AMD "Steamroller" - What's the deal?

[BD231]

Lets not pretend better fab tech wouldn't help current AMD arch regardless of R&D spending.

 

[Arachnotronic]

The thing is though, the whole "module" concept actually seemed like a really clever way to try to raise performance/mm^2. We saw that some fairly minor tweaks really helped Piledriver, so it still seems as though Steamroller could narrow the gap further. It's not like Intel will be spending the majority of the process advantage on performance at the high end.

 

[Vesku]

Well there is definitely plenty of room for improvement, Piledriver got a nice bump from targeting the low hanging fruit. At this point though I really think AMD should be focusing on APUs and not high end stand alone CPUs.

 

[IntelUser2000]

What I think is that while AMD will catch up in the market AMD deems most important, the high end desktop market(only because desktops are significant portion of AMD's sales), they won't do it elsewhere.

AMD's volume and revenue are almost evenly split between desktop and laptops. But for Intel, laptops take 65%, and its increasing.

For Intel, it is very worth it to focus on more mobile products like Ultrabooks, Tablets, and Smartphones. So it might end up looking like they are becoming less competitive in the traditional desktop market, when it is actually because they are changing their focus(and money) to elsewhere. Naturally because resources are limited even for a company like Intel, focusing on one area means doing less so on others.

The ONLY market where it deems Intel's newfound focus like power consumption, performance/watt, powerful iGPUs not important is the traditional desktop market.

Everywhere else sees it as important.

 

[frozentundra123456]

Well there is definitely plenty of room for improvement, Piledriver got a nice bump from targeting the low hanging fruit. At this point though I really think AMD should be focusing on APUs and not high end stand alone CPUs.

What I dont understand, given the delays and long time frame to get it out, is why BD come out so poorly optimized.

 

[IntelUser2000]

What I dont understand, given the delays and long time frame to get it out, is why BD come out so poorly optimized.

Because usually delays are the result of things not going well. They can't meet their original expectations and project goals so they push it out. Usually what happens is though the revisions do nothing much and causes them to release a mediocre product later than promised.

 

[blastingcap]

What I think is that while AMD will catch up in the market AMD deems most important, the high end desktop market(only because desktops are significant portion of AMD's sales), they won't do it elsewhere.

I stopped reading after that part. AMD deems desktop market most important since when? I will let others cite to many sources saying that is not true, including Rory Read himself.

 

[Arachnotronic]

What I think is that while AMD will catch up in the market AMD deems most important, the high end desktop market(only because desktops are significant portion of AMD's sales), they won't do it elsewhere.

AMD's volume and revenue are almost evenly split between desktop and laptops. But for Intel, laptops take 65%, and its increasing.

For Intel, it is very worth it to focus on more mobile products like Ultrabooks, Tablets, and Smartphones. So it might end up looking like they are becoming less competitive in the traditional desktop market, when it is actually because they are changing their focus(and money) to elsewhere. Naturally because resources are limited even for a company like Intel, focusing on one area means doing less so on others.

The ONLY market where it deems Intel's newfound focus like power consumption, performance/watt, powerful iGPUs not important is the traditional desktop market.

Everywhere else sees it as important.

Intel is very, very focused on desktops. At the "growth" end, it's all about AIOs. At the "performance" end, Intel's datacenter segment (this includes workstations) is the firm's "bright star" against low PC growth. "Mainstream" desktop goes BGA because it'll all be AIOs and low end Celeron/Pentium/Core type things. The enthusiast stuff will ride the coattails of the workstation market.

X79 and its successors will be what people on this board will be buying going forward. Intel will expand the lineup of high end socket processors to include cheaper chips. Right now, hardly anybody buys X79 because the mainstream socket is just more appealing. This will change.

 

[Vesku]

What I dont understand, given the delays and long time frame to get it out, is why BD come out so poorly optimized.

Only AMD insiders would know for certain, but I'd say a big part of it was lots of compromising at the top level of design decision making, this is not just technical decisions, as the delays piled on. Combine that with depending more and more on the GF node helping rather than hindering the design, not a very rational belief but buying into the rosiest outlooks probably helped delay the inevitable heads rolling.

 

[ShintaiDK]

What I dont understand, given the delays and long time frame to get it out, is why BD come out so poorly optimized.

BD is a server oriented uarch. Thats one of the big problems for AMD. They _wanted_ to be the server people. Ironicly enough, the server segment is where they got the lowest marketshare.

 

[peonyu]

They still have a great GPU division. Hopefully when the ship sinks it doesn't take that down with it.