Some Notes About Our Recommendations

• This list is for gamers who want to get the most for their money. If you don't play games, then the cards on this list are more expensive than what you really need. We've added a reference page at the end of the column covering integrated graphics processors, which is likely more apropos.
• The criteria to get on this list are strictly price/performance. We acknowledge that recommendations for multiple video cards, such as two Radeon cards in CrossFire mode or two GeForce cards in SLI, typically require a motherboard that supports CrossFire or SLI and a chassis with more space to install multiple graphics cards. They also require a beefier power supply compared to what a single card needs, and will almost certainly produce more heat than a single card. Keep these factors in mind when making your purchasing decision. In most cases, if we have recommended a multiple-card solution, we try to recommend a single-card honorable mention at a comparable price point for those who find multi-card setups undesirable.
• Prices and availability change on a daily basis. We can't base our decisions on always-changing pricing information, but we can list some good cards that you probably won't regret buying at the price ranges we suggest, along with real-time prices from our PriceGrabber engine, for your reference.
• The list is based on some of the best U.S. prices from online retailers. In other countries or at retail stores, your mileage will most certainly vary.
• These are new card prices. No used or open-box cards are in the list; they might represent a good deal, but it's outside the scope of what we're trying to do.

Best PCIe Card: Under $85

Best PCI Express (PCIe) Card For ~$50:

Radeon HD 4650 (Check Prices)

Great 1280x1024 performance in most games, 1680x1050 with lowered detail

You will not find a card that packs more punch than ATI's Radeon HD 4650 under the alluring $50 price point. With solid stock performance and an overclockable GPU, this card is an excellent starting point for our recommendations, and a wholly worthwhile upgrade if you're currently stuck using a motherboard with integrated graphics.

Best PCI Express (PCIe) Card For ~$65: Tie

Radeon HD 4670 (Check Prices)

Good 1680x1050 performance in most games

With the release of Nvidia's GeForce GT 240, ATI's Radeon HD 4670 is no longer the most powerful reference card without a dedicated power connector. However, it remains a compelling solution under the $75 price point, which Nvidia's solution simply hasn't hit yet.

Performance is excellent and power usage is very low, making this product an impressive performer all-around. Its accelerated clock rates and modestly-higher price tag are worth considering if you originally had your eye on the Radeon HD 4650.

GeForce 9600 GSO (Check Prices)

Good 1680x1050 performance in most games

The GeForce 9600 GSO is seems to be getting quite hard to find, and is likely being end-of-life'd soon in favor of the new GeForce GT 240. Nevertheless, as long as it is available, the GeForce 9600 GSO remains a powerful competitor compared to the Radeon HD 4670. While the GeForce requires a dedicated PCIe power connector to supply more juice than the Radeon, it does offer better performance in some situations.

Best PCIe Card For ~$85:

GeForce 9600 GT (Check Prices)

Good 1680x1050 performance in most games

The GeForce 9600 GT is a great performer, thanks in part to its high-end 256-bit memory interface and speedy DDR3 memory. It's a great choice on an $85 budget, even if the architecture on which it centers is showing its age.

Certainly, this card's continued presence here is a testament to Nvidia's engineering work dating back almost two years ago. With that said, we'd certainly like to see the company's latest DirectX 10.1 cards drop in price to compete against ATI's strong offerings.

Best PCIe Card: ~$90 To $140

Best PCIe Card For ~$95: Tie

GeForce 9800 GT (Check Prices)

Exceptional 1680x1050 performance in most games, 1920x1200 in most games with lowered detail

The GeForce 9800 GT is essentially a rebadged GeForce 8800 GT, and offers the same great performance it has for years now (that sure sounds funny to say in reference to graphics cards).

With the rising price of ATI's Radeon HD 4850 giving it space to breathe, this legendary card is once again a recommended buy. But once again, we're looking forward to seeing technological progress put new, faster, and cooler products loaded down with more features in this space rather than revisiting history.

Fortunately, there's still PhysX and 3D Vision support to like about this aging board.

Radeon HD 4830 512MB (Check Prices)

Exceptional 1680x1050 performance in most games, 1920x1200 in most games with lowered detail

Just as the GeForce 9800 GT can once again be recommended due to the rising price of the Radeon HD 4850, so can the Radeon HD 4830. While availability is low, this Radeon is still a viable option under the $100 price point if you can find it. You'll discovered that it offers great performance on par with the GeForce 9800 GT, with the added benefit of DirectX 10.1 support.

Best PCIe Card For ~$110:

GeForce GTS 250 512MB (Check Prices)

Good 1920x1200 performance in most games

The dissapearance of the $100 Radeon HD 4850 has not only opened up the GeForce 9800 GT and Radeon HD 4830 for recommended status, but also the GeForce GTS 250.

At $110, the 512MB version of this card offers respectable performance, and nothing else in the price range can compare to it. As fast as the Radeon HD 4850 and new Radeon HD 5750 (and notably cheaper), the GeForce GTS 250 has no real competition from the rest of the sub-$150 market at this time.

Bear in mind that going this route instead of the Radeon HD 5750 will cost you DirectX 11 support and Eyefinity. But in the context of gaming, you'll need to make other quality sacrifices long before trying to enjoy either value-add in the $110 range.

Best PCIe Card For ~$120:

GeForce GTS 250 1GB (Check Prices)

Good 1920x1200 performance in most games

For $10 more than the 512MB version, an interested gamer can get the benefit of a full gigabyte of memory. At the highest resolutions and levels of anti-aliasing, this extra memory might provide a performance boost, though it's unlikely the GeForce GTS 250 is powerful enough to run at those detail levels. Still, many buyers might find the slight $10 price increase worthwhile in something like Grand Theft Auto IV.

Best PCIe Card: ~$150 To $290

Best PCIe Card For ~$155: Tie

Radeon HD 5770 (Check Prices)

Great 1920x1200 performance in most games

While the new Radeon HD 5770 isn't any faster than its older Radeon HD 4870 cousin (we've found that it's even slightly slower in many instances), it does have something the Radeon HD 4870 doesn't have: full DirectX 11 and Eyefinity support. Indeed, while the Radeon HD 5770 doesn't run away with any performance crowns in this category, it does look good from a longevity/value standpoint.

Read our full review of ATI's Radeon HD 5770 for more information on the card and its accompanying architecture.

GeForce GTX 260 (Check Prices)

Great 1920x1200 performance in most games

Like many cards, the GeForce GTX 260 is becoming very hard to find, and may soon be end-of-life'd. In any case, it does offer advantages in titles that run better on Nvidia's GT200 architecture, and it sports some GeForce-only value-added features like PhysX compatibility and support for GeForce 3D Vision.

Once again, a little diligence is required on the part of the buyer to find out which card is best adapted for his or her favorite titles, and whether or not your motherboard supports SLI, CrossFire, or both multi-card technologies.

Best PCIe Card For ~$200:

Radeon HD 4890 (Check Prices)

Excellent 1920x1200 performance in most games

The Radeon HD 4890 is essentially an overclocked Radeon HD 4870. However, the tweaks that AMD made to the newer RV790 die result in much higher overclocking headroom. At stock speeds, this card is worth the $200. But to get the most out of it, some overclocking is in order. And now that the prices on Radeon HD 5850 cards are through the roof, there's not much between this board and ATI's next-fastest solution.

Read our full review of ATI's Radeon HD 4890 for more information on the card and its accompanying architecture.

Best PCIe Card For ~$240:

2 x GeForce GTS 250 1GB in SLI Configuration (Check Prices)

Exceptional 1920x1200 performance, 2560x1600 in most games with lowered detail

Two GeForce GTS 250 cards in SLI pack a punch and make a strong case for multi-card setups. With the Radeon HD 4850s going up in price and down in availability, these GeForce cards replace them as the weapon of choice for sub-$300 brute force power.

Best PCIe Card: ~$300 To $400

Best PCIe Card For ~$310: None

Honorable Mention: Radeon HD 5850

Exceptional 1920x1200 performance, 2560x1600 in most titles

The new Radeon HD 5850 has some definite advantages over a pair of GeForce GTX 260s in SLI or a pair of Radeon HD 5770s in CrossFire. It doesn't need a CrossFire-compatible motherboard, it uses a lot less power thanks to its 40nm manufacturing process, and it sports DirectX 11 capabilities (plus Eyefinity).

Unfortunately, scant availability forces us to relegate the Radeon HD 5850 to Honorable Mention status until it can be purchased without having to fight for it.

Read our full review of ATI's Radeon HD 5850 for more information on the card and its accompanying architecture.

Best PCIe Card For ~$330: Tie

At approximately the same price, these options retain the same advantages of their single-card counterparts: two GeForce GTX 260s offer SLI compatibility, PhysX, and GeForce 3D Vision support, and two Radeon HD 5770s offer DirectX 11, Eyefinity, and high-def audio bitstreaming to CrossFire-compatible motherboard users. A good case can be made for either of these options, and none of them are a poor choice. Just pick your poison.

2 x GeForce GTX 260 in SLI (Check Prices)

Exceptional 1920x1200 performance, good 2560x1600 performance in most titles

Nvidia doesn't have a DirectX 11-class architecture yet, so if you're going to sink $300+ into DirectX 10 hardware, do so knowing that there are competing DirectX 11 boards available in the same price range.

A pair of Radeon HD 5770s won't offer quite the same level of performance as two GeForce GTX 260s in SLI; that's the trade-off for more modern functionality, though.

2 x Radeon HD 5770 in CrossFire (Check Prices)

Exceptional 1920x1200 performance, good 2560x1600 performance in most titles

The new Radeon HD 5770 sports added benefits over its GeForce GTX 260 competition: DirectX 11, triple display outputs, and the ability to bitstream high-definition audio content from Blu-ray movies contribute significant value to ATI's newest mainstream graphics cards. For those seeking long-term DirectX 11 compatibility, this might be the more attractive option.

Read our full review of ATI's Radeon HD 5770 for more information on the card and its accompanying architecture.

Best PCIe Card For ~$400:

Two Radeon HD 4890 cards in CrossFire Configuration (Check Prices)

Good 2560x1600 performance in most games

Two Radeon HD 4890 cards should, on average, perform on par or better than a single GeForce GTX 295, and will definitely beat out a single Radeon HD 5870. Plus, these Radeons cost less than either option. If you have a CrossFire-compatible motherboard and want some serious performance at high resolutions, this route is the way to go.

Read our full review of ATI's Radeon HD 4890 for more information on the card and its accompanying architecture.

Best PCIe Card: ~$400 And Up

Best PCIe Card For ~$410: None

Honorable Mention: Radeon HD 5870

Good 2560x1600 performance in most games

For $10 less, a couple Radeon HD 4890s will easily beat a single Radeon HD 5870 in the titles that matter today (perhaps this will change when DirectX 11 software becomes more pervasive). From a raw price/performance standpoint, this makes the Radeon HD 5870 a hard sell. But that is not to say this card is underpowered: it sports the fastest single GPU on the planet, relatively low power usage (remarkably low at idle), and DirectX 11 support. For folks without a motherboard that supports CrossFire and a hefty power supply, the new Radeon HD 5870 is definitely a more-than-viable option. Unfortunately, availability is still quite rare.

Read our full review of ATI's Radeon HD 5870 for more information on the card and its accompanying architecture.

Best PCIe Card For ~$465: None

Honorable Mention: GeForce GTX 295 (Check Prices)

Good 2560x1600 performance in most games

Despite ATI's new Radeon HD 5970 taking its place as the fastest graphics card on the planet, Nvidia's GeForce GTX 295 (with SLI-on-a-board) remains an extremely powerful graphics card. Essentially two conjoined GeForce GTX 275s, the GeForce GTX 295 offers very notable gains over a single Radeon HD 5870 in the great majority of game titles, although the Radeon will use far less power doing so. The GeForce GTX 295 does have an advantage in that it it still quite easy to find and purchase. Moreover, ATI's release has forced prices on these cards down by a significant chunk.

Read our full review of Nvidia's GeForce GTX 295 for more information on the card and its accompanying architecture.

Best PCIe Card For ~$625: None

Honorable Mention: Radeon HD 5970

Great 2560x1600 performance

3,200 shader processors. There isn't much more we need to say about the brutal grace of execution that characterizes the world's fastest graphics card, the Radeon HD 5970. With two Radeon HD 5870 GPUs onboard, the only things we can complain about are scant availability and an extremely high price tag. Availability should improve over time; the price not so much, but if you're in the market for this card price probably isn't an issue.

Read our full review of ATI's Radeon HD 5970 for more information on the card and its accompanying architecture.

There you have it folks; the best cards for the money this month. Now all that's left to do is to find and purchase them.

Don't worry too much about which brand you choose, because all of the cards out there are close to Nvidia's and ATI's reference designs. Just pay attention to price, warranty, and the manufacturer's reputation for honoring the warranty if something goes wrong.

Also remember that the stores don't follow this list. Things will change over the course of the month and you'll probably have to adapt your buying strategy to deal with fluctuating prices. Good luck!

Both the the Shift: Intel P55 and the Shift: Intel X58 are powered by Intel Core i7 900 processors running Windows 7. ATI Radeo HD and NVIDIA GeForce graphics cards are featured, and I'm sure the discerning gamer expects nothing less.

Storage options include up to six mechanical or 12 SSD drives. The two diverge on memory specs, with the P55 containing up to 8GB DDR3-1600 low latency RAM and the X58 up to 12GB DDR3-2000 or 24GB DDR3-1600. DVD or Blu-Ray drives are options for both rigs while a standard Asetek closed-loop liquid cooling system keeps things chilled.

Pricing begins at an optimistic $2,199 and $2,599, respectively, although with all the options listed above that can (and probably will) climb much, much higher. Since Shift is the only PC that Maingear going's to focus on selling from now on, I hope it works it out for them.

The first GeForce 200M notebook cards—the GTX 280M and 260M—were for crazy gaming rigs, and were actually based on the previous-gen G92 architecture. (Nvidia did pulled some confusing re-branding jujitsu a few months back.) The new 200M cards are based on the "current high-end desktop architecture" (so, actually the G200 architecture) and round out the 200M series, replacing the current 9M series across the board: GTS 260M, GTS 250M, GT 240M, GT 230M, G 210M. Here's how the specs break down:

So to recap in English, all the Nvidia notebook graphics cards that are like "GT 9600" are going to be replaced by ones that are like "GT 240" which are faster but use less power. I don't know why Nvidia went from 9000 to 200, so don't ask me. It's actually kind of a bummer they didn't make it into the new MacBook Pros, though, since they now have officially old and busted graphics chips inside. [Nvidia]

What a Difference a Letter Makes
GPU sounds—and looks—a lot like CPU, but they're pretty different, and not just 'cause dedicated GPUs like the Radeon HD 4870 here can be massive. GPU stands for graphics processing unit, while CPU stands for central processing unit. Spelled out, you can already see the big differences between the two, but it takes some experts from Nvidia and AMD/ATI to get to the heart of what makes them so distinct.

Traditionally, a GPU does basically one thing, speed up the processing of image data that you end up seeing on your screen. As AMD Stream Computing Director Patricia Harrell told me, they're essentially chains of special purpose hardware designed to accelerate each stage of the geometry pipeline, the process of matching image data or a computer model to the pixels on your screen.

GPUs have a pretty long history—you could go all the way back to the Commodore Amiga, if you wanted to—but we're going to stick to the fairly present. That is, the last 10 years, when Nvidia's Sanford Russell says GPUs starting adding cores to distribute the workload across multiple cores. See, graphics calculations—the calculations needed to figure out what pixels to display your screen as you snipe someone's head off in Team Fortress 2—are particularly suited to being handled in parallel.

An example Nvidia's Russell gave to think about the difference between a traditional CPU and a GPU is this: If you were looking for a word in a book, and handed the task to a CPU, it would start at page 1 and read it all the way to the end, because it's a "serial" processor. It would be fast, but would take time because it has to go in order. A GPU, which is a "parallel" processor, "would tear [the book] into a thousand pieces" and read it all at the same time. Even if each individual word is read more slowly, the book may be read in its entirety quicker, because words are read simultaneously.

All those cores in a GPU—800 stream processors in ATI's Radeon 4870—make it really good at performing the same calculation over and over on a whole bunch of data. (Hence a common GPU spec is flops, or floating point operations per second, measured in current hardware in terms of gigaflops and teraflops.) The general-purpose CPU is better at some stuff though, as AMD's Harrell said: general programming, accessing memory randomly, executing steps in order, everyday stuff. It's true, though, that CPUs are sprouting cores, looking more and more like GPUs in some respects, as retiring Intel Chairman Craig Barrett told me.

Explosions Are Cool, But Where's the General Part?
Okay, so the thing about parallel processing—using tons of cores to break stuff up and crunch it all at once—is that applications have to be programmed to take advantage of it. It's not easy, which is why Intel at this point hires more software engineers than hardware ones. So even if the hardware's there, you still need the software to get there, and it's a whole different kind of programming.

Which brings us to OpenCL (Open Computing Language) and, to a lesser extent, CUDA. They're frameworks that make it way easier to use graphics cards for kinds of computing that aren't related to making zombie guts fly in Left 4 Dead. OpenCL is the "open standard for parallel programming of heterogeneous systems" standardized by the Khronos Group—AMD, Apple, IBM, Intel, Nvidia, Samsung and a bunch of others are involved, so it's pretty much an industry-wide thing. In semi-English, it's a cross-platform standard for parallel programming across different kinds of hardware—using both CPU and GPU—that anyone can use for free. CUDA is Nvidia's own architecture for parallel programming on its graphics cards.

OpenCL is a big part of Snow Leopard. Windows 7 will use some graphics card acceleration too (though we're really looking forward to DirectX 11). So graphics card acceleration is going to be a big part of future OSes.

So Uh, What's It Going to Do for Me?
Parallel processing is pretty great for scientists. But what about those regular people? Does it make their stuff go faster. Not everything, and to start, it's not going too far from graphics, since that's still the easiest to parallelize. But converting, decoding and creating videos—stuff you're probably using now more than you did a couple years ago—will improve dramatically soon. Say bye-bye 20-minute renders. Ditto for image editing; there'll be less waiting for effects to propagate with giant images (Photoshop CS4 already uses GPU acceleration). In gaming, beyond straight-up graphical improvements, physics engines can get more complicated and realistic.

If you're just Twittering or checking email, no, GPGPU computing is not going to melt your stone-cold face. But anyone with anything cool on their computer is going to feel the melt eventually.

[Back to our Complete Guide to Snow Leopard]

Again, nothing's confirmed just yet, but preliminary testing has shown users who artificially set the MacBook Pro's cooling fan to 3,000rpm eliminate the screen burn-in issue we told you about earlier this week.

One MacBook Pro user, posting a reply in the Apple Support Forum, said the fans are intended to run at 2,000rpm until default conditions, and then spool up to 3,500 when high-temperature conditions kick in. You know, like what happens when you're using a GPU like the Nvidia 9600M.

That said, the MacBook Pro is having a little trouble getting it up, so to speak, and remains at about 2,050rpm when the temps rise to the point where the screen begins to flake out. That's bad.

What's good is this seems like a simple software fix, at least from the sound of this latest discussion. [Slashgear]

So a quick recap on the original issue and the back and forth between Nvidia and the Inquirer: Nvidia admitted back in July "significant quantities" of notebooks are defective, built and packaged with "weak" materials that are leading to them to overheat and fail at a "higher-than-normal" rate, but declined to state which cards specifically.

The Inquirer said every single G84 and G86 card was affected, and a GPU apocalypse was coming, and indeed, more and more models from different manufacturers came to light with the problem. The previous-gen MacBook Pro was actually the last one revealed to be smitten by the plague, like the final Cylon or something, since Nvidia was reportedly less-than-honest about the problem to Apple. Nvidia continued to reiterate most chips are just peachy.

Nvidia says that its current chips don't use the weak materials, which produce what are called "bad bumps." Obviously, the Inquirer thought they were lying, so they took a new MacBook Pro to a lab, cracked it open, sliced apart the G96 GPU and checked it out under a scanning electron microscope with an X-ray microanalysis system.

The result? The Inq says the same old bad bumps, which were composed mostly of lead, are there in the GeForce 9600 vs. the new, good eutectic bumps that are in the GeForce 9400M (the MacBook and MacBook Pro's chipset/integrated graphics). Or more straight up, "The 9600 is unquestionably using 'bad bumps', directly contradicting the statements from Nvidia...It suggests that there are 15-inch Macbook Pros being sold with 'bad bumps', the same materials that brought down so many HP, Dell and Apple parts, both laptop and desktop."

Naturally, I asked Nvidia for their reaction to Charlie's Inq piece, and a spokesperson reiterated that the GeForce 9600 GT graphics cards in the MacBook Pros "don't have bad bumps at all." He said that, "yes, they're lead bumps" but "hundreds of millions of chips have lead bumps." And it's "a different material set [from the faulty one], one they transferred to earlier" that's used in the 9400, 9600 and 9800.

Of course, the Inquirer's whole point is that Nvidia is lying. So, who to believe? Well, here's what we know for sure. When I talked to Nvidia about the original run of faulty chips, and why we saw it some systems and not others, they told me it was largely a thermal issue, which, in combination with the weak materials, would cause the kiss-of-death cracking—so you'd see it in systems that ran hot, in other words, like some notebooks or slimline desktops with poor circulation. (Which is why the "fix" for the problem were firmware updates that cranked the card's fans sooner.)

We further know that the 9600 GT cards in the MacBook Pro are currently having problems that appear to be heat-related, causing them to lock up and launch into the "black screen of death." Also, Nvidia pointed out that there aren't a whole lot of labs properly equipped to do the kind of analysis the Inquirer commissioned—you can't just walk down to your local Discovery Channel store—though they left it at that.

And that's about as definitive as we can get, for now. Two things bother us: We would've liked a slice and dice of one of the previous-gen bad chips to directly compare to the new, supposedly bad one. And Nvidia's subtle implication that the people with labs equipped to perform this kind of analysis have a vested interest in the outcome also has the magical effect of shielding them from the results.

You can believe Nvidia. You can believe the Inquirer. I just know that given the thermal problems that already clearly exist, I really hope the Inquirer is wrong. If you know something about this you wanna share, email me.

Expected to be unveiled at CES, the GTX 295 (nee GTX260 GX2) actually is made up of two print-circuit boards, and each one has a GTX 200 GPU, 240 stream processors, 448-bit memory bus and 896MB DDR3 memory. It's totally outrageous, in other words, and requires 289W of power, so I hope you've got a behemoth of a power supply.

The price, while unknown, will be proportionately juggernaut-sized, crushing your wallet. Since it's designed to beat ATI's Radeon HD 4870 X2, it'll likely fall in the same price range, probably around or slightly north of $500. It could swing cheaper though, since Nvidia's current high-end card, the GTX 280, is trending south of $400 at the moment. Guess we'll see, but I can't afford it either way.

Oh, and first person to ask "will it run Crysis?" is banned. I'm not kidding. [Expreview via X-bit Labs]

If it pans out, we should be in for some sweet deals—last time Nvidia played hard ball with ATI, they threw bricks, cutting their top-end graphics cards by $200 just a month out the gate, and let loose its GeForce 9800 GTX for around $200 as well. It could be a Merry Christmas after all. [Digitimes via Maximum PC]

This is on the heels of Apple concluding that MacBook Pros manufactured between May 2007 and last month also pack the fail-prone cards, despite Nvidia's assurance earlier that it was unaffected. I'm not one for conspiracy theories, but this is second major contradiction to an earlier Nvidia claim. Nvidia says that there's "no conclusive evidence that the failure mechanism is the same as the notebook issue," but uh, it doesn't look so great, guys.

Nvidia really needs to come clean with the score now—exactly which cards are affected, whose computers they're in, everything. Their near total silence on this while individual computer makers cleanup is absolutely ridiculous. [PC Pro via Slashdot]

The built-in audio and video capabilities in the new Nvidia chipset is supposed to be better than Intel's offering for High Definition video decoding, with 24p mode (progressive 24 frames per second video) and LCPM (Linear Pulse Code Modulation) multi-cannel audio support. [Homemedia via Macsoda]

1) The issue is limited to a few notebook chips only; we have not seen and don't expect to see this issue on any NVIDIA based desktops

2) Only a very small percentage of the notebook chips that have shipped are potentially affected and the problem depends on a combination of environmental conditions, configuration and usage model

3) We continue to work closely with our partners and have taken the necessary steps to ensure that all NVIDIA chips currently in production do not exhibit the problem

[VR-Zone]

All summed up, the Inquirer is saying almost the entire run of 8 and 9-series GeForce cards are defective and "going bad in the field at high rates." More than that, they say that Nvidia will continue to cover it up because the cost of admitting the nature of the problem would be devastating, meaning we'll never know for sure. We'll just have to watch for anecdotes to pile up. We're really hoping they're wrong on this one—that's a lot of graphics cards, and if true, it could basically destroy Nvidia. [Inquirer - Thanks Jagslive!]

While there's no immediate way to tell if Nvidia is in fact lying (which would be a massively boneheaded move), as the Inquirer alleged, we'll know in the coming months. Reports of massive chip failures and an even larger hit to Nvidia's financials than the $150-$200 million it already took would make it super obvious. Hopefully the Inquirer was just being bombastic and British. [Ars Technica]

The root of the problem is the substrate/bumping material, which Nvidia themselves had characterized as "weak." The G84 and G86 chipsets use the same application-specific integrated circuit, which the Inquirer has been told had no changes made to it during its life cycle—besides, Nvidia wouldn't change the assembly process or materials for a single batch, according to engineers. So Nvidia's official claim, that the weak materials were only used for a single batch, doesn't hold so much water. And as the Inquirer points out, their fix, to run the fan more to offseat the heat issue simply covers up the problem while it drains your battery.

If this is all true, Nvidia better have more than $200 million set aside for fixes. [The Inquirer

A Radeon HD 3870 system jumped from an overall 3D Mark Vantage score of P3800 to P4262. Though it won't be available for ATI's latest 4800 cards for a bit, you can download the PhysX enabler this weekend at NGOHQ, which should make your Unreal experience on ATI that much sweeter. [NGOHQ via TG Daily via /.]

Of course, there is a cost to being maybe the best performing GPU on the planet: besides running $650 and $399, respectively (making three-way SLI nearly two grand with the GTX 280), they both require two PCI-E power connections to run and a massive power supply (like 1000w) if you're even thinking about SLI. Yet somehow it actually draws less idle power than the last gen of their ultra high-performance cards. To show you how adept they are at parallel processing here's one more benchmark shot, this time comparing its Foldering@Home performance. Yep, they've got a client coming soon. We should see mid-range cards in the line before too long, for those looking for more affordable next-gen goodness.

Update: Tom's Hardware has a massive novella of a review going over everything in complete, insane detail, but here are the highlights. It never hands down beats the 9800 GX2 in game performance (which is really two cards in one), and in fact, loses more than once, though that might be 'cause the drivers are less optimized. ATI's Radeon HD 3870 X2 gets in its licks too, like on World in Conflict. In the all-important (or maybe overblown) Crysis test, the 9800 GX2 prevails, with the 8800 Ultra not too far behind the GTX 260: Still, the overall raw processing power has doubled from the last gen and is way more efficient than the two-in-one cards. If you're going on price-to-performance, the GTX 260 is the better bet, with only an 18 percent performance lag, despite being nearly 40 percent cheaper. [Nvidia, Tom's Hardware]