NVIDIA GeForce GTX 670M vs NVIDIA RTX PRO 5000 72 GB Blackwell Comparison

NVIDIA
GEFORCE

NVIDIA GeForce GTX 670M

CORE STATE GF114
VRAM 1536 MB
CLOCK SPEED —
TDP 75 W
BUS WIDTH 192 bit
ARCHITECTURE Fermi 2.0
nm
PROCESS 40 nm
LAUNCH DATE 2012
VS
NVIDIA
GEFORCE

RTX PRO 5000 72 GB Blackwell

CORE STATE GB202
VRAM 72 GB
CLOCK SPEED 2377 MHz
TDP 300 W
BUS WIDTH 384 bit
ARCHITECTURE Blackwell 2.0
nm
PROCESS 5 nm
LAUNCH DATE 2025

PERFORMANCE BENCHMARKS

geekbench_opencl
6,513
N/A
3dmark_3dmark_steel_nomad_dx12
N/A
9,579.5
passmark_directx_10
N/A
175
passmark_directx_11
N/A
219
passmark_directx_12
N/A
78
passmark_directx_9
N/A
311
passmark_g2d
N/A
914
passmark_g3d
N/A
24,310
passmark_gpu_compute
N/A
15,667

Analysis: NVIDIA GeForce GTX 670M vs NVIDIA RTX PRO 5000 72 GB Blackwell

NVIDIA’s GeForce GTX 670M and RTX PRO 5000 72 GB Blackwell sit at opposite ends of the GPU timeline, separated by over a decade of architectural evolution. The GTX 670M is an end-of-life mobile part from the Fermi 2.0 era, while the RTX PRO 5000 is an active workstation-class card built on Blackwell 2.0. The data available for direct comparison is sparse—the GTX 670M has a single OpenCL benchmark score, while the RTX PRO 5000 has a broader set of synthetic tests—but the specification sheets and percentile rankings reveal a clear divide in capability and intent.

Where Each One Wins

The GTX 670M’s single benchmark result is a Geekbench OpenCL score of 6,513, placing it at the 38th percentile of all GPUs. Its nearest rivals are the GT 555M (6,493, +0.3%), Quadro M5000M (6,481, +0.5%), and Vega 10 Mobile (6,476, +0.6%), with the Intel UHD Graphics P750 trailing slightly at 6,554 (-0.6%). This places the GTX 670M in a narrow band of mid-2010s mobile and entry-level desktop parts, where a 0.3% to 0.6% margin separates adjacent products. The data shows it wins by the thinnest of margins against its closest competitors, but it loses to the Intel integrated solution by 0.6%. Practically, this GPU wins in scenarios where legacy OpenCL workloads are the benchmark, and it edges out other mobile GPUs of its generation—but the wins are marginal, not transformative.

The RTX PRO 5000 presents a more complex picture. Its average benchmark score is 6,407, which is actually lower than the GTX 670M’s 6,513, yet its percentile ranking is 37—just one point below the GTX 670M’s 38. This counterintuitive result stems from the benchmark mix: the RTX PRO 5000’s PassMark G3D score is 24,310 and its GPU compute score is 15,667, but its DirectX 12 PassMark score is a lowly 78, and its DirectX 10 score is 175. The 3DMark Steel Nomad DX12 score of 9,579.5 is respectable, but the PassMark legacy DX9 and DX11 scores (311 and 219, respectively) drag the average down. The RTX PRO 5000 wins decisively in raw compute and modern graphics tests—its G3D score is over 3.7x the GTX 670M’s entire average—but it loses in any comparison that weights legacy DirectX performance. Its nearest rivals include the GTX 460 SE (6,389, +0.3%) and GTX 580M (6,389, +0.3%), with the Vega 10 Mobile and Quadro M5000M at -1.1% each. This means the RTX PRO 5000 wins narrowly against old mid-range cards, but its average is dragged down by inconsistent legacy scores.

In terms of use-case split, the GTX 670M wins in OpenCL-centric or legacy API scenarios where its Fermi architecture’s simpler pipeline avoids the RTX PRO 5000’s low DirectX 9/10/12 PassMark scores. The RTX PRO 5000 wins in any modern workload—compute, G3D rasterization, or Steel Nomad DX12—by orders of magnitude.

Architecture Differences

The two GPUs share a manufacturer but little else. The GTX 670M is built on the GF114 chip using Fermi 2.0 architecture, fabricated on TSMC’s 40 nm process. It packs 1,950 million transistors into a 332 mm² die, yielding a transistor density of 5.9M per mm². Its memory configuration is 1,536 MB of GDDR5 on a 192-bit bus, with 750 MHz memory clocks translating to 3 Gbps effective and 72.00 GB/s bandwidth. The GPU has 336 shading units, 56 TMUs, and 24 ROPs, with no ray tracing or tensor cores—those features did not exist in Fermi. Its pixel rate is 8.372 GPixel/s, texture rate is 33.49 GTexel/s, and FP32 performance is 803.7 GFLOPS. It supports DirectX 12 (11_0), OpenGL 4.6, but no Vulkan, and connects via an MXM-B (3.0) interface with no power connectors, drawing 75 W.

The RTX PRO 5000 is a radically different machine. It uses the GB202 chip on Blackwell 2.0 architecture, fabricated on TSMC’s 5 nm process. Transistors number 92,200 million—47x more than the GTX 670M—on a 750 mm² die, with a density of 122.9M per mm². Memory is 72 GB of GDDR7 on a 384-bit bus, with 1,750 MHz memory clocks (28 Gbps effective) yielding 1.34 TB/s bandwidth—over 18x the GTX 670M’s. Its 14,080 shading units, 440 TMUs, and 160 ROPs are complemented by 110 RT cores and 440 tensor cores. Pixel rate is 380.3 GPixel/s, texture rate is 1,045.9 GTexel/s, and FP32 is 66.94 TFLOPS, with FP16 at 66.94 TFLOPS (1:1). Base clock is 1,740 MHz with a 2,377 MHz boost. It supports DirectX 12 Ultimate (12_2), OpenGL 4.6, and Vulkan 1.4, and uses a PCIe 5.0 x16 interface with a 1x 16-pin power connector, drawing 300 W and requiring a 700 W PSU. It is a dual-slot card measuring 267 mm by 111 mm by 40 mm, with 4x DisplayPort 2.1b outputs.

Node, transistor count, memory type and size, core counts, and feature sets are all generations apart. The RTX PRO 5000’s RT and tensor cores are absent on the GTX 670M, and the FP16 capability is nonexistent on the older part.

The Verdict

Based strictly on the data, the RTX PRO 5000 is the superior GPU for any modern or compute-heavy task. Its G3D score of 24,310 is nearly four times the GTX 670M’s entire average benchmark score, and its GPU compute score of 15,667 is over twice the GTX 670M’s OpenCL result. The 3DMark Steel Nomad DX12 score of 9,579.5 further confirms its dominance in current-generation graphics. The GTX 670M’s only claim is its higher average score (6,513 vs. 6,407) and its +0.3% lead over the RTX PRO 5000’s closest rivals—but those averages are computed from incomparable benchmark sets. The GTX 670M wins only if the metric is the single Geekbench OpenCL test or if legacy DirectX performance is the sole criterion, which the RTX PRO 5000’s PassMark DX9 score of 311 and DX10 score of 175 do not support as a strength.

Who should pick which? A user constrained to a 2012-era mobile platform with an MXM slot and no external power would pick the GTX 670M, as it is the only physically compatible option. Anyone else—anyone with a PCIe 5.0 slot, a 700 W PSU, and a need for 72 GB of GDDR7, RT cores, or 66.94 TFLOPS—should pick the RTX PRO 5000. The data does not support any scenario where the GTX 670M is preferable on performance alone outside of its narrow legacy benchmark advantage.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The GTX 670M has an average benchmark score of 6,513, which is higher than the RTX PRO 5000’s average of 6,407.

Q: How do their DirectX capabilities differ?

A: The GTX 670M supports DirectX 12 (11_0), while the RTX PRO 5000 supports DirectX 12 Ultimate (12_2). The RTX PRO 5000 also supports Vulkan 1.4, whereas the GTX 670M has no Vulkan support listed.

Q: What is the memory bandwidth difference?

A: The RTX PRO 5000 has 1.34 TB/s of bandwidth from 72 GB of GDDR7 on a 384-bit bus, while the GTX 670M has 72.00 GB/s from 1,536 MB of GDDR5 on a 192-bit bus.

Q: Does the RTX PRO 5000 have ray tracing capabilities?

A: Yes, the RTX PRO 5000 has 110 RT cores. The GTX 670M has no RT cores listed.

Q: What are the transistor counts?

A: The GTX 670M has 1,950 million transistors, while the RTX PRO 5000 has 92,200 million transistors.

Q: Which GPU has a higher FP32 performance?

A: The RTX PRO 5000 has 66.94 TFLOPS of FP32 performance, compared to the GTX 670M’s 803.7 GFLOPS.

Head-to-Head Benchmarks

There are no direct head-to-head benchmark entries in the data, so the comparison relies on individual scores. The GTX 670M’s only benchmark is Geekbench OpenCL at 6,513. The RTX PRO 5000’s strongest result is PassMark G3D at 24,310, which is 3.73x the GTX 670M’s entire average. Its GPU compute score of 15,667 is 2.41x the GTX 670M’s OpenCL score. The RTX PRO 5000’s 3DMark Steel Nomad DX12 score of 9,579.5 is not directly comparable to any GTX 670M test, but it indicates strong modern workload performance. Conversely, the RTX PRO 5000’s PassMark DirectX 12 score of 78 is catastrophically low, and its DirectX 9 score of 311 is also weak—these are the tests where the GTX 670M would likely win if it had scores for them, but it does not. The RTX PRO 5000’s PassMark G2D score of 914 further shows its 2D performance is respectable but not the headline number.

The biggest win for the RTX PRO 5000 is in G3D, where its 24,310 dwarfs the GTX 670M’s 6,513 by a factor of 3.73. The biggest win for the GTX 670M is its average score, which beats the RTX PRO 5000’s average by 1.65% (6,513 vs. 6,407). However, this is misleading because the averages are derived from different test suites. The data shows the RTX PRO 5000’s compute and G3D results are in a different class, while its legacy DirectX scores are anomalously low, possibly reflecting driver or test methodology issues rather than real-world capability. The GTX 670M’s single score is consistent with its nearest rivals, all within 0.6% of each other, indicating a stable, predictable performance envelope.

Specification Differences

The two GPUs differ in nearly every measurable specification. Process node: 40 nm (GTX 670M) vs. 5 nm (RTX PRO 5000). Transistors: 1,950 million vs. 92,200 million. Die size: 332 mm² vs. 750 mm². Transistor density: 5.9M / mm² vs. 122.9M / mm². Memory size: 1,536 MB vs. 72 GB. Memory type: GDDR5 vs. GDDR7. Memory bus: 192 bit vs. 384 bit. Bandwidth: 72.00 GB/s vs. 1.34 TB/s. Shading units: 336 vs. 14,080. TMUs: 56 vs. 440. ROPs: 24 vs. 160. RT cores: none vs. 110. Tensor cores: none vs. 440. Pixel rate: 8.372 GPixel/s vs. 380.3 GPixel/s. Texture rate: 33.49 GTexel/s vs. 1,045.9 GTexel/s. FP32: 803.7 GFLOPS vs. 66.94 TFLOPS. Memory clock: 750 MHz (3 Gbps effective) vs. 1,750 MHz (28 Gbps effective). Base clock: not listed vs. 1,740 MHz. Boost clock: not listed vs. 2,377 MHz. TDP: 75 W vs. 300 W. Slot width: MXM Module vs. Dual-slot. Power connectors: None vs. 1x 16-pin. Bus interface: MXM-B (3.0) vs. PCIe 5.0 x16. Display outputs: Portable Device Dependent vs. 4x DisplayPort 2.1b. DirectX: 12 (11_0) vs. 12 Ultimate (12_2). Vulkan: not listed vs. 1.4. Release date: 2012-03-21 vs. 2025-10-20. Production status: End-of-life vs. Active. Predecessor: GeForce 500M vs. Workstation Ada. Successor: GeForce 700M vs. none. The RTX PRO 5000 also has dimensions (267 mm x 111 mm x 40 mm) and a suggested PSU of 700 W, while the GTX 670M has neither. The only shared specifications are manufacturer (NVIDIA), OpenGL version (4.6), and the absence of a launch MSRP in the data.

DETAILED SPECIFICATIONS

SPECIFICATION
GTX 670M
RTX PRO 5000 72 GB Blackwell
Core Specs
Shading Units
336
14,080 +4090.5%
Shaders
336
14,080 +4090.5%
TMUs
56
440 +685.7%
ROPs
24
160 +566.7%
SM Count
7
110 +1471.4%
Clocks
Base Clock
—
1740 MHz
Boost Clock
—
2377 MHz
GPU Clock
598 MHz
—
Shader Clock
1196 MHz
—
Memory Clock
750 MHz 3 Gbps effective
1750 MHz 28 Gbps effective
Memory
Memory Size
1536 MB
72 GB
VRAM (MB)
1,536
73,728 +4700.0%
Memory Type
GDDR5
GDDR7
Memory Bus
192 bit
384 bit
Bandwidth
72.00 GB/s
1.34 TB/s
Cache
L1 Cache
64 KB (per SM)
128 KB (per SM)
L2 Cache
384 KB
96 MB
Performance
Pixel Rate
8.372 GPixel/s
380.3 GPixel/s
Texture Rate
33.49 GTexel/s
1,045.9 GTexel/s
FP32 (TFLOPS)
803.7 GFLOPS
66.94 TFLOPS
FP64 (TFLOPS)
66.98 GFLOPS (1:12)
1,045.9 GFLOPS (1:64)
FP16 (TFLOPS)
—
66.94 TFLOPS (1:1)
AI/RT
RT Cores
—
110
Tensor Cores
—
440
Power
TDP
75 W
300 W
TDP (W)
75
300 +300.0%
Suggested PSU
—
700 W
Power Connectors
None
1x 16-pin
Architecture
Architecture
Fermi 2.0
Blackwell 2.0
GPU Name
GF114
GB202
Generation
GeForce 600M
Blackwell PRO W (x000)
Process Size
40 nm
5 nm
Transistors
1,950 million
92,200 million
Die Size
332 mm²
750 mm²
Foundry
TSMC
TSMC
Density
5.9M / mm²
122.9M / mm²
API Support
DirectX
12 (11_0)
12 Ultimate (12_2)
OpenGL
4.6
4.6
Vulkan
—
1.4
OpenCL
1.1
3.0
CUDA
2.1
12.0
Shader Model
5.1
6.9
Physical
Slot Width
MXM Module
Dual-slot
Length
—
267 mm 10.5 inches
Height
—
111 mm 4.4 inches
Outputs
Portable Device Dependent
4x DisplayPort 2.1b
Bus Interface
MXM-B (3.0)
PCIe 5.0 x16
Other
Production
End-of-life
Active
Predecessor
GeForce 500M
Workstation Ada
Successor
GeForce 700M
—
View GeForce GTX 670M Details View RTX PRO 5000 72 GB Blackwell Details