
Originally Posted by
Banggugyangu
I'm fully aware that nuts and bolts are a form of hardware... I never contested that. I simply stated that the list you gave as examples had nothing to do with anything from IT whatsoever. As I mentioned, your list was a great list of parts used in electrical engineering. My point was this: if you were in charge of procurement for IT equipment and were trying to list off parts to show your credibility for the topic at hand, you'd have said things such as hard drives, RAID controllers, SANs, heck, even cabling... You gave a single item that's used in IT, but it's such a generic item that it's used in practically every industry. Your list didn't solidify your credibility on the matter, it tarnished it.
To break things down a bit, the semiconductor shortage moniker is a little misleading. It's coprocessors, specifically. Between the pandemic and the tsunamis hitting the manufacturing facilities, coprocessor manufacturers have gotten massively behind. These coprocessors are being shipped to OEMs and retailers. The OEMs get a large majority of them, but OEM products are rarely applicable in something the scale of what we're dealing with, here. Typically companies like SE will have a vendor that custom builds their servers for these types of applications. These vendors have to procure their own products (CPUs) for the builds. They're not typically part of the initial shipments like large scale OEMs are, so they turn to retail partnerships. Those retail partnerships sell out. The vendors are able to acquire a few, sure, but due to the shortage, scalpers are picking them up in droves. These scalpers are charging upwards of 10x per processor. Your original point of "There are other ways to get the parts" would involve purchasing from a scalper. So a CPU that would normally cost $5000 is now going to cost upwards of $50000. Take into consideration that there will be 2-4 of these in each server, and about 30 servers total to resolve the problem. (The ultimate solution is 2-3 more world servers per datacenter, a 3rd data center for EU, and preferably at least 1 more login server per data center.) You'd be looking at almost $1.5m in CPUs alone. The other option? wait. In a couple weeks, this problem will mostly go away. (The login queues, themselves). During the time afforded by the immediate problem resolving itself, SE's vendors will likely be able to secure the necessary parts to fulfill at least part of the needs. Even if SE were to go and purchase the necessary equipment from the scalpers, how long do you think it would take them to get new servers into production? We're looking at a week minimum, but more likely 2-3 weeks even if they're working tirelessly to do it. By then, this issue will have dissipated anyway. There is literally no incentive for SE to spend all of that extra money on something that won't even benefit the situation at hand.