GPU Comparison

The Verdict

The benchmark data presents a close contest between two end-of-life mobile and desktop graphics solutions. The NVIDIA GeForce 920M wins the only direct head-to-head comparison available, scoring 3725 in Geekbench OpenCL against the GT 545's 3594, a 3.5% advantage. However, the average benchmark score tells a different story: the GT 545 averages 3594 across its tested workload, while the 920M averages 3287 when including its Vulkan result of 2849. This discrepancy suggests the 920M's OpenCL win is not representative of overall compute consistency. Buyers should pick the GT 545 for a desktop system with a dedicated slot and higher sustained throughput in OpenCL-centric tasks. The 920M is the choice for portable devices where power draw is critical, as its 33 W TDP is less than half the GT 545's 70 W, and it offers a larger 2 GB frame buffer. Strictly from the data, the 920M edges out the GT 545 in raw OpenCL compute, but the GT 545 holds a higher percentile ranking at 21 versus the 920M's 20, indicating slightly better standing against all GPUs.

Architecture Differences

The two cards represent distinct architectural generations from NVIDIA. The GT 545 is built on the GF116 chip using the Fermi 2.0 architecture, fabricated on a 40 nm process at TSMC. It packs 1,170 million transistors onto a 238 mm² die, yielding a transistor density of 4.9M / mm². In contrast, the 920M uses the GK208B chip with the newer Kepler 2.0 architecture, manufactured on a more efficient 28 nm process. Despite fewer transistors at 1,020 million, the 920M achieves a much higher density of 11.7M / mm² thanks to its compact 87 mm² die size. This architectural shift brings tangible differences: the 920M features 384 shading units versus the GT 545's 144, and 32 texture mapping units compared to 24. The GT 545 counters with 16 ROPs, double the 920M's 8. The 920M also supports Vulkan 1.2.175, a feature entirely absent from the GT 545's API list. Both cards share DirectX 12 (11_0) and OpenGL 4.6 support, but the older Fermi architecture lacks the Vulkan capability that Kepler 2.0 provides.

Head-to-Head Benchmarks

The single direct comparison available is the Geekbench OpenCL test, where the 920M outperforms the GT 545 by 3.5%, scoring 3725 against 3594. This margin is modest but consistent with the 920M's higher FP32 throughput of 732.7 GFLOPS compared to the GT 545's 414.7 GFLOPS. The 920M also leads in pixel rate with 7.632 GPixel/s versus 4.320 GPixel/s, and texture rate at 30.53 GTexel/s against 17.28 GTexel/s. However, the GT 545 shows resilience in average performance. Its average benchmark score of 3594 exactly matches its OpenCL result, while the 920M's average drops to 3287 when factoring in its Vulkan score of 2849. This indicates the 920M's Vulkan performance is significantly weaker than its OpenCL capability, pulling its overall average down by 12% relative to its OpenCL mark. The GT 545's nearest rivals provide context: it sits just 0.1% behind the RTX 5000 Mobile Ada Generation and 0.6% behind the GT 735M, while remaining ahead of the GTX 1050 by 1%. The 920M's nearest rivals show it trailing the GT 730M by 0.9% and the Intel HD Graphics 530 by 1.4%, but leading the GT 640 by 2.4%. These deltas position both cards in the lower quartile of GPU performance, with the GT 545 at the 21st percentile and the 920M at the 20th.

Specification Differences

The two cards diverge across nearly every hardware specification. The GT 545 uses a 192-bit memory bus with 1536 MB of DDR3 memory, delivering 38.40 GB/s of bandwidth. The 920M narrows the bus to 64-bit but doubles capacity to 2 GB, resulting in a much lower bandwidth of 14.40 GB/s. Memory clocks differ as well: the GT 545 runs at 800 MHz (1600 Mbps effective), while the 920M operates at 900 MHz (1800 Mbps effective). Power requirements are dramatically different, with the GT 545 drawing 70 W and requiring a 250 W PSU, while the 920M consumes just 33 W and lists no PSU recommendation. The GT 545 is a single-slot card measuring 145 mm (5.7 inches) with display outputs of 1x DVI, 1x HDMI 1.3a, and 1x VGA. The 920M is an integrated graphics processor (IGP) with portable-device-dependent outputs and no listed dimensions. Bus interfaces differ: the GT 545 uses PCIe 2.0 x16, while the 920M uses PCIe 3.0 x8. The GT 545's base clock is not listed, but the 920M has a fixed base and boost clock of 954 MHz. Release dates place the GT 545 on 2011-05-13 and the 920M on 2015-03-12, with the GT 545 carrying a launch MSRP of 149 USD. The GT 545's predecessor is the GeForce 400 series with successor GeForce 600, while the 920M succeeds the GeForce 800M and precedes GeForce 10 Mobile.

FAQ

Q: Which card has higher raw compute throughput?

A: The GeForce 920M leads in FP32 performance with 732.7 GFLOPS compared to the GT 545's 414.7 GFLOPS, a substantial 76.7% advantage in theoretical compute.

Q: How do their memory systems compare?

A: The GT 545 offers a 192-bit bus with 38.40 GB/s bandwidth, while the 920M uses a 64-bit bus with only 14.40 GB/s. The 920M compensates with double the capacity at 2 GB versus 1536 MB.

Q: What is the power consumption difference?

A: The 920M is far more efficient, drawing 33 W against the GT 545's 70 W. This makes the 920M suitable for integrated portable designs, while the GT 545 requires a dedicated 250 W PSU.

Q: Which card supports Vulkan?

A: Only the GeForce 920M supports Vulkan, version 1.2.175. The GT 545 has no Vulkan support listed, though both cards share DirectX 12 (11_0) and OpenGL 4.6 support.

Q: How do their average benchmark scores differ?

A: The GT 545 averages 3594, while the 920M averages 3287. The 920M's average is dragged down by its Vulkan score of 2849, which is 23.5% lower than its OpenCL score of 3725.

Q: What do the percentile rankings indicate?

A: The GT 545 sits at the 21st percentile of all GPUs, slightly ahead of the 920M's 20th percentile. Both are low-tier performers, with the GT 545's nearest rival being the RTX 5000 Mobile Ada Generation at -0.1% delta, while the 920M's closest competitor is the GT 730M at -0.9%.

Where Each One Wins

The GeForce 920M wins in scenarios prioritizing power efficiency and modern API support. Its 33 W TDP makes it a clear choice for thin-and-light laptops or compact systems where thermal and power budgets are tight. The 2 GB frame buffer provides more headroom for texture-heavy workloads at lower resolutions, and the Vulkan 1.2.175 support enables compatibility with newer graphics APIs that the GT 545 cannot access. In OpenCL compute, the 920M holds a 3.5% edge, and its higher pixel rate of 7.632 GPixel/s suggests better fill-rate-bound performance.

The GT 545 wins in situations requiring memory bandwidth and sustained average performance. Its 38.40 GB/s bandwidth is 2.7 times higher than the 920M's 14.40 GB/s, which benefits memory-intensive operations like high-resolution texture streaming or large data sets. The 192-bit bus provides a wider data path for multi-tasking workloads. The GT 545's average score of 3594 exceeds the 920M's 3287 by 9.3%, indicating more consistent performance across different benchmark types. Its 16 ROPs, double the 920M's count, improve efficiency in pixel-heavy rendering. The GT 545 also offers a single-slot form factor with discrete display outputs (1x DVI, 1x HDMI 1.3a, 1x VGA), making it a drop-in solution for desktop systems with a PCIe 2.0 x16 slot. For users prioritizing OpenCL consistency and bandwidth over portability, the GT 545 is the data-backed choice. For users needing integrated graphics with lower power draw and newer API support, the 920M is the winner.