TI's OMAP 3 hardware doesn't manage to do Windows Mobile 6.5 any favors

We were really impressed with Texas Instruments' OMAP 3-based mobile development platform when we saw it running Android earlier today, and the hardware seems to hold up well to all sorts of video-accelerated magic for a multitude of Linux-based interfaces. Not so much with Windows Mobile 6.5. Granted, this is a very early build of the OS according to Microsoft, and is lacking all sorts optimization mojo that will go into the final product due end-of-year-ish, but at the same time this is basically an interface skin to Windows Mobile 6 that Microsoft has had nearly a year to work on since 6.1, and it's not exaggerating to say that performance is a complete joke at the moment. TI's OMAP 3 was clearly struggling to handle basic scrolling of the home screen and app menu, though "smoothness" was slightly improved in scrolling through a gallery of photos. TI has been working with 6.5 for "a few months," and claims to have added optimizations to interface performance -- and interestingly is running this on a capacitive touchscreen device, something that isn't officially supported as far as we're aware. They're also saying that they plan on achieving full smoothness by the time this launches, but there's really a long way to go, and it's difficult to understand why this seems to be a last priority for Microsoft at the moment. On the other hand, TI is doing great things with its own 3D-accelerated interface, something which Windows Mobile 6.5 brings support for, so perhaps these will all one day (soon) meet in the middle and give us the sort of interface smoothness and usability we'd expect on a phone built a decade into this century. Check out the painful videos after the break.

Gallery: TI's OMAP 3 hardware doesn't manage to do Windows Mobile 6.5 any favors

• 
• 
• 
• 
• 

[Ysleiro @ Feb 18th 2009 11:23AM]

Windows 7 also has compatibility issues, so it's not weird to see another unfinished product have the same problem. It's not done yet, hopefully when it is it will rectify the issue.

But.... lately I have been getting such a bad image of Texas Instruments.

Maybe it's because I'm taking calculus 2 and I'm forced to cough up 150 zingers for something that looks like it should cost 30$.

Maybe I'm wrong but somebody needs enlighten me.

[bhtooefr @ Feb 18th 2009 11:31AM]

That's a completely different division of TI, though.

But, yeah, $150 for a TI calc gets you, what, a slow 68000, and a crappy keyboard?

$150 for an HP gets you a 75MHz ARM9 and a decent tactile keyboard. Downside is, it doesn't have the TI user interface, so you'll have to figure out how to use the thing yourself. Upside is, it has RPN and an SD slot. :)

[Shinigami @ Feb 18th 2009 11:50AM]

Enlightening:
Nvidia Tegra looks like something that should cost $500, promises $100 devices instead.
Texas instruments think they are unique same as their products and nobody can take the market away from them so anything, no matter how good or bad it is, should cost $1000.
In one word: Greed. Thats the word that rules business.

[Macdelaney @ Feb 18th 2009 2:56PM]

@Shinigami
RTFA. The devkit costs $1000, it's not likely that a consumer oriented device will.

[Ike Turner @ Feb 18th 2009 11:27AM]

Being exposed to Paul Millers biased reports is like being exposed to somebody else's fart smell...

[--------- @ Feb 18th 2009 11:55AM]

...like the way he's blaming MS for TI's shortcomings.

[Chris @ Feb 18th 2009 12:06PM]

@--------

Yeah, because android, the pandora, and palm's webOS obviously are so sluggish on TI's crap hardware. It's MS's problem here, not TI's.

[NuShrike @ Feb 18th 2009 12:58PM]

Let's break it down as it has been done before for the Qualcomms:
o MS does NOT have a compiler that can target higher than ARMv5 in the final assembly code. This TI OMAP3 is ARMv7, and the upcoming OMAP4 is ARMv8. The MS compiler is quite behind the times, and there's no other "almost-free" alternative.

o the OS builds released only targets ARMv4 making it portable all the way back to the Wizard, and therefore proving it lacks many of the needed code optimizations.

o the C/C++ compiler natively generates only ARMv4 code without some forced ARMv5 DSP intrinsics, and even that is very clunky code to write.

o the OS (including CE6 in WM 7) was NOT written to automatically support a native FPU. It has to be hacked into the OAL layer by the OEM to enable the ARM fpu, fpu exceptions handler, etc, and we know what a great job OEMs do. A lot of sub-decimal math IS used everywhere in the OS since WM derived from an x86 OS that always had a FPU at that point.

o the whole GDI layer is an antiquated design predating modern OpenGL OSs so does not support delegate-based coding (except in slow slow slow .NET) and hardware acceleration (not that it doesn't have some minimal support for HWA), and other UI design enhancements. Even Qt is more advanced than the Windows Mobile framework at this point.

o We've seen TI OMAP2 on the Motorola Q's and they've topped graphics benchmarks, so we know it's not the cpu.

So it's definitely is a lot of work for anybody to smack WM6.5 into something usable for a modern phone cpu, and I rather have the TI experts do it than HTC, or Qualcomm, n00b hacks.

[Fnuky @ Feb 18th 2009 9:01PM]

@NuShrike - Great analysis!

Couple of questions... My understanding was that OMAP4 / CortexA9 was still ARMv7 instruction set based, not V8?

Any idea on when MS are going to take WinCE SMP? I believe CE 6 is still uniprocessor, so I can't imagine WM7 will be able to take full advantage of the A9 multicore parts.

But I fully agree - these performance problems are not down to a hardware limitation, it's GDI and optimisations, but I'd suggest mostly GDI, based on my current WM6 experience.

[loosely_coupled @ Feb 18th 2009 10:53PM]

@--------/Unabomber

You obviously have no idea what you are talking about. The ARM process in T.I.'s OMAP3 series, the ARM Cortex-A8, is an excellent chip and much faster than an equivalent ARM11.

The sluggishness of WinMo lies on Microsoft, but it obviously has to do with the kernel, drivers, chip level optimizations, debugging code, etc...

[rodan32 @ Feb 18th 2009 11:30AM]

In my experience, those optimizations will never really make it. I've had many Windows devices over the years, and none of them have ever really had ultra-responsive UIs. Even the Axim X51v, which had a pretty quick processor, wasn't a rocket. I want to be a fan, but they've got to prove to me that they care about a responsive UI.

[Electromodo @ Feb 18th 2009 11:34AM]

"...and it's difficult to understand why this seems to be a last priority for Microsoft at the moment..."

Paul, optimizations are done at the end of development lifecycle, that's why they are called "optimizations".

Unless, of course, the application was designed with ultimate performance in mind, which I believe is not a case with Windows Mobile, since it targets functionality, compatibility and third party support.

I know what I said here is debatable (because it is very general), but at least it is clear why Microsoft left any device-specific optimizations for the end of their development.

[Fnuky @ Feb 18th 2009 11:37AM]

Is this really TI's problem or are Microsoft's apps still mostly written to use incredibly old, clunky graphics libraries from the Win95 era that take no advantage of hardware acceleration features and are as smooth as a craggy mountain face?

Windows CE on an x86 platfom doesn't perform any better than that, so I'm guessing it's not the hardware.

[Gnormie @ Feb 18th 2009 11:38AM]

How can 6.5 not be compatible with capacitive but can be with resistive? The hardware doesn't no which is which, it just interprets it as a mouse click, or does it just mean not compatible with multi-touch? Which again is a little odd as it should just be a software thing and available through a software update (which presumably would not be forthcoming).

[Surur @ Feb 18th 2009 11:46AM]

Exactly. Its all just software.

Capacitive screens do lack in precision however.

[Fnuky @ Feb 18th 2009 9:14PM]

Agreed - it's all software. The issue is, as you say, the lack of accuracy of capacitive touch systems.

Apple (oh god, now I'm going to get downranked) accepted the lack of accuracy in capacitive / finger touch systems and traded it off against the benefits of better resilience to scratches and a lighter touch, through not having to push sheets of plastic together to make a contact (as happens in the the resistive screens).

Because they made that design choice at the start of the design process, they were able to optimise their controls and APIs for capacitive / finger touch with massive hit areas that extend beyond the visible control and intelligent guesswork, across the entire O/S, by design.

The clever bit on iPhone OS, and why it seems more accurate than traditional systems is, ironically, because it is presumed you will be less accurate hence you will make a mistake and they apply heuristics to guess what you were trying to do and do that instead, across all UI controls.

Hit 1 pixel left of a control on WM, which is still, at its core, optimised for a stylus (because that's what Palm were doing back in the 90s), and it will fail to operate. Hit 20 pixels to the left of a button on an iPhone, if there is no other button nearby, and it will guess you wanted to hit the button nearest it. Through how well and transparently it works, I can't even begin to imagine how complex the maths they use is.

It's been so successful that most people now associate capacitive touch screens with high accuracy because they have a higher success rate of doing what you want to do when you tap the screen on the iPhone than they did on a WM6 resistive screen whereas, in fact, the phone has very little idea exactly where you were trying to hit on the Apple platform.

So, make a WinMo device that has a capacitive touch screen, but without optimising controls and applying heuristics, and you will end up with a worse experience than a resistive one.

It's like 10,000 spoons, when all you need is a life.

[loosely_coupled @ Feb 18th 2009 10:54PM]

Whether Capactive screens are less accurate, I don't know. What I do know is that the accuracy of my iPod touch is perfectly fine, and I would never trade the amazingly sensitive capacitive screen for a resistive screen akin to my old Treo.

[Joshua Ochs @ Feb 18th 2009 11:52AM]

Looks like it's using software compositing. If they have hardware acceleration in the wings (which it looks like they do based on the video playback), then it's probably just a matter of improved drivers and hardware support.

Based on the excessively negative review above, I thought we'd see them swipe a finger and a few seconds later the device respond. What I saw is completely acceptable for an in-development device.

[Chris @ Feb 18th 2009 11:53AM]

I think it performs a lot better than you give it credit for; the scrolling wasn't horrible and it does look like they need to do a bit of optimization for the video driver. But hey, with 720p playback and a crazy 3D interface, the processor could do a lot of favors for Windows Mobile!

[shravan @ Feb 18th 2009 11:55AM]

The more i see of 6.5, the more im liking it. Put it on a good responsive capacitive screen, add a tegra and give it the specs of the touch HD. I'll snag one right away. One question though. Will backwards compatibility for 6.1 apps be provided ?

[thephonelounge.com @ Feb 18th 2009 1:16PM]

Seriously? This is effectively just a re-skin you know...

[philosopher @ Feb 18th 2009 12:02PM]

A fail straight outta the box. No "touchkiller " here .

[GON @ Feb 18th 2009 12:02PM]

I loved what I saw, though I think this would have been better if demoed on a resistive touch device with a stylus rather than this capacative stuff. Stylus gives the user more precision, and is better to showcase the features since fingers will block some of the action. While Micorsoft has had "nearly a year" to work on 6.5, I think it's important to note that they've also been working on Windows Mobile 7, which is a total revamp of the OS, and I'm sure doing that is no cakewalk. Yes Palm has made WebOS, but Palm was in a much more dire situation than Microsoft.They were on the verge of being beaten to oblivion by their competitors, while Microsoft is thriving at the moment.

[GON @ Feb 18th 2009 12:04PM]

Backwards compatibility could very well be one of the reasons why Microsoft is taking their time with 6.5 and 7.

[TareX @ Feb 18th 2009 12:28PM]

IT.IS.A.POCKET.PC.OS

Once MS realizes they need a SMARTPHONE OS instead, they'll build the architecture right from the ground up. I don't know if that's the case with WM7 (aka photon), but I sure hope it's not just an upgrade to 6.5...

Palm's webOS seems to have gotten it right, although I doubt we'd be able to see truly amazing apps with web tools.... like the case with Android, where it's already hard to re-write emulators in Java... but it's possible nonetheless.

[notty11 @ Feb 19th 2009 6:15AM]

Is there a Apple booth at MWC ? Why is iEngadget even there?

[BradS @ Feb 18th 2009 1:59PM]

If they want to work on the smoothness, they should hire Billy Dee Williams/Lando Calrissian.

[Tim @ Feb 18th 2009 6:11PM]

Cripes, WinCE doesn't give a crap if a touch panel is resistive, capacitive, SAW, or a mouse. Quit saying that "it doesn't support capacitive screens". If you really mean "multi-touch", then say instead.

[Jon @ Feb 18th 2009 5:01PM]

sweet vista for the mobile device

[MaxSMoke @ Feb 18th 2009 5:10PM]

I can't wait to see what the final version looks like.

[MFong @ Feb 19th 2009 3:51PM]

Can it graph a cardioid function?