bit-tech.net

Intel teases 4.5W Haswell Core chip

Intel teases 4.5W Haswell Core chip

Intel's upcoming low-power Haswell chip boasts a 4.5W scenario design power (SDP,) but the company is silent on specifications and full TDP.

Intel has confirmed plans to release an ultra-low power processor in its Haswell fourth-generation Core family, boasting a 4.5W power draw and targeting fanless convertible devices and tablets.

Based around a new microarchitecture, Haswell-based Core processors offer several improvements over their Ivy Bridge counterparts in the fields of performance and power-saving. While these make surprising little difference at the high-end, as our review demonstrates, at the lower end they can make a real impact - and have become central to Intel's battle against Cambridge-based rival ARM in the mobile sector.

Accordingly, Intel has confirmed that it is to launch a Core-branded Haswell part in the second half of this year, aimed directly at convertibles and tablets. The key feature, the company has claimed, is that the new chip will allow for fanless designs - critical to creating slim and light devices that can truly compete with those based around ARM's architecture.

Sadly, the company is being somewhat reticent to release firm details surrounding the part. It hasn't confirmed clock speed or the number of cores, for example, and while it's possible to guess that the processor will be based around a dual-core design it would be exactly that - a guess. What the company has offered is a power draw claim of just 4.5W - a not-inconsiderable saving over the company's current lowest-power Haswell part at 6W.

There's a catch, however: in recent releases, Intel has moved away from the traditional metric of thermal design profile (TDP) to something it calls scenario design power or, sometimes, scenario-driven power (SDP.) While the former offers a figure for peak heat output - critical for designing cooling systems that can cope with sustained load - the latter merely offers an 'average' figure for 'typical' workloads. As you might imagine, there's room for a bit of marketing in the figure: Intel has previously nearly halved a chip's TDP to reach a headline-friendly SDP, as with the Core i5-3439Y which boasts an SDP of 7W but a TDP of 13W.

As a result, and with no firm figures from Intel, it's possible to guesstimate the true TDP of the upcoming chip at somewhere in the region of 7-9W. While that's just about on the border of what can be reasonable cooled by a passive tablet design, it seems likely that if manufacturers want to avoid overheating they'll need to set tight thermal controls - potentially clocking the chip down, or disabling one of the cores, when the workload gets too heavy.

Intel has not yet set pricing or confirmed customers for the part, but claims it will be releasing the processor in the second half of the year - meaning it's possible we'll see the first retail products to use the chip on the market by the end of the year or early 2014.

8 Comments

Discuss in the forums Reply
law99 24th July 2013, 11:56 Quote
Have to look out for this SDP nonsense. It is a shame they feel the need to market it as such... seeing as consumers will no doubt be buying it in an already pre-assembled device which will no doubt be reviewed against its applicable competitors. Ergo, it doesn't matter what you do to your figures to hide the natural results, reviews will either say: "This thing is hot and thirsty like" or "This is cool and everlasting" or a combination of the two.
r3loaded 24th July 2013, 14:27 Quote
It's not complete nonsense - the expectation is that 99% of the time a tablet will not be maxing everything out as a desktop would as the usage demands will be low (because it's a tablet).

What Intel are trying to is allow OEMs to use a far more powerful core design in a tablet enclosure that will stay cool and sip power (<4.5W) most of the time due to the low load. If additional, performance is needed it can be provided for short bursts (up to 11W) until the tablet gets too hot, upon which it will throttle back. The idea is that even when throttled back to 4.5W, a Haswell chip will still beat the pants off Bay Trail as well as any current ARM chip.
law99 24th July 2013, 16:30 Quote
Whilst I partially agree, now we are arguing about what is typical usage. Regardless of what is said, the lower TDP, with equivalent performance, if not better than before, would have put them in the same position.

The last thing I'd think you'd want to encourage is to design the products to be on the edge of their thermal limits using SDP as your metric - which I'm sure no sane product would do unless they are thinking of incorporating their own artificial limits.

Meanwhile, this term is already branded about. So consumers see it without really needing to... they just need a review of the end product. Since reading this, and reading this, you'll notice the headline always read something along the line of "Intel releasing 4.5watt chip" which they aren't - it's the potential to be a 4.5watt chip.

EDIT: It is still great none the less, don't get me wrong. I just think this is becoming a PR/marketing regurgitation exercise. Which if true, is at least vetted by Gareth who points out and explains the 4.5watt SDP figure
Gareth Halfacree 24th July 2013, 16:40 Quote
Quote:
Originally Posted by law99
Which if true, is at least vetted by Gareth who points out and explains the 4.5watt SDP figure
I live to serve.
azazel1024 24th July 2013, 20:51 Quote
Sort of hit on the point. The point being that SDP is what it is going to be using most of the time. Load down the GPU and CPU and it is going to hit much higher power levels. IIRC it was slipped that the TDP of the chips is 11.5w. Again, I think.

So you have two situations here. You can utilize the chips in a chasis that can dissipate just 4.5w of power, because that is what it'll use "normally", but you are pretty much going to force extreme throttling the instant you hit heavy workloads. Or alternately, you can design a chasis that can cool more than 4.5w, which means under normal circumstances you'll have no issues and will be able to use heavier than normal workload situations with any problems as well as some turboing before things have to be pulled back.

Finally you could design a chasis that can handle the full 11.5w TDP, in which case the chip can run pretty freely, but you can also know power management wise, that it is generally only going to be using 4.5w (as most platform designers are concerned more about typical power use when designing battery size, but look at TDP more for chasis heat dissipation).
RedFlames 24th July 2013, 20:58 Quote
how is this different to Turbo on desktop/laptop processors? [kinda genuine question]

it spends the bulk of it's time at a given speed, overclocking itself by a set margin when extra oomph is needed within temp limits...
Gareth Halfacree 24th July 2013, 21:09 Quote
Quote:
Originally Posted by azazel1024
IIRC it was slipped that the TDP of the chips is 11.5w.
Ouch, really? I was guessing 9W, tops.
fdbh96 24th July 2013, 22:39 Quote
The problem I see is that when Im using low load apps on my ipad, it tends to be the case that I never need the extra power. However, if I'm for example playing a game which is fairly high load then Im going to want the power for longer than 'until it gets warm'.
Log in

You are not logged in, please login with your forum account below. If you don't already have an account please register to start contributing.



Discuss in the forums