Anyone has confirmation AMD will use DDR2?

[designit]

As the header says, I would like to see reliable documentation that AMD will use DDR2 next year and if any info about using Rambus XDR instead.
DDR2 is such a short lived because DDR3 just around the corner, plus no performance gain going from DDR to DDR2, I don?t think the move is such a good idea. All reasoning amounts to AMD might skip DDR2.
Your info and input appreciated.

 

[biostud]

The availability and price of DDR2 favors it over XDR x100000

 

[designit]

It is a big mistake to move to DDR2. AMD doesn?t get it. When they follow the footsteps of Intel they stay behind, but when make drastic measures, such as moving to A64 architecture, mem controller in cpu die, they lead. Why not make a move to XDR and force Intel to follow? These days everyone is performance hungry, the combination of AMD mem controller in cpu die+XDR will lead monstrous results in terms of performance. This is what AMD should concentrate on.

 

[jimbob200521]

Originally posted by: designit
It is a big mistake to move to DDR2. AMD doesn?t get it. When they follow the footsteps of Intel they stay behind, but when make drastic measures, such as moving to A64 architecture, mem controller in cpu die, they lead. Why not make a move to XDR and force Intel to follow? These days everyone is performance hungry, the combination of AMD mem controller in cpu die+XDR will lead monstrous results in terms of performance. This is what AMD should concentrate on.

i couldnt agree more. XDR looks so promising if dealt with right, and what happens next should be quite interesting

 

[biostud]

Remember the fate of RDRAM made by Rambus?
The big money aren't made on the top systems but on low- mid end systems so shifting to something that would make these segments to expensive is an economical disaster waiting to break loose.
ATM it doesn't seem like the A64 platform is bottlenecked by memory performance, and they currently have the best performing desktop chips, so no need to take large risks in this area.
While it might sound cool in theory the real world gains would probably be a lot less than you think, and the problems involved would be huge.
And even though you think everyone are performance hungry, 99% of the system builders are cost and availability hungry.

 

[designit]

Originally posted by: biostud
Remember the fate of RDRAM made by Rambus?
The big money aren't made on the top systems but on low- mid end systems so shifting to something that would make these segments to expensive is an economical disaster waiting to break loose.
ATM it doesn't seem like the A64 platform is bottlenecked by memory performance, and they currently have the best performing desktop chips, so no need to take large risks in this area.
While it might sound cool in theory the real world gains would probably be a lot less than you think, and the problems involved would be huge.
And even though you think everyone are performance hungry, 99% of the system builders are cost and availability hungry.

Yes but if DDR2 is less than a year usage and no performance increase- a lot of money wasted by AMD. The implication of that is more severe than making the change to XDR, or not change at all(stay w/ DDR1). AMD took a big risk on A64 and has paid off.
I dont agree w/ you on "real life performance not noticeable", it has already proven otherwise.

 

[redhatlinux]

Sorry guys but you don't get it. I worked as a Consultatnt to AMD and have a good buddy who worked in the Austin fab until a couple
of years ago. AMD doesn't try to force Intel to do anything. They use innovative designs and technology to compete in an extremely tough market place. The CPU's that you are seeing now have been in development for years. In fact 60 nm fab process was available in test labs before my buddy left. Memory benchmarks have always shown that the AMD design does not benefit as much as the long pipe Intel design from increased memory bandwidth. The reason is likely to be the caching scheme that AMD uses. The long pipe of Intel CPU's gets VERY screwed up by a branch miss. IIRC, the entire trace cache is flushed, which means that fast memory has a definite benefit. I don't know if any one has effectively measure cache hit percentages between the two designs but I would guess that AMD has a higher hit rate.

 

[biostud]

Originally posted by: designit

Originally posted by: biostud
Remember the fate of RDRAM made by Rambus?
The big money aren't made on the top systems but on low- mid end systems so shifting to something that would make these segments to expensive is an economical disaster waiting to break loose.
ATM it doesn't seem like the A64 platform is bottlenecked by memory performance, and they currently have the best performing desktop chips, so no need to take large risks in this area.
While it might sound cool in theory the real world gains would probably be a lot less than you think, and the problems involved would be huge.
And even though you think everyone are performance hungry, 99% of the system builders are cost and availability hungry.

Yes but if DDR2 is less than a year usage and no performance increase- a lot of money wasted by AMD. The implication of that is more severe than making the change to XDR, or not change at all(stay w/ DDR1). AMD took a big risk on A64 and has paid off.
I dont agree w/ you on "real life performance not noticeable", it has already proven otherwise.

you have seen AMD processors run with XDR memory? I'm not saying it's not good memory I'm just saying it's not a good solution for AMD, not now atleast.
No offence but I think the engineering and business department of AMD knows a bit more about what will be good for AMD. If they loose money on XDR, even if they make faster systems they're not going down that road. Even though AMD has made some really good products that easily can compete with Intel and even make better solutions do you see AMD gaining fast on Intel in market shares? They would gain less customers than they would loose going to XDR now. It's not the technology there's a problem, it's the concept of earning money that doesn't go well with XDR ATM.
My guess is that when AMD move to the new sockets they saw an opportunity to upgrade the memory controller at the same time. And to be sure not to run into any problems they choose the fastest memory available ATM. If there's one thing AMD cannot risk, it's loosing customers because of lack of XDR modules, or far to high price on these.
It's all about business, not about gaining some extra FPS in UT2K7.

 

[stevty2889]

Originally posted by: designit

Originally posted by: biostud
Remember the fate of RDRAM made by Rambus?
The big money aren't made on the top systems but on low- mid end systems so shifting to something that would make these segments to expensive is an economical disaster waiting to break loose.
ATM it doesn't seem like the A64 platform is bottlenecked by memory performance, and they currently have the best performing desktop chips, so no need to take large risks in this area.
While it might sound cool in theory the real world gains would probably be a lot less than you think, and the problems involved would be huge.
And even though you think everyone are performance hungry, 99% of the system builders are cost and availability hungry.

Yes but if DDR2 is less than a year usage and no performance increase- a lot of money wasted by AMD. The implication of that is more severe than making the change to XDR, or not change at all(stay w/ DDR1). AMD took a big risk on A64 and has paid off.
I dont agree w/ you on "real life performance not noticeable", it has already proven otherwise.

You have to look at it from a differant perspective. DDR2 is the newer technology, weather it's better or not. Eventualy production of the newer technology surpasses production of the older technology. DDR will become less available, and more expensive, while DDR-2 will become mainstream, less expensive, and improve over time. It would be foolish to stick with the older technology thats going to become less available and more expensive, even if the newer stuff does not have any performance increase at the time, eventualy it will get better. XDR wouldn't likely benifit AMD much either, it's a very high bandwidth design, and A64's have never been bandwidth starved, it's not going to be widely available, and it's gonna be expensive.

 

[Pabster]

XDR would be AMD's RAMBUS. Bad idea.

 

[dullard]

AMD does have a licence to use all of Rambus's products. But they've had that license for years and have done nothing with it.

I really don't see why everyone is against Rambus's products. RDRAM was clearly the superior product compared to DDR, was about the same cost for months, and for a good month was the cheaper product. Then DDR manufacturers flooded the market, sold DDR at a loss, was (potentially illegally) supported by local governments to allow the DDR to sell below cost, and DDR became cheaper. But will that scenerio ever happen again for future products? Probably not. Don't count on it in the future.

Honestly no one in the real world really cares if AMD leads or follows. If they produce a good product at a competitive price, then we should all be happy. I just don't see AMD's current chips (and short-term future chips) needing the bandwidth of XDR at the moment. So let AMD follow on the memory front. Don't take that personally. AMD and Intel are just two companies trying to make a buck from you and I. Nothing more.