NVIDIA B300 SXM6 AC vs NVIDIA GeForce RTX 3090 Ti Comparison

NVIDIA
GEFORCE

NVIDIA B300 SXM6 AC

CORE STATE GB110
VRAM 288 GB
CLOCK SPEED 2032 MHz
TDP 1100 W
BUS WIDTH 8192 bit
ARCHITECTURE Blackwell Ultra
nm
PROCESS 5 nm
LAUNCH DATE 2025
VS
NVIDIA
GEFORCE

GeForce RTX 3090 Ti

CORE STATE GA102
VRAM 24 GB
CLOCK SPEED 1860 MHz
TDP 450 W
BUS WIDTH 384 bit
ARCHITECTURE Ampere
nm
PROCESS 8 nm
LAUNCH DATE 2022

PERFORMANCE BENCHMARKS

geekbench_opencl
369,831
174,441
3dmark_3dmark_steel_nomad_dx12
N/A
5,741
geekbench_vulkan
N/A
215,633

Analysis: NVIDIA B300 SXM6 AC vs NVIDIA GeForce RTX 3090 Ti

The NVIDIA B300 SXM6 AC and the NVIDIA GeForce RTX 3090 Ti occupy opposite ends of the GPU spectrum, yet both are recorded in the database with distinct measurement sets. The B300 is a server accelerator built for massive parallel workloads, while the RTX 3090 Ti is a consumer-grade graphics card designed for rendering and gaming. The recorded data shows a single shared benchmark, Geekbench OpenCL, where the B300 SXM6 AC delivers a score of 369,831 against the RTX 3090 Ti’s 174,441. That is a delta of 112%, meaning the B300 SXM6 AC more than doubles the raw compute output of the RTX 3090 Ti in this test. The B300 SXM6 AC also holds the 100th percentile among all GPUs in the database, while the RTX 3090 Ti sits at the 95th percentile, a narrower gap in relative ranking than the raw score suggests because the consumer card still outpaces most other hardware.

Head-to-Head Benchmarks

The only direct comparison available is the Geekbench OpenCL test, which measures general-purpose compute performance across a wide range of tasks such as image processing, physics simulations, and data manipulation. The B300 SXM6 AC scores 369,831, while the RTX 3090 Ti scores 174,441. The delta is 112%, so the B300 SXM6 AC is 112% faster, or 2.12 times the performance of the RTX 3090 Ti. This is a decisive win for the B300 SXM6 AC, and the margin is substantial enough to indicate a different class of hardware rather than a minor generational improvement.

The RTX 3090 Ti has no recorded wins in any head-to-head benchmark against the B300 SXM6 AC. Its other recorded benchmarks, such as 3DMark Steel Nomad DX12 with a score of 5,741 and Geekbench Vulkan with 215,633, do not appear in the shared comparison set. The database shows one win for the B300 SXM6 AC and zero for the RTX 3090 Ti in direct comparisons. The 112% delta in OpenCL is the only numeric evidence of performance difference, but it is consistent with the architectural divide between the two parts.

To put the B300 SXM6 AC’s score in context, the database lists its nearest rivals. The NVIDIA B200, with an average score of 345,482, is 7% slower. The NVIDIA H200 NVL, scoring 334,891, is 10.4% behind. The AMD Instinct MI300X, at 317,994, trails by 16.3%, and the NVIDIA L40S, at 295,763, is 25% behind. These deltas show the B300 SXM6 AC leads its immediate competitor class by a narrow to moderate margin, but the gap to the RTX 3090 Ti is far larger, reflecting that the RTX 3090 Ti belongs to a different tier of performance.

Where Each One Wins

The B300 SXM6 AC wins decisively in raw compute throughput, as evidenced by the OpenCL result. This makes it suitable for workloads that saturate the compute units, such as large-scale parallel calculations, scientific simulations, or data-center tasks that involve processing massive arrays. The 112% lead over the RTX 3090 Ti means that any compute-bound application will complete in less than half the time on the B300 SXM6 AC, assuming the software scales properly. The B300 SXM6 AC also has a 100th percentile ranking, which means no other GPU in the database scores higher on average, so it is the top-tier choice for raw performance.

The RTX 3090 Ti, while slower in OpenCL, is not without strengths in other contexts. Its recorded benchmarks show a Vulkan score of 215,633 and a 3DMark Steel Nomad DX12 score of 5,741, which are not compared against the B300 SXM6 AC because the B300 SXM6 AC has no recorded results for those tests. The RTX 3090 Ti also has display outputs, with 1x HDMI 2.1 and 3x DisplayPort 1.4a, while the B300 SXM6 AC has no outputs. This means the RTX 3090 Ti is the only one of the two that can drive a monitor directly, so it is the practical choice for interactive graphics, desktop rendering, or any task that requires visual output. The B300 SXM6 AC is a compute module with no display capability, so it cannot function as a standalone graphics card for a desktop.

The RTX 3090 Ti also supports graphics APIs that the B300 SXM6 AC does not. The database lists DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4 for the RTX 3090 Ti, while the B300 SXM6 AC has N/A for all three. This means the RTX 3090 Ti can run consumer games and applications that rely on those APIs, whereas the B300 SXM6 AC is not designed for such environments. In the database, the RTX 3090 Ti’s nearest rivals are all within a small band: the NVIDIA L4 at 131,072 is 0.7% slower, the RTX 4000 Ada Generation at 135,218 is 2.4% faster, the A10M at 135,230 is 2.4% faster, and the Radeon PRO W6800 at 135,396 is 2.6% faster. So the RTX 3090 Ti sits right in the middle, with close competition, but the B300 SXM6 AC has no such tight grouping; it stands alone at the top.

Architecture Differences

The B300 SXM6 AC uses the GB110 chip, built on the Blackwell Ultra architecture, manufactured on a 5 nm process at TSMC. The RTX 3090 Ti uses the GA102 chip, based on the Ampere architecture, manufactured on an 8 nm process at Samsung. The process node difference is significant, as the B300 SXM6 AC packs 208,000 million transistors onto a 1,628 mm² die, giving a transistor density of 127.8 million per mm². The RTX 3090 Ti has 28,300 million transistors on a 628 mm² die, for a density of 45.1 million per mm². The B300 SXM6 AC therefore has over seven times as many transistors and a much higher density, which explains its massive compute advantage.

Memory architecture also differs completely. The B300 SXM6 AC features 288 GB of HBM3e memory on an 8192-bit bus, providing 8.19 TB/s of bandwidth. The RTX 3090 Ti has 24 GB of GDDR6X memory on a 384-bit bus, with 1.01 TB/s of bandwidth. The B300 SXM6 AC has 12 times the memory capacity and 8.1 times the bandwidth. This difference is critical for workloads that require large datasets, such as training large models or processing high-resolution volumetric data, where the RTX 3090 Ti would run out of memory capacity or stall on bandwidth.

The compute units also differ. The B300 SXM6 AC has 18,944 shading units, 592 texture mapping units, and 24 render output units. The RTX 3090 Ti has 10,752 shading units, 336 TMUs, and 112 ROPs. The B300 SXM6 AC has more shading units and TMUs, but fewer ROPs. The pixel rate for the B300 SXM6 AC is 48.77 GPixel/s, while the RTX 3090 Ti is 208.3 GPixel/s. That is a notable difference: the RTX 3090 Ti has a much higher pixel throughput, which is important for rasterization and traditional graphics rendering. The texture rate is the opposite, with the B300 SXM6 AC at 1,202.9 GTexel/s versus the RTX 3090 Ti’s 625.0 GTexel/s. So the B300 SXM6 AC is better at texture-heavy compute, while the RTX 3090 Ti is better at pixel output.

The B300 SXM6 AC has 592 tensor cores, while the RTX 3090 Ti has 336 tensor cores. The RTX 3090 Ti also has 84 ray tracing cores, which the B300 SXM6 AC does not list. This suggests the B300 SXM6 AC is not designed for ray-traced graphics, while the RTX 3090 Ti is. The FP32 performance is 76.99 TFLOPS for the B300 SXM6 AC versus 40.00 TFLOPS for the RTX 3090 Ti, and FP16 is the same as FP32 for both, 1:1. The B300 SXM6 AC nearly doubles the FP32 and FP16 throughput.

Power consumption and physical form also differ. The B300 SXM6 AC has a TDP of 1100 W and is an SXM module, while the RTX 3090 Ti has a TDP of 450 W and is a triple-slot card. The B300 SXM6 AC suggests a 1500 W power supply, while the RTX 3090 Ti suggests 850 W. The B300 SXM6 AC uses a PCIe 6.0 x16 interface, while the RTX 3090 Ti uses PCIe 4.0 x16. Process node and foundry differences lead to this power gap, but the B300 SXM6 AC is designed for data center racks with dedicated power, not consumer systems.

The Verdict

From the recorded data, the B300 SXM6 AC is the superior choice for pure compute performance. Its OpenCL score is 112% higher than the RTX 3090 Ti, and it holds the 100th percentile, meaning no other GPU in the database scores higher on average. Any workload that is limited by FP32 or FP16 throughput, memory bandwidth, or memory capacity will benefit from the B300 SXM6 AC’s 8.19 TB/s bandwidth and 288 GB of HBM3e, versus the RTX 3090 Ti’s 1.01 TB/s and 24 GB. The B300 SXM6 AC also has more shading units, more tensor cores, and a higher texture rate, so it is preferable for training large models, scientific computing, or any parallel processing that fits in the GPU’s memory.

The RTX 3090 Ti is the better choice for tasks that require graphics output, because it has display outputs and supports DirectX 12 Ultimate, OpenGL 4.6, and Vulkan 1.4. The B300 SXM6 AC has no display outputs and no API support, so it cannot render to a screen or run consumer graphics applications. The RTX 3090 Ti also has a higher pixel rate at 208.3 GPixel/s versus 48.77 GPixel/s, making it better suited for rasterization-heavy workloads. Additionally, the RTX 3090 Ti’s lower power demand (450 W versus 1100 W) makes it feasible for a desktop system, whereas the B300 SXM6 AC requires a server platform. The RTX 3090 Ti is also smaller in physical size and uses a standard PCIe slot, while the B300 SXM6 AC is a different form factor.

No single GPU wins across all metrics. The B300 SXM6 AC wins in raw compute, memory, and bandwidth. The RTX 3090 Ti wins in pixel rate, display outputs, and graphics API support. The user should select based on the workload: for data-center compute, the B300 SXM6 AC is the clear leader, with no rival in its class. For desktop graphics, rendering, or any task with a monitor attached, the RTX 3090 Ti is the only option because the B300 SXM6 AC cannot produce a video signal. The RTX 3090 Ti’s nearest rivals are all within 2.6% of its average score, so it is competitive in its segment, but it cannot match the B300 SXM6 AC’s performance ceiling.

FAQ

Q: What is the performance difference in the shared benchmark between the two GPUs?

A: In Geekbench OpenCL, the NVIDIA B300 SXM6 AC scores 369,831 while the NVIDIA GeForce RTX 3090 Ti scores 174,441. The B300 SXM6 AC is 112% faster.

Q: Which GPU has more memory and bandwidth?

A: The B300 SXM6 AC has 288 GB of HBM3e memory with 8.19 TB/s bandwidth on an 8192-bit bus. The RTX 3090 Ti has 24 GB of GDDR6X with 1.01 TB/s bandwidth on a 384-bit bus.

Q: Can the B300 SXM6 AC output to a display?

A: No. The B300 SXM6 AC has no display outputs, while the RTX 3090 Ti has 1x HDMI 2.1 and 3x DisplayPort 1.4a.

Q: Which GPU supports DirectX and Vulkan?

A: The RTX 3090 Ti supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4. The B300 SXM6 AC has N/A for all of those APIs.

Q: What is the transistor count difference?

A: The B300 SXM6 AC has 208,000 million transistors on a 5 nm process, while the RTX 3090 Ti has 28,300 million on an 8 nm process.

Q: How does the B300 SXM6 AC compare to its closest rivals?

A: The B300 SXM6 AC is 7% faster than the NVIDIA B200, 10.4% faster than the NVIDIA H200 NVL, 16.3% faster than the AMD Instinct MI300X, and 25% faster than the NVIDIA L40S.

Specification Differences

| Specification | NVIDIA B300 SXM6 AC | NVIDIA GeForce RTX 3090 Ti |

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

| Architecture | Blackwell Ultra | Ampere |

| Process Node | 5 nm | 8 nm |

| Foundry | TSMC | Samsung |

| Transistors | 208,000 million | 28,300 million |

| Die Size | 1628 mm² | 628 mm² |

| Base Clock | 1665 MHz | 1560 MHz |

| Boost Clock | 2032 MHz | 1860 MHz |

| Memory Size | 288 GB | 24 GB |

| Memory Type | HBM3e | GDDR6X |

| Memory Bus Width | 8192 bit | 384 bit |

| Memory Bandwidth | 8.19 TB/s | 1.01 TB/s |

| Shading Units | 18944 | 10752 |

| TMUs | 592 | 336 |

| ROPs | 24 | 112 |

| Tensor Cores | 592 | 336 |

| RT Cores | None listed | 84 |

| FP32 Performance | 76.99 TFLOPS | 40.00 TFLOPS |

| FP16 Performance | 76.99 TFLOPS (1:1) | 40.00 TFLOPS (1:1) |

| Pixel Rate | 48.77 GPixel/s | 208.3 GPixel/s |

| Texture Rate | 1,202.9 GTexel/s | 625.0 GTexel/s |

| TDP | 1100 W | 450 W |

| Suggested PSU | 1500 W | 850 W |

| Bus Interface | PCIe 6.0 x16 | PCIe 4.0 x16 |

| Display Outputs | No outputs | 1x HDMI 2.1, 3x DisplayPort 1.4a |

| DirectX Support | N/A | 12 Ultimate (12_2) |

| OpenGL Support | N/A | 4.6 |

| Vulkan Support | N/A | 1.4 |

| Slot Width | SXM Module | Triple-slot |

| Power Connectors | None listed | 1x 16-pin |

| Release Date | 2025-09-10 | 2022-01-26 |

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

DETAILED SPECIFICATIONS

SPECIFICATION
B300 SXM6 AC
RTX 3090 Ti
Core Specs
Shading Units
18,944
10,752 -43.2%
Shaders
18,944
10,752 -43.2%
TMUs
592
336 -43.2%
ROPs
24
112 +366.7%
SM Count
148
84 -43.2%
Clocks
Base Clock
1665 MHz
1560 MHz
Boost Clock
2032 MHz
1860 MHz
Memory Clock
2000 MHz 8 Gbps effective
1313 MHz 21 Gbps effective
Memory
Memory Size
288 GB
24 GB
VRAM (MB)
294,912
24,576 -91.7%
Memory Type
HBM3e
GDDR6X
Memory Bus
8192 bit
384 bit
Bandwidth
8.19 TB/s
1.01 TB/s
Cache
L1 Cache
256 KB (per SM)
128 KB (per SM)
L2 Cache
126 MB
6 MB
Performance
Pixel Rate
48.77 GPixel/s
208.3 GPixel/s
Texture Rate
1,202.9 GTexel/s
625.0 GTexel/s
FP32 (TFLOPS)
76.99 TFLOPS
40.00 TFLOPS
FP64 (TFLOPS)
1,202.9 GFLOPS (1:64)
625.0 GFLOPS (1:64)
FP16 (TFLOPS)
76.99 TFLOPS (1:1)
40.00 TFLOPS (1:1)
AI/RT
RT Cores
—
84
Tensor Cores
592
336 -43.2%
Power
TDP
1100 W
450 W
TDP (W)
1,100
450 -59.1%
Suggested PSU
1500 W
850 W
Power Connectors
—
1x 16-pin
Architecture
Architecture
Blackwell Ultra
Ampere
GPU Name
GB110
GA102
Generation
Server Blackwell (Bxx)
GeForce 30
Process Size
5 nm
8 nm
Transistors
208,000 million
28,300 million
Die Size
1628 mm²
628 mm²
Foundry
TSMC
Samsung
Density
127.8M / mm²
45.1M / mm²
API Support
DirectX
—
12 Ultimate (12_2)
OpenGL
—
4.6
Vulkan
—
1.4
OpenCL
3.0
3.0
CUDA
10.3
8.6
Shader Model
—
6.8
Physical
Slot Width
SXM Module
Triple-slot
Length
—
336 mm 13.2 inches
Height
—
140 mm 5.5 inches
Outputs
No outputs
1x HDMI 2.13x DisplayPort 1.4a
Bus Interface
PCIe 6.0 x16
PCIe 4.0 x16
Other
Launch Price
—
1,999 USD
Production
Active
End-of-life
Predecessor
Server Hopper
GeForce 20
Successor
Server Rubin
GeForce 40
View B300 SXM6 AC Details View GeForce RTX 3090 Ti Details