- 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.
Yes, because the result would be the same whether it is or isn’t. That means the precedent is murky, which leaves future cases open to consideration.
How so? There’s no assumption of compatibility between devices. If Microsoft changes some APIs on Windows, WINE developers on Linux have no room to claim anti-competitive behavior, they just need to adapt. Assuming that APIs are frozen just because some competitor is basing a product them is bonkers.
The Windows v US anti-trust issue was because was intentionally making performance worse for competition while opening up APIs for only internal use. That’s not what’s happening w/ Nvidia at all, they’re just pushing an API, much like Windows pushes DirectX, and application developers are free to use it, or not. If AMD wants to copy that API, they would have to adapt to any compatibility issues, much like WINE does with Windows APIs.
Other vendors not building compatibility for AMD’s products isn’t “preventing competition.” AMD is free to create a better product, but Nvidia is currently wiping the floor with them hardware-wise, and they’re trying to lock people in on the software side. AMD’s approach should be to:
They’re finally doing #3 for the consumer market with the 9070/9070XT cards. I haven’t been tracking their enterprise offerings, so I don’t know if they’re trying to make inroads there, or just want to expand in consumer first and deal w/ AI later.
AMD is actively prevented from creating a directly competing product. Because Nvidia locks people in… on the software side.
The precedent on copyrightability is that copyright doesn’t even matter. Re-implementing an API is fine, either way.
This hair-splitting over Microsoft’s secret exclusion, versus Nvidia’s in-your-face exclusion, as if that makes Nvidia’s bullshit better, is beyond comprehension.
I am so tired of arguments where I’m just tapping part of someone else’s comment and going, did you read what you just wrote?
Steam isn’t a monopoly because users are free to go elsewhere, they just choose to stay because the experience is better. EGS or GOG could get more market share if they improve their products, but they don’t, so Steam stays dominant.
It’s the same deal with Nvidia. They were first to market, and even today they have a better product. People don’t use Nvidia because they’re locked in, they use Nvidia because their offering is better. That’s why CUDA is dominant and ROCm isn’t.
Here’s a quick comparison:
CUDA:
ROCm:
Maybe. If AMD merely sticks to the published API, they’d probably be fine. However, they’ll always be playing catch-up because they’re not in control of the API.
That’s why they make their own. They get their own form of lockin, control over the API, and they currently can force people to buy the higher end cards since they don’t support ROCm on lower end cards.
The difference is that Microsoft controlled the platform and used better APIs for their own products so they could always win in performance. As in, they knowingly made the competition perform worse on their platform by abusing their control of the platform.
Nvidia isn’t doing anything like that. They just aren’t releasing their API as an open standard, but instead a library. That’s not anti-competitive. AMD can make their own API (they have), or they can try to reverse engineer Nvidia’s (potential legal fight).
Yes.
Steam absolutely has a monopoly. There’s one store that matters and they’re it.
https://sh.itjust.works/post/22820575/12949311
https://sh.itjust.works/post/21848271/12561312
https://sh.itjust.works/post/18613184/11273823
https://sh.itjust.works/post/27285305/14714870
https://sh.itjust.works/post/32680183/16663477
https://sh.itjust.works/post/33084784/16821653
ROCm is open-source. That’s why it’s on some consumer AMD cards. There’s no Nvidia support because CUDA’s market share is “fuck you” territory.
Which is somehow better than competition not performing at all, thanks to arbitrary control of the platform.
Are you saying it would become anticompetitive if Nvidia made CUDA open-source, but it sucked on AMD cards?
A “potential legal fight,” just to let customers run their own code, is a fucking obvious obstacle to hardware competition. What do you think vendor lock-in means?
It’s not a monopoly just because you say it is.
There’s no Nvidia support because it’s an AMD API. There’s also no Intel support because Intel has their own thing as well. It being open source doesn’t mean anything here except maybe that they can offload some maintenance to Linux devs.
It’s not arbitrary. Nvidia created it, so they use it. They’re not obligated to work with competitors.
If AMD had any form of hardware competition for compute over the last 15-20 years, maybe we’d get somewhere. But no, Nvidia created the market, provided the solution, and maintains a lead in the market.
I wish it was different, but AMD missed the boat repeatedly and they’re still not producing interesting enough hardware for compute to get people onto their software platform.
I’m saying if they licensed CUDA to AMD, but throttled performance or just didn’t fix bugs on AMD cards, it would be anti-competitive. That’s much closer to what happened with Microsoft.
The consumer isn’t part of this, the legal fight would be whether AMD can reverse engineer the CUDA API and ABI to work with their own backend solution. That’s a bit more extreme than the Oracle case since the Nvidia libraries were never open, whereas the Java APIs absolutely were.
Yeah fuck all my arguments, right? It must be just because I don’t like it. Swallow that sneering abuse. They have a monopoly because of what words mean.
And anyone could put it anywhere, because it’s open-source. But nobody bothers. Because only CUDA is relevant. CUDA has a monopoly in this space.
“Arbitrary control” is rephrasing what you keep saying, when you insist Nvidia has absolute authority to change shit without telling anybody.
AMD’s compute capabilities have been on-par with their graphics capabilities, which are goddamn near as good as Nvidia’s, except nobody uses AMD for compute, specifically and exclusively because of CUDA’s monopoly on GPU compute.
That’s worse. That’s worse than my obvious bait example. What the fuck?!
Walk me the through how “go to hell” is less anti-competitive than “okay, we’ll cooperate, but poorly.”
WHO DO YOU THINK THEY’RE COMPETING FOR?!
The purpose of CUDA is to let customers run code! They’re the only people who could possibly get locked in! All obstacles to CUDA’s availability, on any other vendor, are obstacles directly for these customers, who are just trying to run their own god-damned code.
How do you write shit like ‘Nvidia’s total enclosure is more extreme than Oracle’ and teleport to pretending that’s a defense of Nvidia? Oracle is the devil. Being worse than them… is not… better. I shouldn’t have to explain this to you - it is what you are saying.
Well yeah, so does AMD. Nvidia having more market share doesn’t change that.
They don’t use AMD for compute because the value proposition isn’t there. Yes, CUDA is part of that, but that’s because Nvidia had been developing and supporting it for almost 10 years before AMD made ROCm.
AMD consistently does a poor job with software support. In fact, ATI/AMD was atrocious on Linux for a long time before AMD decided to embrace FOSS as a competitive advantage, which is why I got an Nvidia card in the first place like 10-15 years ago. AMD is now trying to use “open source” to gain mindshare, but it doesn’t seem to be working at the datacenter.
AMD is late like they always are, and that’s why they’re behind.
Yes, because that’s what Microsoft did. Nvidia is nowhere near that, they’re just refusing to give their IP away.
Agreed.
I never claimed Nvidia was “good,” I claimed they’re not violating anti-trust. Here are actual examples of anti-trust:
You have to go out of your way to violate anti-trust rules, just refusing to share or having a dominant position doesn’t cut it.
Stop snipping sentences in half as if that’s what I fucking said!
“So does AMD” means IT IS ARBITRARY, in direct contradiction to when you declared, “it’s not arbitrary.” I only mentioned “arbitrary control of the platform” referencing your defense of Nvidia’s “control of the API” which you still declare is absolute. Experience cognitive dissonance, god damn you!
AMD is not behind. AMD is excluded.
AMD’s compute capabilities are presumably on-par, dollar for dollar, just like their graphics capabilities. Yet I have to say “presumably” because Nvidia’s spent a king’s ransom ensuring that direct competition does not occur.
Nvidia has reserved the APIs for GPU compute to themselves. CUDA performance on AMD isn’t worse - it’s nothing. It is excluded. AMD does not have a chance to run CUDA programs. It could! - as evidenced by multiple projects which ran CUDA on AMD hardware - but the threat of legal action ensures one vendor controls the de-facto standard for GPU compute.
Having only one vendor that matters is what a monopoly is.
I’m snipping the relevant part so you or someone else can go look it up without it adding too much noise to my comment.
Arbitrary means there’s no reason or purpose for it. They want to iterate quickly, and being forced to stick to a standard gets in the way of that.
If you want a standard, OpenCL is right there. Performance sucks because neither major vendor has cared much about it, but that’s the cross platform API to use.
AMD is only excluded in the sense that Windows APIs don’t work on Linux directly. But just like how WINE exists, HIP exists (AMD created) to run the same code on both. But then you run into the very obvious issue that AMD has a motivation to optimize for their GPUs and not Nvidia’s.
But why deal with that abstraction if you only use Nvidia GPUs anyway?
No, a monopoly has exclusive control of a market. We also use the term to mean a company that abuses its massive market share and therefore acts like a monopoly, such as:
Each of those engaged/engages in anti-competitive behavior with lawsuits, technical barriers, and essentially bribes.
Not sharing your IP isn’t a “technical barrier,” no company is under any obligation to do that. AMD has demonstrated that it can build a bridge to CUDA and ROCm, so unless Nvidia is directly sabotaging users of that or a similar bridge (and not just changing their API for other reasons), they’re not acting like a monopoly.