NVIDIA GB10 vs NVIDIA Tesla V100S PCIe 32 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 V100S PCIe 32 GB

CORE STATE GV100
VRAM 32 GB
CLOCK SPEED 1597 MHz
TDP 250 W
BUS WIDTH 4096 bit
ARCHITECTURE Volta
nm
PROCESS 12 nm
LAUNCH DATE 2019

PERFORMANCE BENCHMARKS

geekbench_opencl
120,137
194,415
geekbench_vulkan
114,648
N/A

Analysis: NVIDIA GB10 vs NVIDIA Tesla V100S PCIe 32 GB

The NVIDIA Tesla V100S PCIe 32 GB and the NVIDIA GB10 represent two distinct generations of NVIDIA's computing hardware, separated by six years of architectural evolution. The data shows a fascinating clash between a mature, high-bandwidth accelerator and a modern, power-efficient system-on-chip. In the single head-to-head benchmark available, the Geekbench OpenCL test, the Tesla V100S delivers a score of 194,415, while the GB10 scores 120,137. This gives the Tesla V100S a decisive 61.8% performance advantage in this specific test, marking its only win in the comparison. This stark difference in raw compute output, however, is only one part of a much more complex story involving memory architecture, power consumption, and intended use cases.

Head-to-Head Benchmarks

The only direct benchmark comparison in the data is the Geekbench OpenCL test, and it is a landslide victory for the older Tesla V100S. Its score of 194,415 eclipses the GB10's 120,137 by a substantial 61.8%. This places the V100S in the 98th percentile of all GPUs, whereas the GB10 sits slightly lower at the 95th percentile. The V100S's nearest rival, the NVIDIA RTX 6000D, scores 195,964, a mere 0.8% higher, indicating that the V100S is at the very top tier of performance in this metric. In contrast, the GB10's score is more aligned with mid-range workstation parts; its closest competitor, the AMD Radeon PRO W7700, scores 118,976, which is 1.3% higher, and the NVIDIA RTX 4000 SFF Ada Generation is just 0.3% behind at 117,088. The delta between the two contenders is enormous, suggesting that for applications heavily reliant on raw OpenCL compute throughput, the older card is in a different league.

However, this singular result does not tell the whole story. The V100S's victory is built on its massive memory bandwidth and dedicated tensor core count. With a 4096-bit memory bus and HBM2 memory delivering 1.13 TB/s of bandwidth, it can feed its 5,120 shading units and 640 tensor cores at a rate the GB10 cannot match. The GB10, with its 256-bit bus and LPDDR5X memory, achieves a bandwidth of 273.2 GB/s, which is a fraction of the V100S's figure. While the GB10 has more shading units (6,144 vs. 5,120) and a higher boost clock (2418 MHz vs. 1597 MHz), its FP32 throughput of 29.71 TFLOPS is actually 81.7% higher than the V100S's 16.35 TFLOPS. This paradoxical result—where a card with lower raw FP32 performance wins a compute benchmark—implies that the Geekbench OpenCL test is heavily influenced by memory bandwidth and the efficiency of the tensor cores, rather than pure shader math. The V100S's HBM2 memory and mature Volta architecture appear to be better optimized for this particular workload.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA Tesla V100S PCIe 32 GB has a significantly higher average benchmark score of 194,415, compared to the NVIDIA GB10's 117,393. This places the V100S in the 98th percentile of all GPUs, while the GB10 is in the 95th percentile.

Q: How does the GB10 compare to its closest rival, the AMD Radeon PRO W7700?

A: The GB10's average score of 117,393 is 1.3% lower than the AMD Radeon PRO W7700's score of 118,976. This indicates that the GB10 is performing at a level just below a current mid-range workstation card in this metric.

Q: Is the Tesla V100S a more powerful card than the GB10 in all aspects?

A: No. While the V100S wins the Geekbench OpenCL test by 61.8%, the GB10 has a higher FP32 compute rating at 29.71 TFLOPS versus the V100S's 16.35 TFLOPS. The GB10 also has a higher texture rate of 928.5 GTexel/s compared to the V100S's 511.0 GTexel/s.

Q: What are the memory capacity differences between the two cards?

A: The NVIDIA GB10 has a significantly larger memory capacity of 128 GB of LPDDR5X, while the NVIDIA Tesla V100S PCIe 32 GB has 32 GB of HBM2. However, the V100S has a much higher memory bandwidth of 1.13 TB/s compared to the GB10's 273.2 GB/s.

Q: Which GPU has a higher boost clock speed?

A: The NVIDIA GB10 has a higher boost clock speed of 2418 MHz, whereas the NVIDIA Tesla V100S PCIe 32 GB has a boost clock of 1597 MHz. The GB10 also has a higher base clock at 1665 MHz versus 1245 MHz.

Architecture Differences

The two GPUs are built on fundamentally different architectural philosophies. The Tesla V100S is based on the Volta architecture, fabricated on a 12 nm process at TSMC, with a massive die size of 815 mm². It houses 21,100 million transistors, resulting in a transistor density of 25.9 million per mm². This design is optimized for high-bandwidth computing, featuring a 4096-bit memory interface paired with HBM2 memory. Its core layout includes 5,120 shading units, 320 TMUs, 128 ROPs, and a substantial 640 tensor cores. The V100S supports a full API suite including DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4.

In contrast, the GB10 is a product of the Blackwell 2.0 architecture, manufactured on a much more advanced 5 nm process. Its die size is considerably smaller at 382 mm², and while the transistor count is unknown, the density is not listed. The GB10 opts for a 256-bit bus with LPDDR5X memory, prioritizing capacity (128 GB) over bandwidth. Its core configuration is different as well: 6,144 shading units, 384 TMUs, only 48 ROPs, 48 dedicated RT cores, and 384 tensor cores. The GB10's API support is listed as "N/A" for DirectX, OpenGL, and Vulkan, indicating it is not designed for traditional graphics workloads but for compute or server tasks. The presence of RT cores in the GB10, absent in the V100S, suggests a hardware capability for ray tracing, even if software support is not exposed through standard graphics APIs.

The process node difference is a key factor in their respective performance and efficiency profiles. The GB10's 5 nm process allows for higher clock speeds—up to 2418 MHz boost—and a lower TDP of 140 W, despite having more shading units. The V100S, on the 12 nm node, is limited to a 1597 MHz boost clock and consumes 250 W. This makes the GB10 a far more power-efficient design per watt of compute, though the V100S compensates with its superior memory subsystem.

The Verdict

Based strictly on the data, the choice between these two GPUs is dictated by workload priorities. For raw compute performance in the Geekbench OpenCL test, the NVIDIA Tesla V100S PCIe 32 GB is the undisputed winner, boasting a 61.8% higher score. Its 98th percentile ranking and proximity to the RTX 6000D (only 0.8% lower) place it in an elite performance tier. The V100S is the correct choice for applications that are heavily dependent on memory bandwidth, where its 1.13 TB/s bandwidth is an overwhelming advantage over the GB10's 273.2 GB/s.

However, the GB10 is not without merit. It offers a significantly higher FP32 compute rating (29.71 TFLOPS) and a higher texture rate, suggesting it can handle parallel shader workloads more quickly. Its 128 GB of memory is four times the capacity of the V100S, which is critical for massive datasets that cannot fit in 32 GB. The GB10 is also a much more power-efficient part, with a 140 W TDP versus the V100S's 250 W. It is also an active product with a launch MSRP of 3,999 USD, while the V100S is end-of-life. For a user needing to process extremely large models or datasets on a single card with lower power constraints, the GB10's capacity and efficiency are compelling. The V100S is a legacy powerhouse for raw speed, while the GB10 is a modern, high-capacity workhorse.

Specification Differences

| Specification | NVIDIA Tesla V100S PCIe 32 GB | NVIDIA GB10 |

| :--- | :--- | :--- |

| Architecture | Volta | Blackwell 2.0 |

| Process Node | 12 nm | 5 nm |

| Die Size | 815 mm² | 382 mm² |

| Base Clock | 1245 MHz | 1665 MHz |

| Boost Clock | 1597 MHz | 2418 MHz |

| Memory Size | 32 GB | 128 GB |

| Memory Type | HBM2 | LPDDR5X |

| Memory Bus Width | 4096 bit | 256 bit |

| Memory Bandwidth | 1.13 TB/s | 273.2 GB/s |

| Shading Units | 5120 | 6144 |

| TMUs | 320 | 384 |

| ROPs | 128 | 48 |

| RT Cores | N/A | 48 |

| Tensor Cores | 640 | 384 |

| FP32 Compute | 16.35 TFLOPS | 29.71 TFLOPS |

| Texture Rate | 511.0 GTexel/s | 928.5 GTexel/s |

| Pixel Rate | 204.4 GPixel/s | 116.1 GPixel/s |

| TDP | 250 W | 140 W |

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

| Suggested PSU | 600 W | 300 W |

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

| Display Outputs | No outputs | 1x HDMI |

| API Support | DirectX 12, OpenGL 4.6, Vulkan 1.4 | N/A |

| Release Date | 2019-11-25 | 2025-10-14 |

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

Where Each One Wins

NVIDIA Tesla V100S PCIe 32 GB: This card wins in scenarios where memory bandwidth is the critical bottleneck. Its 1.13 TB/s of bandwidth is a 4.1x advantage over the GB10. The benchmark data shows it is 61.8% faster in Geekbench OpenCL, suggesting it is superior for compute tasks that are memory-bound, such as certain scientific simulations, large matrix operations, and specific AI inference workloads that require rapid data movement. Its higher pixel rate (204.4 GPixel/s) and ROP count (128) also make it more capable for traditional rasterization if used in a compute context. It also supports standard graphics APIs, which the GB10 lacks.

NVIDIA GB10: The GB10 wins in scenarios prioritizing memory capacity, power efficiency, and raw shader throughput. With 128 GB of memory, it can hold datasets that would be impossible to fit on the V100S's 32 GB. Its FP32 compute rating of 29.71 TFLOPS is 81.7% higher, and its texture rate of 928.5 GTexel/s is 81.7% higher, indicating a significant advantage in tasks that rely on parallel floating-point math and texture fetch operations. Its half the power draw (140 W vs. 250 W) and no external power connectors make it far easier to integrate into dense systems. The presence of RT cores, even without standard API support, hints at a future potential for ray-traced compute tasks. The GB10 is the choice for handling massive, modern AI models or high-throughput FP32 compute where its higher clock speeds and newer architecture can shine.

DETAILED SPECIFICATIONS

SPECIFICATION
GB10
Tesla V100S PCIe 32 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
1597 MHz
Memory Clock
1067 MHz 8.5 Gbps effective
1107 MHz 2.2 Gbps effective
Memory
Memory Size
128 GB
32 GB
VRAM (MB)
131,072
32,768 -75.0%
Memory Type
LPDDR5X
HBM2
Memory Bus
256 bit
4096 bit
Bandwidth
273.2 GB/s
1.13 TB/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
204.4 GPixel/s
Texture Rate
928.5 GTexel/s
511.0 GTexel/s
FP32 (TFLOPS)
29.71 TFLOPS
16.35 TFLOPS
FP64 (TFLOPS)
464.3 GFLOPS (1:64)
8.177 TFLOPS (1:2)
FP16 (TFLOPS)
29.71 TFLOPS (1:1)
32.71 TFLOPS (2:1)
AI/RT
RT Cores
48
—
Tensor Cores
384
640 +66.7%
Power
TDP
140 W
250 W
TDP (W)
140
250 +78.6%
Suggested PSU
300 W
600 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 V100S PCIe 32 GB Details