Intel Core i7 For Laptops — First Benchmarks

Barence writes "PC Pro has benchmarked the first Intel Core i7 processors for laptops. The chips mark the debut of Intel's Turbo Boost technology, which ramps up the speed of the working cores if two or more cores are sitting unused. For the quad-core i7-820QM, this can take the stock speed of 1.73GHz up to a maximum of 3.06GHz. The 2D benchmarks show comparable performance to Core 2 Extreme chips running at 2.53GHz. Power consumption and processor temperature is dramatically lower, which should lead to significant improvements in laptop battery life."

Idle power consumption

[owlstead]

I really like it when chips have small idle power usage, and this chip seems to run pretty cool when it is not taxed. Intel always had the lead in manufacturing capability, and it seems that this is one of the nice results.

I'm really waiting for the day when you (can) just leave your computer on at all times. Most of the times the chips are doing nothing anyway, so why should it use any power? Where is the technology to switch off memory banks when they are not used? Just page the stuff to my SSD (yes, I'm talking about the future here). Why don't processors have a small power efficient core for running the OS and applications at idle? Gigabit ethernet is getting power saving functions as well, and Wifi N has power saving features as well. Having the computer almost idling without having the fan of my PSU or processor switch on should be a killer feature.

One thing missing seems to be software support. I don't like it when my laptop drains much power just because one core is using 100% power because of a friggin flash ad on one of the tabs in my browser. We need more ways of restricting processes to use as many resources. What use is a computer that runs on almost no power when idle when it is never idling? And we'll need OS support for cores with different feature sets as well.

Now give me the dual core...

[mejogid]

The article doesn't seem to suggest that this will really be enough to bring quad core laptops out of their current niche - we're talking an expensive machine which will clock in a bit over 3 hours battery life if you don't use its power, and potentially under an hour if you do. This would presumably be even worse with the higher clocked chip mentioned. I just don't feel there's much demand for such portable workstations - I can't see a good case for doing anything that processor intensive on the go. What does look very interesting is the 32nm dual core version - if they can carry over a comparable power consumption improvement to what they've achieved at the quad-core level that could be a very fast, very power efficient machine.

Re:Goobers still saying Core This and Core That

[uassholes]

It was an attempt to distract from the superiority of the AMD chips at that time, especially the Opteron.
What can you do when you are trying to keep X86 to 32 bits so only your Itanium is the sole 64 bit chip, when along comes AMD and creates a 64 bit x86 chip. You have no choice but to use AMD's 64 bit instruction set in your new 64 bit Pentium, AKA Xeon.
Oh, oh; AMD created a memory controller far more efficient than yours, OK copy that too.
Now Intel had caught up.

We have low end, now the high end...

[tetsukaze]

Where is the middle? Atom based equipment is changing how we define portable computers and is very exciting. These new chips are going to bring amazing power in a portable format. The problem for the average user is that these are two extremes that currently don't help them. The middle of the road laptop that can be used for everyday use has not had any major innovations or significant price drops for some time. I understand diversifying is important, but where is the new tech for that more middle of the road work load?

Re:Who needs that?

[Cornelius the Great]

Laptops have generally kept up with their desktop in terms of speed- as long as you're willing to pay. There are already quad-core laptops with dual GPUs (SLI or Crossfire) that would mop the floor with the majority of desktops. The only problem is, battery life is crap, and they're too hot to actually use on your lap (while gaming at least). Oh, and they cost 2-3 times as much as an equivalent desktop.

Re:Turbo Boost technology?

[Mad Merlin]

Maybe because (from what I can tell), IDA and DDA only boosted one core by ~200 MHz or less, TFS suggests that Turbo Boost can take one core of a 1.73 GHz chip to 3.06 GHz, which is substantially better. Maybe that's why people are noticing now?

Where are the asymetric multiprocessors?

[MMC Monster]

How about pairing one of these with an Intel Atom? The atom turns on cores within the Core I7 when it is pegged, and turns them off (potentially turning off the entire chip) when things quiet down.

Re:Turbo Boost technology?

[click2005]

It reminds me of the old PCs with a turbo button on the front.

Adobe Flash stops Vista from going into sleep mode

[Anonymous Coward]

Flash is even worse for power consumption than you think. On Vista if you have Flash loaded in your browser (maybe you watched youtube or an advertisement or something), it prevents Vista from going to into sleep mode per your power settings. Even if the video is over (e.g. youtube is showing the "replay?" button).

I'm guessing they did this because they had a suspend-resume bug that was hard to debug, and decided to burn lots of (our) coal instead of fixing their bug. Way to go Adobe!

Re:battery life?

[Colonel Korn]

Yep... the Turbo Boost is a great idea for desktops, it always gives you the maximum performance within a given thermal envelope. But to laptops, it's pretty much the anti-steedstep, making it spend as much power as possible when it's almost idle. However, it seems they didn't test the real minimum by disabling turbo. I'm assuming the laptops can control this from software, anything else would be silly. Sure, it'll also drop your performance from 3.06 to 1.73GHz but since power is roughly proportional with frequency squared it should also lower the CPU to about (1.73 / 3.06)^2 = 32% power consumption.

It's a good point that this seems to counter speedstep, but to some extent they work together. A 3.06 GHz frequency allows a particular computing task to be finished faster so the chip can fall back to the idle speedstep frequency (on my ~3 GHz desktop i-7 the idle speed is 1.2 GHz). Also, come to think of it, you mixed up the exponent in the CPU power equation. The power draw is proportional to frequency and to the square of voltage, not the other way around. So assuming CPU task completion time scales to the -1 power with frequency, which seems reasonable, the increased power draw for the higher frequency exactly cancels with the fact that processes will finish faster.

Running single-core

[jones_supa]

Excluding some special cases, I presume you would still get the best real-life performance by just running one core all times at that 3.06GHz speed.

Re:Turbo Boost technology?

[jawtheshark]

With Later models of PC tower units, there would have a couple of LED digits that displayed the clock speed (20/25/33/40/50/60 Mhz).

Actually, they didn't really "display" the speed. They displayed whatever the little jumpers on their back told them to display. Yeah, they were jumper controlled displays. You could set them to anything you wanted. I remember "reprogramming" one into displaying "Hi" and "Lo".

Re:Who needs that?

[TheRaven64]

I think you are failing to take into account economies of scale. For a long time, laptop components were more expensive than desktop equivalents because they had tighter constraints and because desktop components had much higher volumes. Last year, laptop sales passed desktop sales. This means that the highest-volume parts are now made for laptops, not for desktops. If this trend continues, then expect to see a premium on desktop parts to make up for the low volumes in the next few years. You could still build a desktop out of laptop parts, of course, but if all of the components are designed for laptops you won't be able to squeeze much extra power out of them by putting them in a bigger case.