A juicy rumor from SemiAccurate indicates that production for the initial run of Xbox Next devkits has already started, with units expected to arrive at select studios by March or earlier. Although details have not been finalized at this stage, it appears Microsoft will be using some variant of the Radeon HD7000 series in its next box, which is built on 28nm process technology.

Power consumption matters, you can’t lump a 250 watt GPU in a console so it’s likely Microsoft will be using a chip that fits somewhere in AMD’s midrange lineup, rather than a top-end part like the just-released HD7970.

Even if production of dev hardware has started, the release of the Xbox Next is still a ways off. SI expects production on consumer units to begin in late 2012, with a release in 2013. That means we could still see it at E3 2012, so as long as Microsoft’s partners have something to show, and with dev kits reportedly rolling out it seems they will.

Those in attendance at the PDXLAN 16.5 gaming event caught a first-hand glimpse of Nvidia’s future GPU plans, courtesy of technical marketing director Tom Petersen. Nvidia rallied up the crowd with metrics on a new piece of dedicated graphics hardware that is presumed to be the oft-rumored GTX580.

According to Petersen, the unreleased card boasts a robust vapor cooling system, offering quieter operating conditions over Nvidia’s previous-generation behemoth, the GTX 480 — which received quite a bit of flak for its high power consumption levels and heat output.

Petersen topped off the presentation with a tessellation-focused demo of Aliens vs. Triangles, followed a brief segment of silky-smooth Black Ops gameplay on PC.

Rumors point to a November release date for the GTX 580, possibly as soon as next week. AMD is poised to counter with its yet-to-be revealed 6900 series, which will effectively replace last year’s DX11 pioneers, the HD5870 and HD5850.

Hot off the heels of its successful DX11 family of Evergreen GPUs, alleged benchmarks for AMD’s next in line have emerged from the dark corners of the internet. A 3DMark score of the purported HD6870 was recently posted by a forum user over at Chinese site PCinLife, showing off a GPU score of 11,634. While the veracity is questionable, a supplied GPU-Z screenshot does add credence to the score, lining up with internal data that erroneously made its way into ATI’s latest driver release — Catalyst 10.8.

Assuming the score is legit, AMD’s next family of GPU are set to pack a punch. Landing a 11.6K in 3DMark would theoretically place the HD6870 above the HD5970 in raw performance, the latter of which is a dual GPU part. Given that result, there’s little doubt it would also handily compete with Nvidia’s latest flagship — the GTX480.

Additional benchmarks performed on the card indicate an average of 43.55 FPS in Crysis at 1920×1200 with 4xAA and preset very high enabled. An average of 36.6FPS was observed in the DX11-heavy Unigine stress test, with tessellation set to extreme.

HD6000 performance [PCinLife]

Valve’s latest in-house Steam survey reveals a tantalizing statistic – namely that as of May, the vast majority of PC gaming rigs now sport a DirectX 10-capable GPU. More specifically, 56.47 percent of users surveyed on Steam are rocking a DX10 or DX11 GPU.

Interestingly, 21.55 percent of those users are still on Windows XP, which is only DX9 capable. However, that figure is beginning to dwindle as more users skip over Vista and make the transition to Microsoft’s latest offering — Windows 7.

While chip manufacture Nvidia still commands a lion’s share of the GPU market, it’s notable to point out that enthusiasts are turning to ATI’s 5000 series for their DX11 fix. Not exactly surprising, given that Nvidia’s equivalent was late to the party with a higher price point.

Update: There was a bit of erroneous information in our previous report. The GPU in newer PS3 units is manufactured on a 40nm process, not 45nm like that of the Cell processor. We regret the mix-up.

New PS3 Slim units that sport a 40nm die-shrink of RSX — the PS3′s GPU — are out in the wild. In Japan, at least.

Extensive testing performed by Japanese site PocketNews on a CECH-2100A unit reveals that power usage dropped by as much as 29 watts compared to that of the previous slim revision, CECH-2000A. A drop in effective wattage correlates to less heat, an attribute Sony has taken advantage of by including a smaller cooling unit as well as a power supply that weighs less, approximately 412 grams. more »

Graphics chip manufacture Nvidia today officially named the first batch of GPUs that will be based on its GF100 architecture, first revealed by company CEO Jensen Huang at the California-based GPU conference last September under the codename of Fermi. Taking helm at the high end segment is the GTX 480, followed by the GTX 470.

While the exact specifications and pricing of either card still remain unknown, both parts have been confirmed as single GPUs. We can only speculate that Nvidia left wiggle room in there for a possible dual GPU part down the road, similar to the approach ATI took with its HD5000 series.

Though not confirmed, the 300 series will likely be reserved for lower end parts, making it easier for consumers and system builders to differentiate between the enthusiast and mainstream market segments. The first batch of GF100 series based cards are expected to launch later this quarter.

At its GPU Conference in San Jose, California, Nvidia CEO Jensen Huang officially announced the company’s next line of GPU architecture, codenamed Fermi. Described as the “soul of a supercomputer in the body of a GPU,” the specs boast 3 billion transistors, 512 shader cores, eight times the peak performance in double precision situations, and the ability to run C++ code natively.

There’s no word on Fermi will do for gaming just yet in response to ATI’s Radeon 5000 series, with Huang noting that an announcement regarding the company’s Fermi-powered GeForce cards will come at a later date.

According to a DigiTimes report, GPU developer NVIDIA plans to launch its next-generation GeForce 9 series GPU in February 2008.

Inside sources say that the new high-end GPU, dubbed D9E, will use 65nm process technology for reduced power consumption and quieter operation, in addition to fully supporting Microsoft’s Direct X 10.1 spec and Shader Model 4.1.

Following the release of D9E, NVIDIA is set to launch a mid-range GeForce 9 product family in June 2008. The entire mid-range series will use 55nm process, sources say.

Nvidia to launch GeForce 9 series in February [DigiTimes]

A couple weeks ago, we learned that Xbox 360 consoles with the new Falcon motherboard were hitting the market, primarily in Halo 3 LE units. The Falcon motherboard sports a 65nm CPU, but still retains the 90nm GPU from the Zephyr boards.

According to The Mercury News, Microsoft plans to further lower Xbox 360 production costs and improve reliability by introducing Jasper, the codename for the motherboard that will replace Falcon. Unlike Falcon, Jasper will feature a 65nm GPU, which should considerably reduce heat issues, as majority of the 360 heat issues are directly related to the graphics chip itself.

While this all sounds good, the bad news is we’ll have to wait until August 2008 before these new chips start rolling into production.

Xbox 360 secrets: After Falcon comes Jasper [Mercury News]