GPU Comparison
GEFORCE
NVIDIA GB10
CORE STATE
GB20B
VRAM
128 GB
CLOCK SPEED
2418 MHz
TDP
140 W
BUS WIDTH
256 bit
ARCHITECTURE
Blackwell 2.0
PROCESS
5 nm
LAUNCH DATE
2025
VS
GEFORCE
GRID K2
CORE STATE
GK104
VRAM
4 GB
CLOCK SPEED
—
TDP
225 W
BUS WIDTH
256 bit
ARCHITECTURE
Kepler
PROCESS
28 nm
LAUNCH DATE
2013
PERFORMANCE BENCHMARKS
geekbench_opencl
geekbench_vulkan
geekbench_metal
DETAILED SPECIFICATIONS
SPECIFICATION
GB10
GRID K2
Core Specs
Shading Units
6,144
1,536
-75.0%
Shaders
6,144
1,536
-75.0%
TMUs
384
128
-66.7%
ROPs
48
32
-33.3%
SM Count
48
—
Clocks
Base Clock
1665 MHz
—
Boost Clock
2418 MHz
—
GPU Clock
—
745 MHz
Memory Clock
1067 MHz
8.5 Gbps effective
1250 MHz
5 Gbps effective
Memory
Memory Size
128 GB
4 GB
VRAM (MB)
131,072
4,096
-96.9%
Memory Type
LPDDR5X
GDDR5
Memory Bus
256 bit
256 bit
Bandwidth
273.2 GB/s
160.0 GB/s
Cache
L1 Cache
256 KB (per SM)
16 KB (per SMX)
L2 Cache
50 MB
512 KB
Performance
Pixel Rate
116.1 GPixel/s
23.84 GPixel/s
Texture Rate
928.5 GTexel/s
95.36 GTexel/s
FP32 (TFLOPS)
29.71 TFLOPS
2.289 TFLOPS
FP64 (TFLOPS)
464.3 GFLOPS (1:64)
95.36 GFLOPS (1:24)
FP16 (TFLOPS)
29.71 TFLOPS (1:1)
—
AI/RT
RT Cores
48
—
Tensor Cores
384
—
Power
TDP
140 W
225 W
TDP (W)
140
225
+60.7%
Suggested PSU
300 W
550 W
Power Connectors
None
1x 6-pin + 1x 8-pin
Architecture
Architecture
Blackwell 2.0
Kepler
GPU Name
GB20B
GK104
Generation
Server Blackwell
(Bxx)
GRID
(K2)
Process Size
5 nm
28 nm
Transistors
—
3,540 million
Die Size
—
294 mm²
Foundry
TSMC
TSMC
Density
—
12.0M / mm²
API Support
DirectX
—
12 (11_0)
OpenGL
—
4.6
Vulkan
—
1.2.175
OpenCL
3.0
3.0
CUDA
12.1
3.0
Shader Model
—
6.5 (5.1)
Physical
Slot Width
IGP
Dual-slot
Length
150 mm
5.9 inches
267 mm
10.5 inches
Height
51 mm
2 inches
—
Outputs
1x HDMI
No outputs
Bus Interface
PCIe 5.0 x16
PCIe 3.0 x16
Other
Launch Price
3,999 USD
5,199 USD
Production
Active
End-of-life
Predecessor
Server Hopper
—
Successor
Server Rubin
—