I Can Hazwell?

Interesting, this...

http://www.dailytech.com/article.aspx?newsid=29584

Seems to imply that Intel is showing off a 13" reference design using Haswell that is getting 13 hours of battery life in laptop form and 10 hours in tablet mode (detached from keyboard).

That is pretty impressive for the Core architechture chips.

Granted, there's no info from there on the specs of the reference design (how large is the battery? what's the weight of the tablet portion? how fast is the CPU running? etc.)

Regards,
SB
 
10W is still hellishly high for a tablet, but still mighty impressive for a core-series CPU.
 
I wouldn't say hellishly high. Acer with some help from Intel is using a regular Core i5 3317U (17W TDP) in their W700 and still manages to get ~7 hours of video out of that. Which is ~2-3 hours more than most ultrabooks. All in a tablet that weights ~2.1 lbs (it's lighter than my 11.6" Win7 based Atom tablet, ~2.25 pounds, which gets less battery life).

At 10W TDP, that should allow them to increase battery life and reduce the weight.

Granted the benefits will be greatest when compared to Ultrabooks and other vendor tablets. I suspect Intel has had Acer use very specific components in the W700 in order to reduce power consumption. Things that other OEMs will eventually benefit from as well.

Regards,
SB
 
Silent_Buddha said:
At 10W TDP, that should allow them to increase battery life and reduce the weight.
Click to expand...

TDP won't affect battery life in the non demanding loads though. It's only if you are stressing it, which isn't what video playback does.

The W700 has a 54WHr battery, and 7 hours mean 7.7W average power use. That's not very far from idle.

The new Clover Trail(Z2760) Atoms double the battery life over previous Atoms(like the one you have on the Tablet), mostly because the idle power management is improved by leaps and bounds.
 
In fairness, much of Haswell's additional component integration snippets surround major parts of power consumption -- a ton of it around overall chipset consumption and idle characteristics. So again, with TDP having little to no bearing on actual battery life, Haswell has a significant chance of giving us fantastic battery life enhancement.
 
I would think Broadwell is where we will see major changes in power usage. Even Anand has mentioned that the increase in performance with Haswell comes with a higher TDP for those parts .
 
eastmen said:
I would think Broadwell is where we will see major changes in power usage. Even Anand has mentioned that the increase in performance with Haswell comes with a higher TDP for those parts .
Click to expand...

But as Albuquerque said, TDP doesn't matter much to battery life, it's all about idle power.
 
Part of the reason that Haswell might have a higher TDP is their now fully integrated VRM circuitry as part of the package. At full load, the CPU is now going to put out quite a bit more heat simply due to the need for VRM cooling.

Now look at it from an idle perspective; with all that power control directly on-package, you've potentially one-upped the existing power gating capabilities. If your processor directly owns the VRM, imagine simply never even delivering the power across the substrate to an unneeded block, but instead turning it off at the source. You no longer have to suffer any leakage questions at the block that is gated; rather you can leave it at the VRM and call it done.

And for blocks that still need power? The capability to have as many as 320 VRM phases per CPU package means a ridiculous level of granularity and control for all the varying load levels that need to be supported at utmost efficiency. Only running at 5% load? Running at 105% load? The VRM is quite likely to be highly gated itself, so that only the necessary number of phases to deliver the required power need to be active at a given time. Potentially shutting off 19/20ths of your VRM because the CPU is twiddling thumbs makes for some righteous power savings at near-zero load conditions.

If I were a betting man, I would say the most significant portion of the touted "20x reduction in idle power draw over IVB**" claims from Intel can be directly laid at the feet of the integrated VRM.




** What the shit is a 20x reduction? Shouldn't this be 1/20th of the power draw? Or did they reduce it twenty times? I could be cynical and suggest that each time they reduced it, it was by 1/1000th of a watt. I hate stupid people who do this "5x slower" or "20x less" -- you don't multiply by a greater-than-one positive integer to achieve a result of less than you started. ARRRG
 
Last edited by a moderator:
Well the claims on the slide are against sandy bridge now ivy bridge. The cynic in me would even say it could be tech older than sandy bridge since the slide only states core i5 2011 vs 2013 targets. unless there is a new slide
Intel%204th%20Gen.JPG
Click to expand...

Anyway like I said while a reduction in idle power is nice I would rather wait for the die drop + new gpu .

Think about it the surface is out in two weeks. Haswells wont even ship till Q3 according to anand so the earliest I think we'd see a refresh is Q4. I'm sure the same time in 2013 broadwell will be ready. So you can wait almost a year now for a good x86 tablet and then have a much better chip come out a year after that. Or you can buy now and enjoy the tablet for a good 8-10 months before haswell comes and then continue using your tablet for 10-12 months till broadwell comes.

I'm choosing 2013 ivybridge then holiday 2014 / early 2015 broadwell. I can understand some people wanting to wait.

Remember I'm going from a 13.3 inch dual core neo at 1.6ghz with a 3200hd igp and a hd4330 dedicated. I get 3 hours of surfing the web if i'm lucky and maybe an hour and a half gaming. So the i5 will be a huge step up for me. Plus my gf gets my surface when I upgrade to the broadwell one so everyone will be happy.


Anyway intel is starting the 14nm transition and that's what broadwell will be on

http://www.xbitlabs.com/news/other/...tart_Making_14nm_Broadwell_Chips_in_2013.html

So I would expect early 2014 as the date they start rolling out for end users to buy.
 
Last edited by a moderator:
Ah, I see what you're saying. 2011 matches up with the volume production of Ivy Bridge, but actual shipments didn't start until 2012. The quotations floating around the 'net suggest that Intel called out Ivy Bridge in the presentation, I suppose in a verbal way. But the slide you have there is (purposefully?) non-specific in that regard.

There is also some conversation around the power reduction being linked only to standby C3 sleep state rather than a C0 (active) P-state (what we would consider power-on-idle). Your slide seems to indicate that we're comparing C3 sleep states, which is a different type of conversation entirely.

I guess the reality is "Wait and see!"
 
There is also this one

poweroptimizer2sm.jpg
Click to expand...

The clearly state that the 2nd gen (sandy bridge) cores are the base line in which they are comparing the 40th gen parts on. Ivybridge is 3rd gen

I see what your saying about the different states however.
 
Yup, that slide obviously settles that question -- we're talking 28nm Sandy versus 22nm Haswell. In that regard, the ~20% power reduction claim seems paltry to me. I cannot specifically understand why Intel thinks we should be excited for that...

The S3 power savings would be very useful for devices spending large quantities of time in standby, especially combined with claims of significant latency reduction in transitions between S0 and S3. It strikes me as very "phone like", in that I would expect a tap of the power button to provide nearly instantaneous response. Nevertheless, I am skeptical of significance of that advantage on tablet battery life.

I had CDW send me a Samsung ATIV tablets (the version equipped with a core i5, 4GB ram, Win8 Pro 64, digitizer and capacitive 1080P display) for a 30 day eval. As a bit of a Microsoft apologist from time to time, I have defended the Surface Pro's quoted battery life to several other naysayers in my office and in my social network by pointing out that it's a 1080P ultrabook with the addition of a capacitive touchscreen AND digitizer. As such, any rational person should expect a four or five hour battery life.

True to my own expectations, Samsung's version of the Surface Pro did indeed repeatedly survive for four and five hours per charge while using it as my regular PC. Unfortunately as a tablet device, four or five hours really never satisfied me even though my normal Thinkpad laptop likely would never do any better either. Ultimately, the 30 day eval (I give it back next week) convinced me that Surface Pro may not have as much use in my environment as I had hoped.

In light of what you're showing me, I'm far less convinced that Haswell will "save" the battery life situation.
 
Last edited by a moderator:
Albuquerque said:
In light of what you're showing me, I'm far less convinced that Haswell will "save" the battery life situation.
Click to expand...

That would be true for the notebook targetted Mobile-M chips, but the Mobile-U chips are likely the ones targetted at tablets. In that case, even taking the 20x claims with a large helping of salt, that should increase things significantly.

Regards,
SB
 
The battery life will get better that's for sure but we wont go from 5 hours to 10 hours. We might get another 1-2 hours out of it hopefully.

Like I said , I hope that broadwell brings a big reduction in power, we do get a micron drop with it , new gpu design and hopefully some improvements to the core designs .


The 5 hour battery life isn't to bad . the charger should be fast at 48w its not a trickle like an ipad.

The surface rt has a 24w power supply with a 32 Wh battery and takes 2 hours to charge. The surface pro has a 48w supply and a 42w power supply so it seems like charging times wont be bad at all. in comparison the ipad 3 takes 5 hours to charge.

recharge90.png



I believe the surface pro can also charge through usb so perhaps some of those external battery packs can be used too.
 
Last edited by a moderator:
Silent_Buddha said:
That would be true for the notebook targetted Mobile-M chips, but the Mobile-U chips are likely the ones targetted at tablets. In that case, even taking the 20x claims with a large helping of salt, that should increase things significantly.

Regards,
SB
Click to expand...
Yeah, but that "20x" claim was specifically linked to sleep state, not active-idle. That's the piece I was missing until eastmen posted up that slide earlier. I agree there will be power savings, but my estimate would be quite close to eastmen's hour or maybe two rather than a doubling. A 20% increase is "ok" for sure, but 20% on top of 4 or 5 hours is still a little short for a device that wants to be a tablet.

IMO, of course.
 
Albuquerque said:
Yeah, but that "20x" claim was specifically linked to sleep state, not active-idle.
Click to expand...

It won't be 20x reduction, due to the screen, but I know they got big potential there.

I have both a 5-inch Menlow device and a 10.1-inch Clover Trail device. The latter uses a MINIMUM 3.5W in idle with screen on while on the Clover Trail device I've seen it at 1.4W. Despite having similar TDP, the high idle power effectively limits how much battery life it can get. That was the issue with the pre-"new power management" days, even if the TDP was zero for the chip, the improvement on battery life would have been minimal.

The latter uses less power playing back videos in Youtube as the former does when mostly idle!
 
Well, Intel never made claims about the entire device using "20x less power" -- and again, how the hell do you multiply a value by a whole positive number and end up with a smaller number? I hate that shit. Anyway, you're of course exactly right -- the biggest power draw will be the screen, exacerbated by the fact that it's a touch screen.
 
You must log in or register to reply here.
Back
Top