- cross-posted to:
- [email protected]
- cross-posted to:
- [email protected]
This is again a big win on the red team at least for me. They developed a “fully open” 3B parameters model family trained from scratch on AMD Instinct™ MI300X GPUs.
AMD is excited to announce Instella, a family of fully open state-of-the-art 3-billion-parameter language models (LMs) […]. Instella models outperform existing fully open models of similar sizes and achieve competitive performance compared to state-of-the-art open-weight models such as Llama-3.2-3B, Gemma-2-2B, and Qwen-2.5-3B […].
As shown in this image (https://rocm.blogs.amd.com/_images/scaling_perf_instruct.png) this model outperforms current other “fully open” models, coming next to open weight only models.
A step further, thank you AMD.
PS : not doing AMD propaganda but thanks them to help and contribute to the Open Source World.
‘I’d approve this lawsuit’ is not fact.
Devs are not Nvidia’s customers.
Fire and motion is an advantage that only works if you have the lead… like most monopoly power.
CUDA’s competitors do not matter, because their market share is fuck-all. That’s how actual monopolies work. Not even Standard Oil ever had all of the oil. They just had enough that they decided how things works.
ROCm is open-source, so fuck right off with that bullshit. Non-workstation cards support ROCm because of this.
Dominant marketshare is what monopoly meeeeans. The only cogent nuh-uhs are bickering about exploiting a dominant marketshare, which hey look, is what you keep admitting Nvidia does. They just created proprietary way to run your own god damned code, in such a way that you, personally, would totally understand if Nvidia sued the pants off of AMD, if AMD ever dared to use the bridge that you, personally, insist is proof that Nvidia’s not excluding AMD.
Maybe you’re just full of shit.
They absolutely are for datacenter deals. Datacenters need products they can sell, and they’ll buy more of what their customers will use, which is whatever costumers have integrated with. A dramatically lower price can encourage customers to integrate with something else.
It works if you’re a scrappy startup, or a larger company looking to reinvent itself. It honestly works pretty much anywhere with a lot of R&D opportunity.
My company is doing just that right now as we’re trying to capture market share. We’re trying to catch up in one area, while also innovating in others. We’re leaving our flank a bit open by putting the core in maintenance mode, but there’s a pretty high barrier to entry there so we can afford to take the risk.
So? If only AMD is using it, what does that really change? OneAPI is also open source and seems to be getting some traction with other manufacturers.
ROCm is open source for two reasons:
It’s not open source to encourage Nvidia or Intel to join in.
Then you’re using the colloquial definition instead of the legal one (i.e. antitrust).
Exploitation isn’t a problem, anti-competitive abuse of a dominant position is. Every company exploits their unique value proposition, that’s business. It’s only anti-competitive if you cross a certain line, such as collusion.
Equivocation. The topic was game devs. You can’t erase context and respond to how words sound.
The point of saying “devs are not Nvidia’s customers” was pointing out, the thing you said you’d call monopolistic behavior actually happened, and then, you made up an excuse to ignore it.
Oh hi, the point I’m making, what are you doing in someone else’s comment?
Half this conversation is you agreeing with me in an adversarial tone of voice.
Again: Standard fucking Oil fell under the colloquial definition. As did Microsoft when they lost that lawsuit that’s awful close to this situation, which differs only by being worse. Total monopoly almost never happens.
“The law is violated only if the company tries to maintain or acquire a monopoly through unreasonable methods.” Being a monopoly is not the crime. You can just say a business has a monopoly, based on market share. Because that’s what the god dang word means.
CUDA has a monopoly in GPU compute. Nvidia uses that monopoly to prevent competition in hardware sales.
And I likely agree with you on various things, such as:
I put my money where my mouth is. I buy AMD GPUs and CPUs, Intel NICs, etc. I use Linux exclusively, and I used to buy Nvidia when they were better on Linux (back in Catalyst days).
My disagreement is whether Nvidia is a monopoly, and my understanding of antitrust law says they aren’t. They’ve perhaps pushed some ethical boundaries, but every business does that, and I don’t see anything anti-competitive (by the legal standard). They’re not making deals with Microsoft to fumble/delay AMD driver launches, that’s on AMD. They’re not making exclusivity deals with retailers to not stock AMD products. They’re not suing AMD to prevent import and sale of their products with frivolous lawsuits. If you have evidence of any of that, let’s discuss it. But not sharing IP or not protecting IP isn’t anti-competitive.
Nor am I claiming it does. Monopolistic behavior is the standard here. Google is getting hit with antitrust because they’ve been abusing their dominant position. Valve isn’t, probably because they’re not abusing their position (EGS is doing more monopolistic behavior, they just don’t have the market share to matter).
Standard Oil bought all the competition. Microsoft nerfed their competition on their dominant platform.
Nvidia isn’t like those examples. They’re just out-innovating their competition and keeping their IP close. You’d have a very hard time proving that’s anti-competitive to a judge (might have a little luck w/ a jury if the defense is asleep).
Whether something is a monopoly is determined by how close the competition is. AMD is very competitive and Intel is getting there, and Nvidia’s pricing is kept in check due to that competition. So no, I don’t think they have a monopoly because they still have relevant competition and aren’t using illegal means to cut them out of market, they’re just making more attractive products.
The first part is arguably true, but I’d reword it to focus on the hardware. Nvidia isn’t doing anything anti-competitive with CUDA, they’re just keeping their IP to themselves. That’s not illegal or unethical, it’s just not very pro-consumer. Nvidia isn’t doing anything proactively to stop customers from writing portable code.
CUDA has such a monopoly that - as you point out - AMD’s direct alternative is not well-supported even by AMD. OpenCL should matter, but doesn’t.
How we got here is the culmination of efforts to prevent direct competition through a series of dishonest gimmicks buoyed by pick-a-lesser-word-for-bribery of game devs. The pattern of behavior is more important than any specific act.
How Nvidia’s using this status quo is to seize a trillion-dollar bubble. Are their products several orders of magnitude better? Plainly not. So something’s fucky.
It’s so fucky that AMD tiptoes around any connection to running CUDA on their cards, because of the lawsuit Nvidia would obviously attempt. A lawsuit so complex and expensive it might ruin AMD even if they’re completely right about everything. A lawsuit you see no problem with.
Despite a precedent that recreating APIs is fine, and a precedent that withholding APIs can be monopolistic.