Intel announces specs of first Broadwell chips

Quizzical

http://www.anandtech.com/show/8475/intels-core-m-strategy-cpu-specifications-for-9mm-fanless-tablets

The notable news is that they're going to be 4.5 W tablet chips.  For comparison, the AMD A10 Micro-6700T is also a 4.5 W tablet chip.  Intel doesn't announce the TDP of their Bay Trail chips, but the tablet-oriented ones are probably in that range, too.

This isn't the first effort at putting a high power, high performance CPU architecture into a tablet.  There have been various efforts at putting 17 W and 15 W chips into tablets, such as the Microsoft Surface Pro, the Razer Edge, and... well, that's about it.  The Microsoft Surface Pro 1, 2, and 3 have all been ridiculous devices, and the Razer Edge was a "tablet" that was thicker than a lot of clamshell laptops, but such are the sacrifices of putting a 17 W or 15 W chip into a tablet.  Intel tried making 11.5 W versions of Haswell, but no one bit on them for tablets as far as I can tell.

This is, however, the first effort at putting a high power, high performance CPU architecture into a tablet with a reasonable TDP.  So what sacrifices did they have to make?  Let's start with a stock CPU clock speed of 800 MHz.  The stock GPU clock speed is a meager 100 MHz.  Yes, they both have turbo, and if everything else is idle save for a single CPU core, this will give you the best single-threaded CPU performance in a 4.5 W TDP, and probably by a wide margin.  Intel is relying on that as the reason for the chip to exist.

But the idea of turbo is that if a lot of stuff is idle, you can run the other stuff faster.  Try to use the CPU and GPU at the same time and you don't get so much in the way of turbo.  So it's not going to be viable for games, or at least, not as good of an option as the A10 Micro-6700T.

But the chip probably isn't just for tablets.  Intel and AMD commonly have a single die serve several markets clocked differently.  Seeing what the one platform has can be instructive:  the first Broadwell chips will have PCI Express x4 of some unspecified generation, as well as PCI Express 2.0 x4 in the chipset.  x16?  Nope, not even close.  So don't try to attach a video card.

Now, a discrete video card in a tablet is a dumb idea, of course.  But if you clock the same chip higher and put it in a laptop, a discrete video card makes sense.  If you want to sell it in a desktop, a discrete video card needs to be an option or it's a non-starter for gamers.  And this die will be a non-starter for gamers, however it's clocked, as the integrated graphics are very low end and the PCI Express lanes aren't there to add a discrete video card.  Perhaps that's why persistent rumors said that Broadwell wouldn't bother coming to desktops at all.  This initial die sure shouldn't.

Now, that doesn't mean that no Broadwell architecture die will be suitable for desktops.  There will presumably be a quad core coming at some point, and there's no reason why it can't have a PCI Express 3.0 x16 connection built in--just like Ivy Bridge and Haswell.

Intel could also make a separate dual core die with the necessary PCI Express connectivity if so inclined.  They might well do so:  PCI Express lanes burn power, and 16 PCI Express 3.0 lanes trying to fit inside a 4.5 W TDP makes means you have to cut back other stuff that much further.  AMD's Socket AM1 motherboards (desktop version of Kabini) are also limited to PCI Express 2.0 x4, but that hardly means that AMD is trying to kill discrete video cards.

Another notable thing is the graphics API compatibility:  DirectX 11.2 and OpenGL 4.2.  So basically, Intel is still trailing behind Radeon HD 5000 and GeForce 400 series cards on that count, let alone the latest and greatest.  And this matters, too, if you want to play games:  AMD's GCN architecture and Nvidia's Kepler support the OpenGL extensions that allow reducing CPU overhead in rendering, and will presumably support the analogous portions of DirectX 12.  Broadwell doesn't.  If your CPU is 50% faster than the competition, but has to do twice as much work to render the same frame, it still loses.

And then comes the price tag:  $281 for the chip.  You can get an entire laptop for less than that.  Given a choice between an A10 Micro-6700T and a Broadwell Y/Core M chip in a tablet at the same price, you could make a good case for either one.  With the Intel APU costing over $200 more, likely resulting in a retail price gap closer to $300 after subsequent vendors take their markup, this is going to be a tough sell.  You can justify charging an extra $300 if your product is vastly superior to the competition, but Intel's isn't.

Ridelynn

The only issue I have with your write up: All of it is true, except your looking at it from the perspective of a gamer.

These tablets/handhelds/hybrid devices - the market is definitely not video games.
It's people on Facebook, shooting photos and videos, updating blogs, looking up recipes online, making small changes to a business document, showing off a slide show, etc. They only need enough power in order to drive Flash/HTML5-type applications, and the most taxing thing any of these hybrids, with their target audience in mind, will do is play back Youtube HD videos in full screen on whatever resolution screens they happen to get paired with.

These tablets/hybrids make some sense in that context. In that context, the Surface Pro 3 has been well received.

But if you just look at them through the lens as a video game or high performance computing platform, your trying to pound the square peg into a round hole. Just due to physics a tablet-style device will never compete with something you can stick on your desktop. And by the time we have something hand-held that can compete with what we have in a baseline gaming machine today, that baseline will have shifted so much on the desktop it still won't be a fair comparison. Not at least until we have enough wireless bandwidth and availability that can stream/remote run all these applications from desktop/cloud servers.

Quizzical

I agree that most people who buy a tablet buy it for things that aren't terribly demanding.  But if that's the intended use, why pay an extra $300+ to get a Broadwell tablet, when much cheaper tablets can do things that aren't terribly demanding nearly as well?  It might be a nice chip if it were competitive on price, but it isn't.

As for the Surface Pro being well received, how much money is Microsoft's Surface division losing, again?  That wouldn't happen if it were selling well.  Apple and Samsung haven't exactly been losing boatloads of money selling tablets.  It's hard to compete when you're paying $300 for the main chip and your competitors are paying $30--especially when, as you note, for most consumers, the $30 chip will do what they want.  The Surface Pro also needs outlandish measures to handle cooling--measures that aren't necessary in a lower power chip--and that also adds to the cost of production.

Though I'd claim that performance in a tablet still matters to some degree.  I don't think it's a coincidence that Apple, the company usually pushing the highest performance into a tablet-friendly TDP, is also the one selling the most tablets.

Ridelynn

Haswell M being a $250+ part is a mistake - I agree there. The tablet market these are going into - with the exceptions of Apple and Microsoft, is very cut throat. Amazon is selling Kindle Fires under cost, and it's not uncommon to find "free" tablets given away as promotional items. There isn't much margin, and there isn't a lot of room for a premium product when the baseline tablet has become a commodity item.

Surface Pro 3 - It's selling better than the 1 or 2. That isn't saying much - they are marketing it as a Premium item in a market where most people are used to seeing tablets sell for well under $300. That's something only Apple has really cracked, and I don't think Microsoft can replicate it. I just commented that it was well reviewed and received. I don't know that the "Windows on a tablet" market is that big to begin with (and I think right now most of that is due to Windows 8 than the hardware).

Apple being the leading tablet manufacturer - that isn't because they have the highest performance tablet, they very rarely do,a nd when they do, it isn't long until someone comes out with something that's just a hair faster for the sake of saying they have the fastest. They are very much an "Enough Performance" manufacturer, they don't chase the "Top Performance" crown very often. Apple's entire mentality is to meld the entire experience - from the store you buy the thing at to the box the thing comes in to the hardware to the software to the way you get new software, and to polish each one of these aspects as well as they can. They don't get all of it perfect, but the end-to-end experience that Apple offers is what most people buy into. That, or the "cool" factor - there is still a lot of that around, and the Apple brand carries a lot of marketing weight still.

A lot of people have been rumoring that Apple will shift their low end computer line over to their ARM chips soon, and this is the reason they were pushing 64-bit ARM, and are really advancing their A-line... to get something that can compete favoriblity against Intel in the low-end Macbook/Air lineup, and possibly even the iMac/Mini lineup. Like you say - paying $30 for a CPU is a lot better than $250+, particularly if your competing in the low-end market where the margin is single-digit as it is.