Your crypto-mining chips aren’t good for AIHere’s why

[ad_1]

The lucrative world of proof-of-work altcoin mining was a sight to behold, rewarding miners with eye-popping profits…until Ethereum made a detour to proof-of-stake. But one emerging industry is catching the attention of those who were left with hundreds of idle graphics cards after the merger: AI.

This new darling of technology, spurred by the creation of ChatGPT, requires the computation of large amounts of data, similar to Ethereum mining, and the best way to do such heavy work is with GPU slots.

However, it may not be the panacea many were hoping for, as GPUs, the trusty workhorses of cryptocurrency mining, aren’t exactly the MVPs in AI training. . For what? Let’s distil this complex question to its essence.

The story of two needs

Here’s the crux: crypto mining and AI training have divergent appetites for their perfect GPUs. It’s like dating two people with completely different tastes. Proof-of-work cryptocurrencies have a soft spot for cards with tons of hashing power, but care no less about vRAM (Video Random Access Memory). In contrast, the AI ​​is a sucker for cards with abundant vRAM and gives hashing power a cold shoulder.

Think of hashing power as the muscle of the operation, it all depends on how many calculations your GPU can perform per second. The higher the hash rate, the better your chances of unlocking that elusive block and enjoying crypto fortunes. It’s like a speed dating scenario: the more people you meet (calculations you do), the more likely you are to discover a match (mining a block).

Conversely, vRAM is the ability to simultaneously manage and store vast amounts of data. It’s the difference between juggling two balls or twenty. AI training is a notorious data guzzler, requiring GPUs to simultaneously manage and process colossal amounts of data. Using the dating example, going the vRAM route would be like inviting a crowd into a room and simultaneously asking the same questions (data points) to find the perfect match (AI model).

The story continues

And therein lies the paradox. Your hash-powered GPU, perfect for Ethereum mining, is at its wit’s end when it comes to training AI. It’s like asking a world-class sprinter to run a marathon, it’s just not his game.

A good backup plan?

Now, putting the oversimplifications aside, crypto mining does indeed require high hash power, but it doesn’t rule out vRAM entirely. Crypto mining uses vRAM as part of its mining algorithm, although vRAM requirements are generally lower than those needed for AI training. For example, a GPU proficient in crypto mining might have 4GB of vRAM, which isn’t even enough to run (even think about training) an AI image generator like Stable Diffusion properly.

This 23-year-old Snapchat star can be your AI girlfriend for a price

To put things into context, one of the best GPUs for ETH mining was the Nvidia RTX 3060Ti with 60 MH/s of hash power and 8 GB of vRAM. In contrast, OpenAI used Nvidias A100 and V100 models: a V100 comes with 32 GB of vRAM and the A100 models manage up to 80 GB.

But that doesn’t mean the GPUs used for crypto mining are fully redundant for AI training. GPUs with relatively low vRAM can still be used to train smaller AI models or tasks that don’t require significant vRAM.

There are many AI operations and models that work quickly and efficiently on old boards, Scott Norris, CEO and Founder of Optimizer, told Decrypt, adding that entrepreneurs should properly study which AI application they are on. concentrate. GPUs used for crypto mining can work in AI farms with custom models or slightly tweaked models, he said.

Norris mentioned that some crypto mining companies have already moved to AI operations. One example is Omega AI. Hive Blockchain and Hut8 Mining are also testing their luck.

So while this solution may not be as lucrative as your good old crypto mining, it could serve as a viable plan if your proof-of-work alternative altcoin mining plan B looks bleak.

Want to learn more about AI? Check out our latest Decrypt U course, Getting Started with AI. It covers everything from AI history to machine learning, ChatGPT, ChainGPT, and more. Learn more here.

[ad_2]