Originally posted by sophisticles
View Post
This is what i am talking about, incomprehensible nonsense where you think some imagined partnership has something to do with a different company's dominance.
Microsoft is NVIDIA's biggest customer, accounting for 19% of NVIDIA's total revenue last year:
MS and Khronos haven't joined forces against anyone, joining forces would be if they decided to work together to make GPU's to compete against NVIDIA's GPU.
But that's not what happened, you clearly do not understand the article you linked to.
So I apologize. it's not word salad, the problem is earlier in the production pipeline, it's thought salad that is your major malfunction, probably from all the drugs you have admitted you take,
Now in all fairness to you, you are correct about one thing, Microsoft, and the rest of NVIDIA's large AI customers, have been developing their own custom AI solutions for in-house use, in the hopes of saving the costs associated with buying NVIDIA's GPUs, but notice that they are not looking to switch to a competitor such as AMD or Intel, nor are they interested in marketing their in-house chips as competitors to NVIDIA, they are working on trying to eliminate reliance on third party vendors:
Read the section under NVIDIA's Top Customers.
Microsoft is NVIDIA's biggest customer, accounting for 19% of NVIDIA's total revenue last year:
MS and Khronos haven't joined forces against anyone, joining forces would be if they decided to work together to make GPU's to compete against NVIDIA's GPU.
But that's not what happened, you clearly do not understand the article you linked to.
So I apologize. it's not word salad, the problem is earlier in the production pipeline, it's thought salad that is your major malfunction, probably from all the drugs you have admitted you take,
Now in all fairness to you, you are correct about one thing, Microsoft, and the rest of NVIDIA's large AI customers, have been developing their own custom AI solutions for in-house use, in the hopes of saving the costs associated with buying NVIDIA's GPUs, but notice that they are not looking to switch to a competitor such as AMD or Intel, nor are they interested in marketing their in-house chips as competitors to NVIDIA, they are working on trying to eliminate reliance on third party vendors:
Read the section under NVIDIA's Top Customers.
you wrongfully believe that if microsoft does not produce GPU cards that then microsoft does not compete with nvidia...
i have bad news for you the hardware part of nvidia is very small to the software/driver development part of nvidia.
this means it is a fact that DirectX Compute called "Microsoft DirectCompute" does in fact compete with Nvidia CUDA in the market.
if you read wikipedia it competes against: "The DirectCompute architecture shares a range of computational interfaces with its competitors: OpenCL from Khronos Group, compute shaders in OpenGL, and CUDA from NVIDIA."
this alone proof that you are wrong Microsoft does in fact compete against Nvidia in the software field (not in the hardware field)
but you already discovered that even this claim that microsoft only competes in software and not in hardware is nonsense to quote yourself: "you are correct about one thing, Microsoft, and the rest of NVIDIA's large AI customers, have been developing their own custom AI solutions for in-house use, in the hopes of saving the costs associated with buying NVIDIA's GPUs,"
if you face reality then the field of compeition goes down from
"Microsoft DirectCompute"
OpenCL
Vulkan Compute/SPIR-V from
compute shaders in OpenGL/WebGL
CUDA from NVIDIA."
AMD ROCm/hip
Apple Metal
Intel oneAPI
WebGPU compute shaders
to then only:
SPIR-V (Vulkan-compute, Microsoft DirectCompute ,Rusticl-openCL,WebGPU,Zink OpenGL-On-Vulkan)
CUDA/ZLUDA/HIP
Apple Metal
Intel oneAPI
as you can easily see here the Microsoft move to SPIR-V is in fact a big power shift and a danger to the nvidia monopole. this does not mean Nvidia is defeated but look what if apple decide they will switch to SPIR-V with their Apple Metal API ??? such a move is not impossible because apple did this move already with WebGPU...
now some experts here in the forum say and this will maybe happen to is that some people will write a CUDA PTX (Parallel Thread Execution) runtime similar to ZLUDA but instead of ROCm/HIP as target they will target SPIR-V
i know this does not exist yet but just hypothetical speaking then nearly everyone is on SPIR-V but Nvidia is the only one still on CUDA...
but you foolish claim: "MS and Khronos haven't joined forces against anyone"
lol it does not only look like MS and khronox has joined forces it looks like EVERYONE is joining forces against Nvidia...
Leave a comment: