Without making any public announcement (because it’s not meant for retail sale), NVIDIA listed its GeForce GT 420 graphics card. This product is available to OEMs only. The GT 420 is derived from the Fermi architecture, and is fully compliant with the latest PC graphics technologies, including DirectX 11 and OpenGL 4. NVIDIA’s reference design is low-profile and single-slot, it draws all its power from the PCI-Express slot.

Under the hood is a 40 nm graphics core (perhaps GF108), it has 48 CUDA cores, and connects to 2 GB of memory across a 128-bit wide DDR3 memory interface, with 28.8 GB/s of memory bandwidth. The core is clocked at 700 MHz, CUDA cores at 1400 MHz, and memory at 900 MHz (1800 MHz effective). Display outputs include DVI, HDMI (full-size), and a detachable D-Sub connector. The card has a maximum power draw of 50W. Later down the line, one can expect NVIDIA to make a consumer GeForce SKU with the same specifications.

Colorful has come up with beastly-looking GeForce GTX 460 graphics card, that makes use of a shark-inspired GPU cooler design, coupled with an equally capable VRM design to facilitate a core clock speed of 900 MHz (vs. 675 MHz reference). The Colorful iGame GTX 460 custom-design PCB of the card makes use of a 6+1 phase VRM, with voltage-measure points. With two BIOS EEPROM chips installed, a switch allows users to choose between the two BIOS chips when the system is powered down. Apart from helping maintain two clock and fan profiles, the feature protects against overly-ambitious BIOS modifications by adding some redundancy. Apart from 900 MHz (core), the CUDA cores are clocked at 1800 MHz, and the memory at 1050 MHz (4200 MHz effective).

The cooler makes use of aggressive-looking shroud design that is inspired from shark-fins. Underneath it is a dense aluminum fin array that sits right on top of the GPU. The GPU base has four 8 mm thick heat pipes that make direct contact with the GPU. Two of these heat pipes convey the heat to the main aluminum fin array, and two to a smaller aluminum fin array that is positioned right on top of the VRM area. A backplate decks up the reverse side of the PCB. Colorful will release this card to the Asian market in September.

More pictures follow:

Colorful has come up with beastly-looking GeForce GTX 460 graphics card, that makes use of a shark-inspired GPU cooler design, coupled with an equally capable VRM design to facilitate a core clock speed of 900 MHz (vs. 675 MHz reference). The Colorful iGame GTX 460 custom-design PCB of the card makes use of a 6+1 phase VRM, with voltage-measure points. With two BIOS EEPROM chips installed, a switch allows users to choose between the two BIOS chips when the system is powered down. Apart from helping maintain two clock and fan profiles, the feature protects against overly-ambitious BIOS modifications by adding some redundancy. Apart from 900 MHz (core), the CUDA cores are clocked at 1800 MHz, and the memory at 1050 MHz (4200 MHz effective).

The cooler makes use of aggressive-looking shroud design that is inspired from shark-fins. Underneath it is a dense aluminum fin array that sits right on top of the GPU. The GPU base has four 8 mm thick heat pipes that make direct contact with the GPU. Two of these heat pipes convey the heat to the main aluminum fin array, and two to a smaller aluminum fin array that is positioned right on top of the VRM area. A backplate decks up the reverse side of the PCB. Colorful will release this card to the Asian market in September.

More pictures follow:

MSI is readying its newest non-reference graphics card based on the NVIDIA GeForce GTX 460 GPU, the overclocker-friendly N460GTX HAWK 1 GB. This is MSI’s second non-reference design after the N460GTX Cyclone series. It features a beefier VRM that facilitates overclocking close to (or over) 25% of the reference speeds. The card is cooled by MSI’s well known Twin-Frozr II GPU cooler. The cooler is claimed by the company to reduce temperatures by as much as 18°C and run 15.7% quieter compared to NVIDIA reference cooler.

Under the hood is a VRM that makes use of a 7+1 phase design with MSI’s Solid State Chokes (SSCs), which don’t give out high-pitched noise (translating into energy loss) during load. The SSCs are backed by high-C capacitors. The clock speeds out of the box are 780 MHz (core), 1560 MHz (CUDA core), and 900/3600 effective MHz (memory). Like most other 1 GB GTX 460 cards, it features 336 CUDA cores, and a 256-bit wide memory interface. A copy of Assassins’ Creed II game is bundled. MSI’s N460GTX HAWK will release later this week, priced at €226.

Source: TechConnect Magazine

MSI is readying its newest non-reference graphics card based on the NVIDIA GeForce GTX 460 GPU, the overclocker-friendly N460GTX HAWK 1 GB. This is MSI’s second non-reference design after the N460GTX Cyclone series. It features a beefier VRM that facilitates overclocking close to (or over) 25% of the reference speeds. The card is cooled by MSI’s well known Twin-Frozr II GPU cooler. The cooler is claimed by the company to reduce temperatures by as much as 18°C and run 15.7% quieter compared to NVIDIA reference cooler.

Under the hood is a VRM that makes use of a 7+1 phase design with MSI’s Solid State Chokes (SSCs), which don’t give out high-pitched noise (translating into energy loss) during load. The SSCs are backed by high-C capacitors. The clock speeds out of the box are 780 MHz (core), 1560 MHz (CUDA core), and 900/3600 effective MHz (memory). Like most other 1 GB GTX 460 cards, it features 336 CUDA cores, and a 256-bit wide memory interface. A copy of Assassins’ Creed II game is bundled. MSI’s N460GTX HAWK will release later this week, priced at €226.

Source: TechConnect Magazine

Well placed sources seem to have finally pieced together specifications of NVIDIA’s GeForce GTS 450 graphics card, the company’s mid-range offering that competes with ATI Radeon HD 5700 series products. In the run up to this, we’ve come across reports showing that the SKU is based on a GPU codenamed GF106, then we came across NVIDIA reference design PCB drawings, followed by clock speeds, and finally, pictures of the GPU itself revealing quite a bit about the die size. The most recent report affirms the specs we know till now, plus revealed the CUDA core count, which is 192. The CUDA cores are clocked at 1566 MHz, and that the memory is clocked at 902 MHz (3608 MHz effective).

So as a quick recap of the tech specs of the GeForce GTS 450, it’s based on the new 40 nm GF106 silicon, the die area is roughly 240 mm². The GPU has 192 CUDA cores, and a 128-bit wide GDDR5 memory interface holding 1 GB of memory. The core is clocked at 789 MHz, and memory at 902 MHz (3608 MHz effective), translating into 57.8 GB/s of memory bandwidth, CUDA cores at 1566 MHz. It supports 2-way SLI, and gives you access to the latest PC graphics technologies, including support for DirectX 11 and OpenGL 4.x, along with support for a wide range of NVIDIA-exclusive technologies. NVIDIA’s GeForce GTS 450 will be released by mid-September.

Source: Heise.de

Seen on numerous occasions and teaser videos as in either its prototype form or just plain CGI drawings, XFX seems to have finally made a working, ready for release iteration of its HD 5970 Black Edition 4 GB graphics card in its wacky retail packaging, which it hype-marketed through teaser videos. TweakPC got to lay its hands on one of these limited edition packages, and went ahead un”boxing” it. The package is actually a bag meant to carry small sniper-rifles (good luck getting it through customs). Inside, the actual graphics card is enclosed inside a case shaped somewhat like a Belgian P90 submachine gun. If that wasn’t enough, the magazine pockets on the bag contain accessory boxes that look like magazines.

The card itself has a redesigned cooling assembly, that looks different from what the first iteration looked like. It has grooves on its top, and a badge with a limited edition serial number. Internally, the GPUs are pushed far apart, each has its own heatsink. The fan is located in the middle, it blows air to those heatsinks located on either sides. Power is drawn in from two 8-pin PCI-E power connectors. Display connectivity includes 6 mini-DP (DVI adapters included), supporting Eyefinity6. Each GPU carries the same clock profiles as the Radeon HD 5870, 850 MHz core, 1200 MHz (4800 MHz effective) memory, but addresses 2 GB of memory each. XFX seems to have commenced selling, expect it to be priced over $1000.

Pictures of the card follow.

Inno3D released its non-reference design GeForce GTX 465 OC Vapor-Freeze graphics card. This factory-overclocked card uses NVIDIA reference PCB (albeit in green), with its own-design cooling assembly. The cooler makes use of a GPU direct-contact base, from which five heat-pipes convey heat directly from the GPU to aluminum fin blocks on either sides of a 92 mm PWM-controlled fan that spins at speeds of 1000~4200 rpm, with a rated noise output of 18 dBA. Inno3D claims the cooling solution to up cooling efficiency by 15% over reference NVIDIA cooler.

As far as clock speeds go, the GPU is clocked at 750 MHz, 1350 MHz CUDA cores, and 850 MHz (3400 MHz effective) memory, against reference speeds of 607/1215/3206 MHz. The card has 1 GB of GDDR5 memory across a 256-bit wide interface. As with every other GTX 465, the card is based on the 40 nm GF100 GPU, it has 352 CUDA cores, and supports the latest PC graphics technologies including DirectX 11 and OpenGL 4. The higher clock speeds, Inno3D claims, up the card’s performance by 10%. Display connectivity includes two DVI and a mini-HDMI. The card is 4-way SLI capable, it takes power from two 6-pin PCI-E power inputs. Inno3D did not give out and details about the card’s pricing. A cheaper variant that sticks to NVIDIA reference clock speeds is also available, it lacks the “OC” marker on the box.

PC DDR4-SDRAM, the successor standard for DDR3-SDRAM, which continues increasing memory bandwidths while maintaining the same electrical footprints, surfaced on JEDEC roadmaps, at a recent MemCon conference in Tokyo, Japan. Just the target clock speed range for DDR2-SDRAM was 400~1066 MHz, and that of DDR3-SDRAM is 1066~2133 MHz, the DDR4-SDRAM standard will aim for clock speeds between 2133 and 4266 MHz, with DRAM voltages of 1.1~1.2V, the voltages standards-compliant DDR3 memory will ultimately end up with. Some of the first DDR4 memory chips will be built on the 32 nm or 36 nm manufacturing processes. JEDEC expects sampling of the new memory type to start in 2011 for the industry to come up with appropriate memory controllers and deployment platforms, while actual mass production is slated by 2015.

Source: X-bit Labs

PC DDR4-SDRAM, the successor standard for DDR3-SDRAM, which continues increasing memory bandwidths while maintaining the same electrical footprints, surfaced on JEDEC roadmaps, at a recent MemCon conference in Tokyo, Japan. Just the target clock speed range for DDR2-SDRAM was 400~1066 MHz, and that of DDR3-SDRAM is 1066~2133 MHz, the DDR4-SDRAM standard will aim for clock speeds between 2133 and 4266 MHz, with DRAM voltages of 1.1~1.2V, the voltages standards-compliant DDR3 memory will ultimately end up with. Some of the first DDR4 memory chips will be built on the 32 nm or 36 nm manufacturing processes. JEDEC expects sampling of the new memory type to start in 2011 for the industry to come up with appropriate memory controllers and deployment platforms, while actual mass production is slated by 2015.

Source: X-bit Labs