NVIDIA GeForce 945M vs NVIDIA GeForce GTX 650 Ti Comparison

The GeForce 945M and GeForce GTX 650 Ti represent two distinct approaches to GPU design from the same manufacturer, separated by three years of architectural evolution. The data shows an almost perfectly matched pair in overall synthetic performance, with the 945M edging ahead in the sole head-to-head benchmark available. This analysis breaks down the quantitative results, architectural divergences, and the specific workloads where each card holds an advantage.

Head-to-Head Benchmarks

The only direct benchmark comparison available is the Geekbench OpenCL test, which measures general-purpose compute performance. In this test, the GeForce 945M scores 8099, while the GeForce GTX 650 Ti scores 7877. The delta is a narrow 2.8% in favor of the mobile part. This is a slim margin, but it is a consistent win, giving the 945M a 1-0 record in direct comparisons.

Interpreting this score requires context from the nearestRivals data. The 945M’s score of 8099 places it within a tight cluster: it is 0.2% ahead of the NVIDIA GRID K2 (8080), 0.4% behind the AMD Radeon R9 M360 (8129), and 0.4% ahead of the NVIDIA GeForce GTX 650 Ti Boost (8067). The GTX 650 Ti’s score of 7877, in turn, is 0.2% behind the GeForce GTX 880M (8040) and 0.2% behind the Quadro P5000 (8039). The performance envelope here is remarkably compressed; the entire spread between the best and worst of these rivals is less than 1.5%.

The 2.8% delta between the two cards in question, therefore, is actually larger than the gaps seen between either card and its nearest rivals. This suggests that while the 945M and GTX 650 Ti are in the same performance class, the difference between them is measurable and consistent, rather than noise. The 945M’s win is not a statistical tie; it is a genuine, if modest, performance lead. The GTX 650 Ti, however, counters this in the broader benchmark suite with a Vulkan score of 8229, a test the 945M does not have a recorded result for. While this does not count as a head-to-head win, it indicates that the older card is capable of higher peak output in specific API workloads.

The Verdict

From a pure benchmark standpoint, the GeForce 945M is the faster card. Its 2.8% lead in OpenCL is the deciding factor in a matchup where both cards occupy the 42nd percentile of all GPUs. For users prioritizing raw compute performance in OpenCL-accelerated applications, the 945M is the clear choice based on the data.

However, the GTX 650 Ti is not without its merits. Its Vulkan score of 8229 is 4.5% higher than its own OpenCL score, suggesting a strength in modern graphics APIs that the 945M cannot match due to the lack of a recorded Vulkan result. This makes the GTX 650 Ti a more compelling option for scenarios leveraging Vulkan, provided the user has the power budget to accommodate a 110 W TDP versus the 945M’s 75 W TDP.

The choice ultimately depends on the application. For a mobile platform with power constraints, the 945M offers a slight compute edge and a much lower power draw. For a desktop system with a 300 W PSU recommendation and a single 6-pin connector, the GTX 650 Ti offers a higher peak score in Vulkan, albeit at a higher power cost. The data does not support a universal winner; it supports a contextual one.

Architecture Differences

The two GPUs are built on fundamentally different architectures, despite sharing a 28 nm TSMC process node. The 945M uses the GM107 chip based on the Maxwell architecture, while the GTX 650 Ti uses the GK106S chip based on the older Kepler architecture.

The transistor counts reveal a significant design divergence. The GTX 650 Ti’s GK106S packs 2,540 million transistors on a 221 mm² die, yielding a density of 11.5M transistors per mm². The 945M’s GM107, by contrast, has only 1,870 million transistors on a 148 mm² die, but achieves a higher density of 12.6M transistors per mm². This indicates that Maxwell is a more area-efficient design, delivering comparable performance with 26% fewer transistors and a 33% smaller die.

The execution resources are configured differently. The GTX 650 Ti has 768 shading units, 64 texture mapping units (TMUs), and 16 raster output units (ROPs). The 945M has fewer of everything in raw count: 640 shading units, 40 TMUs, and 16 ROPs. Despite having 20% fewer shaders and 37.5% fewer TMUs, the 945M achieves a higher pixel rate of 16.32 GPixel/s versus the GTX 650 Ti’s 14.85 GPixel/s. The 945M’s texture rate of 40.80 GTexel/s, however, is significantly lower than the GTX 650 Ti’s 59.39 GTexel/s, reflecting the TMU deficit.

Memory configurations present a clear trade-off. The 945M uses 2 GB of DDR3 memory on a 128-bit bus, delivering 28.80 GB/s of bandwidth. The GTX 650 Ti uses 1024 MB of GDDR5 memory on the same 128-bit bus, but delivers 86.40 GB/s of bandwidth—exactly three times the 945M’s bandwidth. Clock speeds also differ: the 945M runs at a base of 928 MHz with a boost of 1020 MHz, while the GTX 650 Ti’s core clock is not specified in the data, though its memory runs at 1350 MHz (5.4 Gbps effective) versus the 945M’s 900 MHz (1800 Mbps effective).

The feature sets show generational gains. The 945M supports Vulkan 1.4, while the GTX 650 Ti is limited to Vulkan 1.2.175. Both support DirectX 12 (11_0) and OpenGL 4.6. Physically, the 945M is an MXM-B (3.0) module with no power connectors, while the GTX 650 Ti is a 145 mm (5.7-inch) single-slot card requiring a 6-pin power connector and a 300 W power supply.

FAQ

Q: Which GPU has a higher average benchmark score?

A: The GeForce 945M has an average benchmark score of 8099, while the GeForce GTX 650 Ti has an average score of 8053. The 945M leads by 46 points.

Q: How much faster is the GeForce 945M in OpenCL?

A: The 945M scores 8099 versus the GTX 650 Ti’s 7877 in Geekbench OpenCL, a lead of 2.8%.

Q: Does the GeForce GTX 650 Ti have better memory bandwidth?

A: Yes. The GTX 650 Ti has 86.40 GB/s of bandwidth using GDDR5, while the 945M has 28.80 GB/s using DDR3. The GTX 650 Ti has exactly three times the bandwidth.

Q: What is the power consumption difference?

A: The GeForce 945M has a TDP of 75 W with no power connectors, while the GeForce GTX 650 Ti has a TDP of 110 W and requires a single 6-pin power connector.

Q: Which GPU supports a newer version of Vulkan?

A: The GeForce 945M supports Vulkan 1.4, which is newer than the Vulkan 1.2.175 supported by the GeForce GTX 650 Ti.

Q: Did the GTX 650 Ti have a launch MSRP?

A: Yes, the GTX 650 Ti had a launch MSRP of 149 USD. The 945M has no recorded launch MSRP.

Where Each One Wins

The GeForce 945M wins in scenarios where its architectural efficiency and power economy are paramount. Its 2.8% OpenCL advantage makes it the better choice for compute tasks that rely on that API. Its 75 W TDP, combined with the MXM form factor, is designed for portable devices where power draw is critical. The 2 GB frame buffer doubles the GTX 650 Ti’s 1024 MB, which can be an advantage in memory-constrained workloads that require larger datasets to reside on the card. The Maxwell architecture’s higher transistor density (12.6M / mm² versus 11.5M / mm²) and higher pixel rate (16.32 GPixel/s versus 14.85 GPixel/s) indicate a more modern design that extracts more performance per transistor.

The GeForce GTX 650 Ti wins in scenarios demanding raw memory throughput and texture processing. Its 86.40 GB/s bandwidth is a massive advantage over the 945M’s 28.80 GB/s, a 3x difference that will be felt in resolution-dependent tasks and memory-heavy rendering. The 59.39 GTexel/s texture rate, driven by 64 TMUs, is 45% higher than the 945M’s 40.80 GTexel/s, making it the better choice for texture-intensive workloads. Its Vulkan score of 8229, which is higher than its OpenCL score, suggests better optimization for that modern API, a point in its favor for users on Vulkan-based engines. The higher FP32 throughput of 1,425.4 GFLOPS versus the 945M’s 1,305.6 GFLOPS also gives it a raw compute edge for single-precision floating-point tasks, even if the OpenCL benchmark does not reflect this. Its desktop form factor with a 6-pin connector and 300 W PSU recommendation makes it a suitable drop-in for older systems with ample power headroom.