News and rumor site Fudzilla claims to have received confirmation that AMD's dual-GPU Radeon HD 6990 videocard, otherwise known as "Antilles," won't show up on store shelves until the first quarter of 2011.
As the water cooler banter goes, AMD originally wanted to ship its dual-chip card before the end of the year, and if that's true, it raises of the question of why it's being postponed. It could be that AMD wants to fine tune the architecture and drivers, or maybe there's a problem that requires additional time to address. It's all speculation, with the only thing we know for sure (assuming Fudzilla's sources are solid) is that there won't be any Radeon HD 6990 cards in Santa's sleigh this year.
A recently leaked slide of the HD 6990 part reveals that the card will come with 3840 stream processors, 4GB of GDDR5 memory, and deliver up to 6.0TFLPS. Under load, the HD 6990 is supposed to draw a manageable 300W, and 30W when idle.
As Nvidia often does following the launch of a new videocard, the GPU maker has rolled out updated GPU drivers, version 263.09.
Specific to this release is added support for the GeForce GTX 580 and 460 SE videocards, and presumably the just launched GTX 570 (oddly enough, there's no mention of the GTX 570 in the release notes). Otherwise, there doesn't appear to be much else to the 263.09 driver packages.
For what it's worth, the latest drivers install Nvidia PhysX System Software version 9.10.0514 and HD audio driver version 1.1.9.0, neither of which do we envision making you pump your fist in excitement.
Hey, they can't all be GeForce GTX 580 caliber cards and dual-GPU Cayman killers (which we're still waiting on), and sometimes you have to cater to the professional crowd as well. That's what Nvidia is doing with the launch of its NVS 300 graphics card, a GPU specifically designed for the enterprise with 25 percent more efficient power utilization when compared to the NVS 295, Nvidia claims.
"The NVS is built for demanding enterprises that require high reliability, improved manageability, and tremendous value," said Jeff Brown, general manager, Professional Solutions Group, Nvidia. "The ability to support legacy and current display types provides an upgrade path without disrupting existing, complex installations."
Nvidia is touting versatile connectivity with the NVS 300. The low-profile card supports single and multi-display setups via the nView Desktop Management software and the built-in Mosaic technology, which allows for taskbar spanning and transparent scaling of any app across up to eight displays.
We're slightly perplexed with VisionTek's new "Killer HD 5770" graphics card, which the company touts as the world's first online gaming upgrade card for PCs. We get the marketing behind that statement -- after all, the videocard comes with Bigfoot Networks' Killer E2100 NIC baked in -- but who exactly is this card aimed at?
If you're buying into the hype of a dedicated hardware NIC for lower pings (you can read our review of the standalone Killer 2100 here), then integrating it into a videocard makes perfect sense. Two gaming solutions, one PCI-E slot -- that's a winning combo. Buy why not use a higher end GPU, like the HD 5850, 5870, or even one of the new 6000 series cards?
For those of you interested in a mid-range card from last year with an integrated NIC, the Killer HD 5770 is the "only upgrade card specifically designed to combine high-quality graphics and low-latency networking for gaming," VisionTek says. The combo card comes with 1GB of GDDR video memory and supports DirectX 11, OpenGL 3.1, and everything else you already know about the HD 5770.
The Killer HD 5770 will be available early this month for $200.
Psst, come here. You aren't supposed to know this, but according to a Palit Product Information slide sweclockers.com posted online, it appears Nvidia is readying its GeForce GTX 570 videocard for a December 7, 2010 launch date, the same day as World of Warcraft: Cataclysm.
That's also just a few days ahead of AMD's Cayman launch, barring any last minute surprises. Regardless of how it all shakes out, Nvidia's GTX 570, according to the posted spec sheet, will come with 480 CUDA cores and 1280MB of memory clocked at 3800MHz on a 320-bit bus. The GPU will race along at 732MHz and the shaders at 1464MHz.
The GTX 570 will come with a pair of 6-pin PCI-Express power connectors and carry a 225W TDP, compared to 244W on the GTX 580 and 250W on the GTX 480.
You already know Intel's upcoming Sandy Bridge processors will come with integrated graphics, but there's more to this architecture than just a fused GPU. According to CNet, Intel VP and Director of PC Client Operations Stephen L. Smith confirmed the existence of special media accelerators.
"The other cool thing is dedicated circuitry for media acceleration," Smith said in response to a question at a Wells Fargo Securities conference earlier this month. "All of us in our daily use, whether it's home videos or photos tend to pull things from the Internet, pull things from our own capture devices at home, bring them on to our PC, transform them into different formats...all of that will be dramatically faster if one utilizes this hardware acceleration, media acceleration that we have on Sandy Bridge."
In addition, Smith said Sandy Bridge should allow for slimmer form factors "and potentially longer battery life" on the notebook front.
Looking beyond Sandy Bridge, Smith reiterated that Intel is on track to deliver its 22nm Ivy Bridge architecture by the end of 2011. According to Smith, Ivy Bridge will be a shrink of Sandy Bridge "with some enhancements."
We've already heard that Nvidia is planning to release a dual GF110 GPU videocard to snatch the overall performance crown from AMD (HD 5970), and courtesy of Chinese website enet.com.cn, we now get to see what the upcoming card will look like, Electronista reports.
Bear in mind that this is a prototype part built around Nvidia's reference design, so this isn't necessarily how the shipping part will look. That said, the dual-GPU part comes with 3GB of memory, a single SLI-connector, and two 8-pin PCI Express power connectors. There's also three DVI ports, which hints at support for Nvidia's 3DVision Surround.
Other than that, there isn't much new to report, such as a release date or pricing info. Stay tuned for more details as they become available.
Let's start with the good news. According to Fudzilla, Nvidia is definitely planning to release a dual-GPU videocard, and relatively soon. Ready for the bad? It won't ship in time for the holiday shopping season.
Citing "sources close to the company," Fudzilla says Nvidia wants to wait for AMD's dual-GPU Antilles Radeon HD 6990 before releasing its own dual-chip videocard. Nvidia already owns the single-GPU performance crown (GeForce GTX 580) and wants to go after the dual-GPU crown as well, but is reluctant to do so before knowing what AMD's upcoming part will be capable of, Fudzilla suggests.
The GTX 590, as it's rumored to be card, will almost certainly be faster than AMD's current flagship part, the HD 5970. But if it's not faster than the Radeon HD 6990, Nvidia may opt to tweak the design until it's certain it has the best performing videocard on the planet.
No, this really long card isn't an actual prototype, just a quick and dirty mockup.
Futuremark, the Finland-based maker of several popular benchmarking tools, today announced that 3DMark 11 will be released on November 30, 2010, with pre-orders beginning today.
The free version includes the Performance PC benchmark preset, an audio/visual demo fixed at 720p, the ability to browser, search, and compare results online, and store one result online.
For $20, the Advanced Edition ups the ante with Entry-level and Extreme PC benchmark presets, custom benchmark settings, custom resolutions for the audio/visual demo, benchmark looping to test stability, unlimited online results storage, hide results from public view, advert-free online service, and offline result management.
3DMark is largely a graphics card benchmarking utility with the upcoming release putting a heavy focus on DirectX 11, including tessellation and volumetric lighting created with DX11.
Nvidia isn't making a big deal about its GeForce GTX 460 SE videocard (hit the specifications tab), the latest addition to the GTX 460 line with the least amount of CUDA cores.
The SE version comes with 288 CUDA cores in all, compared to 336 on the standard GTX 460 in both 768MB and 1GB trim. It's also clocked a bit slower at 650MHz core and 1700MHz memory, compared to 1350MHz and 1800MHz on the two other models, respectively.
It does, however, sport the same 256-bit memory bus interface as the non-SE 1GB version, whereas the standard GTX 460 in 768MB form features a 192-bit bus. That provides the SE with more memory bandwidth than the 768GB version (108.8GB/s versus 86.4GB/s) though still less than the non-SE 1GB version (115.2GB/s). And finally, the SE comes with 1GB of GDDR5 memory.
