NVIDIA GeForce GTX 765M vs NVIDIA RTX PRO 6000 Blackwell Server Comparison

FAQ

Q: How do the two GPUs compare in raw compute performance?

A: The RTX PRO 6000 Blackwell Server delivers 126.0 TFLOPS FP32, while the GeForce GTX 765M delivers 1,325.6 GFLOPS (approximately 1.33 TFLOPS). The server GPU is roughly 95 times faster in FP32 throughput.

Q: What architecture generations do these cards represent?

A: The RTX PRO 6000 uses Blackwell 2.0 architecture on a 5 nm TSMC process, while the GTX 765M uses Kepler architecture on a 28 nm TSMC process. They are separated by roughly 12 years of design evolution.

Q: How much memory do they have and what type?

A: The RTX PRO 6000 has 96 GB of GDDR7 memory on a 512-bit bus with 1.79 TB/s bandwidth. The GTX 765M has 2 GB of GDDR5 memory on a 128-bit bus with 64.13 GB/s bandwidth.

Q: What is the performance percentile ranking for each?

A: The RTX PRO 6000 sits at the 34th percentile among all GPUs in the database, while the GTX 765M sits at the 32nd percentile. Despite the enormous specification gap, both rank near each other in the database's percentile system.

Q: What is the average benchmark score for each?

A: The RTX PRO 6000 has an average benchmark score of 5996 (from a single 3DMark Steel Nomad DX12 result), while the GTX 765M has an average score of 5501 across Geekbench Metal, OpenCL, and Vulkan tests.

Q: What are the closest rivals for each card?

A: For the RTX PRO 6000, the nearest rivals are the GTX 770M (6000, -0.1%), RX 6400 (6001, -0.1%), FirePro W4100 (5987, 0.2%), and Quadro K4000M (5986, 0.2%). For the GTX 765M, they are the MX130 (5508, -0.1%), Radeon R7 M440 (5483, 0.3%), Quadro M4000 (5467, 0.6%), and FirePro M4000 (5537, -0.7%).

Architecture Differences

The architectural gulf between these two NVIDIA parts is vast. The RTX PRO 6000 Blackwell Server uses the GB202 chip built on Blackwell 2.0 architecture, fabricated at TSMC on a 5 nm process. It packs 92,200 million transistors into a 750 mm² die, yielding a transistor density of 122.9 million per square millimeter. The GTX 765M uses the GK106 chip on Kepler architecture, fabricated at the same foundry but on a 28 nm process. It contains 2,540 million transistors on a 221 mm² die, for a density of 11.5 million per square millimeter.

The server part is built for compute acceleration, featuring 188 RT cores and 752 tensor cores, hardware that the mobile Kepler chip completely lacks. The GTX 765M has no RT cores and no tensor cores, reflecting its era before ray tracing acceleration and AI tensor operations became standard. The RTX PRO 6000 also supports DirectX 12 Ultimate (12_2) and Vulkan 1.4, while the GTX 765M is limited to DirectX 12 (11_0) and Vulkan 1.2.175. Both support OpenGL 4.6.

The Blackwell GPU includes 24,064 shading units, 752 TMUs, and 192 ROPs. The Kepler chip includes 768 shading units, 64 TMUs, and 16 ROPs. That is a 31x difference in shader count, an 11.75x difference in TMUs, and a 12x difference in ROPs. The pixel rate tells the story: 502.5 GPixel/s for the RTX PRO 6000 versus 13.81 GPixel/s for the GTX 765M. Texture rate is 1,968.0 GTexel/s versus 55.23 GTexel/s.

Power delivery also differs radically. The RTX PRO 6000 draws 600 W through a single 16-pin connector and requires a 1000 W power supply. The GTX 765M draws just 75 W, uses no external power connectors, and comes as an MXM module with a portable-device-dependent display output. The server card outputs through 4x DisplayPort 2.1b, while the mobile card's outputs depend entirely on the host laptop.

Head-to-Head Benchmarks

The head-to-head benchmark list between these two cards is empty, so direct comparisons must rely on the recorded database scores and specification data. The RTX PRO 6000's sole benchmark entry is 3DMark Steel Nomad DX12 with a score of 5996. The GTX 765M has three Geekbench entries: Metal at 2612, OpenCL at 7176, and Vulkan at 6714, averaging to 5501.

The RTX PRO 6000's nearest rivals are telling. The GTX 770M scores 6000, which is 0.1% higher. The RX 6400 also scores 6001, 0.1% higher. The FirePro W4100 scores 5987, 0.2% lower, and the Quadro K4000M scores 5986, 0.2% lower. These are essentially tied results, meaning the RTX PRO 6000's average score places it in a cluster of mid-range GPUs from different eras despite its extreme specifications.

For the GTX 765M, its nearest rivals include the MX130 at 5508 (0.1% higher), the Radeon R7 M440 at 5483 (0.3% lower), the Quadro M4000 at 5467 (0.6% lower), and the FirePro M4000 at 5537 (0.7% higher). The mobile Kepler card sits in a similar tight cluster around the 5500 score range.

The RTX PRO 6000's average score of 5996 is 9.0% higher than the GTX 765M's 5501. That is a modest gap in database terms, but it comes from different test suites. The Steel Nomad DX12 test stresses modern rendering pipelines, while the Geekbench tests measure compute-oriented workloads across Metal, OpenCL, and Vulkan. The GTX 765M actually outperforms the RTX PRO 6000 in the OpenCL metric at 7176, but the average pulls it down.

Specification Differences

The two cards diverge on nearly every measurable specification. Process node: 5 nm versus 28 nm. Transistors: 92,200 million versus 2,540 million. Die size: 750 mm² versus 221 mm². Transistor density: 122.9M/mm² versus 11.5M/mm².

Clock speeds: the RTX PRO 6000 runs at 1590 MHz base and 2617 MHz boost, while the GTX 765M runs at 797 MHz base and 863 MHz boost. Memory clocks are 1750 MHz (28 Gbps effective) for the server card versus 1002 MHz (4 Gbps effective) for the mobile card.

Memory configuration: 96 GB GDDR7 on a 512-bit bus versus 2 GB GDDR5 on a 128-bit bus. Bandwidth: 1.79 TB/s versus 64.13 GB/s. That is a 27.9x bandwidth advantage for the server card.

Compute resources: 24,064 shading units versus 768, 752 TMUs versus 64, 192 ROPs versus 16. The RTX PRO 6000 has 188 RT cores and 752 tensor cores; the GTX 765M has none. FP32 throughput: 126.0 TFLOPS versus 1,325.6 GFLOPS. The server card also lists FP16 at 126.0 TFLOPS (1:1 ratio), while the mobile card has no FP16 entry.

Physical and power specifications: the RTX PRO 6000 is a dual-slot card at 267 mm length, 111 mm height, and 40 mm width, drawing 600 W through a 16-pin connector with a 1000 W PSU recommendation. The GTX 765M is an MXM module with no listed dimensions, drawing 75 W with no power connectors and no PSU recommendation. Bus interface: PCIe 5.0 x16 versus MXM-B (3.0).

Production status: the RTX PRO 6000 is active, released 2025-03-17, with a predecessor of Server Hopper and successor of Server Rubin. The GTX 765M is end-of-life, released 2013-05-29, with a predecessor of GeForce 600M and successor of GeForce 800M.

Where Each One Wins

The RTX PRO 6000 wins decisively in every raw throughput category. It offers 126.0 TFLOPS FP32, 1.79 TB/s memory bandwidth, 502.5 GPixel/s pixel rate, and 1,968.0 GTexel/s texture rate. It supports modern APIs including DirectX 12 Ultimate and Vulkan 1.4, has dedicated RT and tensor cores, and provides 96 GB of GDDR7 memory. For server workloads, AI inference, ray-traced rendering, and large dataset processing, this is the clear choice. The 92,200 million transistor count on a 5 nm process makes it a current-generation compute monster.

The GTX 765M wins in power efficiency and form factor. It draws 75 W compared to 600 W, uses no external power connectors, and fits into an MXM module for portable devices. Its production status is end-of-life, but its 2 GB GDDR5 memory and 64.13 GB/s bandwidth were appropriate for its 2013 era. The Geekbench OpenCL score of 7176 shows it still handles compute tasks reasonably for its age. The GTX 765M also holds its own in the database's percentile system at 32nd, nearly matching the RTX PRO 6000's 34th percentile.

The GTX 765M's nearest rival cluster includes the MX130 and Radeon R7 M440, both modern low-end mobile parts, and it trades blows with them within 0.7%. The RTX PRO 6000's nearest rivals include the GTX 770M and RX 6400, showing that its database score lands in a different performance neighborhood than its specifications suggest.

The Verdict

The data points to a simple conclusion: these are products from different eras serving completely different markets. The RTX PRO 6000 Blackwell Server is an active, current-generation server GPU with 96 GB of GDDR7, 188 RT cores, 752 tensor cores, and 126.0 TFLOPS FP32. It targets workloads that require massive memory bandwidth, ray tracing acceleration, and tensor operations. Its 34th percentile ranking and average score of 5996 place it near the GTX 770M and RX 6400 in the database, an artifact of the single benchmark recorded for it.

The GTX 765M is an end-of-life mobile GPU from 2013. Its 768 shading units, 2 GB GDDR5, and 64.13 GB/s bandwidth were competitive for its time. Its average score of 5501 and 32nd percentile ranking put it in a cluster with the MX130 and FirePro M4000, which means it still performs comparably to entry-level modern parts in the database's metrics.

Anyone building a server or workstation for modern compute, AI, or ray-traced workloads should pick the RTX PRO 6000 based on its enormous resource pool. Anyone maintaining legacy portable systems or needing a low-power MXM module should consider the GTX 765M, which draws only 75 W and requires no external power. The benchmark data shows the RTX PRO 6000 scoring 9.0% higher on average, but the real story is the specification gap: 96x more transistors, 48x more memory, 31x more shading units, and 27.9x more bandwidth. The GTX 765M's efficiency per watt is impressive, but it cannot match the Blackwell server part's raw capability in any category except power draw.