Nvidia Reportedly Looking for Alternative Foundries for 28nm

[deksman]

[citation][nom]fuzzion[/nom]Where are my quantum computers?[/citation]

They are out there... just getting them to the people will take absurd amount of time because capitalism needs to squeeze out every last penny by revising outdated technologies constantly that are being used by consumers.

We could have had diamond based computers decades ago (first man made diamond being made in 1950 - I'd say 10 or 20 years would have been enough time to incorporate the process for computer use).
This wasn't done due to DeBeers holding the diamond market in its fist and keeping the costs way up.

Graphene... could have found its way into electronics 7 years ago (was presented to the public roughly 9 or 10 years ago).
Wasn't done on the excuse the process is 'expensive'. That's garbage talk. From a technological/resource/manpower point of view, we can do it cheap, with quality and fast.
Artificially induced 'costs' are mostly the things that are keeping numerous technologies out for a long time until they finally decide to implement them.

 

[viridiancrystal]

don't know about anyone else, but there seem to be plenty of AMD gpu's in stock right now. NVidia has always had yield problems, now its finally going to hurt them. I really don't see TSMC at much fault here.

 

[ojas]

[citation][nom]Shin-san[/nom]I also heard that nVidia might have contacted Intel.[/citation]
lol and Intel said "go away"

 

[Desert Eagle]

If Nvidia is having problems just getting the GK104 yields up then I imagine they will really have trouble getting the GK110 out in sufficient quantities. People can be impatient and I suspect they are losing quite a few sales to AMD already. Sucks for me because I want the GK110. Maybe I'll just camp out on NewEgg until I can buy a EVGA GTX 680.

 

[tmk221]

[citation][nom]jkflipflop98[/nom]Hi Nvidia. I work for Intel. Let's talk.[/citation]

Never gonna happen. Please tell me why Intel would help major player in ARM business while it's attempting to enter that market with it's own x86 solution? You may answer, well for money? But no. Intel has plenty of spare money, an has plenty of diversified incomes. Intel wants be as competitive as possible. And that's why tthey invest so much money in their own fabs.

 

[JAYDEEJOHN]

nVidia has alot of contracts to fulfill, they need chips, only so much supply, too much demand
Thats reality for you

 

[shin0bi272]

This is actually a good thing. Putting all of your eggs in one basket when it comes to your fabrication process is just asking for delays and price increases. Of course the other solution is to invest in your own fab plant. Sure its expensive but when you need more chips faster its definitely an option to think about. I mean hell nvidia's been around since 93 and they dont make their own chips yet? It might be time to think about investing in their own future.

 

[blazorthon]

[citation][nom]Shin-san[/nom]I now know that Samsung has fabs. I thought every company that wasn't Intel used a third party, or a fab agreement of some sort, usually TSMC or GlobalFoundries.I also heard that nVidia might have contacted Intel.[/citation]

Nvidia either contacted or implied that they wanted Intel's help, but Intel said GTFO noob. It's not going to happen.

Really, why is it that although Nvidia is having such problems, AMD is doing just fine? I mean, sure, AMD didn't do too well when the 7970 first came out, but even then, it wasn't THIS bad and it recovered after a while. Nvidia isn't even using large dies like the GF100 or GF110 (GK100 was supposed to be even bigger than them). In fact, Nvidia's GK104 is smaller than AMD's Tahiti, so why is there such a shortage if TSMC is able to fulfill AMD's demand?

 

[blazorthon]

[citation][nom]shin0bi272[/nom]This is actually a good thing. Putting all of your eggs in one basket when it comes to your fabrication process is just asking for delays and price increases. Of course the other solution is to invest in your own fab plant. Sure its expensive but when you need more chips faster its definitely an option to think about. I mean hell nvidia's been around since 93 and they dont make their own chips yet? It might be time to think about investing in their own future.[/citation]

That would be hard for Nvidia. As things stand now, a fab might cost more than Nvidia is worth.

 

[spookyman]

I guess I will have to stick with buying a 7970 then.

Its nice to know that Nvidia decided to drop production of their GTX 580's.

 

[Blessedman]

Nvidia needs to be bought by Intel. They gain a huge portfolio of IP's and nVidia wins with having the best possible foundries in the world.

 

[madooo12]

[citation][nom]spasmolytic46[/nom]I wonder if AMD still thinks selling the fab was smart?[/citation]
if they hadn't they would've been bankcrupt, besides they woul've been stuck at 40-45nm, they don't have enough money to research smaller nodes that fast

 

[shin0bi272]

[citation][nom]blazorthon[/nom]That would be hard for Nvidia. As things stand now, a fab might cost more than Nvidia is worth.[/citation]
True but you gotta start some time and you cant go back in time to when they were undercutting 3dfx and start one. So its now or some time in the future when they are worth even less and they look back and go damn we should have started 10 years ago.

 

[hannibal]

AMD is better situation because their CPU's are made partly by global foudaries. Nvidia is forsed to use TSMC to produce both GPUs and their mobile CPUs and now they are using it mainly to produce mobile CPUs. If we assume? that both have egual deal with TSMC and Nvidia is using 80% of their capasity to produce mobile prosessors and AMD is using 80 % to produce GPUs then it is easy to see why there are more AMD brand GPUs around. Ofcource these numbers are totally drawn from the air, but we know that mobile parts is area where Nvidia has put a lot of effort lately!

 

[blazorthon]

[citation][nom]Blessedman[/nom]Nvidia needs to be bought by Intel. They gain a huge portfolio of IP's and nVidia wins with having the best possible foundries in the world.[/citation]

Intel does not want to enter the consumer graphics market. It looked like Nvidia wanted Intel to buy them for a while (they might still want that), but Intel wants nothing to do with them. Intel might also just hate Nvidia because Nvidia managed to be even more of an asshole to AMD than Intel did. Intel also probably doesn't want to be competing with AMD in two different markets.

In order to avoid anti-trust laws, Intel needs AMD to stay in the CPU market. AMD is not doing too well as of right now (and was doing worse last year), so Intel seems to be slowing down a little. However, Intel needs AMD to stay alive even if AMD's lifeline is in another market, so how well AMD does compared to Nvidia is paramount to Intel. If Intel buys Nvidia, then they need to put a lot of work into graphics because if Intel loses to AMD, then Intel is losing their name as being better than AMD right now. If Intel swamps AMD, then Intel would need to screw around until AMD catches up and could catch flack for that.

Leaving these problems to another company are in Intel's best interests. With Nvidia competing with AMD in the graphics market, AMD has a decent competitor that really isn't beating AMD and hasn't been for some time. That way, Intel is free to be as lax or aggressive as they need to be in the CPU market. Otherwise, Intel would need to compete in two market, both of which have AMD, who is a fairly variable competitor who sometimes doesn't do well, but some other times does incredibly well. Intel would need to be able to counteract AMD's every move in order to stay ahead, all while not beating AMD too badly that they lose their only competitor in both the CPU and graphics market. If AMD died in both markets for any reason, then Intel would have anti-trust lawsuits in both the graphics and CPU industries to deal with. Intel would not risk that, especially with how stable their position in the CPU market is.

Intel can fairly easily do that in one market, but doing it in two markets is exponentially more complex than a single market, especially with how well AMD does in graphics.

 

[blazorthon]

[citation][nom]shin0bi272[/nom]True but you gotta start some time and you cant go back in time to when they were undercutting 3dfx and start one. So its now or some time in the future when they are worth even less and they look back and go damn we should have started 10 years ago.[/citation]

I think that 2012 will be a time of growth for Nvidia because Nvidia is finally using small, much cheaper to produce dies. That Nvidia is also skimping on the memory sub-system for their graphics cards will only decrease the cost to manufacture them, increasing profits. As it stands, manufacturing the 680 probably costs less than half of the costs of manufacturing the 580.

Building a fab now would put Nvidia in the red for years, at the least. Then Nvidia would need to worry about keeping their fabs up to date with the newest fab tech and that would increase costs more. This works if you already have fabs or can afford them, but Nvidia doesn't seem to be in any position to try that right now. If Nvidia does grow as I think it will and continue growing for a few years after 2012, then maybe Nvidia will be in a position to afford such a venture.

 

[A Bad Day]

[citation][nom]Shin-san[/nom]I now know that Samsung has fabs. I thought every company that wasn't Intel used a third party, or a fab agreement of some sort, usually TSMC or GlobalFoundries.I also heard that nVidia might have contacted Intel.[/citation]

I don't think Intel would be willing to manufacture stuff for Nividia.

 

[shin0bi272]

[citation][nom]blazorthon[/nom]I think that 2012 will be a time of growth for Nvidia because Nvidia is finally using small, much cheaper to produce dies. That Nvidia is also skimping on the memory sub-system for their graphics cards will only decrease the cost to manufacture them, increasing profits. As it stands, manufacturing the 680 probably costs less than half of the costs of manufacturing the 580.Building a fab now would put Nvidia in the red for years, at the least. Then Nvidia would need to worry about keeping their fabs up to date with the newest fab tech and that would increase costs more. This works if you already have fabs or can afford them, but Nvidia doesn't seem to be in any position to try that right now. If Nvidia does grow as I think it will and continue growing for a few years after 2012, then maybe Nvidia will be in a position to afford such a venture.[/citation]

They will have to triple in total worth if you expect them to stay in the black while they build a fab plant. So I guess they'll be building their first fab in 40 more years. But hey god forbid anyone voice their opinion on a website without getting flamed for it right? yeah. whatever.

 

[blazorthon]

[citation][nom]shin0bi272[/nom]They will have to triple in total worth if you expect them to stay in the black while they build a fab plant. So I guess they'll be building their first fab in 40 more years. But hey god forbid anyone voice their opinion on a website without getting flamed for it right? yeah. whatever.[/citation]

Hey, voice an opinion that stands up to logical scrutiny, then it doesn't get flamed. Besides, I simply stated some reasoning for why it was wrong, I didn't insult or flame YOU in any way. But hey, God forbid someone not complaining about something being explained as to why they were wrong. I guess we can't all be right.

Also, Nvidia is worth something like $9B, so they don't need to triple to get a fab. They could get by with a lot less than tripling.

 

[shin0bi272]

[citation][nom]blazorthon[/nom]Hey, voice an opinion that stands up to logical scrutiny, then it doesn't get flamed. Besides, I simply stated some reasoning for why it was wrong, I didn't insult or flame YOU in any way. But hey, God forbid someone not complaining about something being explained as to why they were wrong. I guess we can't all be right.Also, Nvidia is worth something like $9B, so they don't need to triple to get a fab. They could get by with a lot less than tripling.[/citation]
3.5billion total assets according to their wikipedia page. And yes they can build a smaller fab plant. Intel is building a 5 billion dollar one in AZ but they need a lot more capacity since they make a lot more cpus than nvidia makes gpus. Hence why I said they should start now. They could build a 1 billion dollar plant with a loan (since their operating capital is under 100mil) and yes put themselves in the red for a while but that plant would also allow them to make more gpus and make them cheaper which will also allow them to be sold cheaper and that puts them in a better position to compete with amd when it comes to price/performance. I think I saw a blurb on here or anandtech that amd was about to drop the prices of the 7900 series now that nvidia's 680 is out... which isnt surprising. Now do you see why I think they should go ahead and build one now instead of getting delays and lower than expected yields from china?

 

[blazorthon]

3.5billion total assets according to their wikipedia page. And yes they can build a smaller fab plant. Intel is building a 5 billion dollar one in AZ but they need a lot more capacity since they make a lot more cpus than nvidia makes gpus. Hence why I said they should start now. They could build a 1 billion dollar plant with a loan (since their operating capital is under 100mil) and yes put themselves in the red for a while but that plant would also allow them to make more gpus and make them cheaper which will also allow them to be sold cheaper and that puts them in a better position to compete with amd when it comes to price/performance. I think I saw a blurb on here or anandtech that amd was about to drop the prices of the 7900 series now that nvidia's 680 is out... which isnt surprising. Now do you see why I think they should go ahead and build one now instead of getting delays and lower than expected yields from china?

I see your point, but it's more than just the costs of building the fab that need to be taken into account. Nvidia would also need to get the technology down and that's not easy. For example, look at Globo Foundries... Oh sure, they have fabs and enough of them, but Globo has poor quality foundries. If they are having problems and they've been around for a while now (previously part of AMD, but still around nonetheless), how hard will it be for Nvidia? Maybe Nvidia could buyout a small company who has some fabs or take similar action, but they can't just build their own without getting people experienced in this and assets (such as the tech or buying a preexisting fab) that they simply don't have yet. Could they do it? Maybe. However, it would take years before Nvidia actually was able to make good use out of it and until then, it would be costing Nvidia a lot of money.

Perhaps Nvidia could get started now, but it's a very large undertaking. That is why most semi-conductor companies don't have their own fabs. Besides, if AMD is doing so well with TSMC's 28nm fabs, why is Nvidia having problems? Nvidia is probably just stockpiling a supply and only releasing a few of them at a time for now. There also seems to be greater supply of GTX 680s outside of the USA, especially in Asia and Europe...

Also, my $9B number is Nvidia's market cap as listed by Yahoo Finance.