball on linux support. They had everything there to be the open source graphics champions for instance, but they did not care enough.

                      Originally posted by the_scx View Post
                      Fun fact: around 2009 year, NVIDIA had plans to create an ION2 platform compatible with VIA processors. It was a really great opportunity for VIA to conquer the desktop market (netbooks, nettops, barebones, etc.), because their Nano processor was generally better than Intel Atom. Unfortunately, nothing like that happened.
                      This market was also pretty short lived, as then tablets hit, and netbooks are now chromebooks.

                      Also, remember the HP mininote? That was 2008 era, and VIA based, by a reputable big manufacturer. I was at SuSE doing RadeonHD, with SuSE also doing preloads for HP, so i was seeing and hearing some of the fallout of that. Halfway through the product life, the mininote got intel internals, that's how bad it was. It was the typical bad VIA execution. Heck, my mininote died just half a year into testing.

                      Originally posted by the_scx View Post
                      Not exactly. VIA gave up efforts in this regard, it's a fact. But since 2014, Zhaoxin has been working hard to provide Chinese domestic x86 CPU with all the necessary elements, including BIOS, chipset and GPU. And they chose VIA solutions as a base. The ChromE 1000 (C-1000) is probably just a prototype, but the Chrome 320 (C-320) and Chrome 860 (C-860) had to be produced in thousands of pieces, as they are used in ZX-100 chipset and ZX-D SoC.
                      "In 2018, Zhaoxin aims to ship about 100,000 to 200,000 computer chips" (Ye Jun, chairman and chief executive of Zhaoxin)
                      Do you have a handle on where to get any Zhaoxin hw, if at all?

                      Originally posted by the_scx View Post
                      Please note the progress that Zhaoxin has made over these four years:
                      - They replaced AMI (American Megatrends) BIOS with ZD-Tech and BYOSOFT solutions.
                      *sigh* Those are pretty much just old-school BIOS vendors but chinese versions, right? They could've gone for coreboot, as coreboot is aligned even better with the chinese communist party goals. It's a sure sign of copying, not improving, and therefore bad execution again.

                      Originally posted by the_scx View Post
                      - They added support for Chinese Hash Algorithms SM3 and SM4 in the PadLock ACE (Advanced Cryptography Engine)
                      Hrm, that's the first time i hear of those. I think i would trust those even less than existing hash/encryption algorithms.

                      Originally posted by the_scx View Post
                      - They provided new and complete SoC design.
                      - Northbridge has been moved on die.
                      - FSB has been abandoned and replaced by a new point-to-point high-speed interconnect crossbar. New SoC has "8 integrated x86 cores consisting of two clusters of four cores each connected over a new point-to-point crossbar, improving the internal bandwidth and latency considerably" (WikiChip).
                      - They integrated the memory controller into SoC.
                      - DDR4 replaced DDR3.
                      - They integrated the graphics processor into SoC. It is Chrome 860 (C-860) iGPU, derived from Chrome 600 series.
                      - PCIe 3.0 replaced PCIe 2.0.
                      - Added support for USB 3.1 Gen2 (Type-C) ports.
                      - The core has been improved: better out-of-order execution, optimized branch prediction, reduced pipeline, re-balanced execution engines, etc.
                      I read: trying to make it competitive again. All laudable though.

                      Originally posted by the_scx View Post
                      - They decided to produce chips in HLMC (Shanghai Huali Microelectronics Corporation) and SMIC (Semiconductor Manufacturing International Corporation), instead of just TSMC.
                      Originally posted by the_scx View Post
                      - They released new drivers supporting new operating systems: Window 10 government-version provided by C&M Information Tech (a joint venture between Microsoft and China Electronics Technology Group.), Ubuntu Kylin, NeoKylin, NFS China, iSoft Server OS, etc.
                      Now this is where it will all seriously fall down.

                      Originally posted by the_scx View Post
                      The Chinese government already put 250 million USD in the joint venture between Zhaoxin and VIA in 2014. And this year, the Chinese government is said to have launched a fund of over 300 billion yuan (47 billion USD) to support its own semiconductor industry.
                      Today Zhaoxin hardware is available only in China, but tomorrow, who knows? They are already talking about offering their products abroad in the future. And ten years ago, nobody expected that Huawei, ZTE, Xiaomi, Lenovo, and other Chinese brands will be so popular on the worldwide smartphone market. The next decade can be very exciting. Just saying.
                      I have dealt with the likes of allwinner over an extended period of time, just like i have dealt with VIA, as an open source graphics driver developer, seeing their actions, seeing what the results of their decisions are. Allwinner was no better, and perhaps even worse than VIA and that makes me very weary of any actual progress being made where it needs to happen for Zhaoxin.

                      In these two specific cases, both VIA and allwinner had extremely bad execution, and the ended up having cheap throw-away products, making them cheap throw-away companies. One cycle they're big, next cycle they're gone. To beat this, the hw needs to be solid and needs to have solid driver support. And i have not seen solid drivers coming out of this part of the world yet.

                      Comment

                      Working...
                      X