Replies to post #116771 on Intel Corp (INTC)

[chipguy]

Tegra 4 already hits 1168 for SPECInt2k @ 1.9GHz

I'm surprised that ARM's brand new aggressive OOOE A15 core only
achieves about 40% of the single thread performance of a high end
desktop x86 processor from 7 years ago for such a relatively low
memory footprint workload like CPU2000.

With a 30% IPC improvements in A57 and higher frequencies (to allow for the larger thermal headroom in a notebook), you'll get decent performance.

You mean up to the towering heights of say, 50% of the performance
of a decent x86 processor of 5 years ago? :-D

[pilgrim68]

So you think Apple will be singing, Got 20 dollars in my pocket, Popped this chip at the thrift shop down the road, this is AWESOME.

Interesting

[wbmw]

IBC, your post is filled with false equivalency.

Tegra 4 already hits 1168 for SPECInt2k @ 1.9GHz

So what? You think this resets expectations for A57? First off, as you mentioned, it has the extra cache size to help it. It also has 12% extra frequency boost over the projection as well - and as you've seen on the A15 power-frequency curve, you pay for extra frequency with a LOT of extra power. Lastly, the IPC you apply at 30% is at the top of the generous range already provided when comparing SPEC benchmarks, and yet you seem to expect even more IPC on top of this.

Let's say most of what you wish for comes true, and a high powered 8-10W SOC with A57 cores gets the rough SPECint2k score of about ~2000. The best you've been able to argue is that it would get it to the range of a MacBook Air of 2010, when this theoretical machine starts shipping in 2014.

A late-2010 Macbook Air's performance is what I'm claiming is "good-enough" for most people. That was only ~2 years ago.

Well, nice math there, genius, given that A57 is still a couple of years out. It's about a 4 year gap, actually, and that's only achieved by growing the power envelope. It will still be a third of the performance of an equivalent x86 part, and maybe half to 3/4 of the performance of an x86 part scaled down to the same power envelope.

But the question remains - why would anyone in a year and a half from now be interested in spending any amount of money to "upgrade" their MacBook Air/Pro class machine from 2009-2010 to something with ~about the same performance, when there will be machines with Intel processors available at the same time?

You seem to forget the rule of incumbency, which requires that you offer a much better experience than the market leader in order to change the market. ARM can barely hope to catch up to the market leader's product from several years ago, let alone offer a better experience than what they will be shipping in 2014 - let alone, offer a "much" better experience.

That's why I offered AMD as an example. They succeeded for a short time when they offered a much higher performance product, but then they tried to win by offering a much lower priced product. You can view their earnings and stock price in both instances, and see which ended up the better strategy.