- 65 nanometer process at TSMC
- 756 million transistors
- 128 Stream Processors
- 256-bit memory interface
- CUDA
- PhysX
- DirectX 10.0
- Pixel Shader 4.0
- Vertex Shader 4.0
- 2nd generation Unified Shader Architecture
- One 6-pin power connector
- 756 MHz core clock
- 2200MHz memory clock
- 1836MHz Shader clock
- 16 ROPs
- 70.4GB/second memory bandwidth
- 12.1Gigapixel/second fill rate
- 48.4Gigatexels/second Texel fill rate
- SLI
- Height 4.376in-111.15mm
- Length 9.5in-241.3mm
- OpenGL 3.0
- 1024GB GDDR3 memory
|
Brand Name |
EVGA
|
|
Part Number |
01G-P3-1155-TR
|
|
Graphics Chip |
G92
|
|
Core clock |
756
|
|
Shader Clock |
1856
|
|
SPs |
128
|
|
Fabrication Process |
65nm
|
|
Transistors |
754 million
|
|
Memory clock |
2200
|
|
Memory Interface |
256-bit
|
|
Memory bandwidth |
70.4GB/second
|
|
Memory Size |
1024MB
|
|
ROPs |
16
|
|
Texture Filtering Units |
64
|
|
Texture Filtering Rate |
48.4G/Texel's second
|
|
HDCP Support |
Yes
|
|
HDMI Support |
Yes (via adapter)
|
|
Connectors |
Dual Display Port, DVI, TV-Out
|
|
RAMDACs |
400MHz
|
|
Bus |
PCI Express 2.0
|
|
Form Factor |
Dual Slot
|
|
Power Connectors |
6-pin
|
The GeForce GTS 250 is based upon NVIDIA's G92 chip that was launched in 2007 in the guise of the 8800GT. The chip on this particular card was manufactured on July 29, 2008 on TSMC's 65nanometer process and has 754 million transistors. NVIDIA decided to rebrand their 9800GTX+ to the GTS250. I've stated my objections to the rename in other reviews but after some thought I think that a unified naming process makes a lot of sense. After all, with the old naming scheme we would have a GeForce 10200 that does not outperform the GeForce 9800GTX due to it being an integrated platform.
I won't rehash the DirectX 10.0 feature set that we have seen since the GeForce 8800GTX in 2006 except to summarize the features. Suffice it to say, you have a unified Shader architecture, Pixel Shader 4.0, Vertex Shader 4.0, and much more. Developers are shipping games with DirectX 10.0 and DirectX 10.1 support. The big feature that NVIDIA is pushing is PhysX with games from 1C, Capcom, and Electronic Arts using hardware physics to do amazing effects like persistent debris, wind affecting the way a character or cloth behaves, realistic smoke effects, and much more. Games like Cryostasis, Mirror's Edge and others have been announced using PhysX.
EVGA decided to implement several versions of the GeForce GTS 250 as they do with every NVIDIA based launch. Today's review is on the reference clocked 1024MB version which means 756MHz for the core, 1100MHz for the memory and 1836MHz for the Shader clock. There are 16 ROPs and 64 texturing units. The pixel fill rate for the card is 12.1 Gigapixels and the Texel fill rate for the card is 48.4 Gigatexels/second. Memory bandwidth is important to video cards and this card has 70.4GB/second of memory bandwidth. There is also a 512MB version and a Superclocked version of both this 1024MB card and the 512MB card.
