Yoga Pro 3 review: Broadwell is a mixed blessing

The design is ever more exotic, but the performance is worse.

Read the whole story

 

[LinkSatonaka]

Has Lenovo released a laptop/device in the last 2 years that didn't in some way re-arrange the keyboard or change how the trackpad functions? Has anyone ever appreciated these changes? Seriously, reply to me if you are familiar with the older trackpad and like the newer ones, or if you like the scrambled keyboard layouts.

Because I'd been looking for a laptop for the past year and I returned not one, but two Lenovo laptops after a month (each) of using because their "innovations" on the keyboard and trackpad are absolutely intolerable. I can justify buying none of the newer laptops- and anything I could find without these recent "innovations" is just a bit too old (like the first generation carbon X1).

For those wondering why fewer keys is a trend now (and this affects all manufacturers), I'm fairly sure it's because the less things there are, the cheaper it is to mass produce. But, in addition to constantly "reinventing" the keyboard and touchpad, Lenovo doesn't know when to stop cheapening the production (and therefore quality) of their products! The trackpoint is literally unusable without its physical buttons Lenovo! Someone realized that with the old carbon X1- but you still dropped the ball on the carbon ultrabook and several other models...

/rant

 

[name99]

Callitrax[/url]":3gxc0ymq]

melgross[/url]":3gxc0ymq]

Callitrax[/url]":3gxc0ymq]

Joriarty[/url]":3gxc0ymq]
Also - although pricing for the Core M isn't public, I've read it's in the $300 range.

So it seems that A8X is an order of magnitude cheaper, certainly consumes less power, and might also be faster.

ARM is kicking some serious ass. It's a pity that architecture changes aren't an entirely smooth affair.

Is the A8X cheaper? The lowest price you can get one is $499. Yeah you get a lot more with that but you also get the Apple high margin markup at that point. With the Intel based systems the highest margin part is the CPU where the markup is applied earlier in the supply chain. Given that the A8X has 3 Billion transisters, and the die sizes seem to be in the same range, I'm doubting that the production cost of the A8X Chips is markedly different from that of the Core M, the markup is just applied at a different point in the chain. (Short version, analyzing cost structures is a pain, but the chips shouldn't really be materially different in production cost)

And we don't inherently know that it uses less power. I've never seen real power measurements of the IPad SOCs under load, and there are so many other power draws in an operating computer that it can be hard to compare. But Intel chips have been shown to work in the same power ranges as other ARM chips.

The iPad may cost $499, but the SoC is believed to cost about $30. Apple doesn't get a "high end" markup, they make a good profit, unlike most other manufacturers who are barely making any profit on their machines. We've known that for many years, it's why we've seen so much bloatware, which has been where most of the profit for Windows machines have come from. But this is an expensive machine.

We can look at the price of the Surface Pro 3 to see that using an x86 chip raises the price substantially. The low price Surface Pro 3, at $799, sans the required keyboard, uses an ultra low power i3, while the better performing model at $999 uses the i5 version. While they need to be spec'd up because they run Windows, which needs far more RAM and storage than do Android or iOS, a large part of that higher pricing are the x86 chips inside.

My point was that we don't know WHAT the A8X costs, because nobody can buy one. $30 dollars is the number people like to throw around for the high end ARM chips. And I don't know that may even be what Apple pays TSMC to fab it. Also remember its a lot bigger than the regular A8, so if the A8 is $30 this may be more like $40.

But if Apple were to sell them alone? They would charge a lot more than that. Why? Because they could. Same reason Intel charges what they do for their mobile processors. Production wise the Core M should be relatively similar to the I3 (Same CPU and GPU config) which sells for as low as $100. The Core M will be a little more because it is packaged the extra southbridge chip - which is much smaller and on a process intel is really good at by now, and more CPUs won't make the cut for power reasons (although those just get sold as higher TDP units).

But production wise the A8X and the Come M shouldn't be too dissimilar. Apple is different from the other ARM vendors, they make bigger chips to drive lower power at the same or better performance. Since die size is the main driver of cost in the actual production of a CPU trying to judge to cost of Apple chips by using those from Qualcomm or Samsung as analogues is not ideal. The A8X is about 123mm^2. The Core M, well I haven't seen a die size yet, but its probably in the same vicinity - bigger versions of Haswell were in the 177-183mm^2 range with more cores or GPU units, and on a bigger process, so Core M should be much smaller and probably fairly similar to A8X. (See the image of Haswell and Broadwell dies here - the Core M is much smaller).

Basically the cost of producing those chips should be in the same ballpark, but comparing how much they "cost" the consumer is borderline impossible. The companies are very dissimilar, Apple sell systems and not chips, Intel chips but no systems. They've both dumped huge amounts of intellectual capital into developing the chips which is of coarse hard to quantify in the cost contribution. And both companies like high profit margins so they both have big markups on what they sell. Yes tablet and variants with the Intel chips sell for more than a base IPad - but they also have other things driving up costs - large real SSDs, more RAM, etc.

Short version, the A8X and Core M are going to be similar in production costs, the Core M is probably more, but its maybe 50%, not the order of magnitude that some might suggest. After that its all business decisions and differences that influence cost to user. Yes the Intel CPU is a large fraction of the cost of a system using it. But given that Apple sells at large profit margins and their SoCs are one of the prime differentiators from other products (along with IOS) its possible to attribute some of that selling price to the CPU.

You miss the point. Apple can survive on "selling" its chips at $30. (Or more precisely TSMC can survive at that level, and Apple can survive paying the design costs.) It also seems plausible, now that Apple have shown it can be done, that ARM and Qualcomm will strive for the same performance points (nV kinda sorta seems to have already, maybe also AMD).

Meanwhile Intel very definitely CAN NOT survive on selling chips at that price...

Look, the more intelligent commenters have ALWAYS said that Intel's problem is not the area or power cost of x86, it is the design/verification cost. THAT is where Intel has to pay a fortune compared to ARM. That fortune was an acceptable cost of business when there was no serious competition; but those days of no serious competition are over.

 

[Jo King]

Thanks Peter for the article. I'm quite curious about the AC Adapter ~ could you post a picture of it? Are they using USB Power Delivery 1.0? Thanks!

 

[Cognac]

MisterAV[/url]":1qn7za1s]Seeing the hinge I wonder how the communication goes from the screen to the bottom part...

Magic

 

[foxyshadis]

name99[/url]":g9e5qjkb]You miss the point. Apple can survive on "selling" its chips at $30. (Or more precisely TSMC can survive at that level, and Apple can survive paying the design costs.) It also seems plausible, now that Apple have shown it can be done, that ARM and Qualcomm will strive for the same performance points (nV kinda sorta seems to have already, maybe also AMD).

Meanwhile Intel very definitely CAN NOT survive on selling chips at that price...

Look, the more intelligent commenters have ALWAYS said that Intel's problem is not the area or power cost of x86, it is the design/verification cost. THAT is where Intel has to pay a fortune compared to ARM. That fortune was an acceptable cost of business when there was no serious competition; but those days of no serious competition are over.

I think it's more likely that Intel's going to start pumping out more Celerons (or do they just call them all Pentiums now?) as quickly and cheaply as they can, reusing masks and old process as much as possible, to keep the margins up. The main innovations will be continued power-gating to the point that only the few bytes of memory being accessed are powered up.

My concern is, will this lead to a death spiral in processor advancement? Once the low end is good enough, will the small base justify further investment, and will corporate customers be willing to pay for the investment instead of expanding horizontally? I don't want to see another P4 era stagnation.

 

[protomech]

Callitrax[/url]":1zmko5s2]But production wise the A8X and the Come M shouldn't be too dissimilar. Apple is different from the other ARM vendors, they make bigger chips to drive lower power at the same or better performance. Since die size is the main driver of cost in the actual production of a CPU trying to judge to cost of Apple chips by using those from Qualcomm or Samsung as analogues is not ideal. The A8X is about 123mm^2. The Core M, well I haven't seen a die size yet, but its probably in the same vicinity - bigger versions of Haswell were in the 177-183mm^2 range with more cores or GPU units, and on a bigger process, so Core M should be much smaller and probably fairly similar to A8X. (See the image of Haswell and Broadwell dies here - the Core M is much smaller).

M-5Y70 is 82mm^2 per this article.
http://www.pcper.com/reviews/Processors ... roadwell-Y

Apple trades higher die size (3B transistors) and cost for improved power efficiency; the 20 nm A8X is only slightly smaller than the 22 nm i5 haswell, so the 14 nm Core M is substantially smaller both in die size and transistor count (1.3B transistors).

It's curious how poorly the Yoga Pro 3 performs, considering it is actively cooled. I'm curious if a cTDP-up 6W version of Core M would work well in the rumored 12" retina macbook air.

Intel's pricing model is also curious. Core M in other previews appears to have excellent performance and price; but it's also rumored to cost $300 in quantity. That's a pretty big cost constraint for OEMs, who are no doubt eyeing ARM SOCs with somewhat lower performance and an order of magnitude lower price tag..

 

[name99]

foxyshadis[/url]"yju4dez]

name99[/url]"yju4dez]You miss the point. Apple can survive on "selling" its chips at $30. (Or more precisely TSMC can survive at that level, and Apple can survive paying the design costs.) It also seems plausible, now that Apple have shown it can be done, that ARM and Qualcomm will strive for the same performance points (nV kinda sorta seems to have already, maybe also AMD).

Meanwhile Intel very definitely CAN NOT survive on selling chips at that price...

Look, the more intelligent commenters have ALWAYS said that Intel's problem is not the area or power cost of x86, it is the design/verification cost. THAT is where Intel has to pay a fortune compared to ARM. That fortune was an acceptable cost of business when there was no serious competition; but those days of no serious competition are over.

I think it's more likely that Intel's going to start pumping out more Celerons (or do they just call them all Pentiums now?) as quickly and cheaply as they can, reusing masks and old process as much as possible, to keep the margins up. The main innovations will be continued power-gating to the point that only the few bytes of memory being accessed are powered up.

My concern is, will this lead to a death spiral in processor advancement? Once the low end is good enough, will the small base justify further investment, and will corporate customers be willing to pay for the investment instead of expanding horizontally? I don't want to see another P4 era stagnation.

They're already doing that. Original Phi was based on a rehashed Pentium design, Quark is a warmed up 486 design. Hasn't done them any good.
Basically they are trying to design CPUs using roman numerals, while ARM et al are doing the same thing using arabic numerals --- it's just a lot harder for Intel to keep up when they insist on solving a problem that is so much harder than the problem everyone else is solving.

And putting a break on the speed at which they upgrade their CPUs ain't gonna help. The idea of Apple switching to ARM based OSX has moved from crazy to eminently plausible. The A8X would not make a satisfactory MacBook Air CPU, let alone iMac. But it wasn't designed for that task either. More important is that, given the skill Apple has shown in this area, it's not implausible to believe that Apple COULD design an iMac-level chip, happy to burn 5x the power of the an A8X, targeting 3.5GHz or so, 6 cores or so. Right now they have no compelling reason to do so. But Intel failing to deliver a steady stream of improvements DOES give them a compelling reason.

And once again, as goes Apple so (eventually, after a few years and struggles) goes the rest of the Industry. The other ARM vendors will experiment with similar chips, and if MS refuse to ship a "full" ARM Windows, they'll just see Chrome and Android move upstream, starting with their specialty markets like cash registers and medical equipment, where no-one really cares about "real" Windows compatibility.

 

[korg250]

The ThinkPad X300 had the finest keyboard layout of all time. I wish Lenovo would bring it back.

Everybody is complaining about those new keyboard's layouts - specially in the Thinkpad line - and they continue to ignore and change it.

Why Lenovo, why?!!

 

[Kristofff]

I hope they use a different screen than the 3200×1800 screen from the Yoga 2 Pro. That screen has yellows that appear slightly brownish to such a degree that I've seen a handful sent in for repairs because the owners thought the screen was defective. Comparing it to the normal lesser screen versions you can see the difference and it is painfully clear.

The hinge is incredible. It makes a perfect arc depending of the angle it is bent. Whatever engineer designed it is a genius. The LCD cable is a ribbon cable now instead of the rounded cables, which I think will mean less cable failures at the hinge. No more hinges separating from the assemblies, either, there are a ton of screws mounting it to the top cover and the LCD cover.

 

[Shuffy]

Oh goddammit.

Anyone who takes out the F# row needs to be thrown under a bus.

 

[s4a]

They shaved off 1mm in thickness out of still pretty fat device and added $500-700 to the price tag. Nice job

 

[Tiger33]

Price of the Yoga Pro 3 as dropped now

Do you think this machine is good enough for development applications?

I.e Visual Studio 2013/2015 and MS Sql ?

or should i look at something else.

Was looking at the Yoga pro 2 but man does it have wireless issues..

 

[foxyshadis]

Tiger33[/url]":351pcqfs]Price of the Yoga Pro 3 as dropped now

Do you think this machine is good enough for development applications?

I.e Visual Studio 2013/2015 and MS Sql ?

or should i look at something else.

Was looking at the Yoga pro 2 but man does it have wireless issues..

With 8 GB of memory and a huge SSD, you'll have a pretty killer dev system. What it won't be suited for is 3D games (or heavy use of OpenCL) and video encoding.

 

[Tiger33]

Thanks

was just a bit worried about the Intel Core M.

everything else on the Yoga seems fine

even compared tot he XPS (new) - Yoga is now cheaper than XPS... (in the UK anyway)

 

[foxyshadis]

The CPU is as powerful as an i5, except that it'll throttle itself earlier (this is BIOS adjustable, I have no idea if Yoga lets you). The GPU isn't anywhere near as powerful, that's the big downside. Intellisense and compiling need quick bursts of power, but not as often long sustained power.

 

[Tiger33]

Thanks buddy

much appreciated

started working away with the Yoga pro 3 and visual studio and so far so good. no problems really.


just turn on high performance mode and away i go ;-)
*it does take a while to load projects though in visual studio but no biggie (visual studio scales great on the display , as does sql server)


browsing is a little jaggy though.. (two finger scroll not that great ... )

overall pretty happy with it - battery seems fine, amazingly quick boot up time though!!! - like 4/5 secs or so ...