amd is nothing compared to nvda
$NVIDIA (NVDA.US)$
AMD don't even tell me about it I've used their gpus at work I'm an AI engineer i train these models for a living and AMD gpus AMD software RocM it's a terrible piece of shit it's a pain to work with. if i have a deadline to meet real world deadlines our clients need results from our AI models because they actually help their business these models lead to BILLIONS of dollars of profit or cost saving compared to human experts so they trust the AI models it complements the human experts and guess what if I were to use AMD gpus instead of Nvda gpus it's going to slow the project down because I've got to fix all these stupid bugs and issues I've got to install a completely different software stack I've got to run the unit tests and check that i get the same numerical results the same inference speed and that takes weeks to months and that's going to delay the project and the client can't have that, my boss can't have that, so guess what we buy more NVDA gpus we suck it up because the cost's well worth it, we make 10x to 100x the cost the cost is nothing to us and so we pay for it and we get the work done. that's how i know i need to buy more nvda shares.
AMD don't even tell me about it I've used their gpus at work I'm an AI engineer i train these models for a living and AMD gpus AMD software RocM it's a terrible piece of shit it's a pain to work with. if i have a deadline to meet real world deadlines our clients need results from our AI models because they actually help their business these models lead to BILLIONS of dollars of profit or cost saving compared to human experts so they trust the AI models it complements the human experts and guess what if I were to use AMD gpus instead of Nvda gpus it's going to slow the project down because I've got to fix all these stupid bugs and issues I've got to install a completely different software stack I've got to run the unit tests and check that i get the same numerical results the same inference speed and that takes weeks to months and that's going to delay the project and the client can't have that, my boss can't have that, so guess what we buy more NVDA gpus we suck it up because the cost's well worth it, we make 10x to 100x the cost the cost is nothing to us and so we pay for it and we get the work done. that's how i know i need to buy more nvda shares.
i have the channel checks i have insiders in the industry i tell you what, even nvda people are surprised at the fucking demand for their gpus they're sold out all the way they can't meet the demand. even elon said it he says a lot of stuff all the time but he said they have to make their own inferior chips for the time being after buying those 100k H100s because nvda just can't meet that immense demand. they would love to buy even more nvda gpus but they just can't because nvda needs to sell to other customers and everybody wants to buy the gpus.
Disclaimer: Community is offered by Moomoo Technologies Inc. and is for educational purposes only.
Read more
Comment
Sign in to post a comment
MomentumPython1337 OP : amazon, googl, meta, appl they all have or are building their own AI accelerators but guess what. they still use nvda gpus despite all that because some stuff still works better on nvda gpus. with all those custom accelerators you need to do a lot of customizing to maximize the throughput and it takes time. so generally workloads most often the nvda gpus end up giving the best performance for the least time investment the only cost you pay is the price of the GPUs
MomentumPython1337 OP Quesera : happy to share :)
as always, don't buy more than what you can stomach and always manage your risk :)