NVIDIA GB10 vs NVIDIA Tesla V100 PCIe 16 GB Comparison

NVIDIA
GEFORCE

NVIDIA GB10

CORE STATE GB20B
VRAM 128 GB
CLOCK SPEED 2418 MHz
TDP 140 W
BUS WIDTH 256 bit
ARCHITECTURE Blackwell 2.0
nm
PROCESS 5 nm
LAUNCH DATE 2025
VS
NVIDIA
GEFORCE

Tesla V100 PCIe 16 GB

CORE STATE GV100
VRAM 16 GB
CLOCK SPEED 1380 MHz
TDP 300 W
BUS WIDTH 4096 bit
ARCHITECTURE Volta
nm
PROCESS 12 nm
LAUNCH DATE 2017

PERFORMANCE BENCHMARKS

geekbench_opencl
120,137
163,063
geekbench_vulkan
114,648
113,062

Analysis: NVIDIA GB10 vs NVIDIA Tesla V100 PCIe 16 GB

The benchmark data presents a clear generational clash: the NVIDIA Tesla V100 PCIe 16 GB, an end-of-life Volta-era accelerator, and the NVIDIA GB10, an active Blackwell 2.0 part, trade blows in different workloads. The V100 dominates in OpenCL compute, while the GB10 edges ahead in Vulkan, with the final verdict hinging on the specific application and the massive memory and feature differences between the two.

Head-to-Head Benchmarks

The most decisive result is in Geekbench OpenCL, where the NVIDIA Tesla V100 PCIe 16 GB delivers a score of 163,063 against the NVIDIA GB10's 120,137. This is a substantial 35.7% lead for the V100, a gap that underscores the older card's raw compute throughput in this specific API test. The V100's advantage here is so significant that it places it firmly above its nearest rivals, including the AMD Instinct MI100 (139,035) and the NVIDIA Tesla V100 SXM2 32 GB (137,731), further solidifying its position as a compute heavyweight in this benchmark.

However, the tables turn in the Geekbench Vulkan test. Here, the NVIDIA GB10 scores 114,648, narrowly edging out the V100's 113,062. The delta is just 1.4% in favor of the GB10, making it a much closer contest. While the V100 still wins the overall average benchmark score (138,063 vs. 117,393 for the GB10), the Vulkan result shows that the newer architecture is not without its strengths, particularly in graphics-oriented APIs. The GB10's Vulkan score is also competitive with its own rivals, sitting close to the AMD Radeon PRO W7700 (118,976) and the NVIDIA RTX 4000 SFF Ada Generation (117,088).

The data shows a split decision: one win for each card. The V100's OpenCL victory is a dominant one, while the GB10's Vulkan win is marginal. This pattern suggests that the V100 is the superior choice for compute-heavy, OpenCL-based workloads, whereas the GB10 holds a slight edge in Vulkan-centric applications. The average benchmark scores reflect the V100's overall lead, but the Vulkan result is a notable data point for anyone prioritizing that specific API.

Architecture Differences

The two GPUs are separated by two major architectural generations, and the differences are profound. The V100 is built on the Volta architecture using a 12 nm process at TSMC, with a large 815 mm² die housing 21,100 million transistors. In contrast, the GB10 uses the newer Blackwell 2.0 architecture on a 5 nm process, also at TSMC, but with a significantly smaller 382 mm² die. The GB10's transistor count is listed as "unknown," but the die size reduction is stark, and the newer process node allows for a much higher transistor density on paper, even if not directly quantified.

The core configurations diverge significantly. The V100 packs 5,120 shading units, 320 TMUs, and 128 ROPs, while the GB10 offers more shading units (6,144) and TMUs (384) but far fewer ROPs (48). The GB10 also introduces 48 dedicated ray tracing cores, a feature entirely absent from the V100. Tensor core counts also differ, with the V100 having 640 compared to the GB10's 384. Clock speeds are a major differentiator: the V100's base and boost clocks are 1245 MHz and 1380 MHz, respectively, while the GB10 runs much higher at 1665 MHz base and 2418 MHz boost. This clock advantage helps the GB10 achieve a higher FP32 throughput of 29.71 TFLOPS compared to the V100's 14.13 TFLOPS, despite the V100's larger core count in some areas. The FP16 performance is also telling: the V100 achieves 28.26 TFLOPS at a 2:1 ratio, while the GB10 matches its FP32 rate at 29.71 TFLOPS with a 1:1 ratio, indicating a more unified compute path.

Memory is another stark contrast. The V100 uses 16 GB of HBM2 on a 4096-bit bus, delivering a massive 897.0 GB/s of bandwidth. The GB10, however, has 128 GB of LPDDR5X on a much narrower 256-bit bus, resulting in 273.2 GB/s of bandwidth. This is a clear trade-off: the V100 offers far higher bandwidth for memory-intensive compute, while the GB10 offers eight times the capacity for large datasets. The GB10 also features a PCIe 5.0 x16 interface compared to the V100's PCIe 3.0 x16, and it has a single HDMI output, whereas the V100 has no display outputs. Finally, the GB10 is an IGP with a 140 W TDP and no power connectors, while the V100 is a dual-slot card requiring 2x 8-pin connectors and a 300 W TDP.

FAQ

Q: Which GPU has a higher average benchmark score?

A: The NVIDIA Tesla V100 PCIe 16 GB has a higher average benchmark score of 138,063, compared to the NVIDIA GB10's 117,393.

Q: Is the NVIDIA GB10 faster in every test?

A: No. The GB10 wins the Geekbench Vulkan test with a score of 114,648, but the V100 wins the Geekbench OpenCL test with a score of 163,063, a 35.7% difference.

Q: How does the memory capacity and bandwidth compare?

A: The GB10 has significantly more memory at 128 GB of LPDDR5X, but the V100 has much higher bandwidth at 897.0 GB/s over a 4096-bit bus, compared to the GB10's 273.2 GB/s over a 256-bit bus.

Q: What are the key architectural differences?

A: The V100 is based on the Volta architecture on a 12 nm process, while the GB10 uses the Blackwell 2.0 architecture on a 5 nm process. The GB10 also includes 48 ray tracing cores, which the V100 lacks.

Q: Which card has a higher boost clock?

A: The NVIDIA GB10 has a much higher boost clock of 2418 MHz, compared to the V100's 1380 MHz.

Q: What is the power draw difference?

A: The GB10 has a TDP of 140 W and requires no power connectors, while the V100 has a TDP of 300 W and requires 2x 8-pin power connectors.

The Verdict

The data points to a simple conclusion: the NVIDIA Tesla V100 PCIe 16 GB is the superior choice for raw compute performance in OpenCL-based workloads, as evidenced by its 35.7% lead over the GB10 in that specific benchmark. Its higher average score (138,063 vs. 117,393) and its position in the 96th percentile of all GPUs, compared to the GB10's 95th percentile, reinforce its overall performance advantage. For users whose primary application relies on OpenCL compute, the V100 is the clear winner.

However, the NVIDIA GB10 is not without merit. Its 1.4% victory in Vulkan shows it is competitive in graphics-oriented APIs, and its architectural advantages are substantial. The GB10 offers eight times the memory capacity (128 GB vs. 16 GB), a much higher boost clock (2418 MHz vs. 1380 MHz), and double the FP32 throughput (29.71 TFLOPS vs. 14.13 TFLOPS). It also includes ray tracing cores and has a significantly lower TDP of 140 W. The verdict is workload-dependent: the V100 for OpenCL compute dominance, the GB10 for its newer feature set, higher capacity memory, and efficiency.

Specification Differences

This section highlights the key fields where the two GPUs differ.

| Specification | NVIDIA Tesla V100 PCIe 16 GB | NVIDIA GB10 |

|---|---|---|

| Architecture | Volta | Blackwell 2.0 |

| Process Node | 12 nm | 5 nm |

| Die Size | 815 mm² | 382 mm² |

| Transistors | 21,100 million | unknown |

| Base Clock | 1245 MHz | 1665 MHz |

| Boost Clock | 1380 MHz | 2418 MHz |

| Memory Size | 16 GB | 128 GB |

| Memory Type | HBM2 | LPDDR5X |

| Memory Bus Width | 4096 bit | 256 bit |

| Memory Bandwidth | 897.0 GB/s | 273.2 GB/s |

| Shading Units | 5120 | 6144 |

| TMUs | 320 | 384 |

| ROPs | 128 | 48 |

| RT Cores | null | 48 |

| Tensor Cores | 640 | 384 |

| FP32 Performance | 14.13 TFLOPS | 29.71 TFLOPS |

| FP16 Performance | 28.26 TFLOPS (2:1) | 29.71 TFLOPS (1:1) |

| TDP | 300 W | 140 W |

| Slot Width | Dual-slot | IGP |

| Power Connectors | 2x 8-pin | None |

| Suggested PSU | 700 W | 300 W |

| Bus Interface | PCIe 3.0 x16 | PCIe 5.0 x16 |

| Display Outputs | No outputs | 1x HDMI |

| Release Date | 2017-06-20 | 2025-10-14 |

| Production Status | End-of-life | Active |

Where Each One Wins

The NVIDIA Tesla V100 PCIe 16 GB is the clear choice for compute-intensive tasks that leverage OpenCL. Its 35.7% lead in the Geekbench OpenCL test and its higher average benchmark score make it the superior accelerator for scientific computing, data processing, or any workload that can utilize its massive 897.0 GB/s memory bandwidth. Its high percentile ranking (96th) among all GPUs further confirms its strength in this area. If the application is built on OpenCL, the V100's performance lead is decisive.

The NVIDIA GB10 is the better option for a new system or a workload that benefits from its modern features. Its victory in Vulkan, while narrow, indicates better support for that API. More importantly, its 128 GB of memory is a decisive factor for large models or datasets that would not fit in the V100's 16 GB. The GB10's higher FP32 throughput (29.71 TFLOPS) and inclusion of ray tracing cores make it a more versatile and future-proof part. Its significantly lower power draw (140 W vs. 300 W) and lack of external power connectors also make it far easier to integrate into a system. For users starting fresh with a focus on Vulkan, large memory capacity, or modern features, the GB10 is the winner.

DETAILED SPECIFICATIONS

SPECIFICATION
GB10
Tesla V100 PCIe 16 GB
Core Specs
Shading Units
6,144
5,120 -16.7%
Shaders
6,144
5,120 -16.7%
TMUs
384
320 -16.7%
ROPs
48
128 +166.7%
SM Count
48
80 +66.7%
Clocks
Base Clock
1665 MHz
1245 MHz
Boost Clock
2418 MHz
1380 MHz
Memory Clock
1067 MHz 8.5 Gbps effective
876 MHz 1752 Mbps effective
Memory
Memory Size
128 GB
16 GB
VRAM (MB)
131,072
16,384 -87.5%
Memory Type
LPDDR5X
HBM2
Memory Bus
256 bit
4096 bit
Bandwidth
273.2 GB/s
897.0 GB/s
Cache
L1 Cache
128 KB (per SM)
128 KB (per SM)
L2 Cache
50 MB
6 MB
Performance
Pixel Rate
116.1 GPixel/s
176.6 GPixel/s
Texture Rate
928.5 GTexel/s
441.6 GTexel/s
FP32 (TFLOPS)
29.71 TFLOPS
14.13 TFLOPS
FP64 (TFLOPS)
464.3 GFLOPS (1:64)
7.066 TFLOPS (1:2)
FP16 (TFLOPS)
29.71 TFLOPS (1:1)
28.26 TFLOPS (2:1)
AI/RT
RT Cores
48
—
Tensor Cores
384
640 +66.7%
Power
TDP
140 W
300 W
TDP (W)
140
300 +114.3%
Suggested PSU
300 W
700 W
Power Connectors
None
2x 8-pin
Architecture
Architecture
Blackwell 2.0
Volta
GPU Name
GB20B
GV100
Generation
Server Blackwell (Bxx)
Tesla Volta (Vxx)
Process Size
5 nm
12 nm
Transistors
unknown
21,100 million
Die Size
382 mm²
815 mm²
Foundry
TSMC
TSMC
Density
—
25.9M / mm²
API Support
DirectX
—
12 (12_1)
OpenGL
—
4.6
Vulkan
—
1.4
OpenCL
3.0
3.0
CUDA
12.1
7.0
Shader Model
—
6.8
Physical
Slot Width
IGP
Dual-slot
Length
150 mm 5.9 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
—
Production
Active
End-of-life
Predecessor
Server Hopper
Tesla Pascal
Successor
Server Rubin
Tesla Turing
View GB10 Details View Tesla V100 PCIe 16 GB Details