I blame monolithic design

[kobymu]

Does anyone here also think that AMD shot themselves in the foot with that decision?

Because the more I think about it, the more I'm convinced that it was a crucial mistake, and moving away from it (monolithic design) could have single handedly alleviate most if not all the problems that AMD has suffered from in the last six months - the delays and setbacks, the manufacturing problems, cost efficiency, and probably a lot more.

I was a mistake, and I think they (AMD) are paying the price now.

If you have the resources to back up your desire to create the 'elegant' product, then ok. If you have the technological advantage (in the design/architecture), as in enough of it to pool ahead giving your disadvantage at the manufacturing end, by all means. If you have the technological advantage (manufacturing capabilities) to pool it off , go for it.

But if you don?t, then what the hell are you thinking!?

Like most people here I'm an enthusiast, I want the 'elegant' product, but not at the price of seeing AMD go down.

Am I the only one here that blames monolithic design as the source of most AMD current problems?

 

[Accord99]

Personally, I think AMD only really intended for Barcelona at 65nm to be mainly a low-clock, high throughput, high-margin server CPU. Then they caught by surprise by Conroe (like everybody else), which combined excellent per-clock performance, high clockability and low power consumption. Conroe's low power consumption enabled Intel to go ahead with high-clock quad-cores and forced AMD into a battle that they never expected to fight.

 

[dmens]

Originally posted by: kobymu
Like most people here I'm an enthusiast, I want the 'elegant' product, but not at the price of seeing AMD go down.

c2d/c2q is inelegant?

 

[Cookie Monster]

Originally posted by: dmens

Originally posted by: kobymu
Like most people here I'm an enthusiast, I want the 'elegant' product, but not at the price of seeing AMD go down.

c2d/c2q is inelegant?

The quad core design could be considered inelegant since its basically 2 core 2 duos in a MCM configuration. Primitive, but effective thanks to Conroes being elegant to begin with.

However as long as the end results are sound then theres nothing wrong with that. AMD went for the extra mile with the native quad core design but its not really paying off right now. However i do think it will in a few years time as AMD gets more experience in dealing with such designs, and possibly give invaluable help for the next generation products to come

 

[ViRGE]

Originally posted by: dmens

Originally posted by: kobymu
Like most people here I'm an enthusiast, I want the 'elegant' product, but not at the price of seeing AMD go down.

c2d/c2q is inelegant?

Design-wise the C2Q is inelegant, the dual-dice have-to-cross-the-FSB design is very inefficient at times. However the chips are so powerful on the whole and there's so little software held back by this, that it doesn't matter.

It's like a lightsaber versus a blaster; both work, but we all like the lightsaber more.

 

[myocardia]

I agree with the OP, although I'm not 100% sure that that's the only problem. I think that anytime you have someone running a company who knows absolutely nothing about the products it's selling, and they happen to be the performance leader at the time that pompous fool takes over control of the company, things usually end up not working out very well for said company. At least, that's my take on it.

 

[kobymu]

Originally posted by: dmens

Originally posted by: kobymu
'elegant'

c2d/c2q is inelegant?

The straight apostrophes (I think they also called ticks) are there for a reason.

 

[dmens]

Originally posted by: ViRGE
Design-wise the C2Q is inelegant, the dual-dice have-to-cross-the-FSB design is very inefficient at times. However the chips are so powerful on the whole and there's so little software held back by this, that it doesn't matter.

It's like a lightsaber versus a blaster; both work, but we all like the lightsaber more.

how inefficient is the old FSB at 4 cores? is it the bottleneck in even a significant fraction of workloads? i think not.

besides, you made my point. the elegance of the processing core is what really matters, the bus is just the most visible so people make a big deal out of it.

 

[JackPack]

CPU designs take 3-5 years before fruition.

AMD's game plan for Barcelona was devised back when Intel was touting NetBurst. Back then, Intel intended to take NetBurst to the next step via Tejas, which was a monolithic 213mm2 device (Barcelona is 285mm2). AMD probably also genuinely believed they could somehow achieve 30% market share and have the fabs along with it.

Intel suddenly did a right-hand turn 2005/2006 and AMD was clearly unprepared for it.

 

[AlabamaCajun]

Just think is AMD had gotten this out in 2h 2006, the IT world be a different place. Barcelonas are selling faster than AMD can make them so it can't be all that bad. Now lets see how well Phenom and Tolem do.

 

[Aluvus]

Originally posted by: dmens

how inefficient is the old FSB at 4 cores? is it the bottleneck in even a significant fraction of workloads? i think not.

besides, you made my point. the elegance of the processing core is what really matters, the bus is just the most visible so people make a big deal out of it.

When you have 4 quad-core processors in a system, the limitations of that bus become more evident. This specific factor is why forecasts generally anticipate that AMD will hold its share in the server space relatively well.

 

[Accord99]

Originally posted by: Aluvus
When you have 4 quad-core processors in a system, the limitations of that bus become more evident. This specific factor is why forecasts generally anticipate that AMD will hold its share in the server space relatively well.

But there are many ways to alleviate those problems and the HT bus has problems of its own; which is why Tigerton Xeon MPs dominate all of the important enterprise benchmarks for x86 servers.

 

[Regs]

Originally posted by: Accord99
Personally, I think AMD only really intended for Barcelona at 65nm to be mainly a low-clock, high throughput, high-margin server CPU. Then they caught by surprise by Conroe (like everybody else), which combined excellent per-clock performance, high clockability and low power consumption. Conroe's low power consumption enabled Intel to go ahead with high-clock quad-cores and forced AMD into a battle that they never expected to fight.

I agree Accord. The manufacturing costs alone of the Quad Barcelona does not offer much flexibility anywhere else in the market. AMD since the K8, have targeted niche markets while keeping operating costs down, and while doing so has effectively shut themselves off from the majority. Intel, imo, knew this and capitalized with both a very flexible product line and price reductions (price war) which nullified AMD's competitive edge and product line. Targeting niche markets are usually what small businesses go after, but AMD isn't exactly a small business. Especially when they're up against Intel as their only competitor..