Brand Name |
EVGA e-GeForce GTX280 FTW Edition |
Part Number |
e-GeForce GTX280 FTW Edition |
Graphics chip |
GeForce GT200 |
Core Clock |
6700MHz |
Shader Clock |
1458MHz |
SPs |
240 |
Fabrication Process |
65nm |
Transistors |
1400 million |
Memory Clock |
2.214GHz |
Memory Interface |
512-bit |
Memory Bandwidth |
141.7 |
Memory size |
1024 |
ROPs |
32 |
Texture Filtering Units |
80 |
Texture Fillrate |
48.2 Gigatexels/second |
HDCP Support |
Yes |
HDMI Support |
Yes (port included on IO) |
Connectors |
2x Dual-Link DVI, 1 HDMI |
RAMDACs |
400MHZ |
Bus |
PCI Express 2.0 |
Form Factor |
Dual Slot |
Power Connectors |
1 x6-pin, 1 x8-pin |
- GT200 Chipset
- 1.4 Billion Transistors
- 65 nanometer process
- 512-bit memory bus
- 670MHz core clock
- 1215MHz memory clock
- 1458MHz Shader Clock
- 240 Stream Processors
- 21.4 Gigapixels fillrate
- 53.6 Gigatexels texture fillrate
- 1024MB GDDR3 memory
- 155.5GB/second memory bandwidth
- DirectX 10.0
- Pixel Shader 4.0
- Vertex Shader 4.0
- Geometry Shader
- PhysX support
- Beyond Gaming
- Gaming Beyond
- Transcoding
NVIDIA calls the GTX280 their 2nd generation Unified Shading architecture, and as such this new card has a lot of new enhancements to things like General Purpose Graphics Processing Unit (GPGPU), which is using the graphics chip to process mathematical equations and other things than just gaming alone. The GTX 280 is based upon TSMC’s 65 nanometer process with 1.4 Billion transistors. This is in fact the first card to feature over a billion transistors from either ATI or NVIDIA. The new card has 240 Stream Processor cores based of the new GT200 Chipset. Earlier NVIDIA cards like the 9800GTX had 128 SPs so this is nearly doubling the number of processor cores. Each group of 8 Stream Processors is grouped into a Shader Multiprocessor. Each group of 3 Shader Multiprocessors make up a Texture Processing Cluster. The GTX280 has a total of ten TPCs which comprise the sum total of 240 SPs.
NVIDIA says that game developers don’t need DirectX 10.1 as it is an incremental upgrade from 10.0. Therefore they state that they didn’t make the GeForce GTX280 fully DirectX 10.1 compliant. Truthfully, at this moment there are no DirectX 10.1 games available except Assassin’s Creed and the developers took support for it out in the first patch. The GeForce GTX280 supports all of the features of DirectX 10.0, including Pixel Shader 4.0, Vertex Shader 4.0 and the Geometry Shader. NVIDIA says that they improved the Geometry Shading with the GTX280 and there are other architectural improvements as well bringing performance up on the new card. Below is a short evolutionary chart of the advancement of the NVIDIA VGA chipset.
| |
G80 |
G92 |
G94 |
GTX260 |
GTX280 |
Process |
90nm |
65nm |
65nm |
65nm |
65nm |
Transistors |
681 million |
754 million |
505 million |
1.4 billion |
1.4 billion |
Memory bus |
384-bit |
256-bit |
256-bit |
448-bit |
512-bit |
ROPs |
24 |
16 |
16 |
28 |
32 |
SPs |
128 |
128 |
64 |
192 |
240 |
Memory |
768MB |
512MB |
512MB |
896MB |
1024MB |
Memory type |
GDDR3 |
GDDR3 |
GDDR3 |
GDDR3 |
GDDR3 |
Memory bandwidth |
85GB/second |
65GB/second |
58GB/second |
112GB/second |
141GB/second |
Texture fillrate |
37 Gigatexels/second |
40 Gigatexels/second |
21 Gigatexels/second |
40GTexels/second |
5Gigatexels/second |
FLOPS (MADD/MUL) |
518GFLOPS |
600GFLOPS |
|
864GFLOPS |
933GFLOPS |
The GeForce GTX 280 is designed to be both a computing processor and a graphics processor. The CPU of a system and GPU have been converging in terms of their features and flexibility as of late. Intel will launch their own graphics architecture next year based upon their multi-core CPUs in some part, presently code name Larrabee. NVIDIA is heading towards making the graphics processor more flexible and programmable heading towards making the CPU less and less important in the overall graphic equation. EVGA clocks their FTA version of the GTX 280 at an impressive 670MHz, a full 68MHz higher than the reference clock of 602MHz like the default clocked cards that will emerge onto the market. This gives the EVGA GTX280 a pixel fillrate of 21.4 Gigapixels a second and a texture fillrate of 53.6 Gigatexels a second. The EVGA graphics card is capable of over 1 TeraFlop of floating point operations a second, the first available graphics card with this capability enabled. That is an impressive fact just in itself kind of like breaking another sound barrier only quieter.
The memory clock on the EVGA card is 1215MHz. The reference clock on the standard GTX280 is 1100MHz, meaning that the memory clock is much higher than NVIDIA’s and other reference cards will be. NVIDIA’s GTX280 has a 512-bit memory interface, which is a jump from the 256-bit interface on the 9800GTX and a jump from the 384-bit interface of the original GeForce 8800GTX card launched in 2006. ATI is using GDDR5 memory with their next generation cards but on a 256-bit bus, meaning that the memory bandwidth will be more or less equal this generation. NVIDIA is really pushing physics processing with the GeForce GTX280 and GTX260 graphics chips. NVIDIA bought Ageia makers of the PhysX PPU (Physics Processing Unit) four short months ago. With the GeForce GTX280 launch they are pushing the fact that the GeForce GTX280 will be able to run PhysX software with a driver scheduled for a few short weeks from now. PhysX is used by every major platform (PC, Wii, X360, PS3) and upcoming games will allow wonderful effects like flags waving in the wind that have a more natural and realistic look to them. A good 20’’ through 28”” monitor will be the perfect matchup to this card as it is intended for high-resolution gaming that a 19’’ or smaller monitor just plainly cannot support
