NVIDIA GeForce 920MX vs NVIDIA GeForce GT 745M Comparison
NVIDIA GeForce 920MX
GeForce GT 745M
PERFORMANCE BENCHMARKS
Analysis: NVIDIA GeForce 920MX vs NVIDIA GeForce GT 745M
# Head-to-Head Benchmarks: NVIDIA GeForce GT 745M vs NVIDIA GeForce 920MX
The data places the GeForce GT 745M and the GeForce 920MX in a close contest. In the OpenCL benchmark, the GT 745M records a score of 3580 against 4068 for the 920MX, a 12% deficit. The Vulkan test flips the result. The GT 745M scores 5502, while the 920MX manages 2987, a commanding 84.2% lead for the older card. This split outcome highlights how API-specific performance can be. The GT 745M's score of 421.6 GFLOPS FP32 and its 17.57 GTexel/s texture rate place it ahead in raw throughput. The data shows the 920MX is not without merit: its OpenCL result of 4068 is 13.6% above the GT 745M's 3580. However, the GT 745M's Vulkan result of 5502 versus 2987 for the 920MX is the single largest gap in the comparison. The GT 745M also leads in the Geekbench Vulkan test with a 5502 score, while the 920MX trails with 2987. Overall, the average benchmark score for the GT 745M is 3953, slightly above the 920MX's 3528. The percentile ranking shows the GT 745M at 23rd and the 920MX at 21st among all GPUs, respectively.
The Verdict
The GeForce GT 745M is the stronger choice for tasks that leverage the Vulkan API. Its 84.2% lead in the Vulkan benchmark is decisive. For OpenCL workloads, the 920MX is the faster card, with a 13.6% advantage. The GT 745M also counters with a higher FP32 rate of 421.6 GFLOPS and a pixel rate of 4.392 GPixel/s. The 920MX counters with a higher FP32 rate of 508.4 GFLOPS and a pixel rate of 7.944 GPixel/s. The choice depends on the software ecosystem. Vulkan-centric workloads favor the GT 745M. OpenCL-centric workloads favor the 920MX.
Architecture Differences
The GT 745M is built on the Kepler architecture with the GK107 chip, while the 920MX uses the Maxwell architecture with the GM108S chip. The GT 745M uses a 28 nm process at TSMC, packing 1,270 million transistors on a 118 mm² die, with a density of 10.8 million transistors per mm². The 920MX uses a 28 nm process at TSMC with 1,020 million transistors on a 77 mm² die, resulting in a density of 13.2 million transistors per mm². The GT 745M has 384 shading units, 32 TMUs, and 16 ROPs. The 920MX is configured with 256 shading units, 24 TMUs, and 8 ROPs. Both cards come with 2 GB of memory, but the GT 745M uses GDDR5 on a 128-bit bus for 64.00 GB/s of bandwidth. The 920MX uses DDR3 on a 64-bit bus for 14.40 GB/s. Clock behavior differs: the GT 745M runs at a 1000 MHz memory clock, while the 920MX runs at 900 MHz. GT 745M's memory clock supports a 1000 MHz effective 4 Gbps rate. The 920MX's memory is rated at 1800 Mbps effective.
FAQ
Q: Which card wins in Vulkan?
A: The NVIDIA GeForce GT 745M. It scores 5502 in the Vulkan test, 84.2% higher than the 2987 scored by the GeForce 920MX.
Q: Which card is stronger in OpenCL?
A: The NVIDIA GeForce 920MX. Its OpenCL score of 4068 is 13.6% higher than the GT 745M's 3580.
Q: What is the difference in FP32 compute?
A: The GeForce 920MX has a higher FP32 rate at 508.4 GFLOPS, beating the GT 745M's 421.6 GFLOPS.
Q: Which GPU has the higher pixel fill rate?
A: The GeForce 920MX with 7.944 GPixel/s. The GT 745M achieves 4.392 GPixel/s.
Q: What memory type does each use?
A: The GT 745M uses 2 GB of GDDR5 on a 128-bit bus. The 920MX uses 2 GB of DDR3 on a 64-bit bus.
Q: Which card has the higher transistor density?
A: The GeForce 920MX, with 13.2 million transistors per mm², higher than the 10.8 million per mm² of the GT 745M.
Where Each One Wins
The GT 745M is the pick in Vulkan-based workloads. It also takes the texture rate crown with 17.57 GTexel/s. Its 384 shading units and 32 TMUs give it an edge in geometry-heavy tasks. The 920MX wins in OpenCL, FP32, pixel rate, and transistor density. It has the higher pixel rate at 7.944 GPixel/s, and the FP32 crown at 508.4 GFLOPS. It is also the chip to choose for OpenCL and FP32 compute. The GT 745M is the Vulkan and texture-rate pick. The 920MX is the OpenCL and FP32 pick.
Specification Differences
In direct specification comparison, the two cards differ in several key areas. The GT 745M uses the Kepler architecture (GK107 chip), while the 920MX uses Maxwell (GM108S chip). The GT 745M packs 1,270 million transistors on a 118 mm² die, while the 920MX has 1,020 million on a 77 mm² die, giving the 920MX a higher density at 13.2M/mm². Clock speeds: the 920MX has a higher base clock of 965 MHz versus the GT 745M's base. The 920MX also boosts to 993 MHz. The memory configs differ: GDDR5 128-bit at 1000 MHz for the GT 745M (64 GB/s bandwidth) versus DDR3 64-bit at 900 MHz for the 920MX (14.4 GB/s). The 920MX has more TMUs at 32 versus 24, but the GT 745M has more ROPs at 16 versus 8. The GT 745M is 384 shaders, the 920MX 256. Both are PCIe 3.0 x16 or x8, respectively. The GT 745M supports DirectX 12 (11_0), OpenGL 4.6, Vulkan 1.2.175. The 920MX also supports DirectX 12 (11_0), OpenGL 4.6, Vulkan 1.4. The GT 745M uses a 128-bit GDDR5 bus; the 920MX uses a 64-bit DDR3 bus. The 920MX is a MXM module, the GT 745M is an IGP. They share the same 28 nm process node. The 920MX has a higher FP64 at 508.4 GFLOPS. The 920MX is rated at a 16 W TDP, while the GT 745M is rated at 45 W.