AMD Radeon R7 350 vs NVIDIA GeForce 945M Comparison

# NVIDIA GeForce 945M vs AMD Radeon R7 350

The NVIDIA GeForce 945M and AMD Radeon R7 350 represent two competing approaches to mobile and compact graphics from the mid-2010s, built on different architectures with different design priorities. The GeForce 945M edges out the R7 350 in the single available head-to-head benchmark, but the R7 350 counters with a lower power draw and a faster memory subsystem. Neither card is a modern powerhouse — both sit near the 40th percentile of all GPUs — but the data reveals meaningful trade-offs between compute throughput, memory bandwidth, and thermal efficiency.

FAQ

Q: Which GPU wins in OpenCL performance?

A: The NVIDIA GeForce 945M scores 8099 in Geekbench OpenCL, while the AMD Radeon R7 350 scores 7792. That gives the 945M a 3.9% advantage in the head-to-head benchmark.

Q: How different are their memory bandwidth figures?

A: The R7 350 offers 72.00 GB/s of bandwidth with GDDR5 memory, while the 945M provides 28.80 GB/s with DDR3. The R7 350's bandwidth is exactly 2.5 times higher, a substantial gap that matters for texture-heavy workloads.

Q: What is the power consumption difference?

A: The R7 350 has a 55 W TDP, while the 945M is rated at 75 W. The AMD card draws 20 W less, which is a 26.7% reduction in power requirement.

Q: Which GPU has more shading units?

A: The 945M has 640 shading units, compared to 512 on the R7 350. The NVIDIA card also has 40 texture mapping units versus 32, though both have 16 ROPs.

Q: What are their transistor counts and die sizes?

A: The 945M uses 1,870 million transistors on a 148 mm² die, while the R7 350 packs 1,500 million transistors into 123 mm². Both are built on TSMC's 28 nm process, with very similar transistor densities of 12.6M/mm² and 12.2M/mm² respectively.

Q: Do both cards support the same APIs?

A: No. The 945M supports DirectX 12 (11_0) and Vulkan 1.4, while the R7 350 supports DirectX 12 (11_1) and Vulkan 1.2.170. Both support OpenGL 4.6. The R7 350 has a slightly higher DirectX feature level, but the 945M offers a more recent Vulkan version.

The Verdict

The data paints a clear picture for different use cases. If raw compute performance is the priority, the NVIDIA GeForce 945M is the pick — it leads in the only shared benchmark and delivers higher FP32 throughput at 1,305.6 GFLOPS versus 819.2 GFLOPS. That 59.3% compute advantage is substantial and explains its better OpenCL showing.

However, the AMD Radeon R7 350 makes a compelling case for power-constrained or bandwidth-sensitive scenarios. Its 55 W TDP versus 75 W means it generates less heat and suits smaller systems, while its 72.00 GB/s memory bandwidth is a massive 150% higher than the 945M's 28.80 GB/s. For workloads that are memory-bound rather than compute-bound, the R7 350's GDDR5 advantage could invert the performance relationship.

The 945M sits at the 42nd percentile of all GPUs, slightly above the R7 350's 40th percentile. Both are end-of-life products, so neither is a forward-looking investment. The choice hinges on whether the user values peak compute (945M) or efficiency and memory throughput (R7 350). The 945M also offers a more modern Vulkan implementation (1.4 vs 1.2.170), which could matter for newer Linux or Vulkan-based workloads.

Head-to-Head Benchmarks

Only one benchmark directly compares these two GPUs: Geekbench OpenCL. The NVIDIA GeForce 945M scores 8099, while the AMD Radeon R7 350 scores 7792. That's a 3.9% delta in favor of NVIDIA.

Contextualizing this result requires looking at their nearest rivals. The 945M's score sits between the NVIDIA GRID K2 (8080, 0.2% lower) and the AMD Radeon R9 M360 (8129, 0.4% higher). It also trails the NVIDIA GeForce GTX 950M (8135) by 0.4% and beats the GTX 650 Ti Boost (8067) by 0.4%. The 945M is essentially in a tight cluster with these cards, all within a 1% band.

The R7 350's 7792 OpenCL score places it against a different set of rivals. It edges out the Intel UHD Graphics 750 (7441) by 4.7% and the AMD Radeon HD 8850M (7447) by 4.6%. It beats the NVIDIA GeForce GTX 1650 (7472) by 4.3% and the Intel Arc A310 (7550) by 3.2%. This suggests the R7 350 is more clearly ahead of its nearest competitors than the 945M is of its own.

The 3.9% win for the 945M is modest but consistent with its higher compute specifications. The R7 350's lower FP32 throughput (819.2 vs 1,305.6 GFLOPS) likely explains its lower OpenCL score, since OpenCL workloads often scale with raw shader compute. The 945M's 640 shading units at 1020 MHz boost clock deliver more parallel execution capacity than the R7 350's 512 units, even though the R7 350's memory bandwidth could help in some access patterns.

Specification Differences

The two GPUs diverge on nearly every major specification except memory capacity and bus width. Both have 2 GB of memory on a 128-bit bus, but the memory type differs fundamentally: the 945M uses DDR3 at 900 MHz (1800 Mbps effective), while the R7 350 uses GDDR5 at 1125 MHz (4.5 Gbps effective). This yields the stark bandwidth difference of 28.80 GB/s versus 72.00 GB/s.

The compute configurations differ significantly. The 945M has 640 shading units, 40 TMUs, and 16 ROPs, producing a pixel rate of 16.32 GPixel/s and texture rate of 40.80 GTexel/s. The R7 350 has 512 shading units, 32 TMUs, and 16 ROPs, with lower pixel and texture rates of 12.80 GPixel/s and 25.60 GTexel/s respectively. FP32 performance follows the same pattern: 1,305.6 GFLOPS for NVIDIA versus 819.2 GFLOPS for AMD — a 59.3% gap.

Power and physical specifications also differ. The 945M is rated at 75 W TDP and uses an MXM Module form factor with an MXM-B (3.0) bus interface. The R7 350 is a 55 W single-slot card using PCIe 3.0 x16, measuring 168 mm (6.6 inches) in length, and recommends a 250 W PSU. Neither card requires power connectors, and both have no suggested PSU listed for the 945M.

Display outputs differ: the 945M's outputs are "Portable Device Dependent" since it's an MXM module, while the R7 350 offers 1x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2. The 945M has no length listed, reflecting its mobile-oriented MXM design.

Architecture Differences

The NVIDIA GeForce 945M is built on the Maxwell architecture using the GM107 chip, part of the GeForce 900M generation. It uses 1,870 million transistors on a 148 mm² die, produced by TSMC at 28 nm, yielding a transistor density of 12.6M/mm². It was released on 2015-10-26, succeeding the GeForce 800M and preceding the GeForce 10 Mobile series.

The AMD Radeon R7 350 uses the GCN 1.0 architecture with the Cape Verde chip, belonging to the Pirate Islands (R7 300) generation. It employs 1,500 million transistors on a smaller 123 mm² die, also at TSMC's 28 nm process, with a slightly lower density of 12.2M/mm². Its release date is later at 2016-07-05, with the Volcanic Islands series as predecessor and Arctic Islands as successor.

The architectural philosophy differs considerably. Maxwell emphasizes efficiency and higher clock speeds per shader, which explains the 945M's higher boost clock of 1020 MHz and its superior compute throughput despite having only 640 shading units. GCN 1.0, by contrast, focuses on massive parallelism through wider SIMD arrays, but the R7 350's 512 units at an unlisted base clock deliver less raw throughput.

Cache hierarchies and feature sets also diverge. The 945M supports Vulkan 1.4, indicating a more recently updated driver stack, while the R7 350 caps at Vulkan 1.2.170. DirectX support slightly favors AMD (12_1 vs 12_0 feature levels), but both are effectively limited to DirectX 11-era workloads in practice.

The transistor density difference (12.6M/mm² vs 12.2M/mm²) is negligible, suggesting both chips are similarly optimized for their process node. However, the 945M's larger die with 370 million more transistors provides the compute advantage seen in benchmarks.

Where Each One Wins

The NVIDIA GeForce 945M wins in compute-heavy scenarios. Its 59.3% higher FP32 throughput and 3.9% OpenCL benchmark lead make it the better choice for GPU compute tasks like OpenCL-based rendering, physics simulation, or any workload that scales with shader count and clock speed. Its higher texture rate (40.80 vs 25.60 GTexel/s) also benefits texture-filtering workloads. The 945M's Vulkan 1.4 support provides an edge for modern Vulkan applications that can leverage newer API features.

The AMD Radeon R7 350 wins in memory-bandwidth-sensitive workloads. Its 72.00 GB/s bandwidth is 2.5 times that of the 945M, which matters for tasks like large texture streaming, frame buffer operations, or compute kernels with high memory access per operation. The lower 55 W TDP makes it suitable for systems with tighter thermal budgets or smaller power supplies — it explicitly recommends a 250 W PSU, while the 945M's MXM form factor targets laptops where power is at a premium anyway.

The R7 350 also offers more flexible display connectivity with dedicated DVI, HDMI 1.4a, and DisplayPort 1.2 outputs, making it a drop-in desktop card. The 945M's display outputs depend entirely on the host device, limiting it to the laptop or MXM-compatible system it's installed in.

For gaming, the data is ambiguous. The 945M's higher compute and texture rates suggest better raw performance, but the R7 350's memory bandwidth could help at higher resolutions or with anti-aliasing, where bandwidth often becomes the bottleneck. The 945M's 42nd percentile versus the R7 350's 40th percentile suggests the NVIDIA card holds a slight overall edge, but the R7 350's GDDR5 memory makes it a stronger contender in scenarios that favor memory throughput over shader compute.

Ultimately, the 945M is the better all-rounder for compute and modern API support, while the R7 350 excels in efficiency, memory bandwidth, and desktop flexibility. Users prioritizing raw performance should choose NVIDIA; those needing low power consumption or high bandwidth should choose AMD.