Today NVIDIA introduces the latest member of their DirectX 11 lineup. Their new GeForce GTX 460 is based on the all new 40 nm GF104 GPU which is based on the Fermi architecture introduced earlier this year. The GTX 460 is positioned at the lower end of the mid-range performance segment around the $200 price bracket. NVIDIA offers two variants of the GeForce GTX 460, one with 768 MB of GDDR5 memory and one with 1 GB. Due to the GPU architecture this change in memory size not only affects the actual memory but also other performance relevant figures. The reduction of memory size is achieved by installing less memory chips on the card which reduces the bus width of the GPU from 256-bit to 192-bit on the 768 MB version. Since the ROPs are coupled to the memory interface this also results in less ROP units. Combined all those changes reduce the fillrates and memory performance of the card by 25%. One important aspect of this review is how much of a difference this can make in real life.
We have five GTX 460 reviews for you today, making this the most covered VGA card launch in TPU history:
- AXLE GeForce GTX 460 768 MB: Reference design 768 MB
- MSI GeForce GTX 460 Cyclone OC 768 MB: Factory overclocked 768 MB
- ZOTAC GeForce GTX 460 1 GB: Reference design-like 1 GB
- MSI GeForce GTX 460 Cyclone OC 1 GB: Factory overclocked 1 GB
- NVIDIA GeForce GTX 460 SLI: Two 1 GB cards running in SLI.
NVIDIA's GeForce Fermi (GF) 104 GPU comes with 384 shaders (CUDA cores) in the silicon but NVIDIA has disabled 48 of them to reach their intended performance targets and to improve GPU harvesting. Unlike with GF100, the GF104 has more populated streaming multiprocessors (SMs), 48 cores per SM vs. 32 cores per SM on GF100. I marked the disabled units in red in the diagram above, please note that any of the eight SM blocks can be disabled, not only the "last" one. On the 768 MB version, there are also less ROP units and L2 cache as in the original architecture diagram, as indicated above.
|Memory Size||1024 MB||1024 MB||896 MB||768 MB||1024 MB||1024 MB||1024 MB||1024 MB||1280 MB||1024 MB||1536 MB|
|Memory Bus Width||128 bit||256 bit||448 bit||192 bit||256 bit||256 bit||512 bit||256 bit||320 bit||256 bit||384 bit|
|Core Clock||850 MHz||800 MHz||602 MHz||675 MHz||675 MHz||607 MHz||648 MHz||725 MHz||607 MHz||850 MHz||700 MHz|
|Memory Clock||1200 MHz||1000 MHz||1107 MHz||900 MHz||900 MHz||802 MHz||1242 MHz||1000 MHz||837 MHz||1200 MHz||924 MHz|