NVIDIA GeForce GT 745M vs NVIDIA GeForce GTX 650 Comparison

The NVIDIA GeForce GT 745M and the NVIDIA GeForce GTX 650 are both Kepler-architecture parts from the same generation, but they target very different segments of the market. The GT 745M is a mobile integrated graphics processor (IGP) designed for laptops, while the GTX 650 is a single-slot desktop card with its own power connector. Benchmark data shows a split decision: the mobile part wins in two of three compute tests, while the desktop card takes a decisive victory in one. This creates a nuanced comparison where the intended use case matters more than raw average scores.

Where Each One Wins

The data reveals a clear division of labor between these two NVIDIA parts. The GT 745M dominates in the Geekbench Metal and Vulkan tests, posting scores of 2777 and 5502 respectively. These are compute and graphics API workloads that often reflect real-world rendering and general-purpose GPU tasks. The GTX 650, meanwhile, wins the Geekbench OpenCL test with a score of 4545, which is a significant margin over the GT 745M’s 3580. This suggests the desktop card is better suited for OpenCL-accelerated compute tasks, which are common in scientific computing, video encoding, and some productivity applications.

The overall average benchmark scores tell a slightly different story. The GT 745M has an average score of 3953, placing it in the 23rd percentile of all GPUs. The GTX 650 has a lower average of 3823, placing it in the 22nd percentile. Despite the GTX 650 winning the OpenCL test by a large margin, its losses in the other two tests drag its average below the mobile part. This means that in a broad mix of workloads, the GT 745M is slightly more consistent, while the GTX 650 is more specialized. The GT 745M’s nearest rival is the AMD Radeon R5 M420, which scores 3956, a delta of -0.1%, indicating these two are effectively tied. The GTX 650’s closest competitor is the NVIDIA GeForce MX110 at 3834, a -0.3% delta, showing it also sits in a tight competitive cluster.

Architecture Differences

Both GPUs are built on the same 28 nm process node at TSMC, but they use different chips. The GT 745M uses the GK107 chip, while the GTX 650 uses the GK106. This is a substantial difference in physical size and complexity. The GK107 has 1,270 million transistors on a 118 mm² die, resulting in a transistor density of 10.8 million per mm². The GK106 is much larger, with 2,540 million transistors on a 221 mm² die, achieving a density of 11.5 million per mm². This means the GTX 650 has exactly double the transistor count of the GT 745M, which should theoretically give it more headroom for performance.

Despite the transistor difference, both parts share identical core configurations. Each has 384 shading units, 32 texture mapping units (TMUs), and 16 raster output units (ROPs). This is unusual, as the larger chip would typically accommodate more execution resources. The clock speeds and memory configurations differentiate them. The GT 745M runs its memory at 1000 MHz, translating to 4 Gbps effective, while the GTX 650 runs at 1250 MHz, or 5 Gbps effective. Both use a 128-bit memory bus, but the GTX 650 achieves 80.00 GB/s of bandwidth compared to the GT 745M’s 64.00 GB/s. The memory size also differs: the GT 745M comes with 2 GB of GDDR5, while the GTX 650 has 1024 MB.

The power and physical characteristics are starkly different. The GT 745M has a TDP of 45 W and is listed as an IGP, meaning it is designed to be soldered onto a laptop motherboard. The GTX 650 has a TDP of 65 W, requires a single 6-pin power connector, and is a single-slot card measuring 147 mm (5.8 inches) in length. The desktop card also has a suggested PSU rating of 250 W, while the mobile part has no such requirement. Both support PCIe 3.0 x16, DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. The display outputs differ as expected: the GT 745M is "portable device dependent," while the GTX 650 offers 1x DVI and 2x DisplayPort 1.2.

Head-to-Head Benchmarks

The Geekbench Metal test shows the GT 745M ahead with a score of 2777 versus the GTX 650’s 2400. This is a delta of 15.7%, a substantial margin favoring the mobile part. Metal is Apple’s graphics API, and while it is typically used on macOS, this benchmark result indicates the GT 745M’s architecture is better optimized for that specific workload. The GTX 650 is not far behind, but the 377-point gap is not negligible.

The Geekbench OpenCL test reverses the result dramatically. The GTX 650 scores 4545, while the GT 745M manages only 3580. The delta here is -21.2% from the perspective of the GT 745M, meaning the GTX 650 is more than a fifth faster. This is the largest single-test margin in the comparison. OpenCL is a cross-platform compute standard, and the GTX 650’s higher memory bandwidth and doubled transistor count likely contribute to this advantage. The 965-point difference is significant for any workload that leverages OpenCL.

The Geekbench Vulkan test returns to the GT 745M’s favor. It scores 5502, beating the GTX 650’s 4524 by a delta of 21.6%. This is the highest absolute score in the entire comparison, and the largest percentage win for either card. Vulkan is a modern low-overhead graphics API, and the GT 745M’s strong showing here suggests it handles draw calls and rendering pipelines efficiently. The 978-point gap is substantial, making this the most lopsided result in the head-to-head data. Overall, the GT 745M wins two tests, the GTX 650 wins one, but the magnitudes of the wins are comparable.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce GT 745M has an average benchmark score of 3953, while the NVIDIA GeForce GTX 650 scores 3823. The GT 745M is 130 points higher, placing it in the 23rd percentile versus the GTX 650’s 22nd.

Q: Is the GTX 650 always faster than the GT 745M?

A: No. The GTX 650 wins the Geekbench OpenCL test with 4545 against 3580, a 21.2% advantage. However, the GT 745M wins the Metal test (2777 vs 2400) and the Vulkan test (5502 vs 4524).

Q: How do these cards compare to their nearest rivals?

A: The GT 745M is nearly tied with the AMD Radeon R5 M420 (3956, -0.1%) and the NVIDIA GeForce 830M (3957, -0.1%). The GTX 650 is close to the NVIDIA GeForce MX110 (3834, -0.3%) and the Intel UHD Graphics 710 (3792, 0.8%).

Q: Do these GPUs have the same core configuration?

A: Yes, both have 384 shading units, 32 TMUs, and 16 ROPs. The difference lies in the chip size and memory: the GTX 650 uses the GK106 chip with 2,540 million transistors, while the GT 745M uses the GK107 with 1,270 million.

Q: What are the memory bandwidth differences?

A: The GTX 650 has 80.00 GB/s of bandwidth, while the GT 745M has 64.00 GB/s. The GTX 650 also has faster memory at 5 Gbps effective versus 4 Gbps, but the GT 745M offers 2 GB of VRAM versus 1024 MB.

Q: Which card is more power-efficient?

A: The GT 745M has a TDP of 45 W and is an IGP, while the GTX 650 has a TDP of 65 W and requires a 6-pin power connector. The mobile part uses less power, but the desktop card is physically larger.

The Verdict

The data supports a clear choice based on workload. For users running OpenCL-based compute tasks, the NVIDIA GeForce GTX 650 is the superior option. Its 4545 score in Geekbench OpenCL versus the GT 745M’s 3580 represents a 21.2% advantage, and its higher memory bandwidth of 80.00 GB/s versus 64.00 GB/s is a tangible benefit. The desktop card also has double the transistors (2,540 million vs 1,270 million), which explains its raw compute strength. This is the card for users who prioritize compute throughput over other factors.

For anyone focused on modern graphics APIs or working in environments where Metal or Vulkan are the primary interfaces, the NVIDIA GeForce GT 745M is the stronger choice. It wins the Metal test by 15.7% and the Vulkan test by 21.6%, with the latter being its peak score of 5502. The mobile part also offers 2 GB of VRAM, which is double the GTX 650’s allocation, providing more headroom for texture-heavy workloads. Its lower TDP of 45 W makes it suitable for portable devices, whereas the GTX 650’s 65 W TDP and single-slot form factor require a desktop chassis.

The average benchmark scores slightly favor the GT 745M (3953 vs 3823), but this masks the GTX 650’s specialization. The GT 745M is the more balanced performer across the tested APIs, while the GTX 650 is a specialist that excels in one domain. Neither card is a clear overall winner; the choice hinges on the user’s software ecosystem and performance priorities. The GT 745M’s 23rd percentile and the GTX 650’s 22nd percentile both indicate entry-level positioning, but the GT 745M offers broader API compatibility with its two wins. The GTX 650, despite its lower average, provides the highest peak output in a single test. Ultimately, the GT 745M is the better all-rounder for graphics API workloads, while the GTX 650 is the pick for OpenCL compute performance.