The MSI Windpad 120W is the Cedar Trail version of the Windpad U110W – an AMD Fusion device.  We reported on the 120W in June where it showed at IFA with a Cedar Trail 1.86Ghz CPU. We saw it at IDF a few weeks ago and it had a low-power platform inside. We had the chance to give it a few tests.

The Windpad 120W shown at IDF used the Intel Atom N2600 with 32nm CedarView-M with dual-core, 2-threads per core (contrary to specs) and 1.6Ghz clockrate. The PowerVR SGX545 graphics operates at 400Mhz and forms part of the GMA graphics unit. The TDP is 3.5W and it is coupled with a communications chip (NM10) that uses 1.5W.

We only had time to do some CPU and battery tests but in Cinebench R10 we get a good idea of the CPU performance. A multi-core result of 1507 matches Atom N550 and very closely.

The 120W comes with 2GB of RAM

And after a good 10-15 minutes of testing, we saw over 4hrs left on the battery (97%) which is an indicator of working time. Given the low TDP of the Cedar Trail platform we expect it to idle down well to add a few hours to that.

One thing we still haven’t worked out is the screen resolution although the 110W has 1280×800. If that’s the case, this could make the perfect entry-level Windows 8 tablet.

Still outstanding from Cedar Trail tests are definitive 3D graphics performance tests. We’ve seen an indication that the N2800 with the graphics clocked at 650Mhz will be twice as good as the graphics on the Pine Trail generation but remember, the N2600 graphics are clocked lower. Don’t expect anything ground-breaking here.

I’ve had a theory about web-page loading speed for about three years now. When a web page takes more than 12 seconds to load, a user considers it a slow experience. Anything less than 10-12 seconds and everything is in the green zone. Smartphones are now entering the green zone.

The mobile Internet experience needs to be compatible and quick and if either one of those elements fails, then the customers Internet experience fails.

In my search for the full internet experience in my pocket I’ve spent hundreds of hours testing, analyzing, questioning, discussing and writing about the subject and I have a very clear picture of what is satisfying for the end user. In Sept 2007 I wrote about the extremely poor ARM-based internet experience. Really, it was embarrassing. The best phones of the day were taking nearly twice as long as the slowest of X86-based platforms with small screen sizes and low-quality rendering adding to the overall disappointment.

In October 2009, 2 years after my ARM-bashing article, the ecosystem had moved forward a long way and we saw what was probably the fastest ARM-based browsing experience ever. The Archos 5 was even beating low-end UMPCs. The iPhone 3GS was doing well too. Take another 5 months step into the future to the present day and you now have smartphones beating low-end Ultra-Mobile PCs. Pocketables, always reliable for a good set of browsing speed test results, shows us that there are at least three phones out there that are in the green zone now with the Nexus one and iPhone 3GS breaking the 10-second barrier. If JKK’s excitement about the Milestone and the Android 2.1 upgrade is anything to go by, we can expect the Droid/Milestone to be in that category too. The HTC desire will be joining the club in just a few weeks and following closely will be the Dell Mini 5 and Sprint EvO.

Just to re-cap, that’s pocketable, always-on, fast-internet devices with mobile-focused operating systems, mobile photo and video cams, high speed 3G, GPS, sensors, touchscreen, multi-GB’s of storage and access to thousands of apps costing under 400 Euros.

Related: The Full Internet Experience of 2010

Related: ARM’s lock-in opportunity

It’s not just the CPU.

You might think that ARM and their silicon partners are responsible for the advance but fortunately for Intel, that’s not quite true because in the last 2 years we’ve seen some amazing progress in browser software performance. Javascript processing speed, delayed script processing and other tricks and optimisations mean we’re also seeing improvements on desktop browsers too. I haven’t had time to do a complete set of tests but after disabling Flash on my desktop browsers I’m seeing something like 20-30% speed improvements over results of two years ago.

I’m not going to sit here and defend X86 though because its the ARM ecosystem that deserves the praise here. The fact is that the ARM ecosystem of hardware and software developers has moved forward quickly and shows no signs of stopping. In fact, as ARM tablets and always-on netbooks enter the market, that momentum could increase. With Moorestown and MeeGo on the horizon for X86 there’s a ray of hope for X86 but if that hardware/software platform isn’t good enough (most of us in the industry agree that it needs to move on another generation before it’s ‘ripe’ for smartphones) and the speed of development and investment doesn’t ramp-up quickly enough, Intel will never be able to catch up with the smartphone market.

If you are a regular reader of sites in our end of the blogosphere I’m almost sure that you are aware that the iPhone 3GS has now been released by Apple. While not a drastic change from the original, the 3GS boasts speed improvements that claim to make it twice as fast as the iPhone 3G. Chippy mentions that the iPhone 3GS is ‘the most important MID’ out there right now, and it is inevitable that people will compare many of the upcoming Moorestown devices to the iPhone 3GS. MIDs run lots of different software, amd while we won’t be able to do direct application comparisons, one of factors that we can most easily compare is the speed in which a MID can load a webpage. AnandTech put together a table which stacks the iPhone 3GS up against some of best smartphones out there.

It is easy to see the ability of the iPhone 3GS to quickly render websites, and the iPhone 3GS has set the bar that upcoming Moorestown MIDs will match and hopefully surpass.

Check out the full article at AnandTech for additional tests.

A quick look around the iPhone 3GS ‘Tech Specs’ page reveals plenty of info about battery life, screen size, resolution, and a number of other info, but we have yet to see any info on the RAM or CPU speed which claims to make the iPhone 3GS “twice as fast’. Even when asked directly, they refused to say exactly.

It seems a bit strange for them to be doing this, but I have a theory. You may have read my article a short while back questioning the rumors that said that Apple would release a touch oriented slate style device at WWDC. In that article, I mentioned that the App Store has been a huge part of the iPhones success, and Apple wouldn’t be releasing anything that runs the iPhone OS with specs that would ruin app compatibility. So, I feel that Apple has done some work to ensure that apps stay cross compatible with every device of their touch series.

I’ll have to mention that I haven’t yet been able to test the iPhone 3GS, but here is my initial theory. Apple may have doubled the RAM in the iPhone but kept the CPU speed the same; they want to keep the double RAM a secret. Why would they do this? The previous iPhone 3G has 128MB of RAM that gets allocated around the system to where it is needed. If a heavy app uses too much RAM, it crashes because the phone doesn’t have any more RAM to offer it, as it is being used up by the rest of the system. By doubling the RAM, they have plenty of space to allocate to the system, and a much larger space to load and run applications. The RAM speed remains the same (and the CPU) which means that app performance (while it is actually running) will stay the same, but with the increased quantity of RAM, more data can be loaded to the RAM at once, increasing the speed in which applications load. This is important because Apple doesn’t want any of the touch devices to run applications better than any other, ensuring compatibility between all.

Apple understands the success of their App Store. In the WWDC keynote, they mentioned that they have 50,000 applications, over 1,000,000,000 downloads of apps, and an install base of 40,000,000 users, across all generations of the iPod Touch and iPhone. Those numbers are insanely impressive, and as I stressed in the aforementioned article, they are not about to ruin compatibility of their precious app store. If they just doubled the CPU speed and double the RAM, developers would have a new performance bar that they could develop for. This means that an application might be designed specifically for a theoretical, more powerful iPhone, and yet it wouldn’t run well on their older devices, totally breaking the trend that the current app store has had since it was released: complete interoperability between all devices regardless of generation. But here is the really tricky part… why keep it a secret?

(continue reading on page 2…)

I’ve had the Vaio P for a few weeks now and it’s time to look at the performance of the unit. You may recall that this is the top-end model: 1.86GHz Atom CPU, 128GB SSD, 2GB of RAM, and running Windows Vista (see more detailed information on the Portal page). This particular configuration is only available in Japan at the moment, but can be bought from places like Dynamism.

The Vaio P runs Vista, which most people recognize to be quite the resource hog, and feel that it isn’t optimal for mobile computers because of limited resources. Luckily the Vaio P’s lightning fast SSD, 2GB of RAM, and 1.86GHz CPU, handle Vista pretty well. The OS itself is responsive and doesn’t show any signs of hanging. Even Windows Media Center, which runs a graphically complex GUI, runs acceptably on the VAIO P, even if it doesn have some slight issues with the VAIO P’s super-wide screen (more on that in the upcoming full software impressions section).

Video Playback

It amazes me how much more video content we see online these days, than compared to a few years ago. Back in the day, usually only small videos would be uploaded to the web, and playing them back often required downloading them then watching them. With many different Flash players out there today, we see lots and lots of video content, and video playback through Flash players is a big deal now, as it is often a frequent web activity. So let’s see how well the VAIO P can play back Flash video. Using an SD quality video from Gametrailers.com (which uses a pretty efficient Flash player), I saw full framerate playback with no issues. Below you can see the CPU usage while the video was playing:

Unfortunately, the VAIO P can’t handle the same video in 720p HD through Flash playback. The image below represents the CPU usage during Flash HD playback of the video. One thing to note is how the CPU graph on the left (representing the virtual second CPU from Hyperthreading) is a bit higher than the right. While the CPU isn’t being maxed out, the video playback was very choppy, to the point that I would call it unwatchable.
Flash HD playback might not work to well, but 720p, h.264 encoded WMV, and similar formats play back without much issue in Windows Media Player. For some reason or another, videos were a bit choppy, and showed some horizontal tearing when using my preferred video player of choice, VLC. I would imagine this has something to do with VLC’s implementation in Vista, as I’m used to using the XP counterpart. The image below shows a 720p h.264 encoded video playing through Windows Media Player, and the resulting CPU usage.

HD playback performance is particularly important on the VAIO P because the screen can actually display the content. In most cases, we test HD video playback on netbooks simply for performance’s sake, however when you really think about it, it doesn’t always make sense to watch a 720p (1280×720) video on your netbook which only has a resolution of 1024×600. However, in the case of the VAIO P, the high res screen can really display full 720p content with 1:1 pixel mapping and even has room to spare. The VAIO P’s screen has a resoultion of 1600×768, which is even wider than the widescreen HD standard of 16:9. For comparison’s sake, 16:9 (standard widescreen format) breaks down to 1.78:1, while the VAIO P’s screen has an aspect ratio of 2.08:1, meaning that the screen is slightly more than twice as wide as the height.

(continue reading on page 2…)

I’ve been testing the HP Mini 1000 MIE [Portal page] netbook for a few weeks now and I’m ready to give some performance information. As you’ll recall, the ‘MIE’ in Mini MIE stands for Mobile Internet Edition. This is HPs way of denoting their custom Ubuntu operating system which runs in place of a typical Windows OS. Unfortunately, because of the OS on the unit I can’t bring you the typical benchmarks. Luckily some benchmarks are already available out on the web from the XP versions which should give you a rough idea of what you will see, and I’ll point you in their direction toward the end of the post. What I can bring you however are some good old fashioned numbers (and maybe a chart or two)!

Video Playback

Video playback within a web browser usually isn’t considered a big deal on a desktop computer, but when working with the relatively small resources of a netbook, it can become a bit more serious of a task. In the screenshot above (click to enlarge) I showed the CPU usage graph while playing back a flash video from Gametrailers.com. This is SD footage. Flash HD video playback is too choppy for realistic viewing on the Mini 1000 MIE. I wouldn’t expect to be multi-tasking while listening to music through YouTube in the background. The CPU was around 80% usage during playback.

Boot Times

All tests are done with no apps open. Open apps will theoretically add a small amount of time to hibernate and standby.

It is clear to see that standby is going to be your best option in most cases, offering a pretty quick 5-seconds into standby, and bringing you right back to your home screen in 6 seconds. Startup and shutdown times seem fairly normal, but something is a bit odd with the hibernation times on the Mini 1000 MIE. In general, going into hibernate should take about half as long as shutdown, or at least be equal in its speed. In the case of the Mini 1000 MIE, going into hibernation takes the longest of any of the boot options.

A few things could be causing this. Let’s quickly define the difference between hibernation and standby. As the computer enters standby, it keeps all of the important OS files, and any running applications in the RAM, and powers down to a very low-power state, using only enough power to keep the data on the RAM alive. This means that it consumes a little bit of power while it is in standby. Hibernation consumes no power because the RAM data is backed up to the HDD (in this case, an SSD), and the computer shuts down completely. When turning back on, the data is rewritten to the RAM from the HDD, and the unit returns to the same exact state that was saved when it went into hibernation. Now that we know a bit about standby vs. hibernate, let’s theorize. One potential cause of poor hibernation times is a bottleneck in communication between the RAM and the SSD. Another reason could be due to the way that Ubuntu handles hibernation.

Either way, it is clear that a full shutdown or standby are your best options. In my testing I’ve seen that the Mini 1000 MIE should provide 70 hours of standby time which is pretty good. Put your computer into standby and come back 7 hours later, and you’ll only be missing 10% of your battery life. Considering you are only 6 seconds away from a full computing experience, I’d say that is a pretty good deal.

Battery Life

The battery on the Mini 1000 MIE is pretty darn slim. Take a look at the photos above. The computer itself says that the 3-cell battery will power it for 2 hours and 35 minutes from a full charge. The netbook is good at delivering this amount of time while you are using the computer for web browsing. While not available just yet, a 6-cell battery should be available from HP in the near future. Keep an eye on their accessories page for availability.

Benchmarks

As I mentioned, the non-standard OS doesn’t allow us to run our usual benchmarking programs, but the XP flavored Mini 1000 has been benchmarked by others, I’ll drop the SSD benchmark below, but also swing by this page to see some additional ones.

Check out the HP Mini 1000 MIE coverage roadmap for more coverage on this netbook.

The first of the MIDs are arriving at the doors of MID Moves doorsteps and one of the first to get attention is the OQO 2+ with it’s OLED screen and 1.8Ghz Atom CPU. Jenn has pushed out some benchmarks already and you can instantly see the advantage over lower-clocked Atom CPUs. In fact, that ALU figure is higher than I expected. Disk results there look fairly average so I’m wondering if Jenn has the HDD version. Actually I’m hoping she has the HDD version!

As I type, Jenn is twittering [follow her mutterings] in the background so although it must be past her bedtime, I know she hasn’t been able to put it down yet!

Source: MID Moves

P.S. Jenn has the Benq S6 too. It’s the cheapest MID on the planet. Euro299!

Specifictions and details in the database: OQO 2 Plus. Benq S6.

I just cant wait any longer for that Kohjinsha SC3 to arrive. The videos and feedback from early tests have proven to me that this could be one of the best UMPC / mini-notebooks yet. Its tiny, light, well-specified and looks great. The question we are all asking though is…”How good will Vista run.” We’re one step closer to the answer today because I’ve just recieved a set of CrystalMark test results from DFJ. I believe this is the first ever set of test results for a production Menlow (Silverthorne/Pulsbo) device so lets take a quick look and I’ll make some comments below.

Detailed results here.

The tests were done on a stock device (Vista Home Premium. 1GB RAM) and overall, its looking good. The CPU figures are quite impressive. Better, in fact, than any other UMPC platform. Only the Intel notebook-based CPU’s used in the Q1 Ultra Premium and the Sony UX are better. Memory speed is also good. Top marks! Looking at the disk (Spinpoint N, HS06THB, 60G, 4200 RPM, PATA(ZIF), 8MB CACHE) I see a high average which, if you look at the detailed results, translates to a max read speed of 30MB/s and a max write speed of about 27MB/s. For a 1.8″ drive, this is about as good as it gets right now.

Looking at the graphics scores shows some very low-end results. They dont quite tally with the Vista performance score we saw so for anyone looking at gaming on this – dont! For the 3D components in Google Earth and Itunes, we’ll have to see how it performs when we do tests. I suspect it will be OK.

For a big list of comparable CrzstalMark results, see the list that Frank keeps over at Tweaks2K2.

Note that the CrystaMark test doesn’t show us any video performance results. The Poulsbo chipset contains hardware acceleration for many common codecs and we wont see those results until someone does tests with a media player. In theory, results should be good.

So will it run Vista? The jury is still out on that one. Ive seen Vista Home Basic running acceptably on the HTC Shift but it wasnt exactly fast.  With some optimisation, i think Vista is going to be acceptable for most people. For others, upgrading the RAM to 2GB might be the answer and for a few people, it simply wont be good enough. The flip-side of the coin is that we know it would run XP extremely well. I hope, after time, a downgrade to XP will be possible. Ive taken the precation to include Vista Ultimate in mz purchase so at least i’ll have the license when it’s a possibility!

Thanks to Direct From Japan for the test results.

Guess what the first word of this Nvidia video is? a) Welcome b) Today c) Intel

The Tegra demonstration is impressive, showing HD video content being decoded in around 1W instead of 12W but they clearly didn’t cover all bases. For a start, you can’t even run a big screen backlight at 1W and if I’m doing my research correctly I can see that Tegra doesn’t include a communications and radio subsystem. Even so, what exactly are they trying to compare here? They are putting a general purpose CPU (i’m guessing its Celeron there) against a dedicated video codec in a video decoding test. Not only that, they are showing their new tech against Intel’s old tech. Booo!

I’m sure the NVidia board is impressive but come on guys, lets see you doing some general purpose computing with Wifi, BT, 3G, storage, USB and compare it against a Silverthorne/Poulsbo combo that will do the same in a MID-sized board in under 5W. Bring out the MIDs Nvidia. Or better still, send them to me and I’ll perform a special selection of Chippy tests on them.

Yesterday, I completed the final part of a series of upgrades that have been done to my Q1 Ultra to bring it up to what is close to my ideal all-round UMPC device. In this article I’ll outline what I’ve done and give you some detail about my most recent upgrade, swapping the 60GB hard drive out for an 8GB flash drive. The difference is quite remarkable. Not only do test results show up to 4x the read speed and 2x write speed but boot-times are quicker and the overall responsiveness of the system is improved dramatically. If you need convincing that SSD’s are fast before you take the jump, check out the video at the end of the (long and media-rich) post. Note that much of this applies to other devices that use 1.8" drives. The Sony UX, Fujitsu U810, HTC Shift and others.

Read the full story

Most, if not all of you reading this will have heard of the Firefox browser and many of you will have tried Firefox 3.0 beta. I held back from using it for a long time because it was beta software but the latest release candidate seems stable and has me converted on all platforms now. Firefox 3.0 is fast. Firefox 3.0 is memory efficient. Firefox 3.0 has great features and overall its a clear winner on UMPC and netbook platforms, especially when using online applications.

Like Safari, it appears from my test results that Firefox 3.0 can process java-heavy pages on a Ghz-class UMPC faster than the data arrives over my 6mbps Internet connection which means that for rich Internet applications, the bottleneck is at the remote server and there’s very little else you can do to speed up the experience. Apparently, java processing in FF3 is many many times faster than in version 2 so this explains the big improvement with online applications. Not only is the speed improved but there are some great features that will appeal to UMPC users too. But first, here’s some test results. I took 5 devices and ran speed tests on 3 browsers [*1] using reader.google.com as the target page. It’s a java-heavy page and there’s no flash or major numbers of images to process but its typically my slowest-loading browser application. It represents a typical online application and for web-workers, its a good, tough benchmark.

More info after the jump…

Read the full story

UMPCFever continue their hands-on with the Aigo MID and have been playing with the device running under Windows. While they were there they did a CrystalMark test which gives us the first ever look at how various parts of a Silverthorne-based MID might perform. However, I would take these results with a large pinch of salt as they appear to be on the low side. The memory speed and disk speed figures look especially low so there could be some driver issues. Certainly the graphics figures should be ignored at this stage.

I’d advise to wait for real-life browsing and video figures from the system while running the Atom-optimised Moblin build rather than a vanilla XP build that has never heard of Poulsbo or Silverthorne!

On the image at the top right, from the top are the total score, the ALU test score, the FPU test score the MEM, HDD, GDI, D2D and OGL scores from the Crystalmark test. For a list of UMPC crystalmark scores, see Ctitanics table of results from various sources.

UMPCFever Aigo MID CrystalMark test video.

Update: A more detailed discussion is continuing on Atom performance details in the forum here.