Google just shook the entire Silicon chip market, Silicon chip manufacturers from SK Hynix to Samsung all saw a sharp drop in their stock prices. People saw a 30% drop in RAM prices across different retailers. Amazon shoppers found that some Corsair Vengeance DDR5 RAM is now slightly more affordable. Google’s TurboQuant algorithm may allow AI data centers to use fewer modules, increasing supply for gamers!!!! Or there is another side to it? Let’s discuss.
Since the AI revolution, high memory costs have wreaked havoc for console and PC gamers. The prices of building PC has shot up significantly cause modern title especially AAA and highly competitive games, require a lot of RAM, not just PC builders are struggling, but console buyers are struggling too, the prices console have also shot up. Read this blog from PlayStation itself.
But this is the first time since the jump in RAM price hype we’ve seen a drop in prices. Google’s TurboQuant AI compression algorithm could be a good news for consumers.
Notebookcheck recently reported cheaper memory prices across Europe. Even if it is mainly limited to a single manufacturer.
One 32GB RGB version of Corsair Vengeance DDR% RAM is not $379.99, down from it’s peak $439.99.
Unfortunately, most DRAM’s remain unaffordable for most buyers, as the memory shortage continues and is showing no sign of relief other than the new Google’s new algorithm, though in long term this announcement has provided some relief but there is another side of the story too.
Impact of TurboQuant on memory prices
TurboQuant addresses the growing key-value(kV) cache memory demand from AI models. Every time we talk to a chat bot, it creates a memory cache of that chat, as the chat grows, the memory grows too. TurboQuant uses a specialized compression technique which shrink memory cache up to 6 times. In theory, complex models would need less DRAM and HBM, easing memory shortage. Stocks of major suppliers like SK Hynix and Samsung dropped after investors heard the news.
But still, TurboQuant may not reduce the prices of memory for consumers. Emboldened data centers could easily begin relying on even larger AI models that require additional modules. As a result, Manufacturers would just continue to prioritize enterprise clients over gamers.
This can also be explained with a phenomenon that history as proven many times, called the classic efficiency paradox. The more efficient something is, the more people tend to use it. Enterprises will just improve their AI models was earlier limited because of memory, and build even bigger AI models.
Without any effective solution to this burst in the AI revolution, it is unlikely that DRAM prices will fall anytime soon. Along with DRAM GPU prices have also gone up, becoming even a bigger challenge for gamers and PC builders. And the same with Console players too, the price hikes in consoles seem permanent as of now.
