The new Single PCB Geforce GTX 295 has been one of the most frequently discussed graphics cards lately. PC Games Hardware just received a sample from Zotac and tested the card for its capabilities.
GPU-Z doesn't fully recognize the new GTX 295 yet, but confirms the memory frequency of 1,008 MHz. [Source: view picture gallery]
Early this year Nvidia released the Geforce GTX 295 in order to conquer the performance crown. As our review back then revealed this was achieved, but since then the dual PCB layout of the card had been annoying Nvidia's board partners. The sandwich design is of the original GTX 295 is complex and expensive to produce. A redesign that converts the GTX 295 into a single PCB card - including a modified cooling solution - is supposed to improve the production costs. PC Games Hardware tests if you as the customer receive a benefit from the new revision. Nvidia Geforce GTX 295 Single PCB reviewed: Specification
The new GTX 295 runs - with one cosmetic exception - with the same frequencies as the original model. The two graphics chips are clocked to 576/1,242 MHz (GPU/Shader) and communicate via SLI. Both GPUs are fully-fledged GT200b chips with 240 ALUs and 80 texture units each. Only one ROP partition is deactivated causing the reduction of the memory interface from 512 to 448 bit. Each chip has access to 896 MiByte GDDR3 VRAM and the video memory isn't running with the original 999 MHz has it did on the first GTX 295 but at 1,008 MHz. This matches an increase of only 0.9 percent. In other words: The gaming performance of the single PCB Geforce GTX 295 is practically identical to the original design. Therefore we don't run benchmarks with the old GTX 295 and rather compare the new version to more interesting multi GPU setups.
Naming the card is up to the board partners. Most of them call the single PCB GTX 295 by its old name, while Evga for example adds the tag "Co-Op”. In order to make it easier to distinguish the two GTX 295 versions we call the new card "Geforce GTX 295 2.0
”. * Single values are not necessarily doubled for multi GPU setups.
** MUL performance not usable to full extend