Hi, this is Wayne again with a topic “Nvidia may be stuck with a TON of 40 series GPUs”.
Nvidia, how bad do you feel for them right now, not at all, i think it’s hilarious nvidia had ordered five nanometer wafers from tsmc before crypto winter hit. What that means is that, while uh nvidia and amd, let’s be clear, neither of them is some kind of white knight, no savior for gamers. Here, no they’re giant corpos at all, they will sell gpus to whoever is most profitable to sell gpus to it’s. That simple – and we saw it proven over the last couple of years, both of them, so it’s pretty amusing for us now to see that they forecasted their their five nanometer wafer production or their wafer allocation with tsmc, based on the kinds of volumes that they were Moving selling to miners and nvidia has already been fined for not properly disclosing what percentage of their sales were. The result of the gpu mining, craze and nvidia is really funny.
They’Ve gone, they flip-flopped on the issue. They’Ve gone from, we know exactly where every one of our chips goes to well, we have no, we have no way of knowing how many of them are being mined on versus game. Oh, you so do you so do miners do not install g-force now, for instance, so you would know if your sales triple and your g-force now instances go up by 10 that probably some of it went to mining. I would think right right, nvidia, you can do data analysis right and as much as i tried to get people to not install geforce now because of all the telemetry crap, i’m sure the vast majority of people do so that’s convenient. I like it anyway. The point is: nvidia um booked five nanometer wafers with tsmc before they realized that their sales were going to go [ Applause ], which means that the upcoming rtx 4000 series gpus could have um a very good supply, extremely good supply, potentially to the point where nvidia Was trying to back out of a pretty significant amount of the wafers as far as i can tell, or else i don’t think this would be in the news they they tsmc. Just straight up was like nope. They are apparently unwilling to make any major concessions.
They have agreed – oh, i i think this. This has got to be allegedly i’m not actually sure if this is confirmed, so i’m just going to say allegedly, because i’m not double checking it, but they’ve apparently agreed to delay first shipments by one quarter or even to q1 2023. But if nvidia wants to bail on any of these orders, they will apparently be responsible for finding replacement customers.
Our discussion question. This is funny our discussion question is: who could nvidia find to take their orders? Well, let’s see there’s intel’s gpu, which, as far as we can tell so far, is an absolute dumpster. Fire intel isn’t going to need any more of those um all right wait is that even no is that even on tsm c5 nanometer, i don’t even think so.
Tsmc5 nanometer, i mean there’s the upcoming iphone, but isn’t apple, actually forecasting a reduction in demand as well, so apple probably has as much if not more allocation than they need. I know samsung is sitting on apparently piles of smartphones like way more than they usually would be because of the uh weakening of demand heading into what appears to be a potential global economic recession. Um yeah, i don’t know i’ll wait. I i don’t know i don’t think anyone’s really going to be clamoring at the bit to bail them out. Yeah, okay, uh humbled to f over on twitch, says automakers, that’s a good question, but actually no that’s exactly it. Automakers, don’t need that kind of performance, so they don’t operate on the cutting edge node.
They don’t even operate a node or a couple of nodes back unless their name is maybe tesla is on. Oh, it’s going to be modern. There might be some like really expensive. Automated driving cars yeah, but that was a whole that was a whole drama around the whole automotive chip shortage.
Where chip foundries were telling the automakers hey. You guys got to update your crap, because we can’t we can’t keep these ancient manufacturing processes running like this is actually not feasible. We can’t ramp up more of this garbage. You guys need to redo your designs on modern nodes so that we can take advantage of the economies of scale at these modern nodes that everything else is freaking running on, and i mean, of course, the smaller transistors and smaller chip sizes and better power efficiency. Like guys get with the program, so no they would be, they would be the last ones to step in and and take over five nanometer production. Um says drone manufacturers. No, not even not even five nanometer they don’t. They don’t need it.
Um yeah yeah new nintendo switch asks whiffs. I mean that would probably be awesome. A nintendo switch with five nanometer performance and power efficiency. I’M super down.
Nintendo enjoys releasing on last generations, uh last generations. A low performance yeah, fair enough hardware yeah. I don’t think so. Flying crocodile says, amd yeah, but you got ta understand amd, probably has the same overbooking issue the exact same problem, whatever is being reported about. Nvidia almost certainly applies to amd as well like there’s yeah, cheap chips for gamers someone’s saying data centers, but data centers would buy completely x, yeah yeah yeah they’re going to buy the gpus they’re going to buy whatever they’re not going to buy the the wafers, because These are these are they’re trying to sell wafers they’re, not selling like they’re selling, fab capacity right yeah, that’s really expensive, yeah yeah, it’s rough! Oh, that is rough. .