Article: AMD's Mobile Strategy
By: Doug Siebert (foo.delete@this.bar.bar), December 20, 2011 6:39 pm
Room: Moderated Discussions
Bill Henkel (noemail@yahoo.com) on 12/16/11 wrote:
---------------------------
>AMD should be trying to fill the gaps in Intel's product line instead of competing
>head-to-head with Intel. For tablets and notebooks, they could put four 4 Gbit
>DRAM die plus a wide I/O DRAM (for graphics) stacked in the same package as one
>of their CPU+graphics chips. For desktops, they could make a 6 GHz dual core chip
>that would be water cooled. This desktop chip should not have integrated graphics.
>They could put some GDRAM or SRAM chips on the same package as their processor
>for an L4 cache. They could add 512-bit or 1024-bit SIMD extensions. Any of these
>would be more profitable than making chips that go straight into the bargain bin
>the day they are introduced, which is what AMD is doing today. I hope AMD is not
>wasting their limited resources on a combined x86 plus ARM chip. That would make
>as much sense as adding an IBM 360 core just in case someone wants to run some old COBOL code.
Let's say AMD is able to produce a 6 GHz water cooled dual core chip tomorrow. What stops Intel from countering it? They are about one process generation ahead of AMD, an advantage which isn't going away. They could take their top products, laser off all but two cores, crank up the voltage, and slap on a nice water cooling system. Maybe they don't hit 6 GHz, but they'd get high enough that with their IPC advantage they'd easily match or beat AMD.
Now if you're saying AMD should forget about the low end and design something from the ground up targeted at the high end, that's another matter. They'd freely use lots of transistors in the cores where this would help performance, knowing it only needs two, design it with a 6 GHz frequency target, and design it to have better IPC. What happens if they do all that, and built something that hit their frequency target and matched Sandy Bridge's IPC, only to watch Intel tweak a few knobs on their processes to produce a limited run of CPUs that tilt toward the "higher frequency but higher power" end of the process spectrum? That extra power use would be fine as they'd be built for water cooling. Suddenly Intel is selling 6.5 GHz CPUs, and AMD is screwed. AMD would once again be left selling bargain basement CPUs, but in tiny quantities. Bankruptcy would be months away.
They may have leadership issues, but at least they aren't letting you lead them directly to their doom. Basically your option is a total gamble, and unless everything went exactly according to plan, and Intel was not able to use their process generation and cash advantage to take back the performance lead, AMD would be bankrupt before they could get another design out the door.
---------------------------
>AMD should be trying to fill the gaps in Intel's product line instead of competing
>head-to-head with Intel. For tablets and notebooks, they could put four 4 Gbit
>DRAM die plus a wide I/O DRAM (for graphics) stacked in the same package as one
>of their CPU+graphics chips. For desktops, they could make a 6 GHz dual core chip
>that would be water cooled. This desktop chip should not have integrated graphics.
>They could put some GDRAM or SRAM chips on the same package as their processor
>for an L4 cache. They could add 512-bit or 1024-bit SIMD extensions. Any of these
>would be more profitable than making chips that go straight into the bargain bin
>the day they are introduced, which is what AMD is doing today. I hope AMD is not
>wasting their limited resources on a combined x86 plus ARM chip. That would make
>as much sense as adding an IBM 360 core just in case someone wants to run some old COBOL code.
Let's say AMD is able to produce a 6 GHz water cooled dual core chip tomorrow. What stops Intel from countering it? They are about one process generation ahead of AMD, an advantage which isn't going away. They could take their top products, laser off all but two cores, crank up the voltage, and slap on a nice water cooling system. Maybe they don't hit 6 GHz, but they'd get high enough that with their IPC advantage they'd easily match or beat AMD.
Now if you're saying AMD should forget about the low end and design something from the ground up targeted at the high end, that's another matter. They'd freely use lots of transistors in the cores where this would help performance, knowing it only needs two, design it with a 6 GHz frequency target, and design it to have better IPC. What happens if they do all that, and built something that hit their frequency target and matched Sandy Bridge's IPC, only to watch Intel tweak a few knobs on their processes to produce a limited run of CPUs that tilt toward the "higher frequency but higher power" end of the process spectrum? That extra power use would be fine as they'd be built for water cooling. Suddenly Intel is selling 6.5 GHz CPUs, and AMD is screwed. AMD would once again be left selling bargain basement CPUs, but in tiny quantities. Bankruptcy would be months away.
They may have leadership issues, but at least they aren't letting you lead them directly to their doom. Basically your option is a total gamble, and unless everything went exactly according to plan, and Intel was not able to use their process generation and cash advantage to take back the performance lead, AMD would be bankrupt before they could get another design out the door.