Fermi Arrives - NVIDIA GeForce GTX 480
- < Prev
-
Page 3 of 13 - Introducing the GF100 (part 1)
Page 3 of 13
- Next >
Introducing the GF100 (part 1)
Meet the GF100
All in all, the new architecture is one that has clearly been setup to perform well for HPC needs, but how will it perform on your normal games? Well, that's what we are here to find out.
NVIDIA's latest top dog is the GeForce GTX 480 and in it will be the all-new GF100 GPU, the first GPU to make use of the new Fermi architecture. The GF100 chip will boast 480 CUDA cores, 60 texture mapping units and 48 raster operating units. In terms of clock speeds, the core will operate at 700MHz, whereas the shaders and memory are at 1401MHz and 3696MHz DDR respectively.
Observant readers will no doubt notice that 480 CUDA cores is 32 short of what NVIDIA claims will be found in Fermi. Does this mean a possible 512-core variant of the GeForce GTX 480 in future? When quizzed, NVIDIA did not say why the GF100 features 480 cores instead of the 512 as stated earlier, but many are speculating that this could be because of poor yields due to the complexity of the new chip. To add, NVIDIA also did not explicitly deny the possibility of a 512-core variant in future - now this is interesting. Factor in NVIDIA's penchant for releasing an updated top model of the same GPU architecture and it becomes apparent that an even more capable version is likely in the making.
Anyhow, launching along side the GeForce GTX 480 is the GeForce GTX 470. The GeForce GTX 470 has a scaled down version of the GF100 chip, with an entire SM disabled. This means 448 CUDA cores, 56 texture mapping units and 40 raster operating units. It gets lower clock speeds as well. Its core is running at 607MHz, while shaders and memory at 1215MHz and 3348MHz DDR respectively. We'll share with you our findings of the GTX 470 in the coming weeks, but for now here's how the GF100, specifically the GeForce GTX 480 stacks up against the competition.
Model | NVIDIA GeForce GTX 480 1536MB GDDR5 | ATI Radeon HD 5870 1GB GDDR5 | ATI Radeon HD 5970 2GB | NVIDIA GeForce GTX 285 1GB GDDR3 | NVIDIA GeForce GTX 295 1729MB GDDR3 |
---|---|---|---|---|---|
Core Code | GF100 | Cypress XT | Hemlock XT | GT200 | GT200 |
Transistor Count | 3200 million | 2150 million | 4300 million | 1400 million | 2800 million |
Manufacturing Process | 40nm | 40nm | 40nm | 55nm | 55nm |
Core Clock | 700MHz | 850MHz | 750MHz | 648MHz | 576MHz |
Stream Processors | 480 Stream Processors | 1600 Stream processing units | 3200 Stream processing units | 240 Stream Processors | 480 Stream Processors |
Stream Processor Clock | 1401MHz | 850MHz | 725MHz | 1476MHz | 1242MHz |
Texture Mapping Units (TMU) or Texture Filtering (TF) units | 60 | 80 | 160 | 80 | 160 |
Raster Operator units (ROP) | 48 | 32 | 64 | 32 | 56 |
Memory Clock | 3696MHz GDDR3 | 4800MHz GDDR5 | 4000MHz GDDR5 | 2484MHz GDDR3 | 1998MHz GDDR3 |
DDR Memory Bus | 384-bit | 256-bit | 256-bit | 512-bit | 448-bit |
Memory Bandwidth | 177.4GB/s | 153.6GB/s | 256GB/s | 159.0GB/s | 223.8GB/s |
PCI Express Interface | PCIe ver 2.0 x16 | PCIe ver 2.0 x16 | PCIe ver 2.0 x16 | PCIe ver 2.0 x16 | PCIe ver 2.0 x16 |
Molex Power Connectors | 1 x 6-pin, 1 x 8-pin | 2 x 6-pin | 1 x 6-pin, 1 x 8-pin | 2 x 6-pin | 1 x 6-pin, 1 x 8-pin |
Multi GPU Technology | SLI | CrossFireX | CrossFireX | SLI | SLI |
DVI Output Support | 2 x Dual-Link | 2 x Dual-Link | 2 x Dual-Link | 2 x Dual-Link | 2 x Dual-Link |
HDCP Output Support | Yes | Yes | Yes | Yes | Yes |
Street Price | Launch Price: US$499 | ~US$399 | ~US$599 | ~US$350 | ~US$500 |
- < Prev
-
Page 3 of 13 - Introducing the GF100 (part 1)
Page 3 of 13
- Next >