The Tesla S2050 was an enthusiast-class professional graphics card by NVIDIA, launched on July 25th, 2011. Built on the 40 nm process, and based on the GF100 graphics processor, the card supports DirectX 12. The GF100 graphics processor is a large chip with a die area of 529 mm² and 3,100 million transistors. Unlike the fully unlocked GeForce GTX 480 Core 512, which uses the same GPU but has all 512 shaders enabled, NVIDIA has disabled some shading units on the Tesla S2050 to reach the product's target shader count. Tesla S2050 combines four graphics processors to increase performance. It features 448 shading units, 56 texture mapping units, and 48 ROPs, per GPU. NVIDIA has paired 12 GB GDDR5 memory with the Tesla S2050, which are connected using a 384-bit memory interface per GPU (each GPU manages 3,072 MB). The GPU is operating at a frequency of 574 MHz, memory is running at 773 MHz (3.1 Gbps effective). Being a dual-slot card, its power draw is rated at 900 W maximum. This device has no display connectivity, as it is not designed to have monitors connected to it. Tesla S2050 is connected to the rest of the system using a PCI-Express 2.0 x16 interface. Its price at launch was 11999 US Dollars.
Based on TPU review data: "Performance Summary" at 1920x1080, 4K for 2080 Ti and faster.
Performance estimated based on architecture, shader count and clocks.
Clock Speeds
GPU Clock
574 MHz
Shader Clock
1147 MHz
Memory Clock
773 MHz
3.1 Gbps effective
Memory
Memory Size
3 GB
Memory Type
GDDR5
Memory Bus
384 bit
Bandwidth
148.4 GB/s
Render Config
Shading Units
448
TMUs
56
ROPs
48
SM Count
14
L1 Cache
64 KB (per SM)
L2 Cache
768 KB
Theoretical Performance
Pixel Rate
16.07 GPixel/s
Texture Rate
32.14 GTexel/s
FP32 (float)
1,028 GFLOPS
FP64 (double)
513.9 GFLOPS (1:2)
Board Design
Slot Width
Dual-slot
TDP
900 W
Suggested PSU
1300 W
Outputs
No outputs
Board Number
P1030 SKU 202
Graphics Features
DirectX
12 (11_0)
OpenGL
4.6
OpenCL
1.1
Vulkan
N/A
CUDA
2.0
Shader Model
5.1
GF100 GPU Notes
PureVideo HD: VP4
VDPAU: Feature Set C
GF100 has 4 GPCs each capable of 8 pixels per clock. This limits complete GPU to 32 pixels per clock and because of that it can't feed all 48 ROPs when they all require data at the same time.
Additional ROPs however can be used for MSAA (because it doesn't require additional data from rasterizers, while giving more work to ROPs.)