I received a tip yesterday with proof that OCZ's Everest controller was actually a Marvell 88SS9174 controller (the same controller used in Crucial's m4, Intel's SSD 510, etc...) with a custom Indilinx firmware. After a bit of digging, it turns out that this is indeed the case (OCZ confirmed it to me earlier today). Although OCZ is working on non-Marvell based solutions, the Everest 1 (Octane) and indeed the Everest 2 (Vertex 4) are both based on Marvell hardware. The firmware is entirely Indilinx's own development, but the hardware is from Marvell. The hardware implementation isn't completely identical as OCZ claims its solutions run at higher clock speeds than the standard off-the-shelf Marvell components. 

This doesn't really change anything but it does explain how OCZ was able to bring two revisions of Everest to market as quickly as it did after the Indilinx acquisition. OCZ and Marvell have been working very closely together for a while now and even announced a native PCIe controller they collaborated on at CES this year called Kilimanjaro. As even Intel has admitted to in the past, the value in delivering an SSD isn't always in controller hardware but rather the firmware and validation.

Update: Just to clarify, my information says the Everest 1 (Octane, Petrol) is a higher clocked Marvell 88SS9174. The Everest 2 (Vertex 4) could very well be the new Marvell 88SS9187 given its significant performance enhancements. If it is the 9187 that could tell us a lot about just how close OCZ and Marvell are, as the Vertex 4 started shipping less than a month after Marvell announced the new controller.

Through Chipworks, we recently learned that Apple's revised A5 SoC (S5L8942) is built on a 32nm Samsung HKMG process. While its presence in the Apple TV (3rd generation) which we've reviewed and inspected is an absolute certainty, up until recently it hasn't been as common knowledge that the iPad 2,4 ($399 cost reduced Wi-Fi model) also contains the same S5L8942 SoC but with both A9 cores enabled, as opposed to the single core A9 configuration of the Apple TV 3.

To quell any possible naysayers, I grabbed the "iPad2,4_5.1_9B176_Restore" image a while ago, opened it up, and inspected the kernelcache for references to S5L8942, which there are. Note that the iPad2,4 is k93aap, there are numerous other references to S5L8942 elsewhere, including the buildmanifest plist, eg "Firmware/all_flash/all_flash.k93aap.production/glyphcharging.s5l8942x.img3." There's no question about S5L8942 (A5R2) being inside iPad2,4.

The end result is that if you're an iPad2,4 owner, you've got Samsung 32nm HKMG silicon inside. There's no doubt about iPad2,4 being out and in the wild either, as a number of geekbench 2 runs have popped up since release on their online result browser, which I've been watching like a hawk. As an aside, performance between the S5L8942 (A5R2) and S5L8940 (A5) is virtually unchanged between the iPad2,4 and iPad2,1 if you compare runs on the result browser.

The interesting other question is whether iPad2,4 owners have improved battery life compared to those with iPad2,1 (WiFi), though admittedly that's not going to be a common upgrade path for existing iPad 2 owners to check out. I no longer own an iPad 2 but am going to set out to measure and compare to see just how much of a difference there is, which in turn might say something interesting about Samsung's 32nm HKMG process.

Overall, Apple porting the existing A5 design to Samsung's 32nm HKMG process makes a lot of sense to both test the waters and figure out any design issues, as well as not jeopardize production for the high end. At this point seeing another future Apple product using Samsung 32nm HKMG would not be very surprising.

32nm A5 in iPad 2,4 (Source: Chipworks)

Update: Chipworks has also updated and confirmed with a teardown of an iPad2,4 that the expected 32nm A5 S5L8942 is indeed inside. 

The 3rd generation Apple TV ships with what Apple tells us is a single-core A5 SoC. After delayering the chip to the transistor gate layer in the new Apple TV, Chipworks just confirmed that the SoC is actually a 32nm dual-core part - presumably with one core disabled.

The iPhone 4S' A5 and the A5X used in the new iPad are both built on Samsung's 45nm LP process. This new A5 in the Apple TV is built on Samsung's 32nm High-K + Metal Gate (gate first) process. When transitioning to a new process node, it's always advisable to have a "pipe-cleaner" part. A small, not overly complex design that you can use to test the process and use to discover any bugs. It also helps if this is a lower volume part as there's always a risk of the new manufacturing process being unable to deliver high enough yields. Apple testing Samsung's 32nm process in the new Apple TV makes a lot of sense. There are far fewer Apple TVs sold than iPhones or iPads, so any troubles on the manufacturing side shouldn't really matter. Furthermore, Apple could also ship die-harvested (1 core disabled) 45nm A5s into Apple TVs if things get really bad.

Either way, it's clear that Apple is testing Samsung's 32nm process and this is likely the node we'll see debut in the next iPhone. As our own Brian Klug pointed out, this is the same part that's used in the new $399 version of the iPad 2 (iPad 2,4). Assuming Samsung's 32nm HK+MG process isn't horribly leaky at this point, we should actually see somewhat better battery life out of this new iPad 2 vs the older 45nm version.

In the short time I spent with our Nokia Lumia 900 review unit, I never encountered any loss of data connectivity issues either on AT&T LTE, AT&T HSPA+. Nevertheless, shortly after the Lumia 900 went on sale this Sunday, some users began reporting issues with the device losing data connectivity on both HSPA+ and LTE. Since those early reports began trickling in, Nokia has now acknowledged a memory management issue is at fault and will issue a software update on April 16th. 

Users who don't want to wait for the Lumia 900 software update on the 16th (which will be made available through Zune like other Windows Phone updates) can also optionally swap the phone at an AT&T store for an updated device. In addition, Lumia 900 buyers from launch through April 21st will receive a $100 credit on their AT&T statement, effectively making the $99 on-contract device free. 

Personally I find the turnaround from device release on April 8th, to acknowledging this data issue on the 11th and providing a patch on the 16th incredibly fast by smartphone ecosystem update standards. Although I didn't encounter any issues with data connectivity on the Lumia 900 we were sampled during the course of preparing our review (and I take pride in usually uncovering these bugs), I did use Nokia's Network Setup application to change APNs back and forth between the supplied LTE-provisioned SIM and my own HSPA+ only SIM, which might have been a factor. Interestingly enough during the pre-NDA cycle some other reviewers encountered data connectivity issues which we were told was the result of a provisioning error. Either way, it's good to see Nokia quickly mitigating these issues with its flagship device.

Source: Nokia Conversations Blog

Just a few months after announcing the IdeaPad S2, Lenovo has brought the Android 4.0 ICS powered 9.7-inch IdeaPad S2109 to light.

In a YouTube video, Lenovo teased some of the specs of the tablet; it wields a 8.9mm-thick gunmetal unibody case wrapped around a 9.7-inch 1024x768 IPS screen. That's not a very impressive resolution, but it does match the iPad 2--which of course has now been superceded by the new iPad with it's 2048x1536 resolution LCD.

The S2109 comes with the now fairly standard setup of a micro-HDMI port, micro-SD expandable memory card slot, and a micro-USB port. Lenovo opted to only equip the S2109 with a 1.3-megapixel front-facing camera, with no rear camera present. Lenovo also claims the battery life for the S2109 will reach 10 hours. The final bit of info we’ve been given is that the back of the device features four SRS TruMedia-branded speakers. Just to leave us eager in anticipation, Lenovo has not yet revealed the SoC nor the onboard memory capacity for the S2109.

With Lenovo having just recently announced the IdeaTab S2, comparisons are begging to be made between the two. The IdeaTab S2 is just a hair larger than the S2109 at 10.1 inches and packs the brand new Qualcomm S4 MSM8960. That’s just a fact to keep in mind while we wait for further details about what the S2109 may possess. For now price and availability are unknown, but we’ll keep you updated as further details arise.

For a while now, a regular feature in our smartphone reviews has been comparison of smartphone voice quality on different devices, air interfaces, and codecs. Until recently, improving voice quality has been something of a secondary objective for wireless carriers, and improvements in voice codec efficiency has been used to increase total call capacity rather than improve cellular voice quality. 

Today however Sprint announced the introduction of its HD Voice initiative which will launch in conjunction with the HTC EVO 4G LTE. Our own Jason Inofuentes is at the event and will have impressions of the device soon, which is based around a 1.5 GHz MSM8960 dual core Krait SoC. For the HTC EVO 4G LTE, HD Voice consists of one part common mode noise rejection using two microphones (something we've seen ship on high-end smartphones for a while now), and one part 1x-Advanced.

Part of the CDMA 1x-Advanced specification is inclusion of support for higher quality EVRC voice encoding. At present, virtually all the CDMA2000 carriers in the USA use EVRC-B for their voice calls, which has adaptive bitrates of between 4 and 8 kbps and sampling of up to 8 KHz. Sprint has talked rather publicly in the past about upgrading its CDMA2000 1x network to 1x-Advanced, and that brings me to HD Voice.

I reached out to Qualcomm and asked what voice codec was being used in conjunction with Sprint's HD Voice branding on the HTC EVO 4G LTE, and learned that EVRC-NW (Service Option 73) is being used. EVRC-NW (Narrowband-Wideband), as the name suggests, includes both the EVRC-B rates with narrowband 8 KHz sampling, and EVRC-WB rates with 16 KHz sampling all under one umbrella.

The end result is double the frequency bandwidth for voice calls, and after Nyquist you get 50 Hz - 7 KHz  of frequency response. That's still a ways away from the widely cited 20 KHz upper bound for human auditory perception, but dramatically better than the present 50 Hz to 4 KHz of frequency range. In fact, 7 KHz of bandwidth should be sufficient to replicate human voice completely. EVRC-NW is superficially analogous to AMR-WB, and I've listened to a number of AMR-WB versus AMR-NB demos that are pretty compelling. Sprint moving on to wideband audio is definitely a welcome move here, and we look forward to testing it out. The only remaining question is just how aggressive and quick Sprint will be with its 1x to 1x-Advanced update. 

It's been a busy spring for Intel. The Intel SSD 520 series was released in February, and a month later, Sandy Bridge EP was released. Only another month or so after that and the Intel SSD 313 Series was released. The upcoming weeks look at least as busy because Intel is preparing Ivy Bridge and 7-series chipsets, but at the same time Intel has been preparing something else in the SSD front: the Intel SSD 330 Series. It will be the successor of 320 Series, Intel's current mainstream SSD offering, and we have some preliminary specs that we would like to share:

Note: These specifications are not official and have not been confirmed by Intel. These have been taken from resellers' websites and may contain inaccurate information.

If the specifications are correct, the 330 Series will finally bring SATA 6Gb/s support for Intel's mainstream SSDs. This is a very welcome feature because while the 320 Series isn't slow, it's still somewhat expensive when compared with the performance it provides. Other manufacturers' SATA 6Gb/s SSDs can look a lot more appealing as they are often 10% (or more) cheaper and the advertised speeds are over twice as fast.

The actual controller is unknown but the capacities hint towards SandForce. That would make sense because Intel has spent a lot time and money on their SandForce driven 520 Series. The only part of the equation that doesn't make sense is the performance. We are used to seeing 550MB/s read and 500MB/s write on SandForce SSDs but the specifications we have seen so far show something else. The interesting part here is whether this is hardware or firmware related. We will have to wait and see.

As for the pricing, the 330 Series appears to be very competitive. Right now you can get the 80GB Intel 320 Series for $135, which is only $15 cheaper than the 120GB 330 Series. Once again, I would like to emphasize that these are by no means official prices and may change when retailers get some actual stock. Official availability is unknown but Amazon.co.uk is listing April 13 as the release date, which is ten days away. I also emailed the reseller we got the specifications from and will update this article once I receive a reply.

Sources: Jimms.fi, (1) (2)

A bit over a month ago, we reported that Intel's SSD 313 would be launching soon. We don't know  when exactly the launch took place but the 313 has been added to Intel's product database (ARK) and Q1'12 is listed as the launch date. The 313 is the successor of Intel's first generation caching SSD, the Intel SSD 311. We first met the 311 when Intel introduced its Smart Response Technology (see our review). Right now, only Z68 chipset supports SRT but Ivy Bridge will widen the support and bring it to more mobile and business chipsets as well.

As we suspected back in February, the controller is the same as the one found inside Intel's SSD 311. This was confirmed by HWBox.gr. The 3Gbps controller dates back to 2009 when Intel X25-M G2 was released, although it's obviously running much newer firmware.

Courtesy of HWBox.gr

The picture of the PCB also reveals the NAND configuration. HWBox's unit is a 20GB model and it has five NAND devices onboard. Not surprisingly, it's Intel NAND and manufactured using the new 25nm process node. Due to the die shrink, each NAND device consists of a single 4GB die. The previous generation used two 2GB dies per NAND device. It seems most likely that the 24GB model simply has an extra NAND device onboard, giving it a capacity of 24GB. 

The 313 seems to be priced identically to the 311. Price per GB is quite high at ~$6/GB, although that's fairly normal for SLC SSDs. NewEgg is already stocking the Intel SSD 313, making availability immediate. 

Sources: Intel, Intel, HWBox.gr

At the beginning of the year LSI completed the acquisition of SandForce, the SSD controller maker we've been covering extensively since its arrival on the scene in late 2009. Although the SandForce business model hasn't changed as a part of the acquisition, LSI is definitely going to start leveraging its newly purchased assets to deliver more SF-powered, LSI branded solutions - particularly for the enterprise.

The first example of this is the new Nytro family of products. These are all PCIe NAND solutions, aimed at various targets in the enterprise, and based on SandForce's SF-2500 series of controllers.

The first member of the family is also the most obvious: the Nytro WarpDrive. Take either 4 or 8 SandForce controllers, put them behind an LSI 2008 SAS RAID controller, throw in a bunch of NAND, put it all on a PCIe 2.0 x8 card and you've got a Nytro WarpDrive. Capacities start at 200GB (SLC) and go all the way up to 3.2TB (MLC). The WarpDrive will be available in both half-height (4 x SF controllers, 1.6TB max) and full-height (8 x SF controllers, 3.2TB max) flavors. Later this year LSI will offer a PCIe 3.0 version of the Nytro WarpDrive that will enable speeds of up to 4GB/s. Pricing for the new WarpDrive will start at $6600 for the 200GB SLC model. The beauty of using an LSI SAS RAID controller on-board is that any OS with native support for the controller will, by default, support the Nytro WarpDrive.

The next product is a little more ambitious as it combines a Nytro WarpDrive with LSI's own SSD caching software into a solution called the Nytro XD ("xcelerated disk"). If you can't (or don't want to) fit all of your data on a single PCIe SSD, the Nytro XD allows you to use a WarpDrive to cache your direct attached storage (SAS) or SAN array (iSCSI, FC). The hardware is identical to the Nytro WarpDrive, but the XD package comes with LSI's caching software and a support contract. Prices start at $9400 including a 1-year support contract.

Last but not least we have the most interesting product in my opinion: the LSI Nytro MegaRAID. Take a standard MegaRAID SAS controller but combine it with on-board NAND, seamless caching and you've got a Nytro MegaRAID card. Without any additional software, all requests that are made to the RAID array attached to the Nytro MegaRAID card are cached in its on-board NAND. This completely redefines what we've come to expect from a RAID card and it can stand to deliver a pretty significant improvement in performance. Since the solution is entirely self contained you don't even need to install any caching software, the end user only sees a more-expensive MegaRAID card. Once again, as long as your OS has driver support for the MegaRAID controller (Windows, Linux, VMWare should be fine) then the Nytro MegaRAID will work. Pricing on the Nytro MegaRAID starts at $1699, although LSI isn't providing details on just how much NAND is on-board. 

In order to help users decide on the right capacity/NAND type for their Nytro purchases, LSI is making available workload analysis software. It's called the Nytro Predictor software and it will look at the access patterns of your workload and give you recommendations as to what specific Nytro products would be best suited for your needs. As we found in our look at the enterprise performance of Intel SSDs, understanding your workload is absolutely necessary when it comes to picking the right SSD and NAND solution.

Google Chrome 18 is working its way out to users' computers today after its release to the stable channel. As usual, the update brings a number of small feature and security enhancements to all users running the browser.

The new features both relate to graphics acceleration in the browser: it enables GPU-accelerated Canvas 2D to Windows and OS X users with supported GPUs. Canvas 2D is primarily used in games and had previously been implemented in software, but GPU acceleration for the feature had only been enabled in the beta channel. The other change takes a GPU accelerated feature, WebGL, and implements a software version for older computers via a product called Swiftshader. Software rendering is slower than GPU-accelerated rendering, naturally, but this at least extends basic functionality to users of unsupported and older configurations.

On the security side, there were nine bugs patched. While the specific exploits used against Chrome in this year's Pwnium competition had been fixed shortly after they were discovered, the new patches are the first of Google's efforts to harden the browser against similar exploits in the future.

Google Chrome can be downloaded here, and is compatible with Windows XP and higher, OS X 10.5 (Intel) and higher, and many Linux distributions.

Source: Google

AMD has released the latest version of its Catalyst graphics driver package for 32-bit and 64-bit versions of Windows XP, Vista, and 7. This version of the drivers introduces support for the Radeon HD 7800 series under Windows Vista and 7 - the 7700 and 7900 series cards were added in the last release, meaning that the Catalyst drivers now support the complete lineup of 7000-series cards. Windows XP support for these cards is slated to be introduced next month in Catalyst 12.4.

The new drivers also fix issues in a number of games, including a texture corruption bug in Skyrim that some of you mentioned was a problem with the 12.2 drivers. Other resolved issues include bugfixes in games like HAWX, Quake 4, XPlane, Alan Wake, and Far Cry 2, a crashing bug in the Furmark benchmark, and startup issues with the Catalyst Control Center.

The drivers support all Radeon HD 2000, 3000, 4000, 5000, 6000, and 7000 series graphics cards, IGPs, and APUs except where otherwise noted in the release notes, which are linked below. Support for the mobile versions of these cards is provided only to Windows Vista and 7 users.

Source: AMD

Microsoft's Jerry Koh and Jeff Piira gave some insight into the Windows 8 touch experience in yet another Building Windows 8 blog post yesterday. They talked specifically about the type of touch hardware that would be required for Windows 8-certified tablets - the touchscreens in Windows 8 and Windows on ARM tablets will need to recognize at least five simultaneous inputs, have good edge detection, and accurately register 95% of all touch input. 

To ensure a decent experience on Windows 7 tablet hardware, the gestures needed for basic OS navigation require no more than two fingers, though tablets with limited multitouch capabilities may not be able to use apps or features that require more complex gestures. To compensate for tablet hardware with poor edge detection, Windows 8 can use a 20 pixel buffer around the screen to help register edge gestures, but the space used for the buffer cannot be used to register other touch input. Various sensitivity issues may also cause problems with individual taps, swipe to select, swipe and slide, and swipe from edge on Windows 7 tablets.

For more information, including specific Windows 7 tablets that Microsoft has used for internal testing, the full post is linked below.

Source: Building Windows 8 blog

We got a chance to check out Novatel's next generation MiFi at CES this past January, and came away impressed with the update. Officially titled the Novatel Wireless MiFi 4620L 4G Global Ready Jetpack, it has big shoes to fill as the successor to the Wireless MiFi 4510L, which was our favorite LTE hotspot device when we reviewed it late last year. The Jetpack 4620L includes global WCDMA and GSM/EDGE support in addition to CDMA 2000 1x, EVDO and Verizon LTE support. Unfortunately, 5GHz WiFi is still not on board, as 802.11 connectivity is limited to b, g, and 2.4GHz n bands. But the most interesting thing about the Jetpack 4620L is the small OLED screen on the front, giving access to settings menus, device information, connectivity, and battery status. This makes it significantly easier to use and troubleshoot than the previous MiFis, and in our brief use, we found the display to be quite responsive and intuitive. 

At the time, we didn't have any details on pricing or availability, but earlier today, Verizon announced that the Jetpack would go on sale April 12th for $49.99 after a $50 mail-in-rebate, with data starting at $50 a month for 5GB. We'll have a review unit soon, so stay tuned!

Just today, Google announced that it was rolling out Android 4.0.4 to the Android Open Source Project (AOSP) for download, with build IMM76D. The note at that point by JBQ was that the timeline for ROM push for the final build was unknown. 

In a tweet and Google+ post, Google has now noted that Android 4.0.4 images are beginning to push out for the GSM/UMTS Galaxy Nexus and Nexus S, and Xoom WiFi. Absent are the CDMA/WiMAX Nexus S on Sprint and the CDMA/LTE Galaxy Nexus on Verizon, both of which are likely waiting on carrier approval. The update includes bugfixes, stability improvements, faster camera launch (which we've noticed in leaked 4.0.4 builds), and better performance in some areas.

We’ve started rolling out Android 4.0.4, Ice Cream Sandwich, to UMTS/GSM Nexus S, Xoom Wi-Fi, and HSPA+ Galaxy Nexus devices. #android

— Android (@Android) March 28, 2012

We're anxiously waiting for our own GSM/UMTS Galaxy Nexus to get the push so we can test out the final build and see what all has changed. Hopefully Google also updates the factory images for Galaxy Nexus pages for those of us who are impatient. 

Source: Google

If the numbers are true, then most of you have already read our Kepler review, and you know that the card has made quite a splash - it's the highest-performing single-GPU card you can buy today, and it's got solid power consumption and a lower price than the AMD Radeon HD 7970 to boot. Kepler still needs to trickle down through the rest of NVIDIA's lineup, but for now NVIDIA has the high-end sewn up. Let's look at what its partners have put together.

As we've noted in past recaps, you should take these card measurements with a grain or two of salt. Manufacturers haven't standardized on a unit of measurement for their cards - some measure in inches and some in metric. I've done the necessary conversions and presented all measurements in both inches and millimeters, but manufacturers play a bit loose with these measurements and the actual physical dimensions may not exactly match the dimensions given on the spec sheet.

Common to all of these cards is 2GB of GDDR5 on a 256-bit bus and all of Kepler's features - in fact, most of these cards have pretty much everything in common with one another, from the across-the-board stock clocks to the display outputs to the single-fan, dual-slot coolers to the lackluster bundles of accessories. This isn't uncommon with high-end launches of all-new architectures - we saw the same thing happen in our Radeon HD 7970 launch recap, another crop of cards that stuck to the reference design.

As such, there's not a ton to say about them, so I'll just make notes below when there's something about the card that makes it different from the stock card that we reviewed a couple of days ago.

ASUS (Product page)

EVGA (Product page)

Galaxy (Product page)

Gigabyte (Product page)

MSI (Product page)

MSI's graphics cards usually have a 3-year parts and 2-year labor warranty, and this card is no exception.

PNY (Product page)

This card is the only one in the lineup that costs more than $500, and there are a couple of reasons why: one is the lifetime warranty you can get by registering the card, and the other is the bundled HDMI cable. It's the only card in the lineup with anything more than power cables and DVI to VGA adapters. It's also the only card for which I can't find measurements (Amazon lists the length at eight inches, which I find suspect since the rest of the cards are at least ten). The card's dimensions should be similar to the others.

Zotac (Product page)

Zotac's is the only card in this lineup with a 2-year warranty instead of the 3-year warranty shared by most of the rest of them.

It has been a couple of weeks since we reviewed the Radeon HD 7800 series, but as we mentioned earlier this week and in our 7850 recap, that was just a paper launch - the cards hit the street only recently, and as usual we're going to go through all of the stuff from AMD's partners and give you the Facts.

As we've noted in past recaps, you should take these card measurements with a grain or two of salt. Manufacturers haven't standardized on a unit of measurement for their cards - some measure in inches and some in metric. I've done the necessary conversions and presented all measurements in both inches and millimeters, but manufacturers play a bit loose with these measurements and the actual physical dimensions may not exactly match the dimensions given on the spec sheet.

Common to all of these cards is 2GB of GDDR5 on a 256-bit bus, Eyefinity support, two 6-pin power connectors, and all of GCN's features. All but one of the cards also offer identical outputs: two mini DisplayPorts, one HDMI port, and one DVI-I port. The PowerColor PCS+ card also offers a second DVI-I output.

ASUS (Product page)

ASUS again uses its DirectCUII cooler on its 7870 - this cooler has made appearances in many of our other launch recaps, including that for the 7850, where the cooler was actually a good bit longer than the card itself. Since the 7870 is a longer card, that isn't an issue here. As with its 7850, ASUS applies a paltry 10MHz overclock to the core and the memory, but the bundled accessories are nothing to write home about - the biggest reason to choose this card over others is the 3-year warranty.

Gigabyte (Product page)

Gigabyte's 7870 employs a massive three-fan cooler, the better to cool its 100MHz (10%) core overclock, which is the highest clock in our recap - it's tied with one of the PowerColor cards, and while that one is $10 more expensive, it also has a slight memory overclock. The Gigabyte card's memory remains at stock clocks - if you've been following these recaps for awhile, you've probably noticed that factory overclocks tend to focus on the core rather than the memory - only three of the seven cards here have memory cards, and none of them are higher than 4%.

MSI (Product page)

Like many of the cards here, MSI's 7870 has a two-fan cooler with a big heatsink, but otherwise it has a hard time distinguishing itself from the crowd - it's tied for the most expensive card, but it has only a modest 50MHz core overclock and a three-year parts and two-year labor warranty that falls in the middle of the rest of the pack. 

PowerColor (Product page)

As is often the case in these recaps, both PowerColor and Sapphire are offering two versions of the 7870, one with stock clocks and a slightly more expensive model with a factory overclock. This is the stock clocked version, and it's the only card in this lineup that uses AMD's reference cooler for the 7870 series.

PowerColor PCS+ (Product page)

This PowerColor card is $10 more expensive than its lower-end cousin, but it comes with a 100MHz core overclock and 25MHz memory overclock that should net you an increase in frames per second. Its also the only card here with a second DVI port, which it adds to the 7870's standard complement of Mini DisplayPorts and HDMI. If you value warranty length over factory overclocks, though, this one only has a two-year warranty to its name.

Sapphire (Product page)

This card has the same stock clocks and 2-year warranty as the PowerColor card, but it's $10 cheaper (the cheapest card in the recap), includes a better accessory bundle, and uses a two-fan cooler with a more impressive heatsink. 

Sapphire OC (Product page)

This card is identical to the other Sapphire offering in almost every way - the warranty, included accessories, and cooler are all the same. Your extra $10 gets you a 50 MHz overclock on both the core and the memory - if you're not comfortable doing your own overclocks, you can spend the extra $10 and get a few frames per second for it. If you do your own overclocks, save the cash.

It has been weeks since we reviewed AMD's Radeon HD 7870 and 7850 cards, but unlike the 7900 and 7700 series cards, the 7800 series was given the typical middle-child treatment and paper launched. and cards began appearing at retailers just this week.

While Kepler's launch has cast a long shadow over the top end of the graphics market (a GTX 680 recap is coming later today, dont worry), but competition is still fierce, and as we noted in our review the 7850 is a solid performer and the fastest 150 watt card on the market today. Let's look at what AMD's partners have for us.

As we've noted in past recaps, you should take these card measurements with a grain or two of salt. Manufacturers haven't standardized on a unit of measurement for their cards - some measure in inches and some in metric. I've done the necessary conversions and presented all measurements in both inches and millimeters, but manufacturers play a bit loose with these measurements and the actual physical dimensions may not exactly match the dimensions given on the spec sheet.

Common to all of these cards is 2GB of GDDR5 on a 256-bit bus, Eyefinity support, and all of GCN's features. All cards also offer identical outputs: two mini DisplayPorts, one HDMI port, and one DVI-I port. Normally we see a range of prices from different manufacturers due to factory overclocks, longer warranties, or included accessories, but in this case we've got identical prices across the board, making it much easier to make an apples-to-apples comparison among cards. As long as you don't have any particular brand loyalty, just pick the one with the value-added extras that you need the most.

ASUS (Product page)

The ASUS 7850 features a 10MHz overclock on both the GPU and the RAM, but it's so small that it won't increase framerates much at all over stock clocks. Its bundle of accessories is pretty sparse, but its 3-year warranty is tied with the Gigabyte card for the longest of the bunch.

This ASUS card's defining characteristic is the DirectCUII cooler, a huge two-fan cooler that was actually designed for longer cards like the Radeon HD 7950. On the shorter 7850, it hangs over the end of the card by quite a bit, requiring the use of an extension cord to make the 6-pin power connection accessible. This move may make the GPU cooler (and, by extension, get you a better overclock), but it will also require a larger case.

Gigabyte (Product page)

The Gigabyte card has a bit in common with the ASUS card - a big fancy two-fan cooler, a 3-year warranty, a bare accessory bundle - but it's a bit shorter in length, and it features an impressive 115MHz (about 12%) overclock on the core, which should actually net you a measurable increase in game performance. The memory clock , however, is left at stock.

HIS (Product page)

Here's a first in AnandTech Graphics Card Launch Recap History: the dimensions for this HIS card aren't available through Newegg or HIS's product page, or anywhere else that I can find (the product page gives "box dimensions", which is useful if you're shipping the card but not if you're using it). Luckily, its humdrum single-fan cooler means that the card should be unremarkable in this regard - I'd guess it should be close to eight inches long.

Otherwise, HIS doesn't give you much in terms of value-adds - it uses stock clocks, the two-year warranty is the minimum I like to see on components that cost this much, and the DVI to VGA adapter and Crossfire bridge constitute a pretty small accessory bundle. 

MSI (Product page)

With the MSI card, we're back to custom two-fan coolers and big heatsinks. A 40MHz (~4.5%) core overclock is respectable but small, and it uses stock memory clocks. A 3-year parts and 2-year labor warranty splits the difference between the longest and shortest warranties on the list.

Where MSI beats the competition is in its accessory bundle, which is actually worthy of the name - in addition to basics like power cable adapters (the Newegg product image appears to include two of these, though it only has the one six pin power plug on the back) and a DVI to VGA adapter, it also includes a Mini DisplayPort to DisplayPort adapter.

PowerColor (Product page)

The PowerColor card is a lot like the HIS model in its single-fan cooler, 2-year warranty, and stock clocks, but it adds some useful display adapters to the package. PowerColor's card is the only one here that's using AMD's reference cooler for the 7850 series (visible on this page of our review).  

Sapphire (Product page)

Sapphire's take on the 7850, which uses another big two-fan cooler, is the only one in the list with a memory overclock worthy of the name. The 50MHz (4%) RAM overclock along with the 60MHz (6.5%) core overclock should give you a noticeable increase in framerates if you're not comfortable doing your own overclocking. Other benefits include the respectable accessory bundle and other drawbacks include a shorter 2-year warranty.

The road to any new microprocessor design is by no means simple. Planning for a major GPU like NVIDIA's Kepler starts four years prior to the chip's debut. In a world that's increasingly more focused on fast production and consumption of everything, it's insane to think of any project taking such a long period of time.

Chip planning involves figuring out what you want to do, what features you want, what the architecture should look like at a high level, etc...  After several rounds of back and forth in the planning stage, actual architecture work begins. This phase can take a good 1 - 1.5 years depending on the complexity of the design. Add another year for layout and validation work, then a 6 - 9 month race from tape out to products on shelves. The teams that spend years on these designs are made up of hard working, very smart people. They all tend to believe in what they're doing and they all show up trying to do the best job possible. 

Unfortunately, picking a target that's 4 years out and trying to hit it better than your competition is extremely difficult. You can put in an amazing amount of work, push through late nights, struggle with issues, be proud of what you've done and still fall short. We've seen this happen to companies on both sides of the fence, whether we're talking CPUs or GPUs, you win some and you lose some. 

Today NVIDIA unveiled Kepler, a more efficient 28nm derivative of its Fermi architecture. The GeForce GTX 680 is the first productized Kepler for the desktop and if you read our review, it did very well. As our own Ryan Smith wrote in his conclusion to the GeForce GTX 680 review:

"But in the meantime, in the here and now, this is by far the easiest recommendation we’ve been able to make for an NVIDIA flagship video card. NVIDIA’s drive for efficiency has paid off handsomely, and as a result they have once again captured the performance crown."

We've all heard stories about what happens inside a company when a chip doesn't do well. Today we have an example of what happens after years of work really pay off. A trusted source within NVIDIA forwarded us a copy of Jen-Hsun's (NVIDIA's CEO) email to all employees, congratulating them on Kepler's launch. With NVIDIA in (presumably) good spirits today, I'm sure they won't mind if we share it here.

If you ever wondered what it's like to be on the receiving end of a happy Jen-Hsun email, here's your chance:

Microsoft's David Washington has penned another informational tome on the Building Windows 8 blog, this one about Windows 8 and its support for varying screen resolutions. The above chart lists the common (but not the only) resolutions that Microsoft is planning for, and while most of the listed display types won't surprise anyone (wall-to-wall 1366x768 and 1920x1080 for most desktops and laptops), it does appear as though Microsoft is planning for Windows tablets with a DPI that approaches or matches that of the new iPad.

Microsoft is planning for tablets that use both the 1024x768 and 1366x768 resolutions common in earlier and lower-end tablets as well as the high-DPI screens that are being (and will be) ushered in by the new iPad. To scale Windows elements so that they're still comfortable to look at and touch at these resolutions, Microsoft has put together some pre-defined scaling percentages: 100% when no scaling is applied, 140% for 1080p tablets, and 180% for quad-XGA tablets like the new iPad. These percentages were all chosen as "pixel density sweet spots" for 10" and 11" tablets with 1920x1080 or 2560x1440 displays. It should be noted that Washington's blog post focused entirely on Metro scaling - whether the Windows desktop will automatically scale using these percentages is unclear.

Microsoft's attention to these specific resolutions suggests that we will probably see some high-DPI Windows tablets when they launch in the fall, though we still don't know anything about the tablets OEMs are designing for Windows 8 and Windows on ARM. It's also telling that there are no 7" tablets on that chart - we may not see Windows versions of smaller tablets like the Kindle Fire or Nook Tablet.

Washington went on to explain the reasoning behind the minimum resolution requirements for Metro apps that we noticed in our Windows 8 preview review - 1024x768 for Metro apps and 1366x768 for the Metro Snap feature. Both choices were largely developer and data-driven: 1024x768 is a common low-end resolution for web developers and tablet app developers, and Microsoft didn't want to restrict these developers to a lower minimum resolution to account for the small percentage of 800x600 and 1024x600 displays that are currently in use.

As for snapped apps: the size for a "snapped" app is always 320 pixels wide, which was again selected because developers have become used to it in their work with smartphones. A 1366x768 display is the lowest common screen resolution that allows for the 320 pixel width and the 1024 pixel minimum width for regular Metro apps.

Also discussed was the methods by which Metro allows programs to expand to take up all of the pixels in a larger laptop or desktop display: To help dynamically expand content to take up more screen space when the pixels are available, Windows 8 uses the same XAML and CSS3 features that are commonly used to accomplish this on modern web pages - examples of such features include the grid, flexible box, and multi-column CSS3 layouts. App templates provided with Visual Studio 11 all make use of these features automatically. Developers can also scale their apps to fit larger displays, which is useful for games or other apps that don't need to make use of additional pixels.

For more, including Windows 8's support for scalable graphics and the Windows Simulator tool that will provide Visual Studio 11 users the ability to test their apps at multiple screen resolutions, the full post is linked below for your convenience.

Source: Building Windows 8 blog

Asus's Transformer Prime just got some company. Available for pre-order today, the Acer Iconia Tab A510 brings the price of entry for a 10.1" Tegra 3-powered tablet down to a cool $449.99, $50 less than the similarly equipped Asus offering. Like the Prime the A510 features a 10.1" 1280x800 display, the 1.3 GHz Tegra 3 SoC with 1 GB of RAM and 32 GB of storage expandable by microSD. The base battery life on the A510 is an impressive 36.26Whr, not quite as much as the new iPad, but somewhat higher than its prececessor and the Prime. That big battery does lead to a somewhat portly frame, with a thickness cresting a centimeter and weighing nearly 100g more than the Prime. The frame is similar to the A200 we saw in January, but is actually a little thinner and with a textured back for extra grip. 

Android 4.0 is on order for software, complete with Acer's Ring UI, a relatively innocuous skin that mainly seeks to put your most commonly used apps in easy reach. When we took a look at the A500, we were pleased with its display quality, not quite IPS but great for a vanilla LCD; we hope we can expect more of the same from this display. Software pre-load includes the usual branded media players and print software, along with Polaris Office 3.5 for productivity. Gone though, is the full-sized USB port, replaced by microUSB, though it remains compatibile with portable HDD up to 2TB in size. 

There's no shortage of options for tablet buyers right now, and everyday another pops up. But if performance, battery life and price are your main criteria, the A510 may just be the tablet for you. Pre-orders start today for $449.99 at your favorite e-tailers; no ship dates are available.