NVIDIA discussion [2024]

  • Thread starter Deleted member 2197
  • Start date

Micron HBM3E helps reduce data center operating costs by consuming about 30% less power than competing HBM3E offerings
Micron Technology, Inc. (Nasdaq: MU), a global leader in memory and storage solutions, today announced it has begun volume production of its HBM3E (High Bandwidth Memory 3E) solution. Micron’s 24GB 8H HBM3E will be part of NVIDIA H200 Tensor Core GPUs, which will begin shipping in the second calendar quarter of 2024. This milestone positions Micron at the forefront of the industry, empowering artificial intelligence (AI) solutions with HBM3E’s industry-leading performance and energy efficiency.
 
Of course they are/do. Any other company would use their leverage to control their partners/customers as well, if they could.

The difference is that Nvidia is probably dominant enough in their market that them doing it is illegal.

Being a monopolist/having market power is not in itself wrong, it just means that you have to play under different rules. Rewarding brand loyalty (or punishing disloyalty, same thing) is something that is entirely legal in a fully competitive market, but becomes illegal if you have market power as defined by antitrust law. Just looking at the revenue split, I think Nvidia qualifies. Intel did the same thing to AMD back in the Opteron era, and had to pay them a cool B to make the lawsuits go away.
 
The difference is that Nvidia is probably dominant enough in their market that them doing it is illegal.
Similar to AMD's dominance and influence over partners in the game console market?
 
Similar to AMD's dominance and influence over partners in the game console market?

Depends, I don't know how large a market share Nintendo has and whether that is considered comparable. If a game console maker wanted to do orders from both AMD and some other source, and AMD punished them for this, they'd quite possibly get sued and lose.

Again, a dominant market position is not by itself illegal or wrong, and in AI NVidia got there by the purest possible means; by providing a better product, earlier, and helping to build the infrastructure that underlies the entire sector. It just limits the kind of actions that are legal for you to do. Standard Oil didn't become dominant by dirty tricks, Rockefeller was the earliest player in the space who invested heavily into improving refining chemistry which gave them a product that was both more economical to make and better for customers. But once this early advantage gave them enough power to shake the market, they eventually used every dirty trick in the book to maintain that position, and later laws were created to make this impossible.
 
Being a monopolist/having market power is not in itself wrong, it just means that you have to play under different rules.
Nvidia is not "a monopolist" though and their market "power" comes from their products advantage. Saying that you should "play under different rules" here is the same as saying "hey why don't you make your products worse for the sake of competition".
 
Similar to AMD's dominance and influence over partners in the game console market?
these are vastly different scenarios so no, not similar

( but if there are questionable practices to mention wrt AMD's console's participation i think it would make for some interesting posts, they just wouldn't belong in this thread)
 
Last edited:
these are vastly different scenarios so no, not similar

( but if there are questionable practices to mention wrt AMD's console's participation i think it would make for some interesting posts, they just wouldn't belong in this thread)
All such discussions belong in a RSPC thread as it is highly speculative and based on rumor, not fact. Also considering original source of the rumor and any motives or financial incentives involved would garner discussion related any possible involvement by his former employer.

I agree the console discussion would make for some interesting posts surrounding dominate market influence on the software sector and the impact of discouragement of rival software technology on consumer choice in non-dominate markets.

But these all would be RSPC topics at best.
 
The difference is that Nvidia is probably dominant enough in their market that them doing it is illegal.

There are likely “unfair dealing” laws that cover heavy handed behavior like imposing conditions on the sale that hurt the competition. I won’t sell to you if you’re talking to the other guys too probably falls into that category.

However on the monopoly point, Nvidia aren’t dominant in data center or AI the way that Microsoft or Apple are dominant in their markets. It’s a plug-in product with no network effects to speak of. Each customer makes an independent decision to buy what fits their needs without needing to care what other people are buying. The customer facing software stack (PyTorch, tensorflow etc) isn’t owned or controlled by Nvidia and support lots of different hardware. The data and models (GPT, Resnet) aren’t either. Compare that to Apple who wield far more control over their entire ecosystem.

And of course there’s tons of competition. Anything you can run on an H100 you can run on a contemporary Xeon or Epyc stack. Nvidia doesn’t have a monopoly on AI training and inference. The definition of monopoly doesn’t include “more popular because better at doing the same thing as competing products”.

Intel vs AMD isn’t a good analogy as Intel controlled the software (x86) that AMD was running on.
 
Similar to AMD's dominance and influence over partners in the game console market?
I don’t think this is dominance. There’s 0 value for nvidia to engage in space. Little revenue, no benefit. AI space is what balloons their market cap, not consoles. One may even say that it could lower your value.
 
I don’t think this is dominance. There’s 0 value for nvidia to engage in space. Little revenue, no benefit. AI space is what balloons their market cap, not consoles. One may even say that it could lower your value.
Using hardware market dominance to encourage game producers (software) to avoid rival technology in PC ports to increase market penetration where they have a less dominant position.
 
Nvidia doesn’t have a monopoly on AI training and inference. The definition of monopoly doesn’t include “more popular because better at doing the same thing as competing products”.
Your post is correct, the monopoly, if someone were to make the argument, is in CUDA. No cuda, then you’re not going to get any library support. So typically we see cpu based AI libraries or cuda based libraries. The OpenCl unfortunately did not take off. If it had, we would see larger competition in the space. It’s an interesting thing because data science space is completely dominated by open source, except for this one thing: CUDA
 
Using hardware market dominance to encourage game producers (software) to avoid rival technology in PC ports to increase market penetration where they have a less dominant position.
There’s no barrier for entry for nvidia to enter the console space. And nvidia owns 80% of the GPU space on PC.

They don’t bid for deals because the price requirement for silicon in the console space is so low, margins aren’t worth it and I don’t even know if they have chips that can go so low in margin even if they event want to. I don’t see them supporting anything but Nintendo since they have a bunch of Jetson style chips that would fit that market well.
 
Your post is correct, the monopoly, if someone were to make the argument, is in CUDA. No cuda, then you’re not going to get any library support. So typically we see cpu based AI libraries or cuda based libraries. The OpenCl unfortunately did not take off. If it had, we would see larger competition in the space. It’s an interesting thing because data science space is completely dominated by open source, except for this one thing: CUDA
However CUDA IP is owned by Nvidia, not an independent third party service company.
 
However CUDA IP is owned by Nvidia, not an independent third party service company.
Yea. I know. But no one else can deploy cuda on their cards. Thus even if you make AI hardware you have no library support, therefore no one will purchase your stuff unless they are willing to remake libraries in opencl.

It is a bit like windows in that sense. All the applications are made for windows and not Linux. And because of this everyone chooses windows, not because it’s necessarily a better OS.

That dominance may not last forever, but I see why someone would point to similar parallels.
 
Yea. I know. But no one else can deploy cuda on their cards. Thus even if you make AI hardware you have no library support, therefore no one will purchase your stuff unless they are willing to remake libraries in opencl.

It is a bit like windows in that sense. All the applications are made for windows and not Linux. And because of this everyone chooses windows, not because it’s necessarily a better OS.

That dominance may not last forever, but I see why someone would point to similar parallels.
Nvidia products includes both a hardware and software components. It is a moat and always has been so nothing new there except it's popularity.
In the same vein AMD hardware has it's ROCM software component (AFAIK doesn't run on Nvidia or Intel gpus).
 
Nvidia products includes both a hardware and software components. It is a moat and always has been so nothing new there except it's popularity.
In the same vein AMD hardware has it's ROCM software component (AFAIK doesn't run on Nvidia or Intel gpus).
Yep I agree, which is why there hasn’t been any legal battle about it. You’re not likely to be able to prove it and we see a variety of hardware being used to do all sorts of things (from training to just running models) in the AI space. Nvidia is just most dominant in the training space.
 
Back
Top