Nvidia isn't using TSMC for any of their current products, they're using Samsung. They may switch to TSMC for their more high-end cards, but I don't think that is the case right now.

This isn't true. Nvidia is using TSMC right now for all of the compute products, e.g., see: https://en.wikipedia.org/wiki/Ampere_(microarchitecture)

> TSMC's 7 nm FinFET process for A100

> Custom version of Samsung's 8nm process (8N) for the GeForce 30 series[8]

That is, NVIDIA is using both, TSMC and Samsung. Both are probably maxed out.

Right, so what’s their volume on a100s and are they a favored mining tool? If so, I think that might be the disconnect between what people think nvidia is focused on, and what they are focused on.

Plus, I’ll point out that a lot of information can be hidden by public companies. Amazon was pretty quiet about it’s growth in sales of AWS as a dominant force for a number of years.

> Right, so what’s their volume on a100s and are they a favored mining tool?

I don't think there are numbers available for A100. NVIDIA discloses in Earnings how much it makes from each sector: gaming ~50% and data center (A100) ~40% or so. The data-center GPUs are way more expensive, but one can probably estimate the numbers from there.

I'm not sure why would that be relevant, but if someone wants to get some factual information, that's probably the only way to go.

Regarding CMP chips, NVIDIA says that CMP chips are GeForce chips that don't pass QA for gaming, and that GeForce chips can be used for CMP but are artificially crippled to prevent cryptominers from using them, so that they don't No idea. take GeForce cards out of gamers (which makes sense since gamers are a long term customer base, and cryptominers are usually a ~<1year trend).

A lot of people are saying that they don't believe this information (e.g. LTT), but AFAIK nobody has provided any facts yet.

I guess we'll have to wait and see how easy / hard it is to "unlock" GeForce products for cryptomining, and for some independent testing of how good the CMP products are for gaming.

The last time this happened, LTT managed to retrofit CMP products for gaming, so their skepticism isn't unfounded.

But TBH the main argument of LTT analysis is that NVIDIA cares about its shareholders, and is only doing this differentiation to increase their profits.

I am having a hard time following the argument that suggests that this behavior is "outrageous". That's what you would expect of any publicly traded company. That's also what all NVIDIA shareholders would expect..