NVIDIA GeForce GTX 670MX vs NVIDIA RTX PRO 6000 Blackwell Server Comparison

The NVIDIA RTX PRO 6000 Blackwell Server and the NVIDIA GeForce GTX 670MX represent two extremes of the GPU timeline, separated by over a decade of architectural evolution. The RTX PRO 6000 targets modern server workloads with a 5 nm process and Blackwell 2.0 architecture, while the GTX 670MX is an end-of-life mobile part built on 28 nm Kepler. Benchmark data shows a fascinating convergence in raw scores despite their vastly different specifications, but deeper analysis reveals the RTX PRO 6000's dominance is contextual rather than absolute.

Head-to-Head Benchmarks

The most striking finding is the near-parity in their primary benchmark scores. The RTX PRO 6000 Blackwell Server achieves a score of 5996 in the 3DMark Steel Nomad DX12 test, while the GTX 670MX posts an average benchmark score of 5721 across Geekbench OpenCL and Vulkan tests. This 4.8% gap in favor of the RTX PRO 6000 is surprisingly narrow given the generational chasm between them, but the comparison is complicated by the fact that they run different benchmark suites. The RTX PRO 6000's score places it at the 34th percentile of all GPUs, while the GTX 670MX sits just one percentile lower at 33rd, indicating that despite its age, the mobile Kepler part still holds its own in the broader performance distribution.

Looking at the rival comparisons, the RTX PRO 6000's nearest competitors in the database include the NVIDIA GeForce GTX 770M, which posts an average score of 6000, a mere 0.1% delta from the RTX PRO 6000. The AMD Radeon RX 6400 also lands at 6001, again just 0.1% off. On the other side, the GTX 670MX's closest rival is the NVIDIA GeForce GTX 550 Ti with a score of 5731, which is 0.2% higher, and the Intel Iris Pro Graphics P6300 at 5712, which trails by 0.2%. These deltas are so small that they fall within typical run-to-run variance, meaning that in terms of raw benchmark output, these two cards are effectively peers in their respective test environments.

However, the RTX PRO 6000's single benchmark result masks its true capability. The data shows a 126.0 TFLOPS FP32 throughput and a 502.5 GPixel/s pixel rate, figures that dwarf the GTX 670MX's 1,153.9 GFLOPS and 12.02 GPixel/s. The benchmark scores simply do not reflect the massive computational disparity, likely due to the different workloads being tested. The RTX PRO 6000 is designed for server-side ray tracing and tensor operations, while the GTX 670MX's Geekbench scores reflect legacy mobile graphics tasks. The 3DMark Steel Nomad test on the RTX PRO 6000 is a DX12 workload that may not fully utilize the card's 188 RT cores and 752 tensor cores, explaining why its percentile ranking is unremarkable despite its enormous specifications.

FAQ

Q: How do the average benchmark scores compare between the two cards?

A: The RTX PRO 6000 Blackwell Server scores 5996 in 3DMark Steel Nomad DX12, while the GTX 670MX averages 5721 across Geekbench OpenCL (6125) and Vulkan (5316) tests. This represents a 4.8% difference in favor of the RTX PRO 6000.

Q: What are the nearest rivals for each card based on benchmark data?

A: The RTX PRO 6000's closest rival is the NVIDIA GeForce GTX 770M at 6000 (0.1% difference), followed by the AMD Radeon RX 6400 at 6001. The GTX 670MX's nearest rival is the NVIDIA GeForce GTX 550 Ti at 5731 (0.2% higher), with the Intel Iris Pro Graphics P6300 at 5712 trailing by 0.2%.

Q: Which card has a higher percentile ranking among all GPUs?

A: The RTX PRO 6000 ranks at the 34th percentile, while the GTX 670MX ranks at the 33rd percentile. The one-percentile gap indicates nearly identical relative standing in the database.

Q: What is the transistor density difference between the two architectures?

A: The RTX PRO 6000's GB202 chip packs 92,200 million transistors on a 750 mm² die, yielding a density of 122.9M transistors per mm². The GTX 670MX's GK104 chip contains 3,540 million transistors on a 294 mm² die, with a density of just 12.0M per mm² — a 10.2x difference in density.

Q: How do memory specifications differ between the cards?

A: The RTX PRO 6000 features 96 GB of GDDR7 on a 512-bit bus with 1.79 TB/s bandwidth. The GTX 670MX has 3 GB of GDDR5 on a 192-bit bus with 67.20 GB/s bandwidth, representing a 26.6x difference in total memory bandwidth.

Q: What is the production status of each card?

A: The RTX PRO 6000 Blackwell Server is marked as Active production, while the GTX 670MX is designated End-of-life. Their release dates are 2025-03-17 and 2012-09-30, respectively.

Architecture Differences

The architectural gap between these two GPUs is monumental. The RTX PRO 6000 uses the GB202 chip built on TSMC's 5 nm process, featuring Blackwell 2.0 architecture from the Server Blackwell (Bxx) generation. Its 92,200 million transistors are packed into a 750 mm² die, achieving a density of 122.9M transistors per mm². The GTX 670MX employs the GK104 chip on a 28 nm TSMC process, using the Kepler architecture from the GeForce 600M generation. Its 3,540 million transistors occupy a 294 mm² die with a density of 12.0M per mm² — a 10.2x density advantage for the newer part.

Compute resources differ by orders of magnitude. The RTX PRO 6000 packs 24,064 shading units, 752 TMUs, and 192 ROPs, along with 188 RT cores and 752 tensor cores. The GTX 670MX has just 960 shading units, 80 TMUs, and 24 ROPs, with no RT cores or tensor cores at all. This translates to a 126.0 TFLOPS FP32 throughput for the RTX PRO 6000 versus 1,153.9 GFLOPS for the GTX 670MX — a 109x difference. Pixel and texture rates tell a similar story: 502.5 GPixel/s versus 12.02 GPixel/s, and 1,968.0 GTexel/s versus 48.08 GTexel/s, respectively.

Memory architecture is equally divergent. The RTX PRO 6000 uses 96 GB of GDDR7 on a 512-bit bus, delivering 1.79 TB/s bandwidth, while the GTX 670MX uses 3 GB of GDDR5 on a 192-bit bus for 67.20 GB/s. Clock speeds show the RTX PRO 6000 running at a 1590 MHz base and 2617 MHz boost, versus the GTX 670MX's fixed 601 MHz. The RTX PRO 6000 supports PCIe 5.0 x16, while the GTX 670MX uses PCIe 3.0 x16. API support also differs: the RTX PRO 6000 offers DirectX 12 Ultimate (12_2) and Vulkan 1.4, while the GTX 670MX provides DirectX 12 (11_0) and Vulkan 1.2.175.

The Verdict

The data presents a clear but nuanced verdict. For modern server workloads involving ray tracing, tensor operations, or massive memory bandwidth requirements, the RTX PRO 6000 Blackwell Server is the only viable choice. Its 188 RT cores, 752 tensor cores, 96 GB of GDDR7 memory, and 126.0 TFLOPS FP32 throughput are indispensable for such tasks. The GTX 670MX simply cannot execute these workloads, lacking the specialized hardware entirely. The 600 W TDP and dual-slot design of the RTX PRO 6000 reflect its server-class positioning, and its Active production status ensures continued availability.

However, the benchmark data tells a different story for legacy applications. The GTX 670MX's average score of 5721 is within 4.8% of the RTX PRO 6000's 5996, and its 33rd percentile ranking nearly matches the newer card's 34th percentile. For older DirectX 11-era games or lightweight mobile graphics tasks, the GTX 670MX's 75 W TDP and lack of external power connectors make it a far more practical solution, provided the user can accept its End-of-life status and 3 GB memory limit. The GTX 670MX's nearest rivals, such as the GTX 550 Ti, show that it remains competitive within its own performance tier.

Choosing between them depends entirely on workload. The RTX PRO 6000 is the obvious pick for any server deployment requiring modern features, given its PCIe 5.0 interface, DisplayPort 2.1b outputs, and support for DirectX 12 Ultimate. The GTX 670MX, with its portable-device-dependent outputs and PCIe 3.0 interface, is suited only for legacy mobile systems. The benchmark parity is a statistical curiosity rather than a functional equivalence, as the two cards target completely different application spaces. The RTX PRO 6000's 34th percentile ranking, despite its massive compute advantage, suggests that its specialized capabilities are not fully captured by the 3DMark Steel Nomad test — a reminder that raw scores rarely tell the whole story.

Specification Differences

The two cards differ in nearly every measurable specification. The RTX PRO 6000 uses a 5 nm process and GB202 chip, while the GTX 670MX uses a 28 nm process and GK104 chip. Transistor counts are 92,200 million versus 3,540 million, with die sizes of 750 mm² versus 294 mm². Clock speeds differ substantially: the RTX PRO 6000 runs at 1590 MHz base and 2617 MHz boost, while the GTX 670MX is locked at 601 MHz. Memory configurations are 96 GB GDDR7 on a 512-bit bus versus 3 GB GDDR5 on a 192-bit bus, yielding bandwidths of 1.79 TB/s and 67.20 GB/s respectively.

Compute units show the RTX PRO 6000 with 24,064 shading units, 752 TMUs, 192 ROPs, 188 RT cores, and 752 tensor cores, versus the GTX 670MX's 960 shading units, 80 TMUs, and 24 ROPs with no RT or tensor cores. The RTX PRO 6000 delivers 126.0 TFLOPS FP32, 502.5 GPixel/s pixel rate, and 1,968.0 GTexel/s texture rate, while the GTX 670MX achieves 1,153.9 GFLOPS FP32, 12.02 GPixel/s, and 48.08 GTexel/s. Power consumption is 600 W versus 75 W, with the RTX PRO 6000 requiring a 1x 16-pin connector and 1000 W suggested PSU, while the GTX 670MX uses no external connectors. The RTX PRO 6000 offers PCIe 5.0 x16 and 4x DisplayPort 2.1b outputs, versus PCIe 3.0 x16 and portable-device-dependent outputs for the GTX 670MX. API support includes DirectX 12 Ultimate and Vulkan 1.4 for the RTX PRO 6000, versus DirectX 12 (11_0) and Vulkan 1.2.175 for the GTX 670MX. The RTX PRO 6000 measures 267 mm by 111 mm by 40 mm, while the GTX 670MX has no listed dimensions.