We've been covering the issues surrounding Samsung's SSD 840/840 Pro lately. The issue was first discovered when Anand's pre-production review sample died during testing and we also noted that in our initial review. Samsung quickly sent us another drive but it also failed after a couple of days of testing. My SSD 840 managed over a month but ironically enough, it died right after I had completed endurance testing.

Earlier Samsung told us that all review samples including our three shipped with a pre-production firmware that had a bug in it causing the failures (retail units were shipped with a newer firmware without the bug). At the time we didn't know what exactly was wrong in the firmware, but now we do. When the drive was issued a secure erase command, it would clear all table mapping information at the Address Translation Layer (ATL) but not at the Host Interface Layer (HIL). The data in both layers needs to be up-to-date for the drive operate properly, so when a write request came in, the controller wasn't able to map the data correctly, which caused the firmware to hang. An SSD obviously can't operate without a functioning firmware so from a user's standpoint, it looked like the drive had completely died even though only its firmware was broken.

All our three failures support this explanation. Our first 840 Pro sample died during a 128KB sequential write pass that we use to pre-condition our drives for enterprise tests, but the drive was secure erased just before beginning to fill the drive. The second 840 Pro died during power consumption testing but again it was secure erased right before starting the test. The regular 840 actually died when I tried to secure erase it. The secure erase command resulted in an error so I power cycled the drive but it was no longer detected by the system after reconnecting it. 

The good news is that all retail units have shipped with a newer firmware, only reviewers and others who have access to pre-production units were affected by this bug. 

For users considering the SSD 840/840 Pro, this should be reassuring news. The 840 Pro is still the fastest SATA 6Gbps SSD we have tested and it's definitely one of the top choices where performance is concerned today. The TLC NAND based SSD 840 is more mainstream focused but from what I have seen, it seems to be fairly competitively priced. The SSD 830 spoiled many with low prices but that was only to clear stocks. If you can still find a bargain SSD 830, don't hesitate to pull the trigger as those won't be available much longer, but we're more comfortable recommending the 840/840 Pro now.

While long-term reliability is still unknown, if history is any indication the 840/840 Pro are in good company as the 830 was a solid drive. Our third 840 Pro with the new fixed firmware has been going strong for weeks now and we have even recreated the scenarios that killed the earlier samples. We are also waiting for more samples from Samsung to test all capacities of SSD 840 and 840 Pro, so stay tuned!

Sometimes it's easier to take risks, if you aren't moving quite so many units. ARCHOS, long a player in media playback devices has spent a few years exploring Android tablets and we've put them through their paces. Their last offering was the remarkably thin 101 XS, with its impressive keyboard cover. The ARCHOS GamePad takes a rather different turn. One part tablet, one part portable game system, the GamePad brings hardware gaming controls and is available now for just 149.99 EURO, with a US release in Q1 '13. 

Arranged in landscape mode, the GamePad's 1024 x 600 screen resolution isn't going to light the world on fire, but paired with the ARM Mali-400 GPU it should provide excellent frame rates. We've seen a lot of TI silicon in the last several ARCHOS tablets, so it was a surprise to see the processor go unnamed. We're still digging to find more details, but it's described as a 1.6 GHz dual-core processor, with the aforementioned Mali-400 GPU. Given the last gen nature of the GPU, I imagine that the application processor is most likely of the ARM Cortex-A9 variety, somewhat similar to a higher clocked Samsung Exynos 4210. 

On the software front, this will be the first Jelly Bean tablet we've seen from ARHOS. Their software engineers have no doubt pushed themselves to the brink to bring the same media-centric features we've come to expect to, not just a new Android version, but also a new SoC. They've built a scheme for mapping the hardware buttons to soft buttons in games that aren't compatible with gamepads. 

So, is tablet gaming ready for a more dedicated device than we've seen in the past? We'll certainly be glad to find out. 

We're seeing a true shift in the mobile market to including SSDs in more of the market spectrum. What was once the domain of $1500+ laptops is now being pushed well into sub-$1000 territory, and good 128GB SSDs are generally available for under $100 now. If you need capacity, it's still hard to beat hard drives, but for fast storage SSDs are the way to go and have been for a few years. Along with the transition to faster and better SSDs, we're also seeing smaller mSATA SSDs.

Up until now, the largest mSATA SSDs have topped out at 240GB/256GB, but Mushkin is now announcing their 480GB SF-2281 based Atlas mSATA drive. What's particularly interesting is that this may be the first 16GB NAND die we've seen; we're trying to confirm that, but it would make sense considering the capacity and form factor. Update: Nope! It looks like Mushkin is using a stacked daughterboard--thanks to PaulJeff for the image link in the comments to TweakTown's review.

Ultrabooks and ultraportables in general are set to benefit from the presence of higher capacity mSATA SSDs. While many users can get by with 128GB or 256GB of storage, another class of users can routinely fill up 256GB SSDs and then some. A quick look at my Users folder on my primary desktop reveals I'm using 150GB just for my normal data (documents, images, videos, music, email, etc.) Add in my Program Files and Windows directories and that's another 80GB. Obviously I'm not a typical user, but if I were to try and go pure SSD while keeping all of my data on one drive, I'd definitely need more 256GB--and as Anand showed earlier this week, having more spare area available will only help improve the consistency of performance.

Specifications for the new Atlas mSATA drive are what you would expect from a modern SandForce 480GB offering. Trim, ECC, SATA 6Gbps, upgradeable firmware, and all the other usual suspects are present. Mushkin is using "high speed MLC NAND", just like virtually every other modern consumer SSD, and the Atlas comes with a 3-year limited warranty. Max read speed is 540MBps, write is 425MBps, with 78K random read IOS and 28K random write IOPS. It uses the MO-300 mSATA form factor (50.80 x 29.85 x 4.85mm). Availability is expected in January with an MSRP of $500. That's higher than 2.5" 480GB SF-2281 drives by $50-$150, but that's the price we pay for smaller form factors.

TeamViewer is one of those applications that some people have heard about and love, while others have no idea what it is. At its core, TeamViewer is a remote control utility for Windows and OS X computers—and unlike Windows Remote Assistance (WRA), in my experience it works well and it’s extremely easy to get connected.

I first stumbled on TeamViewer when I couldn’t get WRA working to help my mom with a computer problem; trying to explain to a computer neophyte how to start WRA, save an invitation file, find the file they just saved, then email the file is not exactly a simple process. What makes it truly frustrating is when you go through all the steps and then the connection still doesn’t work. When WRA does work, I have no complaints, but my experience has been very hit or miss.

After trying for a few hours to get WRA working, I gave up and started looking for other options, and that’s when I found TeamViewer. Literally 15 minutes later, I had connected to the other laptop, fixed the problem (which involved making some registry edits), and we were done. The quick connect options on TeamViewer was particularly handy, as it skips past the installation process and lets you get straight to helping someone.

Long story short, after having used TeamViewer to help several friends and family members over the past month, I received a press release announcing the new TeamViewer 8 with Windows 8 support (available as of December 4). With Windows 8 adding a few new items to the mix—e.g. the Start Screen and Charms Bar—TeamViewer has been updated to provide access to these items through its menus. It might seem like a minor thing, but given the UI overhaul in Windows 8 it’s something that will likely prove necessary. Windows 8 isn’t the only OS with new features; on OS X, Retina displays are now supported. TeamViewer also has their remote access app available on iOS, Android, and Windows RT

The above features are available in all versions of TeamViewer 8. For the non-free (e.g. Business, Premium, and Corporate) versions, other new additions include an improved user management interface, connection reporting of all sessions, browser-based connections, session handover, improved remote printing (VPN is no longer required), deeper MS Outlook integration, transmission of remote sound and video (video will require a fair amount of bandwidth, naturally), and enhanced session recordings.

TeamViewer’s Magdalena Brzakala (Public Relations Manager), Andre Schindler (Business Development Manager), and Tom Carpenter (Account Manager) took some time today to run me through some of the new features, and basically everything worked as expected. The fully licensed version, while expensive, also adds some functionality that I can certainly see as being useful for IT shops in particular, as well as telecommuters or those who need access to their home data and documents. I wouldn’t go so far as to suggest TeamViewer is unique in any of these areas, but I can say that it has worked well when I needed it and it’s definitely one of those utilities I now keep in my software toolkit.

In my Nexus 4 review, I noted that the device was the first shipping handset to include support for SlimPort via an Analogix ANX7808 SlimPort transmitter. When I wrote that review it wasn't immediately clear what advantages SlimPort offered over MHL, which has become relatively ubiquitous in the year or so since its introduction. Since then I've acquired the SlimPort microUSB to HDMI 1.4 SP1002 and VGA (D-SUB) SP2002 adapters from Analogix for testing out SlimPort on the Nexus 4.

SlimPort is a new video output standard which works over microUSB, and is effectively an MHL alternative based on the DisplayPort standard administered by VESA. SlimPort is offered both license and royalty free, and supports up to 1080p60 or 1080p30 with 3D content over HDMI 1.4 (up to 5.4 Gbps of bandwidth), in addition to support for DVI, VGA (up to 1366x768 and 720p at 60 Hz), and DisplayPort. What's different about SlimPort from a user-facing perspective is that external microUSB power is not required for the adapter to work and is instead only required for optional charging. At the same time, the adapters draw no power from the source to operate. 

SlimPort over USB - From Analogix SlimPort Whitepaper (PDF)

I've spent some time playing around with the SP1002 HDMI adapter, which I believe is based on ANX7730, and the SP2002 VGA adapter, which is likely based on ANX9832. As of this writing, the Analogix SP1002 adapter is available on Amazon for around $30 USD. I haven't seen the SP2002 available online anywhere yet, however.

When connected to the Nexus 4, the HDMI adapter rotates the UI into landscape and mirrors the display just as expected. Resolution is scaled to 1080p60 on the Nexus 4 up from the panel resolution of 1280x768, which results in some letterboxing. As expected there's no requirement to have microUSB power connected to the adapter, I tested it with and without external microUSB power, all this does is just enable charging.

I posted a video review (shot on the Samsung Galaxy Camera) demonstrating the Nexus 4 doing mirroring with the SlimPort SP1002 adapter and received a query about whether the combo enables multichannel audio over HDMI as well. I tested VLC for Android with some .MKVs I had laying around with 5.1 AAC tracks and found that VLC will decode the multichannel audio and ship it over as 5.1 PCM to my A/V Receiver, so this does work, surprisingly enough. There's no bitstreaming support as far as I can tell, however. As an aside I was able to playback a number of 720p MKVs with hardware decoding checked in VLC for Android and 5.1 audio out with no dropped frames. The Nexus 4 seems to be a passable platform if you're looking to play the occasional 720p video over HDMI — provided you have the free storage for it.

I also tested out the SP2002 microUSB SlimPort to VGA adapter, which has a longer cable and correspondingly beefier module for accommodating the D-SUB connector on the end. I tried with a Dell G2410 LCD display and XGA resolution DLP projector and saw an output resolution of VGA (640x480 at 60 Hz), which wasn't quite what I expected. The SP2002 should work just fine with 720p60 over VGA, and I know both monitors expose the proper EDID, it seems as though the Nexus 4 just defaults to this mode. Just like the HDMI adapter there's a microUSB port on the back for optional charging, though the dongle doesn't require external power to operate. 

Thus far I'm fairly impressed with SlimPort. I enjoyed discrete microHDMI connections on devices when that was the norm, though finding a microHDMI to HDMI cable was always the challenge. Later MHL seemed a step in the right direction, but requiring three cables to do display mirroring seemed like a bit much. With SlimPort I think we've finally arrived at something close to ideal with the standard not requiring external power to function. It will be interesting to see how SlimPort, MHL, and Miracast adoption fare in the coming year as mirroring to a larger display is in the cards as a convergence trend for smartphones and tablets. 

Source: SlimPort (SlimPort Connect)

Yesterday, Qualcomm announced a new SoC for its Snapdragon S4 Play category, the MSM8x26, and alongside it a new transceiver, WTR2605. The announcement was a little light on detail and I waited until confirmation of a few details, but now know more about these two new parts geared at the growing entry-level Chinese handset market.

First off, MSM8x26 is a 28nm SoC consisting of four ARM Cortex A7 CPUs running at (1.2 GHz) alongside an Adreno 305 GPU. This is to my knowledge the first Qualcomm SoC using a Cortex A7 for CPU, previously we've seen a lot of Cortex A5 use at Qualcomm in parts like MSM8x25 (dual A5s), MSM8x25Q (quad A5s), and also onboard baseband as an optional AP for managing things like a router. MSM8x26 is the spiritual successor to MSM8x25Q, which was again quad core ARM Cortex A5s at 45nm with Adreno 203 graphics. MSM8x26 should bring a nice jump in performance on both CPU and GPU over that part, in addition to supporting 1080p video encode and decode, and support for 13 MP cameras. MSM8x26 will come in two flavors, 8226 with UMTS and TD-SCDMA, and 8626 with UMTS, CDMA, and TD-SDCMA, consistent with Qualcomm's part numbering scheme. 

The other part of the story is the new transceiver, WTR2605, whose name suggests a wafer-level package (W for wafer) and includes necessary improvements to accommodate dual SIM active and standby modes (DS-DS operation) popular in the entry level Chinese market MSM8x26 is geared at. I don't know anything further about the WTR2605 or how it compares in terms of RF ports to WTR1605L, which is Qualcomm's current flagship transceiver, but suspect it's an evolution of that design with changes to accommodate the dual SIM modes. We'll have a piece ready later in the week about WTR1605 and the state of Qualcomm's modem portfolio.

Source: Qualcomm

Continuing their work on the 12.11 drivers, AMD has recently released their 12.11 beta 11 driver update. This driver fixes several system hangs while including the previous improvements to performance found in Beta 8. In particular, the driver notes call out to a performance upgrade for Far Cry 3. Here are the direct links to the various drivers:

AMD Catalyst 12.11 Beta11 Driver for Windows Vista/7/8
AMD Catalyst 12.11 Beta11 Driver for Windows Vista/7/8 - with .NET 4 Support
AMD Catalyst 12.11 Beta11 Driver for Linux
AMD Catalyst 12.11 CAP2

AMD Catalyst 12.11 Beta Release Notes

• Improves performance in Far Cry 3 (up to 25% with 8xMSAA, SSAO enabled @ 1600p, and up to 15% with 8xMSAA, HDAO enabled @1600p) (AMD Catalyst 12.11 CAP2 must also be installed)
• Resolves a sporadic system hang encountered with a single AMD Radeon HD 7000 Series GPU seen on X58 and X79 chipsets.
• Resolves an intermittent hang encountered with AMD Radeon HD 7000 Series GPUs in a CrossFire Eyefinity setup.
• Resolves image corruption found in certain DirectX 9.0c titles
• Resolve missing fonts issue in XBMC
• Resolves no video issue found in Media Player Classic Home Cinema when using full or half floating point processing
• Resolves stability issues found in the previous AMD Catalyst 12.11 Beta8 driver for Linux
• For users experiencing issues with HDMI Audio under Ubuntu 12.04, users should try installing the “dkms-hda - 0.201211291615~precise1” package from  https://code.launchpad.net/~ubuntu-audio-dev/+archive/alsa-daily/+packages and reboot; this will resolve the HDMI Audio issue found in Ubuntu 12.04
• AMD Catalyst 12.11 CAP2 has just been released, and should be used in conjunction with AMD Catalyst 12.11 Beta11
  • Improves Far Cry 3 performance for single GPU configurations with AA enabled

Seeking a candidate for WHQL certification, NVIDIA has just released their 310.70 beta driver update for desktops and laptops. The driver continues to iterate on the 310.33 beta driver, which focused on improving performance for various games while updating SLI and Ambient Occlusion profiles. This driver update has come less than a week after NVIDIA’s 310.64 update, which focused on improvements for the newly released Far Cry 3.

The driver notes specifically call out improved performance over the 310.64 beta driver for Far Cry 3, Call of Duty: Black Ops II, The Chronicles of Riddick: Assault on Dark Athena, Assassin’s Creed III, and Battlefield 3. The driver also adds TXAA support to Call of Duty: Black Ops II and Assassin's Creed III (while resolving a shadow flickering issue in the latter) and continues to add SLI profiles for several games. Additionally, there are various “Driver Not Responding” errors that should no longer occur. Here are the direct links to the various drivers:

Desktop 32-bit Windows Vista/7/8
Desktop 64-bit Windows Vista/7/8
Mobile 32-bit Windows Vista/7/8
Mobile 64-bit Windows Vista/7/8

In both of our MacBook Pro with Retina Display reviews (13-inch & 15-inch), I pointed out a big downside to the user experience today: UI performance in some applications is significantly reduced compared to non-Retina models. I couldn't find a direct cause for the issue, just that whatever work Apple does to make OS X look like OS X ends up requiring quite a bit of CPU power, and the workload scales with resolution. I've seen this in applications like Mail and Safari, although it's present in more than just that. 

In our 13-inch rMBP review I proposed a couple of solutions: 1) a dramatic increase in single-threaded CPU performance, and/or 2) software improvements (e.g. the move to Mountain Lion for example shifted more animation workload over to the GPU, improving scrolling performance vs. Lion on rMBPs). 

Last week I received a tip (thanks Joan!) pointing me at a Macrumors post claiming that the latest nightly builds of WebKit fixed scrolling performance on the rMBP. I grabbed a build (r135516 - it's no longer the latest build but I assume the later builds also contain the fix) and tried it out on the 13-inch rMBP. Scrolling down my Facebook news feed ended up being one of the best showcases for poor scrolling performance on the rMBPs, so that's obviously the first test I ran. As always I used Quartz Debug to measure UI frame rate. First, here's what the average frame rate looked like using the latest version of Safari on Mountain Lion with the 13-inch rMBP running at the scaled 1440 x 900 setting:

13-inch rMBP, 1440 x 900 scaled setting, Safari Version 6.0.2 (8536.26.17)

Average frame rates end up being around 20 fps, with dips down as low as 17 fps. Now here's the same test but using the r135516 WebKit build:

13-inch rMBP, 1440 x 900 scaled setting, WebKit Nightly r135516 Safari Version 6.0.2 (8536.26.17, 537+)

Performance is more than doubled! Scrolling is so much smoother. I also ran tests on pages that previously worked fine (e.g. the AnandTech front page) and performance hadn't changed there. I haven't managed to figure out exactly what's changed in the codebase to improve performance so much but it's appreciable.

For those of you who are early adopters of Retina MBPs, there looks to be some hope that we might see software solutions to improving UI performance. The real question is when we'll see these types of improvements rolled into OS X. 

This isn't major news, but I find the periodic updates to the 3DMark suite at least interesting if only to see what Futuremark is up to. They have released a trailer showing a portion of the next 3DMark, dubbed Fire Strike. Designed to push the envelope in terms of rendering quality and techniques, Fire Strike makes use of a variety of DirectX 11 features including tessellation, dynamic particle illumination and shadowing, smoke simulation using grid-based fluid dynamics, volume ray casting with shadows, and a wide variety of post processing effects including depth of field, blooms, distortions and various lens effects.

Futuremark didn't indicate the name of the next 3DMark release, but they're promising a release before the end of 2012 so the clock is ticking. It is designed to provide a unified benchmark suite for testing everything from tablets to notebooks to desktops, including support for Windows RT, Android, and iOS.

The Fire Strike benchmark is targeting the high performance desktops rather than tablets, with other benchmarks for DX10 and DX9 level devices. Cloud Gate will handle the DX10 duties, with Ice Storm being the cross platform "DX9" level test. On Windows, it will use a DX11 engine limited to DX9 level features, whereas Android and iOS will use OpenGL ES 2.0. The scores will be comparable cross-platform.

You can see the full 3DMark Fire Strike trailer on YouTube, and if you have a high-res qHD or qWXGA display you can even view it at the original resolution of 2560x1440 (albeit with compression artifacts). We should have the final release for testing and evaluation some time in the next month, and it will be nice to have another graphics test outside of GLBenchmark to add to our suite, especially since we'll have comparison points between the three major mobile OSes.

Today's podcast covers two major topics: the rumored BGA-only version of Intel's Broadwell (2014/2015) CPU and Qualcomm's disclosures at its recent analyst day. The Broadwell BGA topic spawned a pretty big discussion about the future of the DIY desktop PC market, including speculation about the return of software unlockable CPU upgrades. On the Qualcomm side, Brian gives us the first details of the upcoming Krait 300 core.

The AnandTech Podcast - Episode 12
featuring Anand Shimpi, Brian Klug & Dr. Ian Cutress

iTunes
RSS - mp3, m4a
Direct Links - mp3, m4a

Total Time: 1 hour 23 minutes

Outline - hh:mm

As always, comments are welcome and appreciated. 

We're back! Brian and I recorded this one just after the holidays last week. Despite there not being a lot going on release wise, we had a bunch to talk about. Brian gave us updates on his experience with the Lumia 920, Droid DNA and Samsung's Galaxy Camera. I talk a bit about what the future holds for driving smartphone costs down, and we both talked about Nintendo's Wii U.

The AnandTech Podcast - Episode 11
featuring Anand Shimpi & Brian Klug

iTunes
RSS - mp3, m4a
Direct Links - mp3, m4a

Total Time: 1 hour 48 minutes

Outline - hh:mm

As always, comments are welcome and appreciated. 

Last month I reviewed Microsoft's Surface RT tablet, and came away generally impressed. The form factor and attention to detail were both much better than expected from Microsoft. The integration of the touch/type covers into the design was very well executed in my opinion. That being said, Surface RT seemed to me like a great start but not the perfect product. I would love to see a Cortex A15 based version with some minor tweaks. We'll likely get that next year, but before then there's one more Surface tablet that we'll meet: the Surface Pro.

Surface RT is Microsoft's Windows RT (Windows on ARM) launch vehicle, while Surface Pro is based on Intel x86 hardware. Despite the funny wording in today's blog post, Surface Pro uses an Ivy Bridge based Core i5 (ULV) processor with Intel HD 4000 graphics. Contrary to what I assumed initially, Surface Pro will launch with a 17W Ivy Bridge CPU - so this is the same chip you'll find in modern Ultrabooks. Without a doubt we'll see a Haswell version sometime next year, but not at launch. I wondered if we might see Microsoft use Intel's upcoming 10W Ivy Bridge, but at this point that seems unlikely.

Surface Pro keeps the same display size, but increases tablet thickness by 43% over the RT version. Weight is also up by half a pound. Screen resolution goes up as well, at 1920 x 1080. Memory capacity also increases to 4GB, and Surface Pro comes with much more NAND on-board. With a 7-series chipset you get SATA support, so my money is on Surface Pro having a full blown SSD inside instead of something eMMC based.

Battery capacity goes up to 42Wh, an increase of 33%, putting it about on par with the 3rd and 4th generation iPads. Charger size also goes up to 48W compared to 24W with the RT version. Update: Microsoft announced via its Surface Twitter account that the Pro version would offer roughly half the battery life of Surface RT. Without S0ix support, Surface Pro should look a lot like a standard Ultrabook when it comes to battery life. If you want the best of both worlds, Haswell will be what you'll need to wait for.

The big news is we now have pricing for Surface Pro: $899 for the 64GB model and $999 for the 128GB model, both available in January 2013. Both versions come with a Surface pen, but neither includes a touch or type cover. Microsoft's Surface Pro pricing is clearly higher than any other ARM based tablet, but I'd look at it more as an Ultrabook/MacBook Air alternative. I'll reserve final judgement for when I get my hands on a review sample, but I'm pretty interested to see how the Pro does in our tests. This could end up being one of the better Ultrabooks. I do wish Microsoft had thrown in a touch or type cover into the bundle though, that would make it a real alternative to a standard Ultrabook without having to pay for anything else. It is entirely possible that Microsoft is banking on notebook users bringing a more traditional keyboard and mouse for work though.

The other big omission is the lack of Thunderbolt support. I don't know what it is with Microsoft's crusade against Thunderbolt (the port is no longer on Acer's W700 either), but I think that's a big mistake. Surface Pro would be a great platform for Thunderbolt in my opinion.

For full specs check out the Surface Pro on Microsoft's site.

In our review of Samsung's SSD 840 Pro I noted that my drive died shortly after I completed testing. Samsung sent me a replacement, which also stopped working (although it still pulled current) after a couple of days. Kristian's Samsung SSD 840 review sample shared a similar fate.

I spoke with Samsung about this problem a couple of weeks ago and was told that there was a bug in the pre-production firmware (version 2B0Q/5B0Q for 840 Pro/840) loaded on our drives. All retail samples should ship with a newer firmware revision (3B0Q/6B0Q) that have this bug fixed. To confirm what firmware revision is on your drive, look at the end of the hardware id string for the SSD in Device Manager.

Samsung sent me an 840 Pro with the updated firmware and so far I haven't had any issues. I'm trying to retrace my steps in bricking the drive and things are looking good thus far. As always, if things change I will  update you all.

We've been discussing AMD's Enduro 5.5 update for a few months now; when last we checked in, everything was almost in place...everything except fully optimized DX9 support. A little over a week ago, AMD posted one final (maybe?) beta release for Enduro laptops, and since we still have the AVADirect P170EM available we quickly ran it through our benchmarks one last time. Not a whole lot has changed, and some of the games even show minor drops—basically margin of error stuff—but let's just run the numbers.

Note: The Enduro 9.01 Beta is similar but not entirely the same as the Desktop/Mobile 12.11 Beta8 driver. We only tested a couple titles with the latter and saw lower performance, so we recommend the Enduro driver for Enduro enabled laptop users—which makes sense, given the driver name. I've referred to the Enduro 9.01 Beta below as 12.11 Beta8, as I believe they're the same core driver but with some extra Enduro tweaks present.

For many of the games, we see only minor differences; in fact, of the fourteen tested games, only two show more than a 5% difference, but those two show major improvements. Civilization V finally fixes the inexcusably poor performance from the 7970M and boosts frame rates by 50% while DiRT: Showdown also gets a major kick in the pants and is up 65% compared to the previous Enduro beta.

Overall, those two results are enough to give an average performance increase of 5%, but really it's the difference between borderline playable and smooth performance in those specific games. The only game where we're still down a ways from the best result we've seen on 7970M is Portal 2; the DX9 Hotfix we tested still manages about 10% better performance than we're seeing with the 12.11 Beta8 Enduro driver

Looking at the big picture, the 7970M still can't max out details on every single title you're likely to play—Sleeping Dogs and The Witcher 2 in particular are brutal, with Battlefield 3 and Guild Wars 2 also hovering closer to 30FPS than we'd like. Still, you're not going to get much faster should you opt for the NVIDIA GTX 680M at this point. Here's the latest breakdown of how the two top mobile GPUs compare:

NVIDIA still holds punishingly large leads in Batman: Arkham City, Borderlands 2, Diablo 3, and Portal 2 (35%, 57%, 113%, and 43% faster, respectively), which is what keeps the GTX in the overall lead by 17%. AMD on the other hand comes away with sizeable wins in DiRT: Showdown and Sniper Elite V2. The remaining games are all close to the 10% or less range, and if we discount the big wins we just listed, the eight other titles are basically a wash—NVIDIA wins by 1%. But that's painting perhaps too rosy a picture for AMD, considering the four vs. two major victories.

The fact is, NVIDIA is still ahead on the drivers and support side of the equation. The latest Enduro driver certainly closes the gap, but NVIDIA has been better about getting updated mobile drivers out with support for new releases for two years now. I suspect if I were to go out and test Black Ops 2 and Assassin's Creed 3, I'd be more likely to see additional large NVIDIA wins—especially since those titles carry NVIDIA branding. On the other hand, Hitman Absolution, Far Cry 3, and Medal of Honor: Warfighter are all AMD branded games that have released recently (or in the case of FC3, it releases in the next week), so they ought to have a lead in those games. If I get a chance, I'll try to confirm performance on those titles, since all are likely to be popular.

We're still back to the same old refrain: where AMD really needs to work is on their drivers and updates. The 7970M is their halo mobile GPU, so it's getting good support (now). In fact, most of the 7000M products are receiving decent support, but earlier Enduro/Dynamic Switchable Graphics laptops are still a bit finicky about getting drivers installed. My overall feeling is that NVIDIA has also had more high profile games in the past couple years; AMD looks like they have a good list of titles for this holiday season, and hopefully they can continue that.

Personally, I loved Deus Ex: Human Revolution, and the DiRT series has been pretty popular as well. Sleeping Dogs looks pretty cool, but I haven't had a chance to play it much so I can't really comment there. Sniper Elite V2 on the other hand has not received particularly high marks, and neither has Nexuiz. NVIDIA also has a lot of multi-platform (re: console) games on their list, but those are still some of the better sellers. Basically, NVIDIA historically ends up with co-branding on more major releases than AMD, and with the latest talks about AMD sell offs and buy outs circulating, it's a scary time for computer enthusiasts looking at AMD—though their GPU division is definitely a brighter spot on the list.

It's taken roughly a year for Enduro to reach where we wanted it to start, and anyone who bought in early is probably more than a little irked. I'm still looking for improved support of HD 6000M/"5000M" laptops with Dynamic Switchable Graphics, but in another year even that won't really matter as it will be time for an upgrade. If you've already got an AMD-equipped laptop, the latest drivers are definitely worth a download. If you're on the fence, we're to the point where waiting for the next mobile GPUs (e.g. "8000M" and "700M") might be the way to go.

I noted in my review of the LG/Google Nexus 4 that the device included hardware necessary for LTE on at least some of its bands, namely bands 1 (2100 MHz), 2 (1900 MHz), and 4 (AWS 1700/2100) based on what I saw in my teardown. Enabling LTE on a given device requires everything in the cellular chain to include support — the cellular baseband to support it must be present and loaded with the appropriate software, the transceiver must be the appropriate kind, and finally the right power amplifiers (PAs) have to be in place for the wider channel bandwidths that LTE brings (up to 20 MHz) over WCDMA  (5 MHz). 

 
Nexus 4 on Band 4 LTE (Left), Band Preference Setting in *#*#4636#*#* (Right)

In the case of the Nexus 4, the hardware includes the latest and greatest cellular hardware from Qualcomm with MDM9215M, its third generation 28nm Category 3 LTE multimode baseband, and a WTR1605L transceiver. I tore down the Nexus 4 to ascertain whether PAs were present that could work with LTE, and saw indeed that at least bands 4, 2, and 1 did have Avago power amplifiers (A5704, A5702, and ACPM–7251) which noted support for LTE. The remaining piece of the puzzle was software stack, both in Android and inside the version of the AMSS (Advanced Mobile Subscriber Software) running onboard MDM9215M.

Recently some Nexus 4 owners in Canada posted on XDA that they had working LTE support on Band 4 if they enabled the appropriate "Preferred Network Type" in the dropdown menu there. This menu is inside "Phone Info" which has been part of Android forever and can be accessed on almost every Android phone either by dialing *#*#4636#*#* (INFO) or using an app called Phone Info which launches that activity directly. 

Recently, Anritsu graciously loaned me an MD8475A signaling tester and LTE/WCDMA/CDMA2000 base station emulator for me to test and evaluate devices with. I tested the Nexus 4 on DC-HSPA+ for the review but didn't think to try testing LTE on the appropriate bands since Google and the FCC documents are both explicit about only WCDMA/GSM being present. I tested for LTE on all the bands that the Nexus 4 includes UMTS support for, starting with 5 MHz wide LTE channels. It appears that Nexus 4 only has support for LTE on Band 4 (AWS) with bandwidths up to 20 MHz in fact. I put together a table for easier parsing. If a band isn't listed, it isn't supported. Update: Lots of people have asked me to test some additional bands (3, 7, 20) because of their international relevance, these unsurprisingly are not supported given the fact that there are no PAs present for them. I've updated the table to be explicit however.

Because testing the 2x2 MIMO configuration requires cabling up the Nexus 4 to the antenna leads directly, and my LG-appropriate antenna connectors haven't arrived yet, I only can test with a 1x1 configuration. The Nexus 4 does include Rx diversity for every band, and thus it's entirely possible on Band 4 that users will see the full 2x2 MIMO rates (37 Mbps on 5 MHz, 73 on 10 MHz, 100 Mbps on 20 MHz), but I can't confirm it directly. For the 1x1 configuration I could test, however, I saw the expected ~75 Mbps on 20 MHz FDD LTE which is very close to the maximum of 75.376 as shown. 

The conclusion is that the Nexus 4 at present curiously includes the software profile on MDM9215M that enables support for LTE on Band 4 (AWS), and users only need to set the appropriate network type preference in Android to use it. None of the other bands that there are even power amplifiers for have LTE support, which is unfortunate for users in places where carriers aren't running LTE on AWS, such as the USA. For example, in the USA, AT&T previously discussed plans for LTE on Band 4 but has only rolled out LTE on Band 17 to date, and is rumored to be turning to refarming its PCS (1900 Band 2) and Cellular (850 Band 5) holdings for additional LTE capacity, perhaps in the stead of AWS. T-Mobile US however will use AWS for LTE. Nexus 4 LTE support is definitely unofficial (and somewhat surprising) at this point, but if you're lucky enough to be in a place where your carrier has rolled it out on Band 4, it's just a setting away.

 
Still working on Android 4.2.1...

Update: This morning an update started rolling out to the Nexus 4, 7, and 10. I suspected initially that Google would quickly disable LTE on band 4 on the Nexus 4 for regulatory reasons (or file a Class II permissive change with the FCC). Interestingly enough the OTA going out to the Nexus 4 bringing it to build JOP40D doesn't disable LTE on band 4 or change the baseband software version at all. I was able to make the device attach the same was as I was previously. This update does fix the missing December option for birthdays in the People application, however. 

Source: XDA Developers

We've made it to 10 episodes of the AnandTech Podcast! As promised, this week's episode is a bit more PC focused as we discuss the future of AMD. Intel's SSD DC S3700 is up for discussion, as well as the HPC space including the launch of Intel's Xeon Phi (baesd on the architecture formerly known as Larrabee). 

There's a bit of mobile discussion in the second half of the podcast, addressing TI's exit from the market and some final thoughts on the Nexus 4 from Brian.

The AnandTech Podcast - Episode 10
featuring Anand Shimpi, Brian Klug, Ryan Smith & Dr. Ian Cutress

iTunes
RSS - mp3, m4a
Direct Links - mp3, m4a

Total Time: 2 hours 3 minutes

Outline - hh:mm

As always, comments are welcome and appreciated. 

We've been waiting for affordable LCDs to start showing some clear improvements in performance, quality, features, and design for a while. AOC has potentially done exactly that with their new i2367fh monitor, delivering an IPS-based panel with virtually no left/right or top borders. The bezel measures 2mm thick, making it a potentially great option for multi-monitor users. It's currently available online starting at $189 (though it's also currently backordered).

Besides the thin bezel, the 23-inch i2367fh has WLED backlighting for the IPS panel, built-in speakers, VGA and two HDMI inputs to drive the 1080p display, and audio in/out ports. The stand doesn't appear to support any adjustments other than tilt, and there's no VESA mount, so this is definitely a niche product. However, it's a niche product that looks quite nice and should at least match what we've seen from TN panels for years at only a slight premium.

The display is so new AOC doesn't even list the i2367fh on their US/English sites yet, so the best information we can provide is via this ~7MB PDF, or if you prefer here's their Taiwan page. All we need now is for B&H to get more in stock (and for other resellers to show up). I don't know how others feel, but I'd love to see more displays ditch the 1" bezels—including on the bottom, please. There's nothing a thick bezel adds other than size, though obviously making a thin bezel does cost more.

Update: Some are questioning whether the content of the display goes to the edge (or 2mm from the edge), or if AOC is simply using glass over the main LCD to give a borderless appearance. I asked AOC for an answer, but all I had to go on initially is the press release stating it has a 2mm bezel. To me, the bezel means the area between the edge and the display content, so 2mm would be great. Unfortunately, this unboxing page indicates that AOC is mincing words and using the term bezel to mean the distance between the edge and the glass cover. I apologize for the confusion, though an 8-bit IPS panel for under $200 is still a nice change of pace.

Pursuant to our coverage of Aleutia's Relia fanless industrial PC, we received feedback from readers to provide more coverage for the industrial / embedded PC space. In this market, reliability and ruggedness are prime concerns. Habey USA (part of the NORCO group) has been in the industrial PC space since 1991. Recently, they have introduced two fanless PC models, one satisfying the requirements of the higher end market (where a lot of computing power is required), and the other for specialized embedded applications where the processing power provided by the Atom lineup is more than enough.

BIS-6922

The BIS-6922 is Habey's high-end offering with support for both Ivy Bridge and Sandy Bridge CPUs, using the QM77 / QM67 chipsets respectively. The units are passively cooled with Habey's custom ICEFIN technology. The motherboard itself has 2 SO-DIMM slots, two GbE LAN ports and a 2.5" HDD / SDD bay. In terms of expansion slots, we have 2x mini-PCIe, 1 x PCIe x16 and 1 x PCIe x4 connectors. Two simultaneous displays can be driven with the VGA, DVI-D, HDMI and 24-bit LVDS intefaces.

Pricing varies from $416-455 for the barebones version (board and chassis), based on the quantity purchased. Full systems (including CPU, RAM and storage) are priced around $995 for an i3, $1195 for an i5 and around $1250 for an i7 CPU. Detailed specifications are provided below.

BIS-6636

The BIS-6636 is a small form factor Intel Atom based system with dual NICs and up to 6 COM ports. Cedar Trail's PowerVR SGX545 GPU coupled with the VXD and VXE decoder and encoder engines allow Habey to advertise this unit for digital signage and NVR (networked video recorder) applications amongst other things. In terms oS support, we have drivers for Windows 7, Windows 7 Embedded and also embedded Linux OSes such as Timesys Fedora.

Pricing information is not available yet. Detailed specifications are provided below.

When Nintendo announced the Wii U there was a lot of interest in its hardware specs. The new console could very well end up the fastest kid on the block thanks to the sheer age of the Xbox 360 and PlayStation 3. I wasn't all that interested in the Wii U, but I did want to get a look at the silicon inside so I grabbed a unit and went into teardown mode.

The Wii U is available in two versions: the basic and deluxe kits. The basic kit comes in white, while the deluxe is black. Both are glossy. The deluxe console gives you 32GB of on-board storage (eMMC NAND) vs. 8GB for the standard model. The deluxe kit also includes a stand and cradle for the Wii U tablet GamePad as well as a stand for the console itself. The two kits retail for $299.99 and $349.99, respectively.

The Wii U hardware itself looks a lot like a larger Wii. Getting inside the chassis is also quite similar. The CMOS battery door is the first thing you'll have to remove, followed by 8 more screws (a mixture of philips head and tri-wing). The bulk of these screws are behind console-colored stickers, be sure to peel them all off. With all 8 (9 including the CMOS battery door) screws removed, you can slide the left side of the Wii U away from the front of the console, and off all together. This reveals the final three tri-wing screws that you'll need to remove to get inside the chassis (the gallery of all of this is at the bottom of the article).

With all 12 screws removed, pry the top of the Wii U up and away from the body until it separates from the rest of the frame revealing the console's fairly compact internals:

The Wii U optical drive uses a custom format for game storage, but offers a very Blu-ray-like 25GB capacity per disc. Max sequential read speeds are pretty high compared to the current gen consoles at 22MB/s.

Two screws hold the front cover in place, followed by four screws that hold the optical drive in place. Be careful when removing the optical drive as there's a ribbon cable attached to the motherboard for power/data. The same goes for the front cover, although its ribbon cable is really only on light/switch duty.

With the optical drive removed, next up is removing the shielding on the top and bottom of the motherboard and the shround on top of the heatsink. Just go around the perimeter of the motherboard removing screws (you'll be able to remove all but two easily). Once you've done this, the motherboard will be able to separate from the Wii U's lower tray.

Removing the shielding itself requires carefully moving the antenna cables out of the way. As these wires are soldered to the Wii U chassis on one end, be very careful not to pull too hard otherwise you run the risk of needing to break out the soldering iron.

If you've removed all of the screws and freed the antenna wires from their guides a bit, you should be able to pull back the plastic heatsink shroud, revealing...more shielding:

With no screws left to hold it in place however, the shielding is easily dealt with (again pay close attention to the antenna wires). The same is true for the bottom of the PCB.

The two sets of antenna wires go to two independent wireless controllers: one for 802.11b/g/n WiFi, the other to a dedicated 802.11n controller to handle Miracast display streaming between the Wii U and the GamePad display. Thanks to Ryan Shrout over at PC Per for figuring this one out!

Once you've removed all shielding you're left with a pretty clean looking motherboard:

On the top side of the board you'll see the eMMC/NAND package, in this case it's a dual-die Samsung eMMC solution (there's another Toshiba NAND device on the back of the board, not for user storage):

There are four 4Gb (512MB) Hynix DDR3-1600 devices surrounding the Wii U's MCM (Multi Chip Module). Memory is shared between the CPU and GPU, and if I'm decoding the DRAM part numbers correctly it looks like these are 16-bit devices giving the Wii U a total of 12.8GB/s of peak memory bandwidth. (Corrected from earlier, I decoded the Hynix part numbers incorrectly). Our own Ryan Smith found a great reference for the original Wii so we can compare memory frequencies. It looks like the original Wii had a 32-bit wide GDDR3 memory interface running at a max datarate of 1.4GHz for a total of 5.6GB/s of bandwidth (excluding eDRAM).

That doesn't sound like a lot (it's the same amount of memory bandwidth on the Nexus 10 and iPad 3/4), but the Wii U is supposed to have a good amount of eDRAM for both the CPU and GPU to use. Also keep in mind that the Nexus 10 and iPad 3/4 have to drive much higher resolutions than the Wii U does.

The Wii U's MCM is unfortunately covered by a heatspreader, but given that I went through all of this to look at the console's silicon, I was going to look at the console's silicon.

Normally to remove an integrated heat spreader (IHS) you grab a sharp blade and go around cutting the (hopefully) glue around the perimeter of the chip. In the case of the Wii U's MCM, the blades I'd normally use were too thick. A few years ago I decided to give shaving with a double edge safety razor a try. My attempts failed poorly, but I had a stack of unused DE razor blades that were thin enough to get the IHS off.

The trick here is to apply enough pressure to the blade to cut through the glue, as simply trying to saw through the glue will take forever. There are two blobs of glue per side, but if you're trying to remove the IHS be careful not to cut through the glue and scrape any of the actual exposed silicon...like I did.

With the IHS off, we have the Wii U's MCM in all of its glory:

There are actually three components on this single package, made in at least two different microprocessor fabs. The multicore PowerPC based CPU is the smaller of the two larger chips. This die is made on IBM's 45nm SOI process. The RV7xx derived GPU is the biggest die on the package, and I'm presuming it was made on a 40nm process. I'm assuming the very tiny die in the corner is actually some off-chip memory. Both the CPU and GPU in the Wii U are supposed to have some eDRAM, although the bulk of it is likely dedicated for the GPU.

Clockwise from the top: CPU, GPU, off-chip memory?

The approximate die sizes for all components on the MCM are in the table below:

If we assume a 40nm process for the GPU, then we're looking at something a bit larger than the RV740. The Wii U does boast backwards compatibility with games made for the original Wii, which is made possible thanks to a shared ISA with the original PowerPC based Wii.

The size comparison between CPU and GPU die shouldn't be too much of a surprise. When building a dedicated gaming machine it always makes sense to throw more transistors at your GPU. The nearly 5x ratio of GPU to CPU die size here is a bit on the extreme side though. I suspect many of the current generation consoles, including the Wii U, suffered from a lack of a powerful yet affordable CPU solution that could be easily implemented. 

I also took some power measurements on the Wii U. The system is powered by a 75W external power supply, but total system power consumption doesn't even hit half of that (at least with the games I tried):

Rendering the Wii U menu actually consumes almost as much power as playing Super Mario U. Watching a movie on Netflix consumes a bit less power, my guess is a lot of the 3D blocks are power gated leaving only the CPU cores and video decode hardware active.

The Wii U ships with its own web browser based on webkit, the user agent string for the latest version of the Wii U's software is: Mozilla/5.0 (Nintendo WiiU) AppleWebKit/534.52 (KHTML, like Gecko) NX/2.1.0.8.21 NintendoBrowser/1.0.0.7494.US.

Pages load quickly and compatibility is surprisingly decent (HTML5 test: 258 + 4 bonus points). By default you control and view the browser on the GamePad, but you can also choose to display the content on your TV via the console. Scrolling is very smooth and the overall experience is way better than what you'd normally expect from a web browser on what's primarily a game console. It's not quite as good as using a modern tablet, but still usable. And where there is a browser, we will run SunSpider on it:

The browser is using an older version of WebKit, which is likely responsible for some of the not absolutely stellar performance here.