You must log in or # to comment.
Correct me if I’m wrong and I won’t even get into why the chinese government is an absolute hypocrite here once again, but Nvidia is probably the defacto monopoly corporation that deserves an anti trust probe the least because competition isn’t even trying as of late.
the chinese government is an absolute hypocrite here once again
Going tit-for-tat with the US on investigations of rival tech companies isn’t hypocritical, its retaliatory. This is a strategic response to an escalating tech sector trade war.
Nvidia is probably the defacto monopoly corporation that deserves an anti trust probe the least because competition isn’t even trying
NVIDIA has acquired 23 different companies in the last 5 years. Six in 2024 alone. Back in 2020, they straight up bought out ARM for $40B, eliminating an enormous chunk of their domestic competition. This was a strategic prelude to cornering the manufacturing of AI-centric hardware.
These mergers never should have been allowed to take place. They’ve squashed anything resembling competitive pricing and created a choke point in distribution that the bigger tech companies have exploited to crowd out competitors in the nascent AI industrial space. In a sane world, an anti-trust claim would be open-and-shut. NVIDIA is caught red-handed in the act. They’re straight up bragging about it to their investors. Its the singular reason for their skyrocketing stock price.
The only incentive rivals have, at this point, is to get large enough for a company like NVIDIA to buy you out. There is no competition because the market has already been cornered.
They didnt buy ARM. They tried to, but the buy out failed. Biden’s FTC shut that merger down.
Are you implying China didn’t ban most western networks and isn’t a state capitalist economy? Hardly anything reactionairy about it, it’s daily business for them since the dawn of the internet.
Retaliatory is not reactionary. Very different meanings.
Last time I checked the US started all this witch hunt and trade war, as they are worried their global influence is waning with China gaining foothold in Africa, Asia and even South America.
So call me sceptical but the US doesn’t care for the Uighurs, or any other human rights violations that China is involved in, they are just protecting their trade interests.
And yes, this antitrust case should have been open against NVIDIA ages ago.
You should check again and realize most major US platforms have always been banned in China.
CUDA is a proprietary platform that (officially) only runs on Nvidia cards, so making projects that use CUDA run on non-Nvidia hardware is not trivial.
I don’t think the consumer-facing stuff can be called a monopoly per se, but Nvidia can easily force proprietary features onto the market (G-Sync before they adapted VESA Adaptive-Sync, DLSS etc.) because they have such a large market share.
Assume a scenario where Nvidia has 90% market share and Nvidia cards would still only support adaptive sync via their proprietary G-Sync solution. Display manufacturers will obviously want to tailor to the market, so most displays will release with support for G-Sync instead of VESA Adaptive-Sync. 9 out of 10 customers will likely buy a G-Sync display as they have Nvidia cards. Now everyone has a monitor supporting some form of adaptive sync. AMD and Nvidia release their new GPU generation and isolated (in this hypothetical scenario), AMD cards are 10% cheaper for the same performance and efficiency as their Nvidia counterparts. The problem for AMD here is that even though per $ they have the better cards, 9 out of 10 people would need new displays to get adaptive sync working with an AMD card (because their current display only supports the proprietary G-Sync), and AMD can’t possibly undercut Nvidia by so much that the customer can also buy a new display for the price difference. This results in 9 out of 10 customers going for Nvidia again.
To be fair to Nvidia, most of their proprietary features are somewhat innovative. When G-Sync first came out, VESA Adaptive-Sync wasn’t really a thing yet. DLSS was way better than any other upscaler in existence when it released and it required hardware that only Nvidia had.
But with CUDA, it’s a big problem. Entire software projects that just won’t (officially) run on non-Nvidia hardware so Nvidia is able to charge whatever they want (unless what they’re charging is more than the cost of switching to competitor products and importantly porting over the affected software projects).
AMD gpus are just as good as Nvidia CPUs. Intel ones suck, but they’re in the market, too. Gaming market is irrelevant to Nvidia and AMD’s market cap and profits.
Nvidia’s main advantage is their proprietary CUDA software, which makes it so the majority of AI software only runs on Nvidia CPUs, and is incompatible with AMD or Intel CPUs.
Exactly: too many people confuse the monopoly aspect with the consumer gaming stuff, which isn’t even pocket change at this point.
CUDA and AI are the whales in the room, and nVidia has a stranglehold on that market and should be investigated.
(Even though, IMO, this is because AMD did their usual shitty job of software, and basically gave the market away.)
Yes, AMD completely overslept here and their ROCm is much inferior. But at least regulators can force NVIDIA to open their CUDA library and at least have some translation layers like ZLUDA.
Even though I think they will play the same card like Microsoft obfuscating and making it very confusing to hinder the portability.
But at least regulators can force NVIDIA to open their CUDA library and at least have some translation layers like ZLUDA.
I don’t believe there’s anything stopping AMD from re-implementing the CUDA APIs; In fact, I’m pretty sure this is exactly what HIP is for, even though it’s not 100% automatic. AMD probably can’t link against the CUDA libraries like cuDNN and cuBLAS, but I don’t know that it would be useful to do that anyway since I’m fairly certain those libraries have GPU-specific optimizations. AMD makes their own replacements for them anyway.
IMO, the biggest annoyance with ROCm is that the consumer GPU support is very poor. On CUDA you can use any reasonably modern NVIDIA GPU and it will “just work.” This means if you’re a student, you have a reasonable chance of experimenting with compute libraries or even GPU programming if you have an NVIDIA card, but less so if you have an AMD card.
