NVIDIA GeForce GTX 650 Ti vs NVIDIA GeForce GTX 680M Comparison

Head-to-Head Benchmarks

The recorded data shows a single direct comparison between these two GPUs: the Geekbench OpenCL test. In that test, the NVIDIA GeForce GTX 680M scores 9,230 points, while the NVIDIA GeForce GTX 650 Ti scores 7,877 points. The delta is 14.7% in favor of the GTX 680M, making it the clear winner in the only head-to-head metric available. That is a substantial margin for a single workload, and it suggests the mobile part is operating at a different performance tier entirely.

However, the average benchmark score tells a more complicated story. The GTX 650 Ti averages 8,053 points across all recorded tests, while the GTX 680M averages 7,023 points. That is a reversal of the head-to-head result, and it is worth investigating. The GTX 650 Ti also has a higher percentile rank among all GPUs: 42nd percentile versus 39th for the GTX 680M. The desktop card is the more consistent performer across the broader benchmark suite, even though it loses the one direct comparison.

The nearest rivals in the database add context. The GTX 650 Ti sits within a tight cluster: the GeForce GTX 880M scores 8,040 (0.2% higher), the Quadro P5000 scores 8,039 (0.2% higher), the GTX 650 Ti Boost scores 8,067 (0.2% higher), and the GRID K2 scores 8,080 (0.3% higher). Every one of these rivals is within a fraction of a percent of the GTX 650 Ti's average, which means the desktop card is essentially at parity with a group of much more expensive or specialized hardware. The GTX 680M, by contrast, sits in a lower cluster: the T600 scores 7,035 (0.2% higher), the GTX 675M scores 6,946 (1.1% lower), the Radeon R5 M240 scores 6,975 (0.7% lower), and the GTX 970 scores 7,157 (1.9% higher). The GTX 680M is the second lowest in its group, and the GTX 970 is nearly 2% ahead of it.

The implication is that the GTX 680M's OpenCL win is not representative of its overall standing. It beats the GTX 650 Ti in that one test, but the database average places it behind, and its percentile rank confirms it. This is a curious split, and the architecture section below may explain why.

Architecture Differences

Both GPUs are built on TSMC's 28 nm process node and share the Kepler architecture, but they are very different chips. The GTX 650 Ti uses the GK106S chip, which packs 2,540 million transistors into a 221 mm² die, giving a transistor density of 11.5 million per square millimeter. The GTX 680M uses the larger GK104 chip, with 3,540 million transistors on a 294 mm² die, for a density of 12.0 million per square millimeter. The GK104 is the more complex and physically larger design, and it has roughly 39% more transistors, a difference that shows up in nearly every compute resource.

The shading unit count is the most dramatic gap. The GTX 650 Ti has 768 shading units, while the GTX 680M has 1,344, a 75% increase. Texture mapping units follow the same pattern: 64 on the GTX 650 Ti versus 112 on the GTX 680M. Raster operation units double from 16 to 32. These are enormous differences in raw hardware, and they explain why the GTX 680M can post such a strong OpenCL score despite its lower overall average.

Memory is another major divergence. The GTX 650 Ti has 1,024 MB of GDDR5 on a 128-bit bus, with bandwidth of 86.40 GB/s. The GTX 680M has 4 GB of GDDR5 on a 256-bit bus, with bandwidth of 115.2 GB/s. The wider bus and larger frame buffer give the mobile chip a clear advantage in memory-bound workloads, and the 33% higher bandwidth is significant. The memory clock also differs: the GTX 650 Ti runs at 1,350 MHz (5.4 Gbps effective), while the GTX 680M runs at 900 MHz (3.6 Gbps effective). The lower clock on the GTX 680M is compensated by the wider bus, which is why total bandwidth is still higher.

Clock speeds are listed only for the GTX 680M, with a base of 719 MHz and a boost of 758 MHz. The GTX 650 Ti has no base or boost clock in the database, so a direct clock comparison is not possible. The pixel rate favors the GTX 680M at 21.22 GPixel/s versus 14.85 GPixel/s for the GTX 650 Ti, and the texture rate also favors it at 84.90 GTexel/s versus 59.39 GTexel/s. Floating point performance is listed in different units: the GTX 650 Ti delivers 1,425.4 GFLOPS, while the GTX 680M delivers 2.038 TFLOPS, which is roughly 43% higher. The GTX 680M is the faster chip on paper in almost every compute metric.

The form factors could not be more different. The GTX 650 Ti is a single-slot desktop card with a 1x 6-pin power connector and a suggested 300 W power supply. It measures 145 mm (5.7 inches) in length. The GTX 680M is an MXM module with no power connectors, and its display outputs are listed as dependent on the portable device. The GTX 650 Ti has 2x DVI and 1x mini-HDMI 1.4a outputs. The GTX 680M's power draw is actually lower at 100 W versus 110 W for the GTX 650 Ti, which is notable given the GTX 680M's larger chip and higher compute throughput. The GTX 680M also has no suggested PSU rating, as it relies on the host laptop's power delivery.

Both support DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175, so API capabilities are identical. The production status is end-of-life for both, and both belong to the GeForce 600 family, though the GTX 680M is in the 600M mobile generation while the GTX 650 Ti is in the desktop 600 generation. Release dates are close: the GTX 680M launched on June 3, 2012, and the GTX 650 Ti on October 8, 2012. The GTX 680M's predecessor is the GeForce 500M and its successor is the GeForce 700M; the GTX 650 Ti's predecessor is the GeForce 500 and its successor is the GeForce 700.

FAQ

Q: Which GPU has more shading units?

A: The NVIDIA GeForce GTX 680M has 1,344 shading units, while the NVIDIA GeForce GTX 650 Ti has 768, a 75% advantage for the mobile chip.

Q: Why does the GTX 680M win the OpenCL test but lose the average benchmark score?

A: The GTX 680M scores 9,230 in Geekbench OpenCL versus 7,877 for the GTX 650 Ti, a 14.7% win. However, the database average shows the GTX 650 Ti at 8,053 and the GTX 680M at 7,023, meaning the GTX 650 Ti performs more consistently across other recorded workloads.

Q: What is the memory configuration difference?

A: The GTX 650 Ti has 1,024 MB of GDDR5 on a 128-bit bus with 86.40 GB/s bandwidth. The GTX 680M has 4 GB of GDDR5 on a 256-bit bus with 115.2 GB/s bandwidth, a 33% bandwidth advantage.

Q: Are the GPUs the same architecture?

A: Yes, both use the Kepler architecture and are manufactured on TSMC's 28 nm process node. The GTX 650 Ti uses the GK106S chip, and the GTX 680M uses the GK104 chip.

Q: Which GPU has a higher transistor count?

A: The GTX 680M has 3,540 million transistors on a 294 mm² die, compared to 2,540 million on a 221 mm² die for the GTX 650 Ti.

Q: What is the average benchmark score for each card?

A: The GTX 650 Ti averages 8,053 points, and the GTX 680M averages 7,023 points. The GTX 650 Ti also ranks higher at the 42nd percentile versus the 39th percentile for the GTX 680M.

Specification Differences

The two GPUs differ in nearly every specification. The chip is different: GK106S for the GTX 650 Ti, GK104 for the GTX 680M. Transistor count is 2,540 million versus 3,540 million, and die size is 221 mm² versus 294 mm². Transistor density is 11.5M per mm² versus 12.0M per mm². The GTX 680M has explicit base and boost clocks (719 MHz and 758 MHz), while the GTX 650 Ti has none listed. Memory clock is 1,350 MHz (5.4 Gbps effective) for the GTX 650 Ti and 900 MHz (3.6 Gbps effective) for the GTX 680M. Memory size is 1,024 MB versus 4 GB, bus width is 128 bit versus 256 bit, and bandwidth is 86.40 GB/s versus 115.2 GB/s.

Compute resources differ sharply: shading units are 768 versus 1,344, TMUs are 64 versus 112, and ROPs are 16 versus 32. Pixel rate is 14.85 GPixel/s versus 21.22 GPixel/s, texture rate is 59.39 GTexel/s versus 84.90 GTexel/s, and FP32 is 1,425.4 GFLOPS versus 2.038 TFLOPS. TDP is 110 W versus 100 W. The slot width is single-slot for the GTX 650 Ti and MXM Module for the GTX 680M. Power connectors are 1x 6-pin versus none, and the suggested PSU is 300 W for the GTX 650 Ti with no rating for the GTX 680M. Bus interface is PCIe 3.0 x16 versus MXM-B (3.0). Display outputs are 2x DVI and 1x mini-HDMI 1.4a versus portable device dependent. Dimensions are 145 mm (5.7 inches) for the GTX 650 Ti, with no dimensions for the GTX 680M. The GTX 650 Ti has a launch MSRP of 149 USD, while the GTX 680M has no launch MSRP. Generation and series names also differ: GeForce 600 versus GeForce 600M.

The Verdict

The data does not point to a single winner. In the only direct head-to-head test, the GeForce GTX 680M wins by 14.7%, and it has overwhelming advantages in shading units, texture units, ROPs, memory capacity, and memory bandwidth. For compute-heavy workloads that resemble the OpenCL test, the GTX 680M is clearly the stronger part. The GTX 650 Ti, however, has the higher average benchmark score (8,053 versus 7,023), the higher percentile rank (42 versus 39), and a lower TDP is not a differentiator since the GTX 680M draws less power. The GTX 650 Ti also belongs to a tighter cluster of near-identical rivals, while the GTX 680M sits near the bottom of its group with the GTX 970 ahead of it.

The GTX 650 Ti is the safer choice for general-purpose desktop use, where consistency across multiple workloads matters more than peak compute in a single test. The GTX 680M is the choice for scenarios where raw compute throughput is the priority, such as OpenCL-based tasks, and where the mobile MXM form factor is required. The GTX 680M's larger frame buffer and wider memory bus also make it better suited for memory-intensive applications, even though its overall average score is lower.

There is a tension in the numbers. The GTX 680M looks better on paper in almost every architectural specification, yet it loses the average benchmark comparison. The explanation may lie in thermal constraints, driver optimization, or the specific workload mix in the database. The data alone cannot resolve that question, but it does show that hardware specifications do not always translate directly to benchmark averages.

Where Each One Wins

The GeForce GTX 680M wins in the Geekbench OpenCL test, scoring 9,230 against 7,877 for the GTX 650 Ti, a 14.7% margin. It also wins on every architectural compute metric: 2.038 TFLOPS FP32 versus 1,425.4 GFLOPS, 21.22 GPixel/s versus 14.85 GPixel/s, 84.90 GTexel/s versus 59.39 GTexel/s, and higher counts for shading units, TMUs, and ROPs. It wins on memory bandwidth at 115.2 GB/s versus 86.40 GB/s, and it offers 4 GB of VRAM versus 1,024 MB. The GTX 680M is the winner for tasks that stress raw compute throughput, memory capacity, and fill rates, and it does so at a lower TDP of 100 W versus 110 W.

The GeForce GTX 650 Ti wins on the aggregate benchmark average, 8,053 versus 7,023, and on percentile rank, 42 versus 39. It also has a launch MSRP of 149 USD, while the GTX 680M has no listed launch MSRP. Its nearest rivals are within 0.3% of its average score, showing that it performs at parity with a range of other GPUs. The GTX 650 Ti is the winner for consistent general-purpose performance, for desktop builds with a standard PCIe slot, and for users who want a single-slot card with DVI and mini-HDMI outputs. The data suggests it is the more balanced part, even if it cannot match the GTX 680M in peak compute.