AMD Radeon R8 M445DX vs NVIDIA GeForce 840M Comparison

Head-to-Head Benchmarks

The recorded database contains a single direct benchmark comparison between these two mobile graphics parts, and the result is decisive. In the Geekbench OpenCL test, the NVIDIA GeForce 840M scores 5,764 points, while the AMD Radeon R8 M445DX scores 4,727 points. That gives the NVIDIA part a 21.9% advantage, a substantial margin in the context of integrated and entry-level discrete-class mobile GPUs.

The 840M's lead is not a narrow edge; it is a clear performance tier separation. The GeForce 840M's average benchmark score across all recorded tests sits at 5,322, while the Radeon R8 M445DX averages 4,727. This means the NVIDIA part's typical performance is roughly 12.6% higher than the AMD part's only recorded score, even when including the 840M's additional Vulkan result of 4,880, which is lower than its OpenCL score.

Context from the nearest rivals in the database reinforces the 840M's position. The GeForce 840M sits within 0.7% of the NVIDIA GeForce 940M (which averages 5,284), within 0.3% of the much larger NVIDIA GeForce GTX 980M (5,308), and within 0.1% of the GeForce 930A (5,317). It trails the AMD Radeon R7 M445 (5,358) by only 0.7%. In other words, the 840M is clustered in a tight band of mid-tier mobile GPUs, none of which separate themselves by even a single percentage point.

The Radeon R8 M445DX, by contrast, sits in a lower performance neighborhood. Its nearest rival, the AMD Radeon R5 M335, scores 4,752, which is 0.5% higher. The NVIDIA Quadro P400 scores 4,684, which is 0.9% lower. The Radeon R5 M255 scores 4,788, 1.3% higher, and the NVIDIA GeForce RTX 3080 12 GB, a desktop-class part with a far higher performance ceiling, scores 4,791 in this particular workload, also 1.3% higher. The presence of the RTX 3080 12 GB in this rival list is an artifact of the benchmark's scoring, but it does not change the core finding: the R8 M445DX's OpenCL result places it in a lower percentile band.

The percentile data confirms the gap. The GeForce 840M ranks at the 31st percentile among all GPUs in the database, while the Radeon R8 M445DX ranks at the 27th percentile. That four-point percentile difference corresponds directly to the 21.9% OpenCL win for NVIDIA.

Architecture Differences

The two chips come from different architectural generations and fundamentally different design philosophies. The NVIDIA GeForce 840M uses the GM108S chip, built on the Maxwell architecture, and belongs to the GeForce 800M generation. The AMD Radeon R8 M445DX uses the Meso chip, built on the GCN 3.0 architecture, and belongs to the Gem System Hybrid (Rx M400) generation.

Both are fabricated on the same 28 nm process at TSMC, but the dies differ significantly in size and complexity. The NVIDIA chip has 1,020 million transistors on a 77 mm² die, yielding a transistor density of 13.2 million per square millimeter. The AMD chip has 1,550 million transistors on a 125 mm² die, yielding a density of 12.4 million per square millimeter. The AMD chip is physically larger and contains more transistors, but the NVIDIA chip packs them slightly more densely.

The compute configurations tell the opposite story. The GeForce 840M has 384 shading units, 16 texture mapping units, and 8 ROPs. The Radeon R8 M445DX has 320 shading units, 20 TMUs, and 8 ROPs. So NVIDIA fields more shader cores, AMD fields more texture units, and both have the same ROP count. These differences directly affect the theoretical rates: the 840M delivers 17.98 GTexel/s and 8.992 GPixel/s, while the R8 M445DX delivers 20.42 GTexel/s and 8.168 GPixel/s. AMD's higher TMU count gives it a 13.6% texture rate advantage, but NVIDIA's higher pixel rate gives it a 10.1% fillrate edge.

Clock speeds also differ. The GeForce 840M has a base clock of 1,029 MHz and a boost clock of 1,124 MHz. The Radeon R8 M445DX has a base clock of 780 MHz and a boost clock of 1,021 MHz. NVIDIA's higher clocks, combined with its larger shader count, produce a substantial FP32 throughput advantage: 863.2 GFLOPS versus 653.4 GFLOPS. That is a 32.1% gap in raw floating-point compute, which explains much of the OpenCL benchmark delta.

Memory architecture is a major divergence. The GeForce 840M has 2 GB of dedicated DDR3 memory on a 64-bit bus, with a bandwidth of 16.02 GB/s and a memory clock of 1001 MHz (2 Gbps effective). The Radeon R8 M445DX has no dedicated memory at all: its memory size, type, and bus width are all reported as "System Shared," and its bandwidth is "System Dependent." This is a critical structural difference. The NVIDIA part accesses local VRAM, while the AMD part relies on shared system memory, whose performance depends entirely on the host laptop's configuration.

The AMD part does support FP16 at a 1:1 ratio (653.4 GFLOPS), while the NVIDIA part lists no FP16 capability in the database. This could matter for certain compute workloads, but it did not translate into a benchmark win in the recorded OpenCL test.

The bus interfaces also differ. The GeForce 840M uses PCIe 3.0 x8, a standard external interface. The Radeon R8 M445DX is listed with a bus interface of "IGP," meaning it is an integrated graphics processor. Both are classified as "IGP" in slot width, which indicates they are not discrete cards installed in expansion slots, but rather chips mounted on the motherboard or packaged for mobile systems.

The Verdict

The data points in one direction: the NVIDIA GeForce 840M is the faster GPU in this pairing. Its 21.9% OpenCL lead is the only head-to-head result recorded, and it aligns with the raw compute specifications. The 840M's 863.2 GFLOPS FP32 throughput exceeds the R8 M445DX's 653.4 GFLOPS by 32.1%. Its pixel rate is 10.1% higher. Its dedicated 2 GB DDR3 memory with 16.02 GB/s bandwidth is a structural advantage over a "System Shared" configuration that offers no guaranteed bandwidth.

The Radeon R8 M445DX does have genuine strengths in the specification sheet. Its 20.42 GTexel/s texture rate is 13.6% higher than the 840M's 17.98 GTexel/s, which can help in texture-bound workloads. It supports FP16 at full rate, which the NVIDIA part does not list. It also supports DirectX 12 at the 12_0 feature level, while the 840M caps at DirectX 12 (11_0). Its Vulkan support is version 1.2.170 versus NVIDIA's 1.4.

But benchmark results indicate that these advantages did not materialize into a faster OpenCL score. The R8 M445DX's 4,727 points place it at the 27th percentile, while the 840M's 5,764 points place it at the 31st percentile. In the database's nearest-rival lists, the 840M competes with parts like the GeForce 940M and Radeon R7 M445, all within 0.7%. The R8 M445DX competes with parts like the Radeon R5 M335 and R5 M255, which are lower-tier mobile GPUs.

For a user choosing between two mobile GPUs based solely on recorded performance, the GeForce 840M is the safer pick. It delivers higher compute throughput, higher pixel fillrate, and dedicated video memory, and it wins the only benchmark that directly compares them.

FAQ

Q: Which GPU is faster in the recorded OpenCL benchmark?

A: The NVIDIA GeForce 840M scores 5,764 points versus 4,727 points for the AMD Radeon R8 M445DX, a 21.9% advantage.

Q: Does the AMD Radeon R8 M445DX have any specification advantage?

A: Yes. It has 20 TMUs versus 16, giving it a 20.42 GTexel/s texture rate versus 17.98 GTexel/s. It also supports FP16 at 653.4 GFLOPS (1:1) and DirectX 12 (12_0), while the 840M supports DirectX 12 (11_0).

Q: How much memory does each GPU have?

A: The GeForce 840M has 2 GB of dedicated DDR3 memory on a 64-bit bus with 16.02 GB/s bandwidth. The Radeon R8 M445DX has "System Shared" memory with "System Dependent" bandwidth.

Q: What are the clock speeds of the two GPUs?

A: The GeForce 840M runs at 1,029 MHz base and 1,124 MHz boost. The Radeon R8 M445DX runs at 780 MHz base and 1,021 MHz boost.

Q: Which GPU has more shading units?

A: The GeForce 840M has 384 shading units, while the Radeon R8 M445DX has 320.

Q: What is the percentile ranking of each GPU?

A: The GeForce 840M ranks at the 31st percentile among all GPUs in the database. The Radeon R8 M445DX ranks at the 27th percentile.

Where Each One Wins

The GeForce 840M wins in raw compute throughput. Its FP32 rate of 863.2 GFLOPS is 32.1% higher than the R8 M445DX's 653.4 GFLOPS. Its pixel rate of 8.992 GPixel/s is 10.1% higher. These figures directly support its 21.9% OpenCL benchmark victory.

The Radeon R8 M445DX wins in texture throughput. Its 20.42 GTexel/s is 13.6% higher than the 840M's 17.98 GTexel/s. This could make it preferable for workloads that are heavily dependent on texture sampling rather than general compute or pixel output.

The Radeon R8 M445DX also wins on API feature level. It supports DirectX 12 at the 12_0 feature level, while the 840M supports DirectX 12 at the 11_0 level. Its Vulkan version is 1.2.170, versus 1.4 for NVIDIA. The AMD part also lists FP16 support at 653.4 GFLOPS, which the NVIDIA part does not specify.

The GeForce 840M wins on memory architecture. Its dedicated 2 GB DDR3 pool with a fixed 16.02 GB/s bandwidth is a predictable resource. The R8 M445DX's "System Shared" memory has no fixed bandwidth and depends entirely on the host system, which introduces performance variability.

In terms of physical design, both are IGP-class parts with portable-device-dependent display outputs. The GeForce 840M uses a PCIe 3.0 x8 interface, while the Radeon R8 M445DX is listed as "IGP." Neither part requires external power connectors, and the 840M has a TDP of 33 W, while the R8 M445DX's TDP is not recorded in the database.

Specification Differences

| Field | NVIDIA GeForce 840M | AMD Radeon R8 M445DX |

|---|---|---|

| Chip | GM108S | Meso |

| Architecture | Maxwell | GCN 3.0 |

| Generation | GeForce 800M | Gem System Hybrid (Rx M400) |

| Transistors | 1,020 million | 1,550 million |

| Die Size | 77 mm² | 125 mm² |

| Transistor Density | 13.2M / mm² | 12.4M / mm² |

| Base Clock | 1,029 MHz | 780 MHz |

| Boost Clock | 1,124 MHz | 1,021 MHz |

| Memory Clock | 1001 MHz (2 Gbps effective) | System Shared |

| Memory Size | 2 GB | System Shared |

| Memory Type | DDR3 | System Shared |

| Memory Bus Width | 64 bit | System Shared |

| Memory Bandwidth | 16.02 GB/s | System Dependent |

| Shading Units | 384 | 320 |

| TMUs | 16 | 20 |

| ROPs | 8 | 8 |

| Pixel Rate | 8.992 GPixel/s | 8.168 GPixel/s |

| Texture Rate | 17.98 GTexel/s | 20.42 GTexel/s |

| FP32 | 863.2 GFLOPS | 653.4 GFLOPS |

| FP16 | Not listed | 653.4 GFLOPS (1:1) |

| TDP | 33 W | Not recorded |

| Bus Interface | PCIe 3.0 x8 | IGP |

| DirectX | 12 (11_0) | 12 (12_0) |

| OpenGL | 4.6 | 4.6 |

| Vulkan | 1.4 | 1.2.170 |

| Release Date | 2014-03-11 | 2016-05-14 |

| Production Status | End-of-life | End-of-life |