Which is why I say that Nvdia is (or became long ago) a vertically integrated software company (like apple)
Whereas AMD is still a hardware company.
What Nvdia (& Apple) do will work only for them and not for others
Nvidia was amongst the first to go for unified shaders and Taiwanese/TSMC outsourced manufacturing. They've been able to distort the market enough to make it consume their supply.
I don't know their history enough, but clearly they are a-ok with high volume/high losses while the penny pinchers across the street are afraid of losing a single mm².
Nvidia's perfectly fine with wasting inordinate amounts of money to get partnerships and/or software stacks that pay off, or as you said, vertically integrate.
AMD sees itself as a hardware vendor that wants to respond to market demands. But Nvidia CREATES market demand because they do so much more than just wait for others to request work towards something.
So AMD's always saving every penny and being several steps behind while Nvidia is ok with wasting money to be market and technology leader.
Which is perfectly fine IF you want to be the eternal N°1 and N°2, but isn't going to work if you want to actually overtake the other one...
Why would they risk decreasing their wafer yield for parts that most gamers consider the budget option?
Really REALLY shouldn't have provoked me on this.
That isn't how that works at all LOL. I bought a 7900 XT, I guarantee you I didn't consider the 1000€ bill "budget". I guarantee that any price above $300 isn't budget. AMD doesn't have special commercial rules that don't apply to Nvidia.
And what do you mean "decrease their yields" lol?
You're speaking as if AMD had a necessity to maximise yield while Nvidia somehow would have a Glove of Free Yields like Jensen is a JRPG protagonist.
They're both in the same boat, they even buy at the same Fab lel.
Jensen, the Hero of Raytracia
Actually since both of them have the same yield issues, riddle me this:
- Nvidia makes the 4090, a 144 SM (24 deactivated) 600mm²+ die. All monolithic. All wasting ~10% of their area. Those that had 140 SMs functional? They have 120. Those that had 130? 120. 120? 120. Nvidia wastes area like it's free.
- AMD makes the 7900 XTX, a 96 CU, fully activated, ~500mm² die. It's chipletized. The GCD itself is only 300mm², rest is 6 24mm² MCDs. If the die is imperfect, it gets SKU'd down into a 7900 XT, so you minimise your losses to only dies that have less than 84 CUs functional.
So since the yields are much better on 300mm² than 600mm², and the 300mm² can be maxed or cut down, tell me, which one is obsessed with margins?
I'm not talking "compare final selling price of a 4090 to a XTX and calculate the better margin". Frankly even if you had a full BOM cost of an XTX straight from an AMD paper and the same from Nvidia for the 4090, you'd probably find AMD to have quite better margins but not miraculously better.
I'm talking about comparing the value which AMD places on margins, yield, cost saving, versus the value that Nvidia places on it.
AMD is not "lowering costs". They're "maximising margins by lowering costs". Nvidia puts their money towards next versions of DLSS, Raytracing, cool new things? AMD puts its money towards researching how to penny pinch every mm². And if you're doing a very budget oriented strategy, I got nothing against it. But that is NOT what AMD supporters claim they're doing, on the contrary. They claim that AMD is "doing that so they can compete and beat Nvidia"...except they keep outputting borked stuff, or low risk, low expenditure stuff.
RDNA 3's a failure?
"Oh, respins are TOO EXPENSIVE, fixing it was just impossible, and yes we still lied bold faced about the perf and power draw".
RDNA 3 is a failure, and doesn't get fixed?
"Well, RDNA 4 planned for everything to be chiplets chiplets chiplets to save AMD's precious money, and guess what, it's also all borked for some reason. So instead, you'll get the one monolithic chip that we have. It's 240mm²."
Versus Nvidia calmly fabbing 600mm² monolithic dies.
We know that Raja Koduri screwed them over with his ambitions and then did the same thing to Intel's GPU effort too by overhyping and over-promising. Maybe he's a double agent working for Nvidia. Screwed over the only two viable competitors.
There's several dimensions between Ponte Vecchio and ending up with a 240mm² die as "the best this gen". I find it frankly insulting to my intelligence that AMD supporters defend this as even remotely acceptable. The level of effort here is nothing. For the first time that I can think of, a "new generation" is so lazy that it won't even attain the raw raster power of the last generation's top die. WHILE THAT GENERATION WAS BORKED IN THE FIRST PLACE, HOLY HECK.
"But the die is so smol and all because people want budget things and they didn't have a choice and poor wittle Radeon deserves our love and care and patience and" no. Shut.
The die is small because Nvidia's ok with outputting a 600mm² giant fat hog of a die. Even if they lose area. Even if it's silly expensive.
AMD's not. So they pour all this effort in magical omega chipletized 144WGP monsters, the monsters all get borked like Frankenstein's monsters and then a manager goes "well, let's just completely give up and go for selling only our tiny monolithic die".
And no, it's not "so utterly borked that it's unsellable". RDNA 3 was sold by AMD, you can trust that RDNA 4 could've been sold above the ridiculous mini die. And you can trust that they cancelled anything above that die because the margins wouldn't have been satisfactory enough for Lisa's lofty penny pinching. It's always don't lose yields, don't take risks, SAVE THE PENNIIIIIIIIIIIIIIIIES...
It's always borked things, cancelled things, or cheap things. All the while asking for "understanding their circumstances". All the while we hear more promises of "next gen it'll finally be great".
All the while Nvidia will just make another super expensive 600mm² and the Dragon of Yields eats another spear in the belly.
If we want AMD to start taking risks, first we need to start taking risks ourselves by plopping down full price for their GPUs (good or bad). I still have an unopened ASUS 6800 XT LC card (coasting by on Gigabyte RX 6800). I believe DrMrLordX also has an unopened 7900 XTX. We need more of us to support AMD.
Hello, AMD supporter here, bought the 7900 XT, believed in the marketing when they said "50% better performance, 50% better power efficiency", and when the reviews came out and all the shills said "they have a temporary problem, the drivers will fix this", I believed them too. Less so every month, and ultimately I bought it knowing it was not that good. But it was still 1000€ all the same. It's not budget. It's not something I bought thinking "I'm saving money buying AMD". I bought because it was the best thing to buy at the time. Period.
I am very much supporting Radeon. I am however getting extremely annoyed at the absurd position that their vocal supporters have:
- AMD is our friend that needs our support to fight Nvidia, but AMD is a business that needs to maximise margins
- AMD tries to make great stuff for cheap, but AMD is a business that needs to maximise margins, so expect the production to be cheap, not the final price
- AMD is all about open source and the public interest, but AMD is a business that needs to maximise margins by lowering costs, so they invest a ton less in software than Nvidia
- AMD is "helping fight Nvidia", but AMD is only going to bother putting out a 240mm² die this gen, anything else would be too expensive for their penny pinching goals
- AMD is "trying to lower costs for the public to get good GPUs for cheaper", and will price it as high as Nvidia's competition will allow them to because MUH MARGINS
I have had way, WAAAAY enough of that narrative. I'm extremely sick of it.
You can't be a semi-charity and "this is a business not a charity". You can't be "for the public" and "margins are what we're about". You can't be about lower costs and sell at the highest possible. And more than anything, you can't be the company that "deserves our support" while giving up or derailing everything. RDNA 3 borked? RDNA 4 cut down all the way to a 5700 xt/6600 xt size die? I mean what is that? THAT's what we gotta support so they can "fight Nvidia"?
I am so absolutely sick of these excuses. Either they want to penny pinch so much because it's about lowering the general costs and getting a "good enough" product out for good enough margins(not viable in this industry), either they
get their fat margins in and try to get a real, costly, risky high end product out. But you can't have the cake and eat it too. Right now what AMD's been doing since the RDNA 3 failure is not fix anything, not respin, not have a high end product.
And do not dare throw a "well, they couldn't", YES THEY CAN. They supposedly fixed the RDNA 3 power draw problem, so put out a danged 7950 XTX with 20% more clocks. Plan for a die, even if it's monolithic and more expensive, that's not a 240 ridiculous mm² low-end thing. Do not output a failed gen only to follow with a gimped gen! I'd not have said a thing if the top die was just 300-350 mm², but 240 is straight up not even trying.
Remember Thermi? Jensen threw an entire generation with an arch he knew was terrible and yields that were atrocious. Did that stop him? No! Because he takes risks and PUTS HIS MONEY WHERE HIS MOUTH IS!
AMD with TSMC's backing can't even get at least a die half the size of what Nvidia will output?!!! And everything above 240mm² was supposedly "so borked that it wasn't even worth putting on the market"???? F off with that story for babies, they couldn't get their ultra-optimised margins so they just gave up. Just put out a tiny die, say that it's about "helping the budget buyers", price it as high as it'll go, and your supporters will Defend your Honour from any criticisms.
AMD has to take its money and margins, and do something with them, or stop being so margin obsessed.
But the AMD supporters have to stop doing THIS:
Beggar in the lab, gouger in the shop. Failure after failure, disappointment after disappointment, 10 years of the same supporter narrative.