No one who does AI seriously gets AMD. AI market is extremely dominated by Nvidia.
No one who does AI seriously gets AMD. AI market is extremely dominated by Nvidia.
Your lifetime is nearly 80 years. Companies lasting 80 years is ultra rare in history, large behemoths included. I bet you can already name several behemoth IT companies that’s already come and gone.
I wouldn’t trust even larger behemoths like google and MSFT to last another 80 yrs. It’s just too statistically unlikely.
I don’t know if steam does this since I have no experience selling on steam, but generally when you sell anything anywhere the sales channels will often demand that you give them the lowest retail price. Most commonly done by ones that give the most exposure since they have that much more power. Failure to do so will result in some penalty (Amazon prevents your offer from being in buy box) or just outright refusal to take your product (such as Walmart).
Additionally, customers complain too when you sell at two different pricing elsewhere. If you’re a company that gives virtually no support (like you sell pickles or whatever), you prob don’t care. But for things like games, you’ll get bombarded with demands that they got ripped off by buying from one place and ask for difference in pricing or submit a refund request. Refunds are more expensive to sellers than not selling at all since you still have to pay transaction/refund fees by payment processors. Or if physical product, cost of shipping as well.
Different sales channels having different pricing isn’t really an option. It’s not really worth it. You’ll get problems left and right.
I don’t really care for size. What’s cheap and fast?
At that point, it probably starts costing steam more money in support agents.
I get your point… But I feel like people in this thread doesn’t know how cake making works…
20 people will make a single cake faster. Not 20x faster, but faster. There are multiple part of the work that can be divided out to different people. Like you can have one person make batter while other makes icing. Fancy cakes actually do take multiple people to make simultaneously.
There are many different niches of ML. 99% of hobbyist would use consumer grade hardware. It’s quite frankly more than good enough.
Even in commercial usage, consumer GPUs provide better value unless you need to do something that very specifically require a huge vram pool. Like connecting multiple A100 GPUs to have hundreds or tens of thousands of gigabyte vram. Those use cases only come up if you’re making base models for general purpose.
If you’re using it for single person use case, something like 4090 is actually the best hardware. Enough ram to run almost anything and it’s higher clock speed than enterprise GPU means your results come back faster.
Even training doesn’t require that much vram. Chat models are generally more vram heavy but if you’re doing specific image training like stable diffusion for how to render your face, or some specific fetish porn, you only really need like 12GB of vram to do it. There are ways to even do it at lower like 8GB but 12 is sweet value spot where even 3060 or 4060ti can do. Consumer GPUs will get that trained in like 30min to 24hrs depending on settings and model.
I relate with the The Inner Light episode so much it hurts. I had a phase where I had dreams where I lived an entire life from childhood, getting married, having kids, to playing with my grandkids and eventually facing my death of old age. And then I’d wake up completely drenched in sweat from exhaustion. Knowing that none of these people I deeply cared about and loved are real; that they’re mere figment of my imagination completely devastated me. Though I didn’t suddenly learn to play a flute, I can’t help but feel that each of these dreams permanently changed me in some way. Even decades after having these dreams, I still feel and it still affects me. I salute to the writer of this episode who probably has had a similar experience as I did.
This is my first time making use of crosspost feature!
I would assume that they are saying in a bigger scope and just happen to divide down to a ratio of 1 to 32.
Like rendering in 480p (307k pixels) and then generating 4k (8.3M pixels). Which results in like 1:27, sorta close enough to what he’s saying. The AI upscale like dlss and fsr are doing just that at less extreme upscale.