The Tesla K40t was an enthusiast-class professional graphics card by NVIDIA, launched on November 22nd, 2013. Built on the 28 nm process, and based on the GK110B graphics processor, the card supports DirectX 12. The GK110B graphics processor is a large chip with a die area of 561 mm² and 7,080 million transistors. It features 2880 shading units, 240 texture mapping units, and 48 ROPs. NVIDIA has paired 12 GB GDDR5 memory with the Tesla K40t, which are connected using a 384-bit memory interface. The GPU is operating at a frequency of 745 MHz, which can be boosted up to 876 MHz, memory is running at 1502 MHz (6 Gbps effective). Being a dual-slot card, its power draw is rated at 245 W maximum. This device has no display connectivity, as it is not designed to have monitors connected to it. Tesla K40t is connected to the rest of the system using a PCI-Express 3.0 x16 interface. The card measures 267 mm in length, and features a dual-slot cooling solution. Its price at launch was 7699 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
Base Clock
745 MHz
Boost Clock
876 MHz
Memory Clock
1502 MHz
6 Gbps effective
Memory
Memory Size
12 GB
Memory Type
GDDR5
Memory Bus
384 bit
Bandwidth
288.4 GB/s
Render Config
Shading Units
2880
TMUs
240
ROPs
48
SMX Count
15
L1 Cache
16 KB (per SMX)
L2 Cache
1536 KB
Theoretical Performance
Pixel Rate
52.56 GPixel/s
Texture Rate
210.2 GTexel/s
FP32 (float)
5.046 TFLOPS
FP64 (double)
1.682 TFLOPS (1:3)
Board Design
Slot Width
Dual-slot
Length
267 mm
10.5 inches
TDP
245 W
Suggested PSU
550 W
Outputs
No outputs
Graphics Features
DirectX
12 (11_1)
OpenGL
4.6
OpenCL
3.0
Vulkan
1.2.175
CUDA
3.5
Shader Model
6.5 (5.1)
Card Notes
some specs unknown
GK110B GPU Notes
NVENC: 1st Gen
NVDEC: 1st Gen
PureVideo HD: VP5
VDPAU: Feature Set D
L1 Cache is configurable from 16 KB up to 48 KB per SMX
GK110B has 5 GPCs each capable of 8 pixels per clock. This limits complete GPU to 40 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.)
Latest Drivers:
Windows XP / Server 2003 x64:
Quadro Release R319 U2 (321.01)