Michael Brown EVGA e-GeForce 8800 GT SSC

Previous generations of Nvidia GPUs (AMD’s, too) presented buyers with a difficult choice: You could get great 3D performance for gaming or you could offload high-definition video decoding from the host CPU, but you couldn’t have both. Nvidia’s 8800 GT not only changes that situation, it does so at a competitive price.

The 8800 GT delivers stronger 3D performance than the industry’s previous sweet spot (the 8800 GTS with a 320MB frame buffer), it delivers more memory than that board, and it clubs AMD’s far-more expensive Radeon HD 2900 XT over the head for good measure (although AMD has responded with the Radeon HD 3870, see page 80). Nvidia managed to cram 754 million transistors into this beast thanks to a die shrink and a 65nm fabrication process (previous 8800-series GPUs were manufactured using a 90nm process).

The new part packs 112 stream processors, 512MB of GDDR3 memory, and a 256-bit memory interface into a GPU that requires a single-slot cooler (the fan howls like a banshee on startup but goes whisper-quiet as soon as Windows launches). Reference-design boards will run their cores at 600MHz and their memory at 900MHz; EVGA pumps these numbers to 700MHz and a cool 1GHz, respectively. The company also commands a premium price for the speed boost: While the average price for more typical boards was running around $270, this SSC Edition was fetching $330 at press time.

As mentioned above, the new GPU is capable of offloading the entire HD decode process from the host CPU, and it also provides HDCP decryption on both DVI links. This latter feature renders the chip capable of displaying Blu-ray and HD DVD movies at the native resolution of a 30-inch LCD. It’s also compliant with PCI Express 2.0 (see the White Paper on page 72 for more details). We didn’t test this card in that type of motherboard—no one’s shipping one in an SLI configuration just yet—but the new architecture offers double the bandwidth of PCI Express 1.1 (8GB/s in each direction).

AMD moved to GDDR4 memory several iterations back, but Nvidia continues to stick with GDDR3—and the decision doesn’t seem to be costing its cards anything in terms of performance. Interestingly, AMD has retreated from its 512-bit memory interface, building a 256-bit interface into the 3870 (same as the 8800 GT). But there are still two other features that could hold the 8800 GT back when it comes to competing with AMD’s Radeon 3870: First, these cards have only one SLI connector. Nvidia’s other cards, from the 8800 GTS on up, have two SLI connectors, even though only one of them is used in dual-card mode. Why worry about it? Nvidia will inevitably debut an SLI version that enables you to run more than two GPUs on one motherboard (remember quad SLI?), and that’s why the other cards have two SLI connectors. You’ll never be able to run more than two 8800 GTs in one box.

The other advantage AMD will soon offer is the ability to run more than one monitor in CrossFire mode, although that will likely require AMD’s new RD790 chipset (which hasn’t been released). Nvidia’s SLI system shuts off the second monitor when running in SLI mode (as does the current version of CrossFire).

While we’re looking at the future, we should also consider the fact that, unlike the Radeon HD 3870, the 8800 GT does not support Microsoft’s Direct3D 10.1 and Shader Model 4.1. Given the tepid response that most developers have given Windows Vista, and Microsoft’s continued insistence on tying DX10’s fortunes to its new OS, we don’t think this shortcoming matters much at all.

It’s been a long time since we’ve been able to unabashedly recommend a videocard priced this low. We just couldn’t get excited about the anemic 8800 GTS; and until now, AMD has had nothing meaty to offer. But the 8800 GT is an absolutely fantastic value, delivering great gaming performance and features that can’t be found in Nvidia’s higher-end boards. If you can swing the price, you’ll get a better gaming experience from a GeForce 8800 GTX or an Ultra (although we don’t think the latter is worth its premium); but if you’re rolling with a lower budget, the 8800 GT is a slam-dunk winner.

EVGA pumped up the 8800 GT’s core and memory clocks, but it’s sticking with a reference-design cooler.

EVGA e-GeForce 8800 GT SSC
www.evga.com

Old Potrero

Scorching gaming performance; complete solution for HD-video decoding; fabulous price.

Old Granddad

Only one SLI connector; no support for DX10.1.

CrossFire and SLI Performance Fix for Vista, Part Deux

Back in July, Microsoft rolled out a fix for a nasty bug in its Windows Vista SLI and CrossFire multi-GPU support: Windows Vista was using only one of the GPUs! However, the Inquirer.net website reported yesterday that a more subtle bug is still plaguing some multi-GPU setups: a problem with load-balancing between GPUs.

Load-Balancing Woes and System Slowdowns

If one of the GPUs in your system is very busy, and and other idle GPUs receive a new DMA packet, you may experience scheduling latency. In plain language, your system slows down until all GPUs are assigned work to do. If you like to juggle games and other applications at the same time on your SLI or CrossFire system, you're a prime candidate for this type of slowdown.

Meet LDA, a Way to Avoid Saying "SLI" or "CrossFire"

This update isn't going to show up in Windows Update, at least not for awhile. And, if you ran into it during a casual scan of recent downloads, you might not even recognize it. That's because Microsoft is now referring to multi-GPU configurations as Linked Display Adapter (LDA) setups. SLI and CrossFire fans, make a note of this: every time you see a reference to LDA at Microsoft, think SLI (nVIDIA) or CrossFire (ATI/AMD).

Learning More About the Update...and Getting It

Microsoft has information about the hotfix, available for both 32-bit and 64-bit Windows Vista, at Knowledge Base article 945149. You must request the hotfix from Microsoft, rather than download it directly from the KB article. Microsoft promises a download link within 8 business hours, as we discussed in our article Hotfixes by Email - Just What the Windows Doctor Ordered.

-------------------------------------------------------------------------------

Help Windows Vista newbies learn how to use BitLocker to stop data theft, get organized with Windows Calendar, and back up data and systems the easy way: give Mark's new book Maximum PC Microsoft Windows Vista Exposed. It's now available at Amazon.com and other fine bookstores.

CrossFire and SLI Performance Fix for Vista, Part Deux

Back in July, Microsoft rolled out a fix for a nasty bug in its Windows Vista SLI and CrossFire multi-GPU support: Windows Vista was using only one of the GPUs! However, the Inquirer.net website reported yesterday that a more subtle bug is still plaguing some multi-GPU setups: a problem with load-balancing between GPUs.

Load-Balancing Woes and System Slowdowns

If one of the GPUs in your system is very busy, and and other idle GPUs receive a new DMA packet, you may experience scheduling latency. In plain language, your system slows down until all GPUs are assigned work to do. If you like to juggle games and other applications at the same time on your SLI or CrossFire system, you're a prime candidate for this type of slowdown.

Meet LDA, a Way to Avoid Saying "SLI" or "CrossFire"

This update isn't going to show up in Windows Update, at least not for awhile. And, if you ran into it during a casual scan of recent downloads, you might not even recognize it. That's because Microsoft is now referring to multi-GPU configurations as Linked Display Adapter (LDA) setups. SLI and CrossFire fans, make a note of this: every time you see a reference to LDA at Microsoft, think SLI (nVIDIA) or CrossFire (ATI/AMD).

Learning More About the Update...and Getting It

Microsoft has information about the hotfix, available for both 32-bit and 64-bit Windows Vista, at Knowledge Base article 945149. You must request the hotfix from Microsoft, rather than download it directly from the KB article. Microsoft promises a download link within 8 business hours, as we discussed in our article Hotfixes by Email - Just What the Windows Doctor Ordered.

-------------------------------------------------------------------------------

Help Windows Vista newbies learn how to use BitLocker to stop data theft, get organized with Windows Calendar, and back up data and systems the easy way: give Mark's new book Maximum PC Microsoft Windows Vista Exposed. It's now available at Amazon.com and other fine bookstores.

Nvidia Beta 256 Drivers Now Available for Download

We're not sure why exactly Nvidia chose to make the leap from version 197 to 256 for its latest drivers, but whatever, designation aside, the first beta release for the 256 driver-set is available now, and it includes a whole bunch of goodies.

Release 256 are the first drivers to incorporate Blu-ray 3D acceleration for the PC (see here for system requirements), just in time for the first 3D Blu-ray movie due out June 1st, Nvidia says. The new driver also includes new settings in the Nvidia Control Panel for 3-way SLI customers that enable 48x SLI Anti-aliasing (96x for 3-way GTX 400 series PCs).

You'll find a lot of buzz words accompanying Release 256, including CUDA 3.1 and OpenGL 4.0 support, as well as a new SLI and PhysX configuration page. What you won't find, however is support for Nvidia's 3D Vision Surround, which the GPU chip maker promises to include in the second release targeted for the end of June.

Driver Download (includes notebook platforms)

Rumor: Nvidia to Dub Dual-GPU Fermi Card "GTX 490"Rumor: Nvidia to Dub Dual-GPU Fermi Card "GTX 490"

The latest chatter from the graphics rumor mill is that Nvidia is hard at work readying a dual-GPU fermi part the company plans to call "GTX 490."

According to kitguru.net, which "continue[s] to look under the rocks in Taiwan," the GTX 490 will in all likelihood use a pair of GTX 470 processors. From what they gather, the power draw will be around 375W. That means you can probably expect some downclocking under the hood, as the GTX 470 draws 215W as a standalone card.

There really aren't any other details to go on, though that could (and probably will) change by the time Computex rolls around. It will be interesting to see how the GTX 490 stacks up against ATI's recently introduced 4GB iteration of its dual-GPU HD 5970 card.

Nvidia GTX 465 Details Leaked to the Web

Computex is less than two weeks away, and it's during that time many are expecting Nvidia to announce a new graphics card, the GTX 465. Well, thanks to Turkish site Donanimhaber, the same site that accurately revealed GTX 470 and 480 specs ahead of launch, we have a pretty good idea of what to expect from Nvidia's upcoming GPU.

According to Donanimhaber, the GTX 465 is being built around the same GF100 core as the 470 and 480, but with five out of the 11 stream clusters turned off. That brings the stream processor count down to 352.

The core clockspeed will race along at 607MHz, the same as the GTX 470, while the stream processors will come clocked at 1,215MHz. You can expect 1GB of GDDR5 on a 256-bit memory bus, along with two 6-pin PCI-E power connectors.

Pricing is estimated to be around $300.

NVIDIA’s Optimus Hybrid Graphics Chip Streamlines Energy Efficiency

PIQOO | www.piqoo.blogspot.com

It’s getting easier to have your cake and eat it too, or so suggests NVIDIA in the announcement of its new Optimus technology. According to NVIDIA: “Consumers no longer have to choose whether they want great graphics performance or sustained battery life. NVIDIA Optimus gives them both - great performance, great battery life and it simply works.”

Optimus, at its heart, allows the automatic routing of graphics demands on a laptop to either a discrete NVIDIA GPU (which sucks up power), or to an Intel integrated graphics chip (which conserves power)--without the hassle of a reboot. Optimus directs the workload, as appropriate, to the most “efficient processor for the job”, thus extending battery life by up to two times.

According to Electronista, Optimus differs from Hybrid SLI in that it doesn’t need drivers to cooperate, but rather relies on the presence of an Intel chipset--be it the Core 2 Duo, Core i7 or Atom. NIVIDA also makes use of a new software model that ensures automatic switches occur when needed--in as little as 300ms. (Users, however, will be able to tweak these settings.)

NVIDIA says Optimus will first be available on an Asus line of notebooks (the UL50Vf, N61Jv, N71Jv, N82Jv, and U30Jc), which should be available soon.

Image Credit: NVIDIA