GPU Comparison

Head-to-Head Benchmarks

The only shared benchmark between these two mobile GPUs is Geekbench OpenCL, and the result is decisive. The NVIDIA GeForce 920MX scores 4,068, while the NVIDIA GeForce 820A scores 2,983. That is a 26.7% advantage for the 920MX, a substantial gap that will be felt in any compute-oriented workload. The 820A trails by more than a quarter, which is a meaningful margin for a class of chips typically found in entry-level laptops.

Looking at the broader benchmark landscape, the 820A's average score of 2,983 places it in the 19th percentile of all GPUs. Its nearest rivals in the database are the GeForce GTX 860M at 2,967 (0.5% behind), the GeForce GTX 750 Ti at 2,953 (1% behind), and the Quadro P600 at 2,923 (2.1% behind). This clustering is notable: the 820A is essentially on par with those desktop and mobile parts, despite being an integrated-style solution. The gap to the RTX 4060 Ti 8 GB is only 2.4%, which shows that in this particular OpenCL test, the 820A holds its own against much newer hardware.

The 920MX, by contrast, averages 3,528 across its two benchmark results (OpenCL at 4,068 and Vulkan at 2,987). That places it in the 21st percentile of all GPUs. Its nearest rivals are the GeForce GT 545 at 3,594 (1.8% ahead), the RTX 5000 Mobile Ada Generation at 3,596 (1.9% ahead), and the GeForce GT 735M at 3,616 (2.4% ahead). The 920MX is within striking distance of those parts, and it beats the Quadro 2000M by 2.7% (3,434). The 920MX's Vulkan score of 2,987 is nearly identical to the 820A's entire OpenCL score, which tells you the 920MX has a more modern feature set to go with its raw speed advantage.

The win count is simple: the 920MX takes the sole head-to-head benchmark, so it leads 1–0. There is no test where the 820A comes out ahead. That does not mean the 820A is useless, but on pure performance, the 920MX is the clear winner. The 26.7% delta in OpenCL is the single most important number in this comparison. For any task that uses GPU compute, video encoding, image processing, or even some lighter rendering work, the 920MX will finish the job noticeably faster.

The Verdict

From the data, the NVIDIA GeForce 920MX is the stronger part. It wins the only shared benchmark by 26.7%, has a higher average score (3,528 vs. 2,983), and sits in a higher percentile (21st vs. 19th). If you are choosing between these two for a laptop, the 920MX is the one to get. The performance gap is large enough to matter in real-world use, and the Vulkan support adds future-proofing that the 820A simply lacks.

The 820A is not without merit. Its OpenCL score of 2,983 puts it right next to the GTX 860M and GTX 750 Ti, which are respectable parts. In a vacuum, the 820A is a capable entry-level GPU. But in a head-to-head, it loses. The 920MX's 256 shading units and higher clocks give it a decisive edge in raw throughput, and the numbers confirm it.

Who should pick the 820A? Only if you are constrained by a system that specifically requires the 820A's IGP form factor (integrated graphics package) and you cannot use a MXM module. The 820A is an IGP, while the 920MX is an MXM module, so physical compatibility may dictate the choice. If your laptop accepts either, the 920MX is the obvious pick.

Who should pick the 920MX? Anyone who wants the better-performing mobile GPU. The 26.7% lead in OpenCL is too large to ignore. The 920MX also has double the memory (2 GB vs. 1 GB) and a higher transistor density (13.2M per mm² vs. 5.0M per mm²), which speaks to a more efficient design. For gaming, light content creation, or general GPU acceleration, the 920MX is the superior option.

FAQ

Q: Which GPU is faster in OpenCL?

A: The NVIDIA GeForce 920MX scores 4,068 in Geekbench OpenCL, while the NVIDIA GeForce 820A scores 2,983. That is a 26.7% advantage for the 920MX.

Q: Does the 820A beat the 920MX in any benchmark?

A: No. In the only shared head-to-head test (Geekbench OpenCL), the 920MX wins. The 820A has zero head-to-head wins.

Q: What is the average benchmark score for each GPU?

A: The 820A averages 2,983 across its single OpenCL result. The 920MX averages 3,528 across its OpenCL (4,068) and Vulkan (2,987) results.

Q: How do these GPUs compare to their nearest rivals?

A: The 820A is within 2.4% of the RTX 4060 Ti 8 GB (2,913) and within 0.5% of the GTX 860M (2,967). The 920MX is within 2.4% of the GT 735M (3,616) and beats the Quadro 2000M (3,434) by 2.7%.

Q: Which GPU has Vulkan support?

A: The 920MX supports Vulkan 1.4. The 820A has no listed Vulkan support, with its API list capped at DirectX 12 (11_0) and OpenGL 4.6.

Q: What is the memory difference between the two?

A: The 820A has 1,024 MB of DDR3 on a 64-bit bus with 16.02 GB/s bandwidth. The 920MX has 2 GB of DDR3 on a 64-bit bus with 14.40 GB/s bandwidth. The 920MX has more capacity but slightly lower bandwidth.

Specification Differences

The two GPUs differ in nearly every major specification. The 820A uses the GF117 chip with a Fermi 2.0 architecture, fabricated on a 28 nm process by TSMC. It packs 585 million transistors on a 116 mm² die, giving a transistor density of 5.0M per mm². The 920MX uses the GM108S chip with a Maxwell architecture, also on 28 nm TSMC, but with 1,020 million transistors on a much smaller 77 mm² die, yielding a density of 13.2M per mm². That is a 2.6x density improvement, reflecting the architectural leap from Fermi to Maxwell.

The shader counts tell a similar story. The 820A has 96 shading units, 16 texture mapping units, and 8 ROPs. The 920MX has 256 shading units, 24 TMUs, and 8 ROPs. That is 2.7x more shading units and 1.5x more TMUs. The pixel rate jumps from 2.500 GPixel/s on the 820A to 7.944 GPixel/s on the 920MX. Texture rate goes from 10.00 GTexel/s to 23.83 GTexel/s. FP32 compute is 240.0 GFLOPS on the 820A versus 508.4 GFLOPS on the 920MX, more than double.

Memory clocks differ as well. The 820A runs at 1001 MHz (2 Gbps effective), while the 920MX runs at 900 MHz (1800 Mbps effective). Despite the lower clock, the 920MX's bandwidth is 14.40 GB/s versus 16.02 GB/s on the 820A, so the 820A actually has a slight bandwidth advantage. The 820A has 1 GB of memory; the 920MX has 2 GB. Both use DDR3 on a 64-bit bus.

The bus interface differs: the 820A uses PCIe 2.0 x16, while the 920MX uses PCIe 3.0 x8. The 920MX has a base clock of 965 MHz and a boost clock of 993 MHz; the 820A has no listed base or boost clocks. TDPs are close: 15 W for the 820A and 16 W for the 920MX. The 820A is an IGP (integrated graphics package), while the 920MX is an MXM module. Both have no power connectors and portable-device-dependent display outputs.

API support diverges: the 820A lists DirectX 12 (11_0) and OpenGL 4.6, with no Vulkan. The 920MX lists DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.4. Neither has RT cores or tensor cores. Neither has a launch MSRP listed.

Architecture Differences

The architectural gap here is generational. The 820A is built on Fermi 2.0, an architecture that dates back to 2010's GeForce 400 series. Fermi was designed for compute-heavy workloads but is inefficient by modern standards. The 920MX uses Maxwell, which was a major redesign focused on power efficiency and higher shader throughput per clock. The transistor counts reflect this: 585 million on Fermi versus 1,020 million on Maxwell, even though the 920MX's die is 39 mm² smaller.

The 28 nm process node is identical for both, so the difference is purely architectural. Maxwell's 13.2M transistors per mm² versus Fermi's 5.0M per mm² shows how much more densely Maxwell packs logic. This translates directly to the 2.7x shading unit advantage (256 vs. 96) and the 2.1x FP32 throughput (508.4 GFLOPS vs. 240.0 GFLOPS). Maxwell also brings Vulkan 1.4 support, which Fermi lacks entirely, the 820A is stuck with DirectX 12 (11_0) and OpenGL 4.6.

The 920MX's boost clock of 993 MHz is modest, but with 256 shaders, it generates far more work per clock than the 820A's 96 shaders. The 820A has no boost clock listed, which suggests it runs at a fixed rate, typical of older IGP designs. The 920MX's MXM module form factor also implies it is meant to be replaceable, while the 820A is soldered as an integrated part. That is a practical difference: the 920MX can be swapped, the 820A cannot.

Memory architecture is similar on paper, both use 64-bit DDR3, but the 920MX's 2 GB capacity is double the 820A's 1 GB. This matters for modern workloads that exceed 1 GB of VRAM. The 820A's slightly higher bandwidth (16.02 GB/s vs. 14.40 GB/s) is offset by the capacity deficit. In practice, the 920MX's larger frame buffer will be more useful.

Where Each One Wins

NVIDIA GeForce 920MX wins: Compute performance is the 920MX's domain. The 26.7% OpenCL lead is the headline number, and it extends to any GPU-accelerated task. Video encoding, image filters, and OpenCL-based physics will all run faster on the 920MX. The 256 shaders and 508.4 GFLOPS of FP32 throughput make it more than twice as fast as the 820A in raw math. The Vulkan 1.4 support also gives it a future-proofing edge for Vulkan-based games and applications, which the 820A cannot run. The 2 GB memory capacity is a win for larger textures and datasets. The higher pixel rate (7.944 GPixel/s vs. 2.500 GPixel/s) means faster fill-rate-bound operations like anti-aliasing or post-processing effects.

NVIDIA GeForce 820A wins: The 820A has one narrow technical advantage: bandwidth. Its 16.02 GB/s is 11% higher than the 920MX's 14.40 GB/s. In memory-bound workloads that are not capacity-limited, the 820A could theoretically edge out the 920MX. Its PCIe 2.0 x16 interface also provides more lanes than the 920MX's PCIe 3.0 x8, though the newer PCIe 3.0 standard on the 920MX likely negates that in practice. The 820A's lower TDP (15 W vs. 16 W) is a negligible power savings. Its IGP form factor means it is always present in system, requiring no additional module slot.

Practical split: If you need raw performance, the 920MX wins every meaningful category. The 820A's only wins are bandwidth and a slightly lower TDP, neither of which will be noticeable in real use. For legacy software that only uses OpenGL or DirectX 11, both are equal, they share the same DirectX 12 (11_0) and OpenGL 4.6 support. For anything modern or compute-heavy, the 920MX is the only choice. The 820A is best suited for basic display output and very light 2D workloads, while the 920MX can handle entry-level 3D gaming and accelerated productivity. The percentile ranking (21st vs. 19th) confirms the 920MX is a step up in the overall GPU hierarchy, even if both sit near the bottom of the performance stack.