Haswell Goes Fanless: 4.5W SDP Parts in Limited Volumes Later this Yearby Anand Lal Shimpi on July 23, 2013 3:00 PM EST
At the beginning of the year, Intel introduced the first 7W SDP Ivy Bridge parts in limited quantities with the promise that Haswell would bring even lower power options. We didn't see much OEM adoption of the 7W SDP IVB SKUs, but at the Haswell launch Intel made good on its promise to follow up with an even lower power version. The first Haswell Y-series parts carried a 6W SDP (and 11.5W TDP), and that value included the on-package PCH. Today Intel is announcing an even lower power version, once again in limited quantities. By the end of this year Haswell will be offered in a 4.5W SDP version (still 11.5W TDP).
I went through the SDP discussion earlier this year. But in short, Intel calculates TDP by looking at average power through a set of benchmarks that tend to include some of the worst case offenders on a desktop or notebook PC. The SDP thermal rating is specific only to Y-series parts and uses a lesser set of benchmarks, more appropriate for a thin tablet chassis, to determine average power. If you run Furmark on one of these parts and the OEM building the system does nothing to thermally manage the platform, it'll dissipate 11.5W. If you use it like a tablet, you should see 4.5W (or 6W depending on the SKU) as the average power dissipated. Realistically the 11.5W value can be boosted up to, but not sustained.
Moving from 6W SDP down to 4.5W SDP is a big deal because it enables something huge: fanless Haswell tablet designs. At 6W Intel recommends some sort of active cooling within the chassis, but at 4.5W it's possible to go completely fanless.
Earlier this week I pointed out that Haswell ULT finally got its platform power story in order, enabling ARM-like battery life on some light tablet workloads. The benefit to tossing Haswell into a tablet is the ability to have a greater dynamic range of performance than what's currently available from the ARM ecosystem. Great battery life when you're running tablet workloads, but the ability to get notebook-like performance when you need it as well.
Just because it's possible to get Haswell in a fanless tablet however, doesn't mean that it's going to happen. Uptake of the 7W SDP Ivy Bridge parts was pretty much non-existent from what I've seen. Intel's use of the phrase "limited volumes" gives us an indication that we shouldn't expect anything major here either. In my Haswell ULT tablet experiment I noted that although CPU power efficiency looked good, power consumption during video decode was pretty bad compared to what you can get from modern ARM SoCs.
The other thing worth thinking about is the total board area required by a Haswell implementation vs. effectively a high-end ARM smartphone SoC. Haswell may be capable of fitting into an iPad-like chassis, but can an OEM shrink the PCB enough to be a comparable size? If the answer is no (and I'm assuming that's the case given how much bigger the Haswell ULT/ULX package is than what you get from ARM), then a 4.5W SDP Haswell ULX based tablet would either have to be larger than an ARM equivalent or you'd get it with a smaller battery.
It's good to see Intel continuing to push its Core family of microprocessors down the power scale, but I feel like there's a lot more room for integration to happen before we can really speak about Core in the same breath as an ARM SoC. The good news is that some adventurous OEMs can likely build something iPad sized out of Haswell this year if they want to, but I'm still expecting the best attempts to come with Broadwell (with the exception of the Haswell Surface Pro of course, depending on what Microsoft decides to do).
Post Your CommentPlease log in or sign up to comment.
View All Comments
althaz - Wednesday, July 24, 2013 - linkBecause they perform worse and use more power. Seeing as a lot of the comments you see here are people complaining that it's still too much power/not enough performance (and I agree on both fronts - i5 2Ghz equivelant performance at 2-3w is what we really want), you can see why AMD is not really in the picture at the moment. They are on the way, but simply trail Intel in every area except cost. At the moment, cost isn't the biggest issue for Core, it's power consumption, heat and performance.
HisDivineOrder - Wednesday, July 24, 2013 - linkI imagine a Surface and a Razer Edge showing up with something like this. Otherwise, eh, I doubt it'll get much traction with OEM's.
Krysto - Wednesday, July 24, 2013 - linkI thought Razer Edge was about gaming? This will get not only much lower CPU performance, but lower GPU performance compared to the notebook versions of Haswell, too.
Krysto - Wednesday, July 24, 2013 - linkThis is going to be so annoying. It's yet again another step for Intel in its recent crusade to mislead the public when it comes to comparing its chips against ARM chips.
How the hell are they even calculating the average? By adding 20h of idle time (close to 0W consumption), 4 hours of very low usage on-screen time, and 20 minutes of high usage? And then they get this "average" SDP number?
So what happens if instead of those 20 minutes of high usage that they calculate, I use the tablet for 3 hours continuously playing a 3D game or watching a high quality 1080p movie? Will I get 3rd degree burns on my hands?
Yeah, I'm exaggerating, but I think you understand where I'm going with this. Intel is going to be very misleading about how they calculate this "average", and it may differ a lot from how people will really use the devices with these chips inside, and real world average could be a lot higher.
The TDP is a KNOWN and good metric, because you know exactly how hot the tablet is going to get if it reaches peak performance. Intel is only messing with this to further mislead us about how good their chips are vs ARM chips. They're doing this because they're desperate and see the writing on the wall.
ironargonaut - Wednesday, July 24, 2013 - linkWhat is the TDP of an ARM processor?
ironargonaut - Wednesday, July 24, 2013 - linkTried to answer my own question, since this is a "KNOWN and good metric". Found this "And then there’s the 32nm Exynos 5, based on ARM’s brand new Cortex-A15. From the get-go we’ve known that Cortex-A15 would be a monster — and indeed, it’s the first ARM chip to comfortably outperform an x86 design from Intel — but at what cost? Anand’s testing shows that the Cortex-A15 chews through power at an impressive rate, comfortably consuming more than 4 watts during load — and that’s with Samsung throttling the CPU and GPU when it hits 4W; the actual TDP allows for up to 8W." http://www.extremetech.com/computing/144778-atom-v...
So, Cortex throttles to 4W, and if Intel throttles to 4.5W. Seems fairly even to me. Next question which has better performance at that wattage? Anyone got benchmarks of this?
Alexvrb - Thursday, July 25, 2013 - linkThey can reduce power consumption slightly at best, otherwise they'll suffer from a performance regression.
ant6n - Saturday, July 27, 2013 - linkI'd like to see a fanless ultraportable laptop with Haswell.
Why don't they put the cpu behind the screen so that the screen part of the laptop gets hot. That way the part that's sitting on the lap can hold the battery and stay cold.
Kogies - Saturday, July 27, 2013 - linkI would love one of these stuck on a Thin-ITX motherboard. Low power and the possibility of a ramp up to higher performance would be great in a silent home server, Atom is not yet 64-bit and I can only imagine this may still have a greater dynamic range for performance than new Silvermont.
towusop - Sunday, July 28, 2013 - linkSimple question alert! This article mentions 6W SDP and the 11.5W TDP. What do SDP and TDP stand for?