If I were the CEO of AMD... (rate)

[Phynaz]

Nope, the closet isn't empty. Quite the opposite, if they are going to sell, they will have to sell the meaty parts: IP.

Typically IP only gets sold in bankruptcy. If you sell your IP you might as well be in bankruptcy because you zero advantage over any other company. Especially in AMD's case, because an IP company is what they are now - they don't make anything anymore.

Recent examples - Kodak and THQ.

 

[AnandThenMan]

To me what mrmt is advocating will put AMD out of business for good, maybe that is what he wants? I can't see any other reason to propose such a suicidal business strategy.

 

[HopJokey]

I’d shut it down and give the money back to the shareholders.

There is no money, I believe they have more debt than assets/cash.

 

[AnandThenMan]

You have to look no further than Nvidia to see that GPU isn't a strong differential as you think. Despite being ahead in graphics than any of the other ARM manufacturers, they are struggling in the mobile arena because they can't field a chip with bleeding edge CPU capabilities.

Nvidia has access to the same ARM IP as everyone else. The differentiation is total design and execution, not anything about a "bleeding edge CPU" disadvantage.

 

[mrmt]

To me what mrmt is advocating will put AMD out of business for good, maybe that is what he wants? I can't see any other reason to propose such a suicidal business strategy.

From what business? From the bleeding edge business? Yes, my strategy would, but AMD will leave the bleeding edge business with their current strategy, so this is a non-issue. What I'm proposing is for AMD to survive in other arenas, not close down.

 

[mrmt]

Nvidia has access to the same ARM IP as everyone else. The differentiation is total design and execution, not anything about a "bleeding edge CPU" disadvantage.

But if GPU is such a big deal, where's Nvidia's edge in the ARM arena?

 

[AnandThenMan]

From what business? From the bleeding edge business? Yes, my strategy would, but AMD will leave the bleeding edge business with their current strategy, so this is a non-issue. What I'm proposing is for AMD to survive in other arenas, not close down.

I want to sell that part because AMD needs to tackle a CPU problem, not GPU, and whatever money they can make with GPU is insignificant for the CPU needs. If AMD cannot make bleeding edge CPUs, and it is clear that they don't have the resources to stay in the bleeding edge in the future, what's the point in bleeding edge graphics?

Then I really don't know what you want AMD to do. You don't want them in graphics, say if they can't make the best performing CPU why bother, they should divest themselves of their graphics tech. So what do you want them to do?

But if GPU is such a big deal, where's Nvidia's edge in the ARM arena?

Different market, different needs. But Nvidia HAS GPU tech, what you are saying is AMD should have none, license it and hope for the best.

 

[mrmt]

Then I really don't know what you want AMD to do. You don't want them in graphics, say if they can't make the best performing CPU why bother, they should divest themselves of their graphics tech. So what do you want them to do?

Embedded, very low cost processors applications, licensing IP, etc.

Just get out of the bleeding edge market, they don't have a stick big enough to fight this fight.

 

[AnandThenMan]

Embedded, very low cost processors applications, licensing IP, etc.

I take it you mean no more x86 APUs, no more desktop or x86 attached mobile GPUs. So then ARM only products? That's a pretty crowded area, how long do you think AMD could at least get to the same level of revenue they have now? Keep in mind they don't have much reserve cash left to keep them going during a radical shift in business strategy/product.

 

[mrmt]

I take it you mean no more x86 APUs, no more desktop or x86 attached mobile GPUs. So then ARM only products?

Yes, no more desktop, no more APU. And no, not ARM. AMD can leverage on x86 to bring it to those markets. Intel won't follow as margin is too low and returns too slow, but a leaner AMD could live with that.

That's a pretty crowded area, how long do you think AMD could at least get to the same level of revenue they have now? Keep in mind they don't have much reserve cash left to keep them going during a radical shift in business strategy/product.

Their revenues are falling like a rock, so I don't think this is a issue.

What do you mean by "applications" exactly? And what IP do you propose they license? They can't license x86, Intel won't allow that.

Applications I meant ASIC, and they can license memory controller, develop uncores, etc.

 

[Wall Street]

The low cost market is no panacea. ARM chips have much lower margins, and instead of one competitor in each space you need to fight against nVidia Tegra, Qualcomm Snapdragon, Intel Medfield, Apple Ax, TI OMAP, MIPS, Samsung Exynos.

Also like I pointed out earlier, if you get off the bleeding edge you save R&D now but in 5 years it catches up to you and your company only has 5 year old tech.

 

[AnandThenMan]

mrmt good debate but I really think you're grasping at straws. I don't see anything you're proposing as being even close to a viable business plan, it's tossing hope in the wind in my view.

 

[cytg111]

AMD's only viable path forward is their APU's, every other turn is a dead end. I'd move all-in with that hand. Wait they allready did.

 

[NTMBK]

AMD's only viable path forward is their APU's, every other turn is a dead end. I'd move all-in with that hand. Wait they allready did.

Not entirely- they're still trying to keep the downright ancient AM2/2+/3/3+ platform alive, with its lack of APU and separate north and southbridge. Ditch that platform, is my move. Keep the server side LGA sockets alive for massive-core-count Opterons, and focus on an FM3 socket. Preferably something which breaks compatibility with FM2 and all the old AMD coolers, because they're not going to get a good range of miniITX boards until they shrink the area they steal with socket and cooler mounting. Focus on kicking ass in small form factor PCs. Pair with Valve, and push the "PC as a console" concept. (It may piss off Sony and MS, but they're already using the AMD chips and will have to deal with it.)

 

[mrmt]

Also like I pointed out earlier, if you get off the bleeding edge you save R&D now but in 5 years it catches up to you and your company only has 5 year old tech.

But AMD is slipping from the bleeding edge already, or do you think they x86 CPUs are on par with Intel's or that their vanilla ARM core will be on par with Qualcomm's, Samsung's or Tegra's offers?

As soon as GPU starts to move to GLF, what happens when that subpar foundry misses another deadline or process parameter?

mrmt good debate but I really think you're grasping at straws. I don't see anything you're proposing as being even close to a viable business plan, it's tossing hope in the wind in my view.

I might be, and what I'm outlining might be unfeasible, but I think the right direction is somewhere along the lines I'm proposing here. By the end of 2013 AMD will have lost some 30% of their sales and the revenue left is with far lower margins than before. The old bleeding edge business model is dead.

I don't see AMD surviving if not going for a much learner, smaller profile, and you can only achieve that out of the bleeding edge. Btw, Rory's strategy shares some of these premises.

 

[Vesku]

Yes, no more desktop, no more APU. And no, not ARM. AMD can leverage on x86 to bring it to those markets. Intel won't follow as margin is too low and returns too slow, but a leaner AMD could live with that.



Their revenues are falling like a rock, so I don't think this is a issue.



Applications I meant ASIC, and they can license memory controller, develop uncores, etc.

Where's the value and future market potential of ditching the GPU integration? The low end demand is heavily on the SoC side.

 

[mrmt]

Where's the value and future market potential of ditching the GPU integration? The low end demand is heavily on the SoC side.

Where's the value in ditching? First I think you need to explain where's the value in building this integration in-house in the first place.

Nobody is going for AMD's level of integration, what we have now is most players are putting graphics with 3rd party IP on the same die, and it stops there. Nvidia isn't too keen on that integration too (and when a company with the sense of marketing that Nvidia has isn't going on that direction it says something), and Nvidia's GPU IP isn't propping up their Tegra line on mobile. So yes, I'd like first to see value in the CPU integration and why couldn't this integration must be done with 3rd party IP, but with in-house IP.

Doing this integration in-house is nothing more than a luxury, and I don't think AMD can afford any luxuries now.

 

[Vesku]

Those going AMDs route of integration: http://hsafoundation.com/

One of the most successful ARM SoC vendors is using GPU tech they bought off AMD, even. Qualcomm HSA founder. Undoing years of ATI GPU integration seems like a big cost sink they don't need.

What's been killing them recently (outside a bad bulldozer launch) has been the ugly detachment of their foundry and the resulting wafer agreement with the resulting independent company GlobalFoundries.

 

[mrmt]

Those going AMDs route of integration: http://hsafoundation.com/

One of the most successful ARM SoC vendors is using GPU tech they bought off AMD, even. Qualcomm HSA founder. Undoing years of ATI GPU integration seems like a big cost sink they don't need.

And except for AMD, what members of the HSA foundation are actually developing HSA silicon?

 

[Idontcare]

Suppose AMD did this starting tomorrow, who would step up and use the GPU tech? Be specific. Now this, "bested only by Nvidia and even then it is an even heat at times" are you serious? AMD has many times bested Nvidia, going back to the ATI days same thing. They have both traded performance crowns many times, lately AMD has generally been first with node shrinks, launches, and hardware tech such as GDDR5 which AMD essentially created.

Step back and read what I wrote, you so desperately want to see me as bashing AMD that you cannot even recognize when I am doing the opposite.

Who is capable of besting AMD? Nvidia. And even then, on those few occasions where Nvidia has bested AMD over the years, even then it was basically an even heat (and never a blow-out). All other times it is AMD at the front.

I posted earlier that the OP was conflating the value of creating innovation with that of the need to create profits, and I believe you are doing the same as well.

Innovation alone does not create profits, look at what happened to DEC despite their Alpha CPU. Innovation must come in engineering as well as in marketing, accounting, and management if profits are to come as well.

AMD's problem has never been that of creating innovation. Compare them to Nvidia. ATI and Nvidia were roughly the same size, revenue and market share, when AMD bought ATI.

And in those 6 yrs since the purchase how much profit has the ATI division generated for AMD versus how much profit has Nvidia generated for itself?

AMD can invent quantum GPUs for all I care, be the first to produce them on a 5nm node, and deftly snag the performance crown - innovation after innovation - but if they do all that to simply generate $15m in quarterly profits then they are doing it wrong.

As to who would license the technology - anyone that is looking for something to give them an edge in integrated graphics over their existing rivals. Who wouldn't want to license it to compete against Apple or Samsung?

If AMD could not find anyone who was interested in licensing their IP then that on its own would be very telling in a different way - that would suggest the IP they have is not innovative enough, is not differentiating enough, to compel customers to want to license it. If that is true, if you propose this to be the case, then that says AMD is already not in a competitive enough position now to sustain their place in the market.

The TAM they have access to now is very small between discrete GPU cards and APU's that have maybe 20% marketshare. Their TAM access would be much larger if they licensed their GPU tech out to the big dogs who would pay for access to such IP.

Think of the business argument for why AMD spunoff their fabs and created GloFo. There is no reason why the same argument could not be made of their GPU tech. Unlock that shareholder value.

 

[cytg111]

Pair with Valve, and push the "PC as a console" concept.

+1.

There is another route as well, if the engineering and profitability aspect of it makes sense; the Crazy Ivan move. While the rest of the industry is going lower power and mobile, AMD could create a formfactor for 400 watts+ and just cram ipc logic and cores at it galore ... retake the 'singlethread crown' no matter the cost in watts, and push even more cores at the unknowing consumer.

 

[mrmt]

There is another route as well, if the engineering and profitability aspect of it makes sense; the Crazy Ivan move. While the rest of the industry is going lower power and mobile, AMD could create a formfactor for 400 watts+ and just cram ipc logic and cores at it galore ... retake the 'singlethread crown' no matter the cost in watts, and push even more cores at the unknowing consumer.

Good luck cramming this power hog in small for factor cases or cooling this thing in power efficient data centers.

 

[Vesku]

And except for AMD, what members of the HSA foundation are actually developing HSA silicon?

We'll have to see over time which companies stick with the concept, but most of those founders are SoC sellers who will certainly be interested in increasing the integration of their CPUs with their GPUs.

You still haven't explained what advantage there is for selling out of integration now that AMD is finally seeing some payoff (consoles).

What AMD really needs is some better business relationships. Having a few large OEMs willing to work in tandem with you goes a long way. Also, not being bound by a very unfavorable production contract would help.

 

[cytg111]

Good luck cramming this power hog in small for factor cases or cooling this thing in power efficient data centers.

nvidia is doing it ?

 

[mrmt]

Nvidia is doing it ?

You mean Tesla on data centers? The reason behind GPGPU on data centers is because the GPU is more efficient than the CPU when dealing with highly parallel workloads, so here more than in the consumer market power efficiency is highly prized and thoroughly scrutinized. The push for efficiency is very evident. Take Titan for example. Titan offers more performance per watt in compute tasks than any other Nvidia chip.

Btw, the Crazy Ivan maneuver you are described was tried by AMD before: That's Bulldozer. Bulldozer was "moar cores, moar power, moar units, moar cache, moar die size". It failed, because AMD is so behind in the efficiency race that they may go for 400W and will not pull a significant edge over 150W Intel processors. Crazy Ivan would work only if AMD were in a position to offer a reasonable benefit for the in exchange for the added power consumption, which they aren't.