How/why did Amd/ATI fall behind Nvidia?

[SirPauly]

Imho,

The APU was AMD's future but it may be the reason for its demise. It was the product that was going to differentiate and lead the way but sadly hasn't come to fruition. The focus and resources spent here without garnering substantial rewards has hurt the company.

Simply put: I believe it is R&D expenditures spread too thin to compete with nVidia and Intel, especially with the APU struggling in the market place. Amd banked on the APU, invested heavily for it, this is why they are struggling today.

 

[ShintaiDK]

It didnt help them either that Clarkdale launched over a year before their first APU. And it took 5 years after they bought ATI to make one.

 

[Keysplayr]

You think Adreno is the reason Qualcomm were so successful? Haha, no. It's all about the modem tech. AMD had no modem tech, and no money to invest in creating modem tech. (NVidia went out and bought Icera, and they still couldn't compete.)

Are we to understand then, that you are of the opinion that AMD made a good decision in selling off their mobile IP?

 

[turtile]

AMD doesn't have the cash to create a new architecture on 28nm and then 14/16nm. That's whey they are building on GCN and adding a new memory configuration for the high end.

They should have new one out by the end of next year if they don't get behind...

 

[ShintaiDK]

They should have new one out by the end of next year if they don't get behind...

If we are to believe AMD from their analyst day. It will just be optimizations with shrinks.

 

[Arachnotronic]

Are we to understand then, that you are of the opinion that AMD made a good decision in selling off their mobile IP?

It wouldn't have made a difference in the scheme of things. Qualcomm was clearly in the market for GPU IP, and it was able to get AMD's for a good price. If it wasn't AMD's, it would have been somebody else's (Imagination, for example).

Qualcomm would still be building best-in-class SoCs with strong graphics, and the world would probably be much as it is now. If AMD didn't sell of its mobile GPUs, there is absolutely no guarantee that management would have actually have been successful with it (the fact they sold it indicates that they didn't have the vision that Qualcomm did).

 

[Keysplayr]

It wouldn't have made a difference in the scheme of things. Qualcomm was clearly in the market for GPU IP, and it was able to get AMD's for a good price. If it wasn't AMD's, it would have been somebody else's (Imagination, for example).

Qualcomm would still be building best-in-class SoCs with strong graphics, and the world would probably be much as it is now. If AMD didn't sell of its mobile GPUs, there is absolutely no guarantee that management would have actually have been successful with it (the fact they sold it indicates that they didn't have the vision that Qualcomm did).

So, simply because Qualcomm was in the market for GPU IP is why AMD sold theirs to them?
I'm not asking anyone to make excuses for AMD here, so you don't really need to offer any. What's done is done. You are right that we don't know if AMD could have done anything with the mobile IP on their own that would be worth anything, but then, they sold off their chance to try to do so. True though that I don't think anyone has the vision that Qualcomm has, though Nvidia is up there.

 

[NTMBK]

Are we to understand then, that you are of the opinion that AMD made a good decision in selling off their mobile IP?

I do think so, yes. AMD didn't have the resources to compete in mobile-this is a market which has chewed up and spat out Intel, Nvidia, and TI. It would just be another money losing business division, better to sell the IP.

I do however agree that they undervalued it.

 

[AnandThenMan]

I often wonder why the same people that say the ATI acquisition was a terrible idea also say selling off the mobile IP was a terrible idea. Does that mean AMD should have only bought ultra mobile tech from somewhere versus buying ATI? That's a pretty nutty idea, makes buying ATI look amazing by comparison.

 

[Keysplayr]

I do think so, yes. AMD didn't have the resources to compete in mobile-this is a market which has chewed up and spat out Intel, Nvidia, and TI. It would just be another money losing business division, better to sell the IP.

I do however agree that they undervalued it.

Well, we'll never know now. Also, TI may be finished, but Intel and Nvidia are far from throwing in the towel. They will keep trying and eventually they will succeed. Even if it means teaming up in the mobile sector.

 

[Keysplayr]

I often wonder why the same people that say the ATI acquisition was a terrible idea also say selling off the mobile IP was a terrible idea. Does that mean AMD should have only bought ultra mobile tech from somewhere versus buying ATI? That's a pretty nutty idea, makes buying ATI look amazing by comparison.

I can't follow your logic. AMD was not big enough to afford the purchase of ATI. It was this purchase that forced them to sell all their fabs, and even sell their headquarters building so that they can lease it from the new owners. I don't recall another corporation willingly spreading itself so thin. Many people feel that AMD selling off their IP was AMD forfeiting potential income by competing in the mobile sector. They held the IP for Adreno. That was pretty impressive IP. It had potential. Perhaps not in the hands of AMD as we have been discussing.

 

[NTMBK]

Well, we'll never know now. Also, TI may be finished, but Intel and Nvidia are far from throwing in the towel. They will keep trying and eventually they will succeed. Even if it means teaming up in the mobile sector.

NVidia is entirely done with phones (they even shut down their modem division), and Tegra seems to be moving more towards automotive than tablets tbh. But yeah, Intel seems happy to keep pouring billions down that black hole

 

[guskline]

If you watched the AMD presentation yesterday, I'm not sure whether they are "behind". If the metric is sales, sure, no argument. If it is innovation in memory utilization and VR, perhaps you should watch the presentation and make up your own mind.

BTW, I don't receive "free" products from either Nvidia or AMD so my opinion is ownly mine, not one I make which might be influenced by receiving free products.

 

[Keysplayr]

If you watched the AMD presentation yesterday, I'm not sure whether they are "behind". If the metric is sales, sure, no argument. If it is innovation in memory utilization and VR, perhaps you should watch the presentation and make up your own mind.

BTW, I don't receive "free" products from either Nvidia or AMD so my opinion is ownly mine, not one I make which might be influenced by receiving free products.

Subtle. My opinion always has been my own, guskline.

 

[guskline]

Subtle. My opinion always has been my own, guskline.

Keysplayr: I'll give you credit. You disclose who you are. I respect that AND your opinions :thumbsup:

I particularly like that you disclose, right below your comment, being an Nvidia focus member.

 

[Keysplayr]

Keysplayr: I'll give you credit. You disclose who you are. I respect that AND your opinions :thumbsup:

I particularly like that you disclose, right below your comment, being an Nvidia focus member.

Thank you, sir.

 

[NTMBK]

If we are to believe AMD from their analyst day. It will just be optimizations with shrinks.

So just like Pascal then?

 

[ShintaiDK]

So just like Pascal then?

Yep. 14/16nm cost is a financial killer for GPUs. And unless something changes. The volume at the time could be below 30 million GPUs a year.

 

[NTMBK]

Yep. 14/16nm cost is a financial killer for GPUs. And unless something changes. The volume at the time could be below 30 million GPUs a year.

Chuck a CPU on the interposer and sell it as an APU, easy way to bump up the volume!

 

[ShintaiDK]

Chuck a CPU on the interposer and sell it as an APU, easy way to bump up the volume!

Future is CPU with IGP attached. And no dGPUs. 14/16nm may be the last dGPUs we see.

 

[NTMBK]

Future is CPU with IGP attached. And no dGPUs. 14/16nm may be the last dGPUs we see.

If the dGPU die is already produced as a byproduct of building the APU, why not release it as a separate component too?

I'm 100% expecting split dies on an interposer to be how AMD does its Zen APU. A CPU-only die on a process tuned for CPU performance, a GPU only die on a process tuned for GPU density, maybe even the separate southbridge on some cheaper process (28nm?), and HBM memory.

 

[ShintaiDK]

If the dGPU die is already produced as a byproduct of building the APU, why not release it as a separate component too?

I'm 100% expecting split dies on an interposer to be how AMD does its Zen APU. A CPU-only die on a process tuned for CPU performance, a GPU only die on a process tuned for GPU density, maybe even the separate southbridge on some cheaper process (28nm?), and HBM memory.

2 dies, even if interposer means you need a lot of highspeed interconnect between the 2. Thats both power and die size increase.

 

[SirPauly]

Future is CPU with IGP attached. And no dGPUs. 14/16nm may be the last dGPUs we see.

Maybe the future is GPUs with a CPU attached. I'm just playing around.

 

[BFG10K]

I particularly like that you disclose, right below your comment, being an Nvidia focus member.

You think he has a choice in the matter? Unless something has recently changed, the disclaimer is a requirement from nVidia if you're part of the program, and was only mandated after their hidden shilling program (AEG) was exposed.

 

[AnandThenMan]

I can't follow your logic.

This seems to happen a lot, I apologize for not articulating my posts more in a way you can understand.