NVIDIA GeForce GT 740 vs NVIDIA GeForce GTX 650 Comparison

The GeForce GTX 650 and GeForce GT 740 are both end-of-life Kepler-generation cards from NVIDIA, but the benchmark data reveals they are not equal performers. While they share the same core configuration on paper, the GTX 650 consistently outperforms the GT 740 across all three Geekbench compute tests, with the margin varying significantly depending on the workload. This analysis explores the data behind these two 28 nm cards to determine where the real differences lie and which card holds up better in modern compute scenarios.

Head-to-Head Benchmarks

The head-to-head results are unambiguous: the GTX 650 wins all three benchmark tests. The smallest victory comes in Geekbench Metal, where the GTX 650 scores 2400 against the GT 740's 2363, a marginal 1.6% advantage. This narrow gap suggests that in Metal-optimized workloads, the two cards are nearly interchangeable, and the architectural differences between them have minimal impact on this particular API's performance.

The gap widens dramatically in Geekbench OpenCL. Here, the GTX 650 posts a score of 4545, while the GT 740 manages only 3847. That represents an 18.1% lead for the GTX 650, a substantial margin that indicates a real performance hierarchy. This is not a case of benchmark noise; an 18-point delta percent is a clear, repeatable difference in compute throughput. The data implies that OpenCL workloads, which often stress raw shader and texture processing, favor the GTX 650's higher clock-derived rates.

Geekbench Vulkan tells a similar story, though with a slightly narrower margin. The GTX 650 scores 4524, while the GT 740 trails at 4083, giving the GTX 650 a 10.8% advantage. This result is particularly telling because Vulkan is a low-overhead API that leans heavily on driver efficiency and raw hardware capability. The fact that the GTX 650 maintains a double-digit lead here suggests its advantage is fundamental to the silicon itself, not just a quirk of a single test suite.

Looking at the average benchmark scores reinforces this pattern. The GTX 650's average score is 3823, while the GT 740's is 3431. This places the GTX 650 at the 22nd percentile of all GPUs, with the GT 740 one point lower at the 21st percentile. While neither card is a performance champion in the broader GPU landscape, the GTX 650's higher average and its clean sweep of the head-to-head tests indicate it is the stronger of the two. The GT 740's nearest rival, the NVIDIA Quadro 2000M, scores 3434 with a deltaPct of -0.1%, showing the GT 740 sits essentially at parity with that mobile workstation card.

Architecture Differences

Both cards are built on the Kepler architecture using a 28 nm process at TSMC, but they are fundamentally different chips. The GTX 650 uses the GK106 chip, while the GT 740 uses the GK107. This is the root of their performance disparity. The GK106 is a larger chip, with a die size of 221 mm², nearly double the GT 740's 118 mm². Transistor counts tell a similar story: the GTX 650 packs 2,540 million transistors, while the GT 740 has just 1,270 million, exactly half. The transistor density is also slightly higher on the GTX 650, at 11.5M per mm² versus 10.8M per mm², indicating a more densely packed design.

Despite the larger chip, the execution resources are identical. Both cards feature 384 shading units, 32 texture mapping units, and 16 render output units. They also share the same memory configuration: 1024 MB of GDDR5 on a 128-bit bus, with nearly identical bandwidth figures of 80.00 GB/s for the GTX 650 and 80.19 GB/s for the GT 740. The memory clocks are also effectively the same, with both running at approximately 1250 MHz (5 Gbps effective). This means the performance difference cannot be attributed to core counts or memory bandwidth.

The difference lies in the clock speeds and the resulting fill rates. While the base and boost clocks are not listed in the data, the derived rates tell the tale. The GTX 650 achieves a pixel rate of 8.464 GPixel/s and a texture rate of 33.86 GTexel/s. The GT 740 lags with a pixel rate of 7.944 GPixel/s and a texture rate of 31.78 GTexel/s. Similarly, the FP32 compute throughput is 812.5 GFLOPS for the GTX 650 versus 762.6 GFLOPS for the GT 740. These differences in derived rates indicate that the GTX 650 is running at higher internal clocks, which is consistent with it being the more aggressively binned and higher-tier part. The GTX 650 is also from the GeForce 600 generation, released on 2013-11-26, while the GT 740 is from the GeForce 700 generation, released later on 2014-05-28.

The feature sets are nearly identical on paper. Both support DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175. Neither has RT cores or tensor cores, and both lack FP16 support, making them strictly FP32-oriented parts. The display outputs differ, with the GTX 650 offering 1x DVI and 2x DisplayPort 1.2, while the GT 740 offers 2x DVI and 1x mini-HDMI 1.4a. This could matter for multi-monitor setups, but it does not affect compute performance.

Where Each One Wins

The GTX 650 wins in every compute benchmark, but the margins tell a nuanced story about where each card is strongest. The GTX 650's biggest win is in OpenCL, where its 18.1% lead is substantial. This suggests that for general-purpose compute tasks that scale well with shader throughput and fill rates, the GTX 650 is the clear choice. Its higher pixel and texture rates, combined with higher FP32 output, give it a tangible edge in workloads like image processing, physics simulation, or any OpenCL-accelerated application.

The GTX 650's 10.8% lead in Vulkan is also notable. Vulkan is increasingly used in gaming and professional 3D applications, and this advantage means the GTX 650 will deliver smoother performance in Vulkan-based titles. Its 1.6% lead in Metal is negligible in practice, but it still means the GTX 650 is the better card for macOS-style Metal compute workloads, albeit by a hair.

The GT 740, having lost all three tests, does not have a single benchmark where it wins. However, its performance profile is not entirely without merit. The data shows it is extremely close to the GTX 650 in Metal, within 1.6%, which means for users whose primary workload is Metal, the difference will be imperceptible. The GT 740 also has a lower TDP at 64 W versus 65 W, though this is a negligible difference. Its smaller die size (118 mm²) and fewer transistors (1,270 million) imply it might run cooler in a constrained chassis, though the data does not provide thermal figures. The GT 740 also has a slightly higher memory bandwidth at 80.19 GB/s, a 0.2% difference that is purely academic.

For users who need DisplayPort outputs, the GTX 650's 2x DisplayPort 1.2 is a clear advantage over the GT 740's mini-HDMI. Conversely, the GT 740's 2x DVI outputs might be preferable for legacy DVI-only monitors. Neither card supports modern ray tracing or tensor core acceleration, so those features are not a differentiator.

FAQ

Q: Is the GTX 650 always faster than the GT 740?

A: Yes, based on the data. The GTX 650 wins all three head-to-head benchmarks: Geekbench Metal (2400 vs 2363), Geekbench OpenCL (4545 vs 3847), and Geekbench Vulkan (4524 vs 4083).

Q: How much faster is the GTX 650 in OpenCL?

A: The GTX 650 leads by 18.1% in Geekbench OpenCL, scoring 4545 versus the GT 740's 3847. This is the largest performance gap between the two cards in any test.

Q: Do the cards have the same core configuration?

A: Yes, both have 384 shading units, 32 TMUs, and 16 ROPs. They also share 1024 MB of GDDR5 memory on a 128-bit bus. The performance difference comes from the GTX 650's higher clock speeds, as evidenced by its higher pixel rate (8.464 vs 7.944 GPixel/s) and texture rate (33.86 vs 31.78 GTexel/s).

Q: Which card has a better overall benchmark score?

A: The GTX 650 has an average benchmark score of 3823, placing it at the 22nd percentile of all GPUs. The GT 740 averages 3431, at the 21st percentile.

Q: Are there any benchmarks where the GT 740 comes close to the GTX 650?

A: The closest is Geekbench Metal, where the GT 740 trails by only 1.6% (2363 vs 2400). In this test, the performance difference is small enough to be considered negligible in real-world use.

Q: What generation do these cards belong to?

A: The GTX 650 is from the GeForce 600 generation, released on 2013-11-26, while the GT 740 is from the GeForce 700 generation, released on 2014-05-28. Both are end-of-life products.

The Verdict

The data is clear: the NVIDIA GeForce GTX 650 is the superior card. It wins every benchmark, with leads ranging from 1.6% in Metal to 18.1% in OpenCL. Its higher average score (3823 vs 3431) and identical core configuration with higher derived clocks make it the more capable compute part. For any user running OpenCL or Vulkan workloads, the GTX 650 offers a meaningful performance advantage that will be noticeable in real-world applications. The GT 740's only consolation is its near-parity in Metal, but even there, the GTX 650 edges it out.

The GT 740 is not without its merits. Its smaller die (118 mm²) and lower transistor count (1,270 million) suggest it may be a more efficient part in terms of silicon cost, and its launch MSRP of 89 USD was likely positioned as a more affordable option. However, the data shows the GTX 650 simply does more with the same resources. The GT 740's nearest rival, the NVIDIA Quadro 2000M, scores 3434 with a deltaPct of -0.1%, meaning the GT 740 is essentially tied with that mobile workstation GPU. In contrast, the GTX 650's nearest rival, the NVIDIA GeForce MX110, scores 3834 with a deltaPct of -0.3%, putting the GTX 650 in the company of more modern entry-level parts.

For users choosing between these two end-of-life cards, the GTX 650 is the default recommendation. It delivers higher compute throughput, better Vulkan and OpenCL performance, and a more capable display output configuration with dual DisplayPort 1.2 connectors. The GT 740 should only be considered if the 2x DVI output layout is a hard requirement, or if the negligible 1 W TDP difference (64 W vs 65 W) matters in an extremely power-sensitive build. Otherwise, the benchmark results leave no room for doubt: the GTX 650 is the stronger GPU.

Specification Differences

The following specifications differ between the two cards, with the GTX 650's values listed first and the GT 740's second:

• Chip: GK106 vs GK107
• Generation: GeForce 600 vs GeForce 700
• Transistors: 2,540 million vs 1,270 million
• Die Size: 221 mm² vs 118 mm²
• Transistor Density: 11.5M / mm² vs 10.8M / mm²
• Memory Clock: 1250 MHz (5 Gbps effective) vs 1253 MHz (5 Gbps effective)
• Bandwidth: 80.00 GB/s vs 80.19 GB/s
• Pixel Rate: 8.464 GPixel/s vs 7.944 GPixel/s
• Texture Rate: 33.86 GTexel/s vs 31.78 GTexel/s
• FP32: 812.5 GFLOPS vs 762.6 GFLOPS
• TDP: 65 W vs 64 W
• Display Outputs: 1x DVI, 2x DisplayPort 1.2 vs 2x DVI, 1x mini-HDMI 1.4a
• Release Date: 2013-11-26 vs 2014-05-28
• Predecessor: GeForce 500 vs GeForce 600
• Successor: GeForce 700 vs GeForce 900
• Launch MSRP: None vs 89 USD
• Dimensions: 147 mm (5.8 inches) vs 145 mm (5.7 inches)