Announcement

Collapse
No announcement yet.

Tiny Corp Changes Course Yet Again With Plans To Offer AMD Radeon GPUs

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

  • qarium
    replied
    Originally posted by Forge View Post
    That's actually why I have an AppleTV in the living room. Does 4K, HDR, HFR, whatever, it's completely closed, and if it stopped working, I could throw it away and buy something else.
    I really don't care about HDCP on my PC anymore. I used to, but realized I was doing it wrong.
    you are absolutely right this closed source firmware stuff only to protect HDCP on the pc becomes a burden of Legacy

    realistically dGPUs do no longer need this DRM/copyprotection HDCP stuff the reason as you say is standalone devices and also all the CPUs have a iGPU with HDCP inside so why in hell would anyone need a dGPU with that crap inside ?

    AMD is just a slow thinker here if they would put a small iGPU inside of every CPU then they could cut down their dGPU of all these useless transistors.

    Leave a comment:


  • Forge
    replied
    Originally posted by qarium View Post

    you can buy a standalone bluray player or a standalone netflix/youtube box or use the standalone android app function of your smart TV...

    means then nothing ever touch your PC/workstation/laptop but people want to perform this task in their PC/Laptop/workstation.

    no joke 1 time it was a long time ago bridgman even told me to just buy a bluray box to watch DRM/copyprotection media he even said hey dude on this box runs linux... it is just nailed and closed down so that you can do nothing with it.

    "It would be a separate ASIC with its own driver, distinct from the GPU driver."

    if in the end you have a complete standalone second device with no conection to your pc/laptop/workstation

    so whats the point ?
    That's actually why I have an AppleTV in the living room. Does 4K, HDR, HFR, whatever, it's completely closed, and if it stopped working, I could throw it away and buy something else.

    I really don't care about HDCP on my PC anymore. I used to, but realized I was doing it wrong.

    Leave a comment:


  • qarium
    replied
    Originally posted by novideo View Post
    It would be a separate ASIC with its own driver, distinct from the GPU driver. However it would probably need its own video decoder, bringing up cost, as it can't just send unencrypted video to the GPU without opening a massive vulnerability. I didn't think of that before.
    you can buy a standalone bluray player or a standalone netflix/youtube box or use the standalone android app function of your smart TV...

    means then nothing ever touch your PC/workstation/laptop but people want to perform this task in their PC/Laptop/workstation.

    no joke 1 time it was a long time ago bridgman even told me to just buy a bluray box to watch DRM/copyprotection media he even said hey dude on this box runs linux... it is just nailed and closed down so that you can do nothing with it.

    "It would be a separate ASIC with its own driver, distinct from the GPU driver."

    if in the end you have a complete standalone second device with no conection to your pc/laptop/workstation

    so whats the point ?

    Leave a comment:


  • novideo
    replied
    I recall reading something about a vulnerability in WebKit which was fixed upstream within a day of its discovery, but Safari didn't get the fix until macOS's annual release cycle.
    It would be a separate ASIC with its own driver, distinct from the GPU driver. However it would probably need its own video decoder, bringing up cost, as it can't just send unencrypted video to the GPU without opening a massive vulnerability. I didn't think of that before.

    Leave a comment:


  • qarium
    replied
    Originally posted by novideo View Post
    I trust it, I just would've liked to get all the details, so would've liked to read his post.
    That would be interesting. Would it still be possible with openSIL?
    Isn't the CPU microcode still closed source even in OpenSIL ? as i unterstand it OpenSIL only makes everything around that opensource
    with OpenSIL the microcode of the cpu does not become open-source and also the firmware part of the iGPU will also not become opensource.

    its really like you make a opensource UEFI but the firmware and the microcode is still closed source.

    Originally posted by novideo View Post
    What if there was just a standalone chip/chiplet for DRM/HDCP that was its own ASIC, without the CPU/GPU needing to implement DRM?
    the question is when and how does it touch the operating system and for example the browser ?

    the data from netflix as they want it goes encrypted from the website directly inside the ASIC on the black box inside the gpu and then send to the monitor stil encrypted because the monitor still has DRM/HDCP...

    well that the iGPU of the CPU has all the DRM/HDCP bits and by this the dGPU no longer need this bits is really our best option.

    Originally posted by novideo View Post
    Now that is a really interesting idea. AMD could release pre-RDNA/CDNA design minus the DRM/sensitive bits under the (A)GPLv3-only, and the strong copyleft would ensure that nobody could use it against them. Projects like LibreSOC could integrate it while reducing budget and focusing on the CPU part only. A fully libre OpenPOWER/RISC-V core with Vega GPU. AMD could observe from the sideline and use it to decide how to approach opening up their own firmware.
    amd could even be more conservative and use some GCN1.0 design anything with Vulkan driver support would do for a start.

    AGPLv3 or later clause would be needed to protect AMD from commercial copy cats yes but keep in mind chinese manufacturers don't care about the license.

    my opinion is AMD does lose nothing and would get positive PR if they release GCN1.0 design without DRM/HDCP as open hardware design.

    the reason why they lose nothing is that intel and chinese manufacturers already have better designs than GCN1.0

    we could even argue that Vega64 design is already obsolete because a intel arc 750 is already faster.

    Vega20 chip of the radeon7 would be the first chip design who would have any value in the market the radeon7 only had 3500 shaders vega20 in apple products had 4048 shaders and its 7nm design so its not hopeless outdated like vega64...

    if they want to enter the open hardware computer hardware anything older than vega64/radeon7 would be poitnless because ROCm/HIP works on vega64 with blender and so one and older designs like polaris do not work.

    from past discussion this has only this problem: evil lawyers could argue that a open hardware vega64/radeon7 design could help to crack DRM/HDCP on other older non-open hardware vega64/radeon7 because it is the same architecture.

    but to prevent this to mitigate this evil lawyers dream they could make minimal changes to make it look like it is different ISA/architecture.

    then people could fix the source code for this driver easily because the changes are only minimal

    Leave a comment:


  • novideo
    replied
    Originally posted by qarium View Post

    this is info from years many years ago. i would say its to much to give you the exact quote or even only the article. you can try google it but... its maybe better to just ask bridgman.

    or else just trust my memory. you could try to google the requirements for HDCP but its very simple anything what HDCP touches in meaning of data need to be in black box outside of the black box anything need to be encrypted to be secure from COPY the data.
    I trust it, I just would've liked to get all the details, so would've liked to read his post.

    then only then AMD could produce dGPU PCIe cards without any "HDCP copy protection" support and because of this they could make a open-source firmware.

    if the iGPU all the time covers all DRM/CopyProtection/H264/h265/AV1/AV2/VP9/HDCP then there would be no need to but any of this poison in the dGPUs.

    and people who want a full open system would use a OpenPOWER 10 mainboard and would put in a dGPU with open-source firmware.​
    That would be interesting. Would it still be possible with openSIL?

    What if there was just a standalone chip/chiplet for DRM/HDCP that was its own ASIC, without the CPU/GPU needing to implement DRM?

    AMD also could always EOL end of life designs as Open-Source and open hardware like they did abolish polaris and Vega64
    AMD could release the Vega64 design without DRM/CopyProtection/HDCP as open-source

    AMD could generate massive positive PR for this and their risk that a vega64 design would tank their newest CDNA3/RDNA3 designs is zero.

    i use a vega64 myself and i am pretty sure people would buy a open-hardware Open-Firmware Vega64 even if the card is only 14nm or 12nm.

    some prople could argue that such a card is already obsolete but i see a real market for open-hardware Open-Firmware GPUs

    and it would compete agaist designs like libresoc/redsemiconductor

    with that move amd could test the market without risk for their current CDNA/RDNA designs.​
    Now that is a really interesting idea. AMD could release pre-RDNA/CDNA design minus the DRM/sensitive bits under the (A)GPLv3-only, and the strong copyleft would ensure that nobody could use it against them. Projects like LibreSOC could integrate it while reducing budget and focusing on the CPU part only. A fully libre OpenPOWER/RISC-V core with Vega GPU. AMD could observe from the sideline and use it to decide how to approach opening up their own firmware.
    Last edited by novideo; 29 March 2024, 07:31 PM.

    Leave a comment:


  • qarium
    replied
    Originally posted by novideo View Post
    Thanks, would you mind sending a link to bridgman's full post? Or giving a hint as to what article it was under?
    there is one possibility where HDCP copy protection for dGPUs becomes defacto irrelevant and by this makes it possible to be more open and make open-source firmware.
    lets assume all future computer systems are like ryzen 7000 and they all have iGPU inside the cpu with "HDCP copy protection" support.
    we are not here yet because the newest threatripper line does not have iGPU sinde the cpu and also AMD do release ryzen7000 without iGPU...

    but this is the possible way for AMD and the linux/Open-Source and AI community is see if AMD say well for the future all CPUs have iGPU and they do not release any variant without iGPI and all iGPU have "HDCP copy protection" support inside and also have h264/h265 and VP9 and AV1 and AV2 decode support.

    then only then AMD could produce dGPU PCIe cards without any "HDCP copy protection" support and because of this they could make a open-source firmware.

    if the iGPU all the time covers all DRM/CopyProtection/H264/h265/AV1/AV2/VP9/HDCP then there would be no need to but any of this poison in the dGPUs.

    and people who want a full open system would use a OpenPOWER 10 mainboard and would put in a dGPU with open-source firmware.

    but also there is the problem what with open-source firmware the competition could copy this AMD design but in a fast market this would not care at all because as soon as the competition copy the design it could already be obsolete and replaced by better design.

    also people could assume reverse logic in that because if libresoc/redsemiconductor and cheap chinese manufacturers would be successfull in the open-source-firmware market this could cause trouble for amd this means if amd covers this market this could make sure the competition can not find a market gap to enter the market. this means competition like libresoc/redsemiconductor could vanish just because amd close the market gap.

    AMD also could always EOL end of life designs as Open-Source and open hardware like they did abolish polaris and Vega64
    AMD could release the Vega64 design without DRM/CopyProtection/HDCP as open-source

    AMD could generate massive positive PR for this and their risk that a vega64 design would tank their newest CDNA3/RDNA3 designs is zero.

    i use a vega64 myself and i am pretty sure people would buy a open-hardware Open-Firmware Vega64 even if the card is only 14nm or 12nm.

    some prople could argue that such a card is already obsolete but i see a real market for open-hardware Open-Firmware GPUs

    and it would compete agaist designs like libresoc/redsemiconductor

    with that move amd could test the market without risk for their current CDNA/RDNA designs.

    and keep in mind amd could use the lastest Vega20 chipdesign as a base in 7nm and unlike radeon7 with 3500 shaders they could use the chip they used only for apple workstations with the full 4086 shader cores. this card should have at minimum 16gb vram.

    bridgman AMD really should try it there is really a market for this.

    Leave a comment:


  • qarium
    replied
    Originally posted by Forge View Post
    That's on the GPU alone. Each GPU has 96GB of HBM. The CPU has 512GB of LPDDR5X. Any GPU can access any GPU's memory and the CPU's memory at 900GB/s. 512GB+96GB+96GB = the 700GB I was talking about.
    https://developer.nvidia.com/blog/nv...ture-in-depth/
    the MI300A can also be coubled with other MI300A and can also access the ram of the other coupled MI300A

    Leave a comment:


  • NM64
    replied
    Originally posted by Quackdoc View Post

    reminds me of the USB devices for portable HDDs that had two usb type-A male adapters for power

    Originally posted by pixelherodev View Post
    The funny thing is, USB ports even as far back as my thinkpad put out enough power that this is _totally unnecessary_.
    There is actually one computer electronics device that definitely does need those two USB type-A male plugs for power, but it's not an x86 computer:

    The Wii U.

    For whatever weird reason, its USB ports are a bit notorious for being a bit low on amperage; a single USB port works fine for SSDs* but it commonly needs two USB ports for a mechanical hard drive, and this is all from first-hand experience as well.

    The funny thing is, the Wii tends to be perfectly fine powering a mechanical hard drive with a single USB port despite it not even officially supporting USB drives and it only being useful for homebrew stuff, yet the Wii U officially supported the use of USB drives for storing purchased digital downloads (...or unofficially storing homebrew stuff of course).


    *presumably most modern SSDs at least; I've noticed the Crucial M4 from over a decade ago requires more power and is my go-to test as to whether a given USB connection will be able to power an SSD or not; I have an ~6ft external power switch extension that also has two USB 2.0 ports that isn't able to power the Crucial M4 for example, but is able to do newer SSDs. Just be warned that Crucial SSDs newer than the M4 like to permanently lock itself to read-only if it doesn't receive enough power.​

    Leave a comment:


  • Forge
    replied
    Originally posted by qarium View Post

    GH100 has only 80GB HBM2e VRAM
    GB200 has only 141 GB HBM3E

    the MI300A is 192GB of HBM3
    That's on the GPU alone. Each GPU has 96GB of HBM. The CPU has 512GB of LPDDR5X. Any GPU can access any GPU's memory and the CPU's memory at 900GB/s. 512GB+96GB+96GB = the 700GB I was talking about.

    The NVIDIA Grace Hopper Superchip Architecture is the first true heterogeneous accelerated platform for high-performance computing (HPC) and AI workloads. It accelerates applications with the…

    Leave a comment:

Working...
X