Solid state storage is getting too fast for SATA. The most common interface for both SSDs and hard drives is serial ATA--it's been a computer mainstay since it replaced IDE about a decade ago. SATA revision 3.0 offers data speeds of approximately 600 MB/s, far faster than any 7200 RPM hard drive can read or write. But SSDs are now pushing up against that limit. They can go faster. What do we do? Turn to a new interface.

Judging by Apple's new Mac Pro, 2013 MacBook Airs and the way Ultrabooks are headed, that new interface will be PCI Express. PCIe and a new SSD product line called NGFF, or Next Gen Form Factor, are going to make Ultrabooks and flash memory-equipped laptops dramatically faster over the next two years. Let's look at how and why.

Photo credit: iFixit

A presentation from 2012's Flash Memory Summit does a good job of addressing why the industry is transitioning away from the SATA interface, at least in thin laptops. It's the data rate, as we already said--but why hasn't there been a new iteration of SATA to keep ahead of flash memory speeds? The presentation explains that "Enabling SATA beyond 600 MB/s is a long term development effort. Single lane scaling beyond ~8Gbps is challenging & requires trade-offs. Multi-lane SATA requires a new connector and modified chipset SATA controllers to make multi-lane software transparent."

PCIe doesn't have those issues. "To enable higher speed client SSDs in near term ('13/'14), PCIe is the only choice," the presentation states. "PCIe has bandwidth lead (1GB/s with Gen3). PCIe has multi-lane for scalability (x2, x4, ...). Software compatible PCIe SSDs can be built as a single port AHCI device.

Until now, mSATA has been a common SSD solution for Ultrabooks and other laptops. But beyond the speed concerns, mSATA presents some problems. At a whole five millimeters, it's too thick, and it's difficult to add extra NAND to. That's why the inudstry is moving to the Next Gen Form Factor, also called SATA Express, which is designed to support a number of sizes and interface via PCIe.

Photo credit: Anandtech

Anandtech wrote a bit about NGFF last fall, noting that it supports two sockets, one for SATA and PCIE x2 interfaces perfect for lower-end SSDs and other add-in cards like Wi-Fi. The second socket supports PCI Express up to 4 GB/s, which is crazy fast. Four gigs a second. Perhaps even crazier: SATA Express cards are even smaller than mSATA cards.

The Flash Summit Presentation also shows how this standard is going to allow for much more flexibility than mSATA, so while all NGFF cards will be thinner and narrower than mSATA, some will be longer. That's to support a standardized approach to SSD sizes. The smaller cards will be for minor storage and caching and will be either 42 or 60 millimeters long. 80 and 110 millimeter cards will hold more flash storage.

At the WWDC opening keynote on Monday, Apple announced a new Mac Pro. People were surprised--because Apple has ignored the Pro line for years, and because the new computer looks less like a computer and more like a classy trashcan. Ars Technica has taken a smart, critical look at the Mac Pro to talk about what's new, exciting, and not-so-exciting about the freshly engineered computer.

Ars' Dave Girard praises the engineering that went into the cylindrical body, which is cooled by a single large fan near the top. That's impressive, given the CPU power and dual-GPU setup of the Mac Pro. You can see how the whole thing fits together in a cool animated page on Apple's site. Girard also praised the PCIe storage in the system, which will offer speeds approximately double current SATA-based SSDs.

So where does the criticism come in? Expandability. The new Mac Pro is clearly engineered around external, rather than internal, expansion. Apple's website says as much. The lack of internal drive bays means the system will be relying on that PCIe memory or external hard drives, likely RAID arrays or network-attached storage. Granted, the professionals who buy Mac Pros will often have RAID setups already, but some of them will still pine for internal storage.

"The four internal drive bays of the existing Mac Pro enclosure became a comfy standard for me and my work," writes Girard. "Anything more seems like too many—but zero extra drive bays is, to put it mildly, too few. Now I will be forced to replace my existing eSATA RAID enclosure since eSATA/Thunderbolt adapters are stupidly expensive and there are no PCI slots in the machine to accommodate an eSATA adapter card. Considering the still-high price of external Thunderbolt enclosures, the price of the Mac Pro better be reasonable because it’s clear that many of us will be forced to take this route as well."

Girard makes another strong point: the focus on external expansion may encourage Thunderbolt adoption, both by the companies making hardware and the professionals deciding between PC and Mac for video editing. If the Thunderbolt market thrives, it's very good for Mac.

Two more expandability problems: the dual workstation GPUs are built onto the Mac Pro's motherboard, and there are only 4 USB 3.0 drives. Again, Apple's obviously pushing Thunderbolt over USB 3.0, but the GPU situation is troubling. It looks like Apple is putting some really powerful cards in the Mac Pro, but not being able to upgrade them 3-4 years down the road limits the computer's longterm usability.

Check out the rest of Girard's post for more on Apple's implementation of GPUs, Crossfire support, and the lack of an Nvidia option.

Is it a trash can? No. Is it a time capsule? Nope. Is it a torpedo? Not even. Surprise--it's a Mac Pro! After several years with no major updates, Apple showed off a brand new Mac Pro desktop at this year's WWDC keynote, and it may possibly the most un-computer-looking computer any major company has ever released. Apple's ditched the classic blocky tower design for a rounded cylindrical body, though they've kept (and significantly improved on) all the power inside the workstation, which is aimed at professional video editors, photographers and designers.

The Mac Pro got the audience hyped, but it wasn't Apple's last hardware announcement for the keynote. In fact, the average consumer has something much better to look forward to--new models of the 11-inch and 13-inch MacBook Air running on Intel's Haswell platform. We recently touched on the hardware details of Haswell and what the new processor line means for laptops, but here's the short version: Sweet, sweet battery life.

Apple announced that its 11-inch MacBook Air is gaining 4 hours of battery life, jumping from 5 hours to 9 hours of use. Even better, the 13-inch model is increasing its 7 hours of battery life to 12 hours. All-day battery life, meet the Mac. It's taken a few years, but the great battery performance of netbooks has finally made its way into much, much more usable computers.

Those estimates are likely on the generous side, as battery estimates almost always are, but those numbers are still close to double what last year's Ivy Bridge platform delivered. Along with drastically improved power usage, Haswell is granting 40 percent faster graphics performance and better low-power states, meaning a MBA can sit in standby for a solid 30 days.

Apple's also throwing 802.11ac Wi-Fi support into the MacBook Air, which will come in handy in the next couple years as the router market moves to the new standard. Apple also pointed out that the new MacBook Airs come with 45 percent faster flash memory. Coupled with that low-power Haswell performance, the system can pop awake from sleep mode in about 1 second.

And more good news: Pricing has come down on the MacBook Airs, though only slightly. The 128GB 11-inch Air starts at $999 (last year's model started at 64GB), with the 256GB model costing $1199, a hundred bucks cheaper than last year's. The 128GB model of the 13-inch Air starts at $1099 (again, a hundred bucks off) and the 256GB model costs $1299.

Also, they're on sale as of today. As in, right now, on Apple's website.

Now, about that new Mac Pro: It looks weird. And nice--but weird. It's also impressively smaller than the old Mac Pro tower--Apple says it's an eight the volume of the previous generation. But what's it got inside?

As I said in my very first PC building story, I think 8GB is the sweet spot for RAM right now, across AMD and Intel platforms. To quote that post:

"RAM is cheap these days. 8GB is plenty for gaming, and an 8GB kit of DDR3/1600 costs around $40. There’s no need to cut corners farther than that. Get a kit with two 4GB sticks, so you can utilize dual-channel mode on your motherboard. You’ll be able to add another 2 x 4GB kit later, or even a 2 x 8GB kit, as you’ll be getting a motherboard with four RAM slots. That’ll let you upgrade to 24GB of RAM before you even consider tossing the RAM you have."

I could stop this guide right there, but it bears fleshing out a little bit. So we'll take it step by step: first we'll talk about how much RAM you need, then what speed. We'll dip into multi-channel mode, speeds and timings, and making sure your RAM works well with your system. Don't worry--we'll talk about the just-announced Haswell chipset from Intel, too.

Photo credit: Flickr user jjackowski via Creative Commons.

Eight Is Enough (For Now)

Why 8GB? I'd consider 4GB the rock-bottom, barrel-scraping minimum amount of RAM necessary to run a modern computer--say, an ultrabook--in such a way that it doesn't usually feel like it's getting hung up on everyday tasks. Even then, I'd try my damnedest to get more RAM. But if you're building a desktop, it's likely because you want more computing power than you get from an ultrabook--either for gaming or some other high-performance task. 8GB is enough for even very intensive games, though video editors, programmers, and people who do a lot of memory-intensive work will want more RAM. 8GB is a good starting point, and you can build from there if you find it isn't enough.

Just by way of comparison, here's my PC using 4GB of RAM while only running CrashPlan, Dropbox, Rdio, and about 15 Chrome tabs (including several Google Docs tabs). Imagine if that's all the RAM you had to work with.

For most general-purpose and gaming rigs, 8GB is the price/performance sweet spot, and this is likely to remain the case with the next generation of AMD and Intel chipsets as well.

Interestingly, DDR3 prices have gone up in the past few months: A good 8GB DDR3/1600 kit now costs around $60, rather than the $40 it cost when I wrote my first column in December. Acer's JT Wang says it's because DRAM factories are prioritizing smartphone DRAM over desktop DRAM.

Let's start with a simple question: how accurate is your gaming mouse? If your answer is in terms of DPI--maybe the number you've read off the side of the mouse' package--you're omitting a lot of attributes and variables that affect the accuracy and performance of a typical mouse. That's part of the reason it's difficult to objectively evaluate a gaming mouse. So much of user experience lies in subjective factors: the physical sculpting and weight of a mouse, your preference for button surface textures, etc. These are the things that you notice immediately and affect your day-to-day use, while sensor quality more often than not just has to pass a threshold of acceptable responsiveness and accuracy for non-professional gamers.

When I visited Logitech's Borel Innovation Center facility last week, I spoke with the company's engineers about the process of designing a gaming mouse and learned about the tracking variables that they care about when testing mouse accuracy.

Full disclosure: Logitech paid for my trip to their laboratories in Lausanne, Switzerland, but we were under no obligation to produce video or write about anything I saw there. The information I learned from Logitech's engineers is genuinely interesting to me from both a consumer and product reviewer's perspective, and the insights about mouse tracking variables are applicable any gaming mouse, whether it's made by Logitech or a competitor.

Intel's Haswell CPUs are coming. On the desktop side, we've already dug into what that means--a new socket, lower power draw, and some nice performance improvements for heavy users. But what about Ultrabooks? On Monday Intel released details on the U-Series processors designed for those lightweight laptops Intel is so fond of. These dual-core, 15 watt TDP chips will be powering Ultrabooks later this year. So what do we have to look forward to? You may sense a theme, but lower power usage is a big draw for this mobile year's chips.

Ars Technica covered the details of the Haswell mobile processors, pointing out that the ultra low voltage chips this generation are even more power efficient than they first appear. The Ivy Bridge ULV chips ran with a 17 watt TDP, but that figure didn't include the chipset, which added another three watts. Ars writes "The 15 watt TDP figure for the U-series Haswell chips includes the chipset, making for a 25 percent combined reduction in TDP. Combined with the "active idle" power state, which allows Haswell laptops do more when idle and transition between active and idle states more quickly, Haswell has the potential to enable significantly better battery life for general-usage workloads."

Since the chipset is integrated into the CPU this time around, it frees up a little space for other Ultrabook components. Specs-wise, the U-series CPUs don't offer much surprising. They're dual-core processors with Hyperthreading and clock between 2.9GHz and 2.3GHz, between the high-end i7 and low-end i5. Two bottom of the barrel i3 processors are stuck around 1.8GHz and drop Intel's Turbo Boost which enables even higher clock speeds. The i3s and i5s feature 3MB L3 cache, while the i7s bump that up a bit to 4MB.

The big difference lies in the GPUs, which we wrote about in early May. Here's a refresher: Intel estimates its new HD Graphics 5000 will deliver 1.5 times the graphics performance of last year's GPUs. Intel's top-of-the-line integrated graphics, the Iris 5100, will offer about double the performance of last year's GPUs, but is too power hungry for these ULV processors. You'll only see it in bigger laptops.

The two fastest i7 and i5 processors in the U-Processor line will run on Graphics 5000, while the lower entries in each category will stick with Graphics 4400, which are lower-power GPUs. Ars Technica writes that the faster GPUs (coupled with lower power draw) could give Ultrabooks a big competitive edge compared to larger laptops--they're still thin and light, but now they have some real graphics power. CPU performance may feel identical to last year's Ivy Bridge laptops, however:

The massively deep reviews on Haswell, Intel’s fourth generaton Core CPU, have now hit the web. Making sense of all the data is more difficult, though. So let’s begin with a chart, rather than burying it later in the article.

Whoa, look at that idle power number!

This is the system power usage for three different systems. The key takeaway is the idle power number: 24 watts less than Ivy Bridge at the same clock speed. Most desktop PCs run at idle or near idle far more than they run under load. For example, when you’re typing in Microsoft Word, your system is really idling for the most part. So keep that idle power number in mind as we try to answer question of whether desktop PC users should upgrade to Haswell.

The Modern Desktop CPU

Haswell is considered a new architecture by Intel – a “tock” in their tick-tock product development model. If you look at just the CPU cores, the overall enhancements are evolutionary. Branch prediction has been beefed up a bit, the cache has been improved. Most of the changes to the actual core are minor; even Intel admits that.

Note the phrase “No change in key pipelines.” The CPU cores themselves have only been tweaked in minor ways.

The main changes to Haswell have been in two areas: power and graphics.

The most important change on the power side is moving the voltage regulators onto the CPU itself, which Intel dubs “FIVR” – fully integrated voltage regulator. It’s just another step in the road to moving increasing functionality to the CPU itself. Building the voltage regulator onto the CPU reduces platform power, and also enables Intel to more closely couple the built in System Agent to the voltage regulator. Additionally, Intel has tweaked overall power management inside the CPU and GPU, enabling much faster switching between PC power states.

All that taken together enables the exceptional idle power performance. And this is just the desktop CPU, Intel’s Core i7 4770K, which lacks the deepest S0ix “active idle” state that will be part of the mobile versions of Haswell.

On the graphics side, the main Intel building blocks have been beefed up.

While you were sleeping, Computex 2013 announcements went underway at the annual computing convention in Taiwan. Asus made a ton of product announcements, including updates to its Zenbook, Transformer, and Memo Pad product lines for the rest of the year. Here are the highlights for the products that are the most interesting to us. First is the Memo Pad 7, an update to the poorly-reviewed Memo Pad ME172V. This $130 tablet (8GB, $150 for 16GB) is strikingly similar to the hardware that Asus made for Google's Nexus 7, but adds a microSD slot for storage capacity and a 5MP rear camera. Unlike the Nexus 7, it doesn't use Nvidia's Tegra 3 processor, but runs on MediaTek's quad-core MT8125 CPU and PowerVR's SGX 544 graphics chip. It also doesn't run stock Android. But with no new Nexus 7 announced at I/O, this could be a attractive budget alternative to the already affordable $200 Nexus 7.

Asus's big notebook announcement was the Transformer Book Trio, which actually has two CPUs, two storage units, two batteries, and can run two operating systems. Basically, it's a Haswell laptop and an Android tablet crammed into one device, sharing a 1080p touchscreen. On the PC side, it runs a Haswell i7 processor for Windows 8 and a 1TB HDD with a 33Whr battery, with that hardware on the base/keyboard section. Under the display is an Atom processor, 64GB SSD storage, and a 19.5Whr battery to run Android. Asus is calling it a "three-in-one" device since it can also dock to an external monitor and peripherals to act as a dedicated desktop PC. No word on pricing or availability.

For PC traditionalists, the Zenbook Infinity will make more sense. It's an Ultrabook with a Haswell processor, but is significantly thinner than Asus' previous Zenbooks (15.5mm). And while Asus hasn't confirmed this, some reports are indicating that the Zenbook will have a display resolution of 2560x1600 on its 13.3" screen, with touchscreen as an option too. Intel released its Haswell embargoes over the weekend as well, and we'll have more on that later today.

If you’ve been told that PC gaming is all about (insert booming voice here) maximum performance, you’ve been told wrong. Most people, even hard core PC gamers, have budgets. So buying the latest, greatest and absolutely fastest piece of hardware sounds great, until you digest the price.

Last week, we tested Nvidia’s new GeForce GTX 780. It’s certainly true that the GTX 780 is one fast card. It’s also true that it costs $649 or more. That’s a pretty serious amount of cash. Ideally, you want to maximize your performance per dollar. Enter the GTX 770.

The GTX 770 reference card looks almost identical to the GTX 780, but the GTX 770 uses a GK104 GPU.

I’m not going to talk much about the GTX 770 architecture, because the 770 is really a beefier GTX 680. That is, it’s a GK104 chip, just like the GTX 680, with the same number of shader cores. But it also boasts two or four gigabytes of 7 gigabit per second GDDR5 – the fastest graphics memory you can get today.

Nvidia is targeting a $399 price point for the 770. That’s only a little more than the current 670, cheaper than the existing 680s and much less than the new GTX 780. It’s also about on par with average price for Radeon HD 7970 GHz Editon cards, and about $75 more than the average Radeon HD 7950. However, the higher speed memory and higher clock speeds do come at a cost. Let’s first parse the specs, comparing the GTX 670, 680 and 770.

It’s safe to say that Nvidia is really competing with itself at this point in time. The current GeForce GTX 680 is pretty much even in performance to AMD’s Radeon HD 7970 GHz Edition, but much quieter and uses less power. The GeForce Titan outperforms AMD’s single-GPU flagship by a wide margin, but costs a cool grand, so it’s out of reach of most users.

Enter the GeForce GTX 780. At first blush, it seems like a “Baby Titan”, but that would be inaccurate. Let’s look at the base specs, compared to both the Titan and the GTX 680.

Given that the GTX 780 uses the same GPU chip as the GTX Titan, but with roughly 15% fewer shader cores and half the memory, the GTX 780 offers about 80% of the gaming performance of a Titan, as we’ll see shortly. Take a look at that memory speed, too: 7000 MHz (effective), or 1gpbs faster throughput than the Titan or GTX 680. There’s no lack of memory bandwidth with the GTX 780. However, Nvidia told us that the GTX 780 would only have about a quarter of the double precision floating point performance of Titan. In other words, the GTX 780 will be a great gaming card, but won’t come close to Titan for high end GPU compute.

Digging a little deeper into the features of the GTX 780 card itself, Nvidia’s made some interesting design decisions in the reference design. The cooling subsystem is tweaked from Titan to run even quieter. Nvidia accomplished this by managing fan speeds to run closer to a steady state, rather than ramping the fan speeds up and down rapidly.

The GTX 780’s cooling system minimizes noise by keeping the average speed within a narrow band, avoiding fast spin rate ramps.

The GTX 780 will cost substantially less than a Titan, at about $649 for reference grade cards, but that's nearly $200 more than a 2GB GTX 680. However, 4GB GTX 680s still cost nearly $600, so the price differential between a GTX 780 and GTX 680 4GB card isn’t as large, while new new card offers quite a bit more performance. Still, $649 is a pretty steep price for a video card, and it’s partly a result of AMD’s inability to compete on single GPU performance. The lack of competition puts Nvidia in the enviable position of being able to set higher prices than they might have if competition had been stiffer. I included a GTX 680 4GB card for comparison, but it’s likely that performance differences with a 2GB card will be minor.

With this sobering thought in mind, let’s take a look at performance.

Here's a situation that should sound familiar. You're using your computer. An application locks up. You click on the window, anyway, to see if it does anything. The application hangs. It's not responding. Instead of giving up, and walking away, you keep clicking--maybe on that application, maybe elsewhere on the desktop. Click. Click. The longer you wait, the angrier you get. Click click click. Your computer is no longer under your control. Clickclickclickclickclickclickclickclickclick.

The most likely answer to "Why do users click randomly and rapidly when an application hangs?", recently posed to StackExchange, is irrational frustration. After the first couple clicks, we know an application isn't going to respond, so we keep clicking in frustration. It's an expression of powerlessness. The question garnered some interesting answered that actually dig into the psychology of man-machine interaction.

Photo credit: Flickr user juditk via Creative Commons.

"People tend to think of their interface in physical terms," says one response. "You think of a 'window' not a 'rectangle of lights on a matrix'. And so, when an application hangs, people revert to interacting with it in the way they might do with a physical object when it stops working.Shaking things seems to be the way that many people try to 'fix' a physical object that isn't working, and the digital equivalent to this is to beat it with your clicking might."

Another cites the psychology of learned behavior and quotes the book You Are Not So Smart:

"Take, as an example, a pigeon that has been reinforced to peck an electronic button. During its training history, every time the pigeon pecked the button, it will have received a small amount of bird seed as a reinforcer. So, whenever the bird is hungry, it will peck the button to receive food. However, if the button were to be turned off, the hungry pigeon will first try pecking the button just as it has in the past. When no food is forthcoming, the bird will likely try again ... and again, and again. After a period of frantic activity, in which their pecking behavior yields no result, the pigeon's pecking will decrease in frequency."

But the most interesting response focuses on a reaction that's slightly different from repeatedly clicking a mouse. It talks about the escape key, and how it has evolved into a button that means Stop! Abort! Back out! Get me outta here!

The New York Times magazine wrote about the history of the escape key in 2012; surprisingly, it wasn't originally designed to close out of programs:

Logitech’s series of Bluetooth Easy-Switch Keyboards are the best keyboards to use with most devices, due to the fact that they're a pleasure to type on, are widely loved by reviewers, boast extremely long battery lives, and come in Apple and Microsoft hardware-friendly configurations. They can be paired with up to three different devices at a time and switch back and forth between those paired devices at the push of a button.

Single Use Keyboards Are Lame

A lot of people bite the bullet and buy multiple keyboards to use with the various devices in their lives. They’ll have a keyboard to use with their computer (if they don’t own a laptop,) an external keyboard or a keyboard case to use with their tablet, and maybe one to use with their set top box or HTPC. That kind of thing gets expensive, and wastes money you could be putting towards something awesome.

The smart money’s on owning a keyboard that’s designed to work well with as many of the devices you own as possible.

One of the most interesting messages Google tries to get across in its Chromebook campaign is the idea that the hardware is disposable. If your Chromebook falls into a volcano or gets run over or stolen, you're out the cost of the hardware, but that's it. You don't lose any data, and the crook/volcano god doesn't get access to it either. All you have to do is grab a new Chromebook (or any PC that can run the Chrome browser) log in, and you're back in business.

Photo credit: Alex Washburn/Wired via Creative Commons.

Most of us can't use a Chromebook full-time. We use programs that don't yet run in a web browser, we play games that require local asset files and don't sync to the cloud, and we have a lot of data we need to hold onto--more than will fit onto a few lousy gigabytes of local storage. But we can take a page from the Chromebook, as it were, and make our data resilient and flexible--resilient, so a hardware loss doesn't mean data loss, and flexible, so that we can pick up pretty much any computer with an Internet connection and be able to work. After all, if you lose your Chromebook, you don't need to find another Chromebook to access your data; you just need to log in to your Google Account from anywhere.

In order to get Chromebook-level data security on our "real" computers, we need two things: good backup software, and good syncing software. All of your data deserves to be backed up, but not all of it needs to be immediately accessible. With a good backup, your data is safe, and with a good sync setup, you can have near-instant access to whatever subset of that data you deem worthy. The good news is that this is now really easy.

I'm not just idly pontificating; I just did some spring cleaning, including a clean Windows install on my desktop, and this is how I prepared, backed up, and synced my data.

Note that this guide is written from the perspective of a Windows user, but the main points are valid for Linux and Mac OS X users as well.

Late last week, Intel unveiled some features and performance data for the graphics cores in their upcoming Haswell CPU. Most of the hoopla revolved around Haswell’s graphics performance on laptops, but Intel also disclosed some interesting bits about desktop processors. Before diving into that, it’s worth considering how integrated graphics typically plays out on desktop PCs.

First Puzzle Piece: Performance CPUs Rarely Use IGPs

On the mobile side, most Intel-based laptops currently include their highest end HD 4000 GPU. Laptops are increasingly becoming closed systems, making user upgrades more difficult–and graphics upgrades impossible. So Intel has been fairly smart, integrating its best GPU into all Core class processors. Even Ultrabooks, with their tightly constrained chassis and limited airflow, utilize CPUs with Intel HD 4000 graphics.

People who build PCs tend to be pretty smart about how they’re going to use a system. Building a small, shared living room PC for web access and light office chores? Integrated graphics may be fine, but so is a lower end CPU. Someone who picks up a higher end CPU – a Core i7 3770K, for example – is unlikely to use the integrated GPU. Usually, that system will end up with at least a mid-range graphics card, like a GeForce GTX 660 or AMD Radeon HD 7870.

Intel knows this, and doesn’t really want to spend the die space on putting a higher end integrated GPU into a performance-oriented CPU where the integrated graphics will mostly go unused. A better integrated GPU requires more die space, which increases the overall cost of the processor. That makes sense when you realize that even a relatively low end graphics card, like Nvidia’s GTX 650 or AMD’s HD 7790 substantially outperforms the HD 4000.

Intel's next processor line, Haswell, is closing in on a summer release. Like last year's bump from Sandy Bridge to Ivy Bridge, the most exciting upgrade in this new processor line is the integrated GPU, aka Intel's HD Graphics. Competitor AMD blew Intel's first shot at integrated graphics out of the water, but last year's Ivy Bridge GPU was a big step forward for Intel. And this year things are looking even better--Intel's claiming comparable performance to an (obviously low end) discrete GPU, and they've given their new graphics a name to go with its new performance. Meet Iris.

More specifically, Iris Graphics 5100 is Intel's new high-performance integrated GPU. But it's not the only name to learn this processor cycle: There are five different tiers of GPUs, but the bottom three (which all stick with HD 4000-level denotations) will likely be reserved for low-performance systems. Iris Graphics and the second-tier GPU, HD Graphics 5000, are the ones to keep an eye on.

Intel's 3DMark tests show Iris doubling the performance of last year's HD Graphics 4000, but there's a price to be paid for that performance. Ars Technica writes:

"The Iris 5100 is confined to chips with a 28W thermal design power (TDP), which is a fair bit higher than the 17W TDP used by both Sandy Bridge and Ivy Bridge CPUs. We've talked before about how Intel's TDP ratings (and the newer SDP ratings) are a bit nebulous, but it may be the case that these chips are confined to slightly larger (think 13-inch) Ultrabooks because of power or thermal constraints."

The new HD Graphics 5000, which delivers approximately 1.5 times the performance of last year's chip, will be better suited to laptops with harsher power restrictions. Iris 5100 and HD 5000 are the two chips we'll likely see in most laptops and Ultrabooks in the coming year, but there's also another, even faster Iris GPU with a higher power draw that Intel will offer.

It's taken AMD a few months to plan its counterattack to Nvidia's fastest graphics cards, the dual-GPU GTX 690 and monstrous single GPU GTX Titan. But now that response is here in the form of the Radeon 7990, a dual-GPU card that essentially slaps together two of the (already speedy) 7970GEs, which sell for at least $400 by themselves. AMD's press release proudly proclaimed this was the card EA used to premiere Battlefield 4, and that it's the only one around that can run Crysis 3 and Tomb Raider at 4K resolution.

AMD's hailing the new 7990 as the world's fastest graphics card thanks to its count of 2048 stream processors per GPU, 950MHz core clock, and 6GB of GDDR5 memory clocked at 6GHz. Total, the 7990 is packing about 8.6 billion transistors. None of this comes cheap, of course--the 7990 will launch online in a couple week at $999.

Anandtech has some great analysis of the new card that digs into the advances AMD has made with dual-GPU tech. It's important to note that this technically isn't the first 7990 model to hit the market--ASUS and PowerColor released their own versions before AMD, but it looks like this will be the one to get. Writes Anandtech:

"AMD’s 7990 has an official TDP of just 375W, which although common for official dual-GPU cards, is quite a bit lower than the TDPs of the unofficial 7990s. As the GPU manufacturer AMD has the ability to do finely grained binning that their partners cannot, so while Asus and PowerColor have essentially been putting together cards that really are two 7970s on a single card – right down to the TDP – official 7990s get the advantage of AMD’s binning process, significantly reducing power consumption. The end result is that while an unofficial 7990 would be a 450W+ part, AMD can deliver the same or better performance while consuming much less power, putting the 7990 within the all-important 375W envelope that OEMs and boutique builders look for."

Two other important things to note. One: AMD's new power technology, which they've dubbed ZeroCore, can turn off a slave GPU when not in use. So for your normal day-to-day, the 7990 can shut off one of its GPUs and bring its idle power consumption down to about 20 watts. And that other important thing: AMD is giving away a ton of games with this card.

Your $999 will also buy you copies of BioShock Infinite, Tomb Raider, Crysis 3, Far Cry 3, Far Cry 3 Blood Dragon, Hitman Absolution, Sleeping Dogs and Deus Ex Human Revolution. If spending a grand on a graphics card leaves you too poor to buy anything else for a month, at least you'll have plenty of games to play.