To what end though? Create a superior product and not ultimately sell very many? Those 40+ margins are likely insufficient in this case. The problem being GPUs aren't inherently high margin!
It's still essential to their current processor model so I don't know why anyone would consider killing it off. Far more useful in the future, just not this moment.
The issue isn't about losing money, but not making as much.
That's part of it, but more that only the pro cards have margins currently warranting production. Market demands 1k CPUs and 1k GPUs and you can produce 1k chips. CPUs sell for 1k and GPUs for 500. Which chips do you make? That's the whole point of GPU performance currently being irrelevant for gamers. Even if the cards worked well, gamers couldn't find them for a reasonable or competitive prices without Pascal sucking big time.
GF isn't building fabs because AMD is going out of business. The issue isn't that Vega sucks, but that Ryzen/Threadripper/Epyc are kicking Intel's butt. Just think about that BoA analyst report with AMD going from 11% to 40-50% CPU market share. Unless GF was happily operating at less than 20% capacity there's a likely problem. The CPUs and GPUs use the same fab. Makes sense with GF expansion, WSA renegotiations, Mubadala raising cash, product releases, etc. A supply side constraint makes perfect sense. Until that eases, GPUs don't have sufficient margins.
Volume vs profit margins is what its all about. Its not so one sided or simple as you put it, there is much more to it then just that. This is economics 101, and if you ever taken higher level business courses, you will understand having marketshare leadership can increase your margins greatly, there are many reasons for this (not just monopolistic tactics, but B2M relationships). Something Intel and nV have been capitalizing on to improve their margins.
Why do you think 3DFx failed where AMD and nV succeeded?
Right now AMD is not in a position to still take on Intel, they have DECENT products that match up with Intel both winning some and loosing some it varies. (what you say is KICKING ASS, is equality in the market place, pretty one sided thinking if you can't see that) They are no where in the graphics card markets any of them and Gaming they are losing marketshare, where currently gaming is VOLUME sales. They loose gaming, they lose game dev support, then modeller dev support because why buy two different products when one can do it all? Then they lose semi custom just like what happened with wii. All of these are interrelated. After that, well they start loosing OS support (not support from an API side since they do that, but influence of API features in upcoming API's).
What you miss is nV created the markets besides the gaming market (even that they had a really big hand in it), they were the innovators and creators of those markets, and it took years to make those markets, it didn't happen over night. AMD is along for the ride when they don't have the funds to push into those markets and by the time they do have the funds (maybe 2 or 3 years down the road, if they keep up with Intel), by then its a much tougher road then it is now.
Can't forget history and the metrics these companies are built on and try to lay a new foundation based on whimsical explanations of one product makes more margins than another while having 1/10 or less of the volume, its actually much less than 1/10. This is the type of thinking management at 3Dfx failed, they lost their OEM's and lost their consumer base and tried to build a beast of a graphics card with multiple processors which on paper looked better than anything nV or AMD had, but that is failure when you lose your volume sales. Who are they going to sell it to? The hobo around the corner who doesn't have the software to run it? That is where Vega is right now for anything in HPC or DL markets, NO DAMN software. So how are they going to sell at higher margins?
Might look good in compute benchmarks, but benchmarks aren't software.
Did you get to the part of that video when JHH stated, the toughest times to keep nV going? It wasn't after making a good product or making the next product, its when they needed to create a new product for a new market. That is when risk outways potential. Because there is no defining those markets, once the markets are defined its easy to see which is the best way to go, but before that its all risk. AMD without the software components to drive those higher margin products is NO WHERE. They need to make their own market just like nV did with CUDA, of course they might get a helping hand with the other companies that are pushing with Open CL, RoCm, HSA, but right now they are not there yet.
Last edited: