Sapphire launched its Radeon HD 7870 Toxic 2 GB graphics card, based on the 28 nm "Tahiti LE" Pitcairn silicon. The card uses a swanky new dual-fan cooler that features a design unlike anything launched by Sapphire so far. It uses a single aluminum fin stack that spans the entire length of the card, to which heat is conveyed by four 8 mm-thick heat pipes that make direct contact with the large die. The card also features a back-plate. It appears to use a high-grade VRM with DirectFETs. The HD 7870 Toxic from Sapphire packs 2 GB of GDDR5 memory across a 256-bit wide memory interface. According to our Sapphire contact, the card is exclusive to the Chinese market.Update 28/11
It turns out this card is a damp squib for two reasons:

• It's still based on Pitcairn and not the superior Tahiti LE chip
• It's exclusive to the Chinese market

ASUS slipped in a new high end GeForce GTX 680 graphics card, the GTX680-DC2-4GD5, which features a dual-slot DirectCU II cooler, and 4 GB of memory. The company's DirectCU series lineup for the GTX 680, till date, only included triple-slot cooling solutions. The dual-slot cooler has been extensively used by ASUS on several performance-segment GPUs. The card ships with NVIDIA-reference clock speeds of 1006 MHz base, 1058 MHz GPU Boost, and 6008 MHz (GDDR5-effective) memory. The card features memory chips on either sides of the PCB, and so an aluminum back-plate comes pre-installed. Expect the card to be priced around US $520. A variant called GTX680-DC2G-4GD5 is available, which packs a game bundle that includes Assassin's Creed III, I Am Alive, R.U.S.E, Splinter Cell: Conviction, and Rainbow Six Vegas 2. Its price is unknown.

MSI launched the GeForce GTX 660 HAWK, which was first pictured in mid-September. The card makes use of a custom-design PCB with 10-phase VRM, and features factory-overclocked speeds of 1085 MHz core, 1150 MHz GPU Boost, and 6.00 GHz memory (GDDR5-effective). It packs 2 GB of memory across a 192-bit wide memory interface. Another major component is its Twin Frozr IV cooler, which uses a dense aluminum fin stack to which head directly draw from the GPU is fed by a number of nickel-plated copper heat pipes. The card uses an external GPU Reactor module that conditions power for the GPU, stabilizing OC. It is expected to be priced around US $250.

A 3DCenter.org report shed light on what NVIDIA's GPU lineup for 2013 could look like. According to the report, NVIDIA's next-generation GPUs could follow a similar path to previous-generation "Fermi Refresh" (GF11x), which turned the performance-per-Watt equation around back in favor of NVIDIA, even though the company's current GeForce Kepler has an established energy-efficiency lead. The "Kepler Refresh" family of GPUs (GK11x), according to the report, could see significant increases in cost-performance, with a bit of clever re-shuffling of the GPU lineup.

NVIDIA's GK104 GPU exceeded performance expectations, which allowed it to drive this generation's flagship single-GPU graphics card for NVIDIA, the GTX 680, giving the company time to perfect the most upscaled chip of this generation, and for its foundry partners to refine its 28 nm manufacturing process. When it's time for Kepler Refresh to go to office, TSMC will have refined its process enough for mass-production of GK110, a 7.1 billion transistor chip on which NVIDIA's low-volume Tesla K20 GPU compute accelerator is currently based.

MSI maintained the R7970 Lightning as its flagship AMD Radeon graphics card from March, even after AMD launched the Radeon HD 7970 GHz Edition in June. The company finally decided to combine its R7970 Lightning board with the new "Tahiti XT2" Radeon HD 7970 GHz Edition ASIC, creating the R7970 Lightning Boost Edition (BE). While the original R7970 Lightning shipped with 1070 MHz core and 5.60 GHz memory, the new R7970 Lightning BE ships with 1070 MHz core, 1150 MHz PowerTune Boost, and 6.00 GHz memory (288 GB/s). It packs 3 GB of GDDR5 memory across a 384-bit wide memory interface.

Most other features remain the same. The R7970 Lightning BE packs an enthusiast-grade 17-phase VRM, dual-BIOS, voltage-conditioning GPU Reactor module, and TwinFrozr IV cooling solution. Quite shockingly, MSI didn't take the opportunity to address the lack of dual-link DVI connectors in the original R7970 Lightning, the Lightning BE continues to have two single-link DVI, and four mini-DisplayPort connectors. The new R7970 Lightning BE could displace the original from its US $499.99 price-point.

PNY announced its XLR8 GeForce GTX 650 Ti graphics card. Based on an identical board design to Palit's offering, the card is likely originally manufactured by it. The XLR8 GTX 650 Ti features near-reference clock speeds, of 928 MHz core, and 5.40 GHz (GDDR5-effective) memory. It packs 1 GB of GDDR5 memory across a 128-bit wide memory interface. The card uses an aluminum fin heatsink that's ventilated by an 80 mm fan. Based on the 28 nm GK106 ASIC, the GTX 650 Ti packs 768 CUDA cores, 64 TMUs, and 16 ROPs. PNY's offering includes download coupons to five movies, and a 3-month subscription to E-Sports Entertainment (ESEA) Premium Video Game Network. It is priced at US $179.99.

Galaxy's upcoming GeForce GTX 650 Ti GC (model: 65IGH8DL7AXX) factory-overclocked graphics card was listed (later deactivated) by American retailer Newegg.com for US $149.99. The sale page revealed quite a few details about the card. The specifications listed beat those which have been doing rounds for the past couple of weeks. According to the specs, the GTX 650 Ti in fact has 768 CUDA cores, and not 576, as previously believed. The GPU core is clocked at 966 MHz, with 5.40 GHz (GDDR5-effective) memory. The card packs 1 GB of memory across a 128-bit wide memory interface. Pictured below is Galaxy's GeForce GTX 650 GC (non-Ti, GK107-based), according to the source, the Galaxy GTX 650 Ti GC (GK106-based) looks very similar.

ASUS is working on a new graphics card based on the upcoming GeForce GTX 650 Ti GPU, augmented with a non-reference design cooling solution. ASUS refers to it as "dual-fan," stopping short of calling it "DirectCU II." One can infer that the cooler could look very similar to dual-slot DirectCU II featured on GTX 660 DirectCU II, with the exclusion of direct-contact heat pipes. The new cooler, according to leaked marketing material will be 11% quieter, and run the GPU 14% cooler than NVIDIA reference cooler. The new marketing materials confirm specifications revealed by an older leak. The GPU is clocked at 980 MHz, with 1350 MHz memory (5.40 GHz GDDR5-effective), with 576 CUDA cores, and 1 GB of memory across a 128-bit wide interface. The GeForce GTX 650 Ti is expected to be priced around US $149, and is slated for October 9.

MSI announced the GX60, an AMD-powered gaming notebook first spotted at this year's Computex event. The notebook is powered by an AMD A10-4600M quad-core "Trinity" APU, Radeon HD 7970M discrete graphics processor with 2 GB GDDR5 memory, and up to 16 GB dual-channel DDR3-1600 MHz memory. Its storage components include 128 GB SuperRAID SSD, 750 GB hard drive, Blu-ray writer or Blu-ray + DVD-RW combo drive. The 15.6-inch anti-glare screen packs 1920 x 1080 pixels resolution. Other gaming-grade features include Bigfoot Killer NIC, keyboard made by SteelSeries, and HD webcam. Pricing differs with optional components.

KFA2 a leading manufacturer of graphics cards based on NVIDIA GPUs, announced today the KFA2 GeForce GTX 660 EX OC and KFA2 GeForce GTX 650 EX OC graphics cards, both cards offering mainstream PC gamers access to the latest advances in GPU technology and unheard of levels of graphics power without breaking the bank. The improved power handling and ultra-efficient cooling of KFA2's custom designs result in incredible overclocking potential with energy efficiency in mind. These aspects combine to deliver even better performance, low power requirements and fantastic overall value.

The KFA2 GeForce GTX 660 EX OC 2GB edition, similar foundation of the previous Ti version, fitted with a newly designed custom PCB, including an improved layout for more efficient signal path, high quality shielded inductors to maintain cleaner signal and eliminate coil whine, and thru-board ventilation to help keep power regulating components cool. The cooler consists of a dual fan system which has been optimized to keep fan noise to an absolute minimum at high speeds. Beneath the fans, a high surface area aluminum heat sink with triple heat pipes maintains very cool GPU temps even under extended run times with heavy workload.

Inno3D's flagship GeForce GTX 660 graphics card, the iChill HerculeZ2000XE smiled for the camera. Videocardz.com scored pictures of the card and its cooler dissembled. To begin with, the card takes advantage of the iChill HerculeZ cooling solution. Its shroud and fans can be easily detached without needing any tools, letting you clean the heatsink underneath. The heatsink uses a pair of aluminum fin stacks, to which heat from the GPU is conveyed by a pair of 8 mm-thick nickel-plated copper heat-pipes.

According to the source, the GTX 660 iChill HerculeZ2000XE from Inno3D is a factory-overclocked graphics card, with a GPU Boost clock speed of 1058 MHz and memory clock speed of 6008 MHz (GDDR5-effective). The card packs 2 GB of GDDR5 memory across a 192-bit wide interface. Based on the 28 nm GK106 silicon, the GeForce GTX 660 is reported to pack 960 CUDA cores. In Mainland China, the card is expected to be priced at 1,699 RMB (US $268).

MSI announced a the R7750 OC V2, a new factory-overclocked graphics card based on the Radeon HD 7750 GPU. The card ships with clock speeds of 830 MHz core, with 4.50 GHz memory. It packs 1 GB of GDDR5 memory across a 128-bit wide interface. Based on the 28 nm "Cape Verde" silicon, it packs 512 Graphics CoreNext stream processors. Unlike the R7750-PMD1GD5/OC, the new OC V2 features a quieter single-fan cooler, which uses a round GPU heatsink with radially-projecting aluminum fins. The new cooler also appears to be cost-effective, giving MSI more room for price-cuts.

A little later this month, NVIDIA plans to launch the GeForce GTX 660, its first desktop GPU based on the 28 nm GK106 silicon. The GTX 660 carries NVIDIA's ASIC codename GK106-400. An internal document detailing other SKUs based on the GK106 in the works, was leaked to the web. According to a 3DCenter.org report, the two ASIC codenames GK106-250 and GK106-200 described in the document could be new SKUs.

The GK106-250, carrying a working title GeForce GTX 655, could have one of the five SMX units of the GK106 disabled, resulting in a CUDA core count of 768. According to the document, it features a 192-bit wide GDDR5 memory interface, holding 2 GB of memory, much like the GTX 660. The GK106-200, on the other hand, with a working title GeForce GTX 650 Ti, could feature just three of the five SMX units on the GK106 silicon, making up 576 CUDA cores. According to the document, it features 2 GB of GDDR5 memory across a 128-bit wide memory interface, much like the GK107-based GTX 650. The two SKUs could help NVIDIA seal gaps in its mainstream desktop GPU lineup.

According to a new 3DCenter.org report, NVIDIA is planning a joint launch of two its key mid-range GPUs, the GeForce GTX 660 and GeForce GTX 650, on September 12, 2012. The two GPUs will let NVIDIA capture key sub-$300 and sub-$200 price-points, with performance levels that could destabilize AMD's Radeon HD 7800 and HD 7700 series. Based on the GK106 silicon, the GTX 660 is said to pack 960 CUDA cores and 2 GB of GDDR5 memory across a 192-bit wide memory interface; while the GK107-based GeForce GTX 650 is said to pack 384 CUDA cores and 1 GB of GDDR5 memory across a 128-bit wide memory interface.

Here is the first picture of what the retail version of NVIDIA's GeForce GTX 650 graphics card looks like. Based on the same GK107 silicon as the GeForce GT 640, the GTX 650 features faster GDDR5 memory. Its reference design board closely resembles that of the GT 640, with the exception of the black PCB and a 6-pin PCIe power connector to draw power from. Interestingly, the new chip has the same typical power draw rating as the GT 640, at 65W. This could give it a significant performance/Watt edge over its predecessor, the GeForce GTS 450. The GeForce GTX 650 packs 384 CUDA cores, 16 ROPs, and a 128-bit wide GDDR5 memory interface, holding 1 GB of memory. It is said to feature similar GPU clock speed to the GT 640, lack of GPU Boost, but with 5.00 GHz GDDR5 memory, compared to the 1800 MHz GDDR3 memory the GT 640 features. The GTX 650 is bound for a little later this quarter.

ASUS' Republic of Gamers MARS III dual-GeForce GTX 680 graphics card may have been extensively covered at this year's Computex event, but very few have had a peek at its innards (PCB). Expreview posted pictures of the card's PCB, sourced from the manufacturer.

The pictures reveal an unusually long and tall PCB, which draws power from three 8-pin PCIe power connectors, a 21-phase VRM that uses Super Alloy chokes and driver-MOSFETs, PLX PEX8747 PCI-Express 3.0 x48 bridge chip, and of course the two GK104 GPUs with a total of 32 individual GDDR5 memory chips (16 for each GPU, 16 on each side of the PCB) wired to them, totaling 8 GB of memory.

Edit: We received an update from ASUS, clarifying that this card will not be released. The design was only displayed during a factory tour, to show ASUS craftmanship.

AMD today launched the AMD FirePro S9000 and S7000, the industry's most powerful dual- and single-slot server graphics cards for compute, virtual desktop infrastructure (VDI) and workstation graphics deployments in data centers. Equipped with next-generation Graphics Core Next architecture and intelligent AMD PowerTune and ZeroCore Power technologies, the AMD FirePro S9000 and S7000 reduce power consumption by up to 95% at idle, meaning greater savings and reduced operating costs in the data center.

By using the AMD FirePro S9000 and S7000 server graphics, information technology (IT) managers can now effectively address a number of daily data center challenges, while empowering IT professionals to manage intense workloads securely from virtually anywhere in the world at virtually any time. By having the flexibility to set up multiple virtual desktops, scientists, engineers and graphics professionals can have access to accelerated productivity suites including computer aided design (CAD), and media and entertainment (M&E) applications, offering them workstation-class performance in a scalable network environment.

ASUS announced a trio of DirectCu V2 series graphics cards, the Radeon HD 7950 DirectCu II V2 TOP (HD7950-DC2T-3GD5-V2), the Radeon HD 7870 DirectCu II V2 TOP (HD7870-DC2TG-2GD5-V2), and Radeon HD 7870 DirectCu II V2 (HD7870-DC2-2GD5-V2). The three cards feature slightly modified cooler and PCB designs over the originals. In case of the HD 7950 DCU2 V2, the cooler is compacted with a denser heatsink, and the PCB reduced in length. In case of the HD 7870 cards, the cooler uses a denser heatsink.

The HD 7950 DCU2 V2 TOP ships with 900 MHz core, and 5.00 GHz GDDR5-effective memory. It packs 3 GB of memory across a 384-bit wide memory interface. The HD 7870 DCU2 V2 TOP ships with 1100 MHz core, and 5.00 GHz memory, while the HD 7870 DCU2 V2 ships with 1000 MHz core and 4.80 GHz memory. The two pack 2 GB of memory across a 256-bit wide memory interface.

Palit Microsystems Ltd, the leading graphics card manufacturer, introduces the latest Kepler architecture: GeForce GTX 660 Ti Series. The GeForce GTX 660 Ti graphics card represents the Kepler "sweet spot" for gamers-the ultimate combination of performance, power efficiency, and affordability.

Continuing with Palit brand new JetStream 風, Palit offers another GeForce GTX 660 Ti JetStream with factory overclocked at boost clock 1084 MHz that is +104 MHz than the standard speed and 2GB GDDR5 memory clock is set at 6108 MHz (+100 MHz higher).- that presents with 1.5x faster and up to 2x more power efficient than the previous-generation GTX560Ti and with innovative custom cooling and optimized overclocking capabilities to deliver the ultimate gaming experience for gamers.

Point of View the leading European manufacturer of an exclusive range NVIDIA based 3D processor boards, advanced netbooks as well as fancy 7" and 10" Tegra tablet computers and additional enthusiast PC products, announces today that it is first to market with a 3D processor board featuring 4GB of GDDR5 memory and a "selected low leakage" GeForce GTX 680 GPU resulting in increased auto-overclocking.

The POV/TGT GeForce GTX 680 4GB UltraCharged "LOW LEAKAGE SELECTION" features POV/TGT's award winning proprietary High-Performance/Wide-Bandwidth hardware design running at a base clock of 1111 MHz. Due to an utmost efficient combination of hand selected low leakage GPUs and TripleFanCooling with a TGT backplate, the auto-overclocked enhanced top-speed of the POV/TGT GeForce GTX 680 4GB UltraCharged "LOW LEAKAGE SELECT- ION" exceeds the performance of many other products with comparable technical specifications.

Apparently, GIGABYTE's GeForce GTX 670 WindForce 3X OC graphics card (model: GV-N670OC-2GD) is riddled with stability issues that can only be corrected with a graphics card BIOS update. The stability issues apparently relate to the way the GPU is addressing overclocked Hynix GDDR5 memory chips. The company released a pair of BIOS files, version F4 based on NVIDIA source BIOS 80.04.31.00.58 and version F12 based on NVIDIA source BIOS 80.04.31.00.09. The two were posted recently, but are date-stamped 13-Jul-2012. The two don't change the clock speeds the card ships with: 980 MHz core, 1058 MHz GPU Boost, and 6.00 GHz memory. Graphics card BIOS updating is a trickier process than motherboard BIOS updates, and you need to be really sure what you're doing. Find the BIOS files, and related documentation in the link below.DOWNLOAD: BIOS updates for GV-N670OC-2GD

MSI's upcoming GeForce GTX 660 Ti HAWK 3 GB graphics card, installed in 3-way SLI configuration was pictured. One of the three was put through 3DMark03 and 3DMark 11. The picture reveals a card that looks similar to the company's GeForce GTX 670 Power Edition. MSI used double the memory amount, by populating the GPU's 192-bit wide GDDR5 memory interface with 3 GB of memory (2 Gbit x 12). The card is cooled by MSI's Twin Frozr IV cooling solution. It is clocked at 1020 MHz GPU core, 1098 MHz GPU Boost, and 6.00 GHz (GDDR5-effective) memory. The GPU-Z screenshot reveals more details.

While three of these cards were pictured in 3-way SLI configuration, only one of the three were put through tests. The cards were driven by a yet-unannounced GeForce 304.87 beta driver. In 3DMark03, the MSI GeForce GTX 660 Ti HAWK scored 18,731 points, and in 3DMark 11 Extreme Preset, it scored X2,862 points. Both these figures are similar to those of a reference NVIDIA GeForce GTX 670.

Several Swedish online retailers have begun listing ASUS GeForce GTX 660 Ti DirectCU II graphics card ahead of its worldwide launch. The card is listed at roughly 3,350 to 3,551 Swedish Kronor (US $492 to $521), including local taxes. The lowest price before taxes is 2,679 Kronor ($393.5). Looking at these prices, we expect US pricing to be in the $380-$420 range, before taxes. None of the retailers detailed the card, let alone post pictures of it. Based on the 28 nm GK104 silicon, the GeForce GTX 660 Ti is expected to feature 1,344 CUDA cores, and 2 GB of memory across a 192-bit wide GDDR5 memory interface.

A little earlier this week, specifications of NVIDIA's GeForce GTX 650 graphics processor were reported. An even newer report by DonanimHaber details its sibling, the GeForce GTX 650 Ti, designed for the sub-$250 market. The new GPU is based on the NVIDIA's newest GK106 silicon, while the GTX 650 sticks to the GK107, and is essentially a beefed-up GeForce GT 640. These are the specifications of the GeForce GTX 650 Ti we're looking at:

• 28 nm GK106 silicon
• 960 CUDA cores
• 192-bit wide GDDR5 memory interface
• 1 GB and 2 GB memory options
• Q4-2012 launch

The newest in the parade of 4 GB GeForce GTX 680 graphics cards is made by Leadtek (model: WFGTX680-4G). The WinFast GTX 680 4 GB, for the most part, sticks to NVIDIA's reference design. Its cooling solution is completely carried over from it. From the looks of it, the PCB, too, is similar to the reference design, with the exception of double the memory amount.

The 4 GB of memory is spread across 16 GDDR5 memory chips. Leadtek's card sticks to NVIDIA reference clock speeds of 1006 MHz core, 1058 MHz GPU Boost, and 6008 MHz GDDR5-effective memory. The card draws power from two 6-pin PCIe power connectors. Slated for 1st August release, the Leadtek WinFast GeForce GTX 680 4 GB is priced at 59,800 JPY (US $765).