So what is the next step for AMD?

[Idontcare]

Originally posted by: Extelleron
You have the die sizes mixed up; Nehalem is 246mm^2, Shanghai is 243mm^2.

Thanks! I corrected my post now.

Originally posted by: Extelleron
AMD is clearly at a disadvantage in the L3 cache; they don't have the resources to design specific SRAM for L3, so they reuse the same SRAM they use for L2. Intel designs SRAM designed for higher density for the L3, so they can stick 8MB L3 in the same area as AMD fits 6MB L3.

I don't quite see how that could be the logic behind why AMD is using their L2$ cell as the L3$ cell. At Texas Instruments (TI) we'd design 8 different SRAM cells for high-density/low-clocks and then downselect to the highest yielding/most-reliable cell for production. Likewise we'd design at least 4 different cells for low-density/high-clocks and test them out in silicon before committing to one or two cells for production.

It was not a resource intensive task to spin a single reticle with a plethora of cell layout schemes to test in parallel the robustness of all designs in one fell swoop.

The argument that AMD simply didn't have the resources to create >1 sram cell just seems unbelievable based on my experience in doing this at TI. There must have been some other reason the L2$ cells were recycled for L3$...maybe AMD intends to operate them both at the same clockspeed (we can hope).

Originally posted by: Extelleron
It seems like AMD has a lot more non-core/cache area on Shanghai compared to Nehalem.... total cache sizes are pretty similar and Nehalem has ~36mm^2 more core area, but the two are virtually the same in total die size.

Makes you wonder how efficient Intel was with their QPI and IMC layout and designs compared to AMD's layout and design of their HT and IMC units.

One thing that strikes me as "odd" is Intel previously stated that once they realized how much Idrive their PMOS xtors were creating thanks to HK/MG that they went to the drawing board and used this to their advantage in designed the next CPU...at the time this set everyone's expectations to mean that Nehalem was going to be designed to capitalize on this PMOS improvement...what I don't see in the xtor numbers versus die-size here with Nehalem versus Shanghai is anything that suggests Intel worked their PMOS magic into the Nehalem design.

 

[rchiu]

If I were AMD, I'd work on some enthusiast platform, like dual/quad socket OC'able with SLI/Crossfire chipset without putting a huge price premium. For the customer, even tho AMD has clock disadvantage, they will be able to run 8/16 way processing and make up the disadvantage with multi-processing. For the AMD, they can sell CPU 2/4 at a time. I know they had that 4x4 crap. It was socket F and they charged a big premium for that.

On the lower end, they have really good package with their new low powered CPU and 780G chipset. They need to market that aggressively, target market looking for low powered consumption setup as well as multi-media setup. They should highlight power saving to businesses running their new combo when 10, 50, 100+ pc is used.

AMD have decent product, they just need to market it correctly. All they have to do is return to profitability, and wait until they are able to transition to 45nm and/or push Phenom to higher clock speed.

 

[v8envy]

I'm not sure a Phenom multi-socket enthusiast platform would be compelling enough to spend R&D dollars on. Unless it could be priced at mainstream levels you'd never recoup that R&D -- the market for such a beast is tiny, and target customer can be reasonably expected to know a skulltrail + QX9770 path is a higher performer.

They'd also risk cannibalizing workgroup server and workstation CPU sales with an enthusiast setup like that. As is high end enthusiast hardware is encroaching on low end server & workstation territory. It's just ECC ram support away from being the same thing.

As far as business PCs -- intel's on board graphics are very thrifty, power wise. Ditto core2 at low clock rates. I doubt you could show a compelling ROI (read: less than 2 years) of sidegrading from e.g. an E2140 box with an IGP to a E4850e + 780G, especially considering both boxes are likely to spend 16 hours a day idle. Heck, the place I'm at now still has most employees on old 17-21" CRTs because they don't feel the ROI of buying new LCDs is justified -- and the power use difference there is huge.

HTPC guys are loading up on Sempron+780G already. Unfortunately the lowest end AMD cpu sales aren't great for their ASPs... Current 780G boards not supporting the higher margin chips was a minor disaster for AMD, profitability wise.

 

[Kuzi]

Originally posted by: Martimus
After Deneb, AMD has Bulldozer scheduled for next year. It will be the new architecture with a 45nm process. They plan to update that with a 32nm process shrink in 2010.

Next year AMD will probably improve their 45nm process and add high-k this should further lower the power draw of shanghai/deneb and also increase frequency 3.2GHz+. They might release 6 or 8 core cpus based on shanghai (they would be huge at 45nm)

I don't think AMD can release Bulldozer next year, maybe Q1 or Q2 2010.

Originally posted by: Idontcare
Kinda crazy how much of a disparity exists between the core's themselves. 24mm^2 vs 15mm^2

If Shanghai even comes close to the IPC of Nehalem it will be a major coupe for the AMD designers. Intel is just throwing tons more logic transistors at pumping up those ALUs/FPUs.

Remember Nehalem has hyperthreading logic for each of the 4 cores. So some of this size difference could be for that reason.

If Nehalem ends up only a little faster than yorkfield in IPC, then I say Shanghai could be competitive in many situations. But if Nehalem is really going to be that much faster than yorkfield (20-25% faster), then Shanghai has no chance.

Thnx for the Nehalem - Shanghai die shots btw. I knew AMD caught up in density to Intel for the most part.

 

[Cookie Monster]

Is Deneb AM2+? or AM3? just what are the difference between the two sockets?

From what ive read, they are electronically the same except AM3 is for DDR3 and AM2+ is still backward compatible with the 45nm Deneb except you can only use DDR2-1066 resulting in a lesser overall bandwidth.

 

[Idontcare]

Originally posted by: Cookie Monster
Is Deneb AM2+? or AM3? just what are the difference between the two sockets?

From what ive read, they are electronically the same except AM3 is for DDR3 and AM2+ is still backward compatible with the 45nm Deneb except you can only use DDR2-1066 resulting in a lesser overall bandwidth.

Dened is AM2+ at first, but in Q1/09 it is supposed to be released on AM3 socket as well.

And yes, AM3 = DDR3 and not much else to distinguish it IIRC aside from some "HT improvements".

http://pics.computerbase.de/2/0/6/2/2/2.png

 

[WTurner]

So what is the next step for AMD?

Simple. Fire Hector Ruiz

 

[bryanW1995]

hector must be a REALLY smooth talker. I mean, seriously, I need to get that guy to come sell cars for me. I can just picture it now:

customer: yeah, I want to compare that pt cruiser for 20k to an accord for 20k.
hector: naw, man, the pt cruiser is WAY better. it's um, er, more stylish, and, uh, just picture how jealous all your neighbors will be when you drive home in it.
customer: but it's PURPLE.
hector: yeah, isn't it great? sign here...


when the ship is running really well the captain often gets an inordinate share of the credit, but when the ship is sinking guess who goes down with it?

 

[rchiu]

Originally posted by: v8envy
I'm not sure a Phenom multi-socket enthusiast platform would be compelling enough to spend R&D dollars on. Unless it could be priced at mainstream levels you'd never recoup that R&D -- the market for such a beast is tiny, and target customer can be reasonably expected to know a skulltrail + QX9770 path is a higher performer.

They'd also risk cannibalizing workgroup server and workstation CPU sales with an enthusiast setup like that. As is high end enthusiast hardware is encroaching on low end server & workstation territory. It's just ECC ram support away from being the same thing.

As far as business PCs -- intel's on board graphics are very thrifty, power wise. Ditto core2 at low clock rates. I doubt you could show a compelling ROI (read: less than 2 years) of sidegrading from e.g. an E2140 box with an IGP to a E4850e + 780G, especially considering both boxes are likely to spend 16 hours a day idle. Heck, the place I'm at now still has most employees on old 17-21" CRTs because they don't feel the ROI of buying new LCDs is justified -- and the power use difference there is huge.

HTPC guys are loading up on Sempron+780G already. Unfortunately the lowest end AMD cpu sales aren't great for their ASPs... Current 780G boards not supporting the higher margin chips was a minor disaster for AMD, profitability wise.

Well, enthusiast platform was never about ROI, it's about bragging right and winning the enthusiast crowd who are often leading the tech direction of average consumer. The point of having multi-socket/OC friendly platform is to show that AMD want to make their product enthusiast friendly, unlike Intel which is rumored to disable OC with some Nehalems.

And about cannibalizing their server/workstation platform, I don't think AMD need to worry too much about it. Beside the ECC memory support, there are things like 16gig+ memory support, PCI-X support, SCSI and many other server related technology that server/workstation crowd is looking for. The support/warranty from the server/workstation product will also serve to separate the enthusiast and server/workstation market. And finally, serious server/workstation crowd won't be interested in OC platform to save some buck for mission critical app anyway.

And looking at Anandtech's 780G platform review, it is about 20% more energy efficient compared to G35/E2200 platform, and I think that's pretty significant. Maybe the sidegrading crowd won't be interested, but it will make great sense for business upgrading. AMD just need to market it well, they have the cheaper cpu/mobo combo, they have the better graphic (which is import for some Webapp with multimedia content, which many company is implementing), and they have lower total cost of ownership from energy consumption point of view. It is a very good value proposition.

As far as ASP, it's a matter of product mix and yield. That's something AMD have to manage better. As long as they can get pass this tough period, they still have great future.

 

[hooflung]

Originally posted by: WTurner

So what is the next step for AMD?

Simple. Fire Hector Ruiz

Well AMD might bleed cash for a while but their company is in a stronger position in the market with ATI in their corner. It has put Intel in a very tough place with Nvidia and Nvidia in a very tough place with chipset design after the current generation of C2Ds.

It also puts more pressure on Intel's Larabee investment to turn a profit and deliver performance since Intel rely's on AMD crossfire to compete with Nvidia on the Intel platform. Something that could end in the future.

AMD is now forced to turn a profit and they aren't doing a bad job of putting themselves in a position they can.