Article: AMD's Mobile Strategy
By: Dan Fay (daniel.fay.delete@this.gmail.com), January 5, 2012 9:03 am
Room: Moderated Discussions
George Baker (george_baker@yahoo.com) on 1/4/12 wrote:
---------------------------
>Intel's announced pricing for dual-socket 8-core SandyBridge-EP ranges from $1106
>(1.8 GHz) to $2057 (2.9 GHz) per chip! There is little benefit of x86 compatibility
>in servers so this market seems ripe for a transition.
Yes, SNB-EP is really expensive, but why does that imply that it's worth switching to ARM? The main reason why Intel charges so much for that part is because there's no real competition in that particular fat-core space.
AMD has a lower-cost offering (Opteron), but it's lower performing. Other server CPUs (SPARC, Itanium, PPC) are some combination of lower performance and higher price.
The Cortex-A15 and Qualcomm Krait are at *best* approaching AMD's K10-series in IPC and still trailing in clock speed. What's the benefit here in switching to ARM again?
Nvidia is working on Project
>Denver, which is an ARM chip for servers. I think other companies will make similar
>chips (perhaps even AMD). As the economic benefit of Moore's Law slows down, companies
>will have no choice but to use more efficient chips to provide continued price/performance
>improvements.
This might be true with really small cores, but I don't think it's particularly true with fat cores. PowerPC was a not-radically-different RISC architecture from ARM, and the last consumer-facing ones (PPC970, MPC74xx, PPC 750) never decisively trounced x86 except perhaps in AltiVec-optimized code.
>It wouldn't surprise me if the majority of annual server >shipments use ARM processors within 5 years.
Five years is an absurdly short timeframe. If this were to happen, it would probably take on the order of 10-15 years at least.
---------------------------
>Intel's announced pricing for dual-socket 8-core SandyBridge-EP ranges from $1106
>(1.8 GHz) to $2057 (2.9 GHz) per chip! There is little benefit of x86 compatibility
>in servers so this market seems ripe for a transition.
Yes, SNB-EP is really expensive, but why does that imply that it's worth switching to ARM? The main reason why Intel charges so much for that part is because there's no real competition in that particular fat-core space.
AMD has a lower-cost offering (Opteron), but it's lower performing. Other server CPUs (SPARC, Itanium, PPC) are some combination of lower performance and higher price.
The Cortex-A15 and Qualcomm Krait are at *best* approaching AMD's K10-series in IPC and still trailing in clock speed. What's the benefit here in switching to ARM again?
Nvidia is working on Project
>Denver, which is an ARM chip for servers. I think other companies will make similar
>chips (perhaps even AMD). As the economic benefit of Moore's Law slows down, companies
>will have no choice but to use more efficient chips to provide continued price/performance
>improvements.
This might be true with really small cores, but I don't think it's particularly true with fat cores. PowerPC was a not-radically-different RISC architecture from ARM, and the last consumer-facing ones (PPC970, MPC74xx, PPC 750) never decisively trounced x86 except perhaps in AltiVec-optimized code.
>It wouldn't surprise me if the majority of annual server >shipments use ARM processors within 5 years.
Five years is an absurdly short timeframe. If this were to happen, it would probably take on the order of 10-15 years at least.