GPU Comparison

The NVIDIA GeForce GTX 650 and NVIDIA GeForce 920MX are both end-of-life mobile-oriented GPUs, but the data shows they are not in the same performance class. Across the shared benchmark suite, the GTX 650 wins every test, with the largest margin coming in Vulkan compute where it leads by 51.5%. The 920MX, however, is a much more power-efficient part, drawing only 16 W compared to the GTX 650's 65 W, and it offers double the memory capacity at 2 GB. This is a classic trade-off between raw throughput and efficiency, and the benchmark results quantify that gap precisely.

Head-to-Head Benchmarks

The Geekbench OpenCL results show a clear but moderate win for the GTX 650. It scores 4545 points against the 920MX's 4068, a delta of 11.7%. This indicates that in general-purpose compute workloads that leverage OpenCL, the older Kepler-based card holds a solid advantage, but it is not a landslide. The 920MX's Maxwell architecture and higher boost clock of 993 MHz help it stay within striking distance, despite having fewer shading units (256 vs. 384) and a much lower FP32 throughput of 508.4 GFLOPS versus 812.5 GFLOPS.

The Vulkan benchmark tells a more dramatic story. The GTX 650 scores 4524, while the 920MX manages only 2987. That is a 51.5% lead for the GTX 650, a massive gap that dwarfs the OpenCL difference. This suggests that the GTX 650's memory subsystem, an 80.00 GB/s bandwidth over a 128-bit bus with GDDR5, provides a substantial advantage in modern, lower-level graphics APIs that are more sensitive to memory bandwidth than to raw shader count. The 920MX's DDR3 memory on a 64-bit bus delivers just 14.40 GB/s, which is a severe bottleneck in Vulkan workloads. The data indicates that if you are targeting Vulkan-based applications or games, the GTX 650 is in a different league entirely.

In terms of average benchmark score across all available tests, the GTX 650 also leads with 3823 points versus 3528 for the 920MX, a difference of roughly 8.4%. This aligns with the head-to-head results and reinforces the conclusion that the GTX 650 is the stronger compute part overall. Interestingly, the GTX 650's nearest rival in the database is the NVIDIA GeForce MX110 at 3834 points, which is only 0.3% higher, while the 920MX sits just below the GeForce GT 545 (3594 points) at a -1.8% delta. This places both cards in a similar low-end tier, but the GTX 650 consistently sits at the top of that tier.

FAQ

Q: Which GPU has the higher average benchmark score?

A: The NVIDIA GeForce GTX 650 has a higher average benchmark score of 3823 points, compared to the NVIDIA GeForce 920MX's 3528 points. This places the GTX 650 at the 22nd percentile of all GPUs, while the 920MX is at the 21st percentile.

Q: How big is the performance gap in Vulkan compute?

A: The GTX 650 scores 4524 in the Geekbench Vulkan test, which is 51.5% higher than the 920MX's score of 2987. This is by far the largest performance delta between the two in any shared benchmark.

Q: Does the 920MX have any memory advantage?

A: Yes, the 920MX has 2 GB of DDR3 memory, which is double the 1024 MB of GDDR5 found on the GTX 650. However, the GTX 650's memory bandwidth is significantly higher at 80.00 GB/s versus 14.40 GB/s.

Q: What is the power consumption difference?

A: The GTX 650 has a TDP of 65 W and requires a 1x 6-pin power connector, while the 920MX has a TDP of just 16 W and needs no external power connectors. The 920MX is designed for ultra-portable devices as an MXM module.

Q: Which GPU has a higher transistor density?

A: The 920MX has a higher transistor density of 13.2M / mm², despite having fewer total transistors (1,020 million) than the GTX 650 (2,540 million). The GTX 650's density is 11.5M / mm².

Q: Are both GPUs from the same architecture family?

A: No. The GTX 650 is based on the Kepler architecture with the GK106 chip, while the 920MX uses the Maxwell architecture with the GM108S chip. Both are manufactured on a 28 nm process by TSMC.

The Verdict

The data is unambiguous: the NVIDIA GeForce GTX 650 is the faster GPU in every benchmark where both were tested. It wins the OpenCL test by 11.7% and the Vulkan test by a dominant 51.5%. Its average benchmark score of 3823 is nearly 300 points higher than the 920MX's 3528. If the priority is compute performance or running Vulkan-based workloads, the GTX 650 is the clear choice based solely on these results.

However, the verdict is not one-sided when considering the full specification picture. The 920MX draws only 16 W, making it dramatically more efficient than the 65 W GTX 650. It also offers 2 GB of memory, which may be preferable for capacity-sensitive tasks, even if the bandwidth is far lower. For a laptop or compact system where power draw and heat are the primary constraints, the 920MX's efficiency profile makes it a more sensible option, despite its lower scores.

The benchmark results indicate that the GTX 650's performance lead comes at a cost: a 6-pin power connector and a 250 W suggested PSU, which are not viable for ultra-thin portable devices. The 920MX, as an MXM module with no power connectors, is built for a different physical environment. So, the verdict is not that one is universally better, it is that the GTX 650 wins on raw performance, while the 920MX wins on efficiency and memory capacity. Choose based on what the system can physically support and what workload matters more.

Specification Differences

The two GPUs differ across nearly every core specification. The GTX 650 has 384 shading units, 32 texture mapping units, and 16 ROPs, while the 920MX has 256 shading units, 24 TMUs, and only 8 ROPs. This translates to a pixel rate of 8.464 GPixel/s and a texture rate of 33.86 GTexel/s for the GTX 650, versus 7.944 GPixel/s and 23.83 GTexel/s for the 920MX. The GTX 650's FP32 performance is 812.5 GFLOPS, compared to 508.4 GFLOPS for the 920MX.

Memory is another major differentiator. The GTX 650 uses 1024 MB of GDDR5 on a 128-bit bus, yielding 80.00 GB/s of bandwidth. The 920MX uses 2 GB of DDR3 on a 64-bit bus, yielding just 14.40 GB/s. The GTX 650's memory clock is listed as 1250 MHz (5 Gbps effective), while the 920MX runs at 900 MHz (1800 Mbps effective). The GTX 650 also has a wider PCIe interface (PCIe 3.0 x16 vs. PCIe 3.0 x8) and a larger die size (221 mm² vs. 77 mm²), with the GTX 650 packing 2,540 million transistors versus 1,020 million on the 920MX.

Power and physical design differ as well. The GTX 650 has a 65 W TDP, is single-slot, requires a 1x 6-pin power connector, and has a 250 W suggested PSU. The 920MX has a 16 W TDP, is an MXM module with no power connectors, and has no suggested PSU listed. The GTX 650 measures 147 mm (5.8 inches) in length, while the 920MX has no listed dimensions. Display outputs also differ: the GTX 650 offers 1x DVI and 2x DisplayPort 1.2, while the 920MX's outputs are listed as "Portable Device Dependent."

Architecture Differences

The GTX 650 is built on the Kepler architecture with the GK106 chip, part of the GeForce 600 generation. The 920MX uses the Maxwell architecture with the GM108S chip, part of the GeForce 900M generation. Both are fabricated on a 28 nm process by TSMC, but the transistor density differs: the GTX 650 has 11.5M transistors per mm², while the 920MX has 13.2M / mm². This higher density on the smaller 920MX die (77 mm² vs. 221 mm²) reflects the newer Maxwell design's efficiency improvements.

In terms of API support, both GPUs support DirectX 12 (11_0) and OpenGL 4.6. However, the Vulkan support differs: the GTX 650 supports Vulkan 1.2.175, while the 920MX supports Vulkan 1.4. This newer Vulkan version on the 920MX does not translate into better Vulkan performance in the benchmarks, the GTX 650 still wins decisively, but it does indicate a more modern driver feature set. The GTX 650 has no base or boost clock listed, while the 920MX has a base clock of 965 MHz and a boost clock of 993 MHz.

The cache hierarchy is not detailed in the data, but the core counts reveal the architectural gulf. The GTX 650's 384 shading units and 32 TMUs are substantially higher than the 920MX's 256 and 24, respectively. The GTX 650 also has double the ROPs (16 vs. 8). The GTX 650's predecessor is the GeForce 500 series and its successor is the GeForce 700 series, while the 920MX's predecessor is the GeForce 800M and its successor is the GeForce 10 Mobile. Both are end-of-life products, with the GTX 650 released on 2013-11-26 and the 920MX released on 2016-03-24.

Where Each One Wins

The GTX 650 wins in every compute benchmark category where both were tested. Its 51.5% Vulkan lead is the standout result, making it the clear choice for any application that leverages Vulkan compute or rendering. Its OpenCL advantage of 11.7% also makes it the better option for general GPU compute tasks, such as OpenCL-accelerated content creation or scientific workloads. The GTX 650's higher texture rate (33.86 GTexel/s) and pixel rate (8.464 GPixel/s) suggest it can handle more demanding graphics workloads, making it suitable for older or less demanding games at higher settings, provided the system can supply 65 W and a 6-pin connector.

The 920MX wins in scenarios where power efficiency and physical footprint are paramount. Its 16 W TDP is a fraction of the GTX 650's 65 W, and its MXM module form factor with no power connectors makes it ideal for thin-and-light laptops where the GTX 650's single-slot, 147 mm length and 6-pin requirement would be impossible to accommodate. The 920MX's 2 GB of memory also gives it an advantage in memory-capacity-bound tasks, such as holding larger textures or datasets in VRAM, even though the 14.40 GB/s bandwidth severely limits how fast that data can be accessed. For a portable device where battery life and thermals are the priority, the 920MX is the only viable option of the two.