Intel Lynnfield CPU Bests Nehalem In Performance/Watt

Vigile writes "Not many people have debated that Intel's Nehalem architecture is the fastest available for consumer desktop computers since it was released last year, but quite a few have complained about the cost of the platform. Intel just released new Lynnfield-based processors under both the Core i7 and Core i5 names and tests are showing the new CPUs beating Nehalem in both performance-per-watt and performance-per-dollar tests to a startling degree. And while raw performance probably still goes to the Nehalem-based Core i7 CPUs, the lower prices of motherboards and memory for Lynnfield processors will likely more than make up for it." Update: 09/08 14:03 GMT by T : There are more eye-wateringly exhaustive examinations of the new chips all over the Web; here's HotHardware's version, and Tom's Hardware's.

Comments

Re:arm

[Kjella]

If you imagine ARM as a women's flyweight newbie and Lynnfield as the men's heavyweight world champion in boxing, you got a pretty good idea how that match will play out. Not nearly the same class and the results are as expected.

Pedantry note

[Kupfernigk]

I think the submitter means "disagreed", or "argued" not "debated". I expect that in the early stages quite a lot of people debated the subject, but when the results become clear they stopped arguing and there was a general agreement

Yes, I know it's pedantry, but some of us like to live in a world where different words mean different things that make a useful distinction. And now, please, do get off my lawn before my dog comes and pees on your shoes.

Nehalem vs. Nehalem

[Demetrius Berman]

Lynnfield is a Nehalem processor, just as the earlier Bloomfield is a Nehalem processor, hence the title to this article makes no sense. The difference is in socket (LGA 1156 vs. LGA 1366), and intended market ... with a couple design differences as well.

Re:Nehalem vs. Nehalem

[Andy Dodd]

By definition, if it isn't a Nehalem die, it's not Nehalem, even if it's just a "tock" variant (die shrunk - see Intel's "tick/tock" roadmap) of Nehalem it's still a different chip design.

In this case, the CPU has significant design differences from a Nehalem CPU. There's a lot more than just removing some pins from the package. The CPU had to be changed significantly (one DDR channel removed, QPI replaced with DMI) in order to allow those pins to be removed.

The removal of QPI in favor of DMI (much slower but simpler/cheaper) is a *significant* difference.

Re:Nehalem vs. Nehalem

[zokier]

There isn't a single Nehalem die/chip. Nehalem refers to the general architecture on which Lynnfield, Bloomfield etc chips are based on.

Re:AES benchmarks

[Neil Hodges]

The VIA Nano has had AES, SHA-1, and SHA-256 acceleration since its inception.

Re:arm

[644bd346996]

There are plenty of ARM processors with a great MIPS/W rating. Just not a great FLOPS/W rating, which is what keeps them out of supercomputers.

Re:What are Intel's naming department on?

[commodore64_love]

>>>The fact that CPUs are still named at all is for the benefit of enthusiasts.

False. If you bothered to learn your history, you'd know the reason why CPUs have names instead of numbers is because the courts ruled companies cannot trademark numbers. Thus the 80586 became the Pentium and that tradition has continued to today. They cannot just go back to calling them 80986 because of legal reasons.

Re:arm

[TheRaven64]

The popular ARM chips are single-core Cortex A8 running at between 600MHz - 1GHz. They perform slightly better than (single-core) Atom clock-for-clock on most workloads, and slightly worse on a few. The next generation chips that are just starting to hit the market are based on the Cortex A9, which does a bit better clock-for-clock and scales up to 4 cores per die. ARM chips also typically have the memory and flash controllers, GPU, and a DSP on die. Something like the OMAP3530 consumes around 250mW in real use or around 15mW when playing back MP3s on the DSP. They are typically limited to around 1GB of RAM, with only about 256MB being available in package-on-package configurations (i.e. not requiring a more expensive motherboard).

In short, they compare like apples and oranges. In terms of performance per watt, the ARM chip most likely wins by an order of magnitude - more if you include the DSP. In terms of absolute performance, the i7 wins by at least an order of magnitude.

Re:LGA 1366 dead now?

[LordKronos]

http://anandtech.com/cpuchipsets/showdoc.aspx?i=3634&p=7

Intel told me something interesting when I was out in LA earlier this summer: it takes at least 3 cores to fully saturate Lynnfield's dual-channel DDR3-1333 memory bus. That's three cores all working on memory bandwidth intensive threads at the same time. That's a pretty stiff requirement. In the vast, vast majority of situations Lynnfield's dual channel DDR3 memory controller won't hurt it.

Move up to 6 or 8 core designs and a third memory channel is necessary, and that's why we'll see those processors debut exclusively on LGA-1366 platforms. In fact, X58 motherboards will only need a BIOS update to work with the 6-core 32nm Gulftown processor next year. P55 looks like it'll be limited to four cores and below.

More Reviews

[tab_b]

Here are a few more reviews for today: The Tech Report, Phoronix, AnandTech, X-bit labs, and Benchmark Reviews. It's all enough to make your eyes bleed. There's a list for the Core i7 870 at 0x6877.com