GPU Comparison

The NVIDIA GeForce 920M and NVIDIA GeForce 820M are two entry-level mobile graphics solutions from the same vendor, separated by a single generation. The data shows a clear overall winner in the 920M, which posts an average benchmark score of 3287 compared to the 820M’s 2784. This places the 920M at the 20th percentile of all GPUs, while the 820M sits at the 18th, indicating both are firmly in the low-end segment, but the 920M holds a measurable edge in raw compute performance.

Where Each One Wins

The head-to-head comparison is one-sided. The NVIDIA GeForce 920M wins the only shared benchmark, Geekbench OpenCL, with a score of 3725 against the 820M’s 2784, a 33.8% advantage. This means the 920M is the superior choice for any workload that leverages OpenCL acceleration, such as video encoding, image processing, or general-purpose GPU compute tasks.

The 820M, by contrast, does not win any benchmark outright. Its only recorded score is the Geekbench OpenCL result of 2784, which lags behind the 920M by that same 33.8% margin. However, the 820M is not without its own merits. Its nearest rival list includes the NVIDIA GeForce RTX 3060 Ti GDDR6X, with an average score of 2795 and a delta of -0.4%, meaning the 820M is essentially on par with that much newer and higher-tier card in this specific benchmark context. This is an anomaly worth noting: the 820M’s performance profile, while low, aligns with a surprisingly broad range of hardware. For a user with a laptop that only has an 820M, the takeaway is that it can handle light, older, or less demanding tasks, but it will not excel at anything compute-heavy.

Architecture Differences

The two GPUs are built on different physical designs despite both being 28 nm parts from TSMC. The 920M uses the GK208B chip with a die size of 87 mm² and 1,020 million transistors, yielding a transistor density of 11.7M per mm². The 820M uses the larger GK107 chip, measuring 118 mm² with 1,270 million transistors, but a lower density of 10.8M per mm².

Architecturally, the 920M is classified as Kepler 2.0, while the 820M is original Kepler. This generational distinction matters. The 920M’s newer architecture allows it to achieve higher clock speeds, 954 MHz base and boost, whereas the 820M has no recorded base or boost clock in the data. The 920M also has a higher pixel rate of 7.632 GPixel/s and texture rate of 30.53 GTexel/s, compared to the 820M’s 6.480 GPixel/s and 25.92 GTexel/s. The 920M’s FP32 compute is rated at 732.7 GFLOPS, versus 622.1 GFLOPS for the 820M. These figures confirm that the 920M is not just a rebadge; it is a more efficient and faster implementation of the same basic idea.

The 820M compensates somewhat with its memory subsystem. It has a 128-bit memory bus width, double the 920M’s 64-bit bus, and as a result, its memory bandwidth is 28.83 GB/s versus 14.40 GB/s for the 920M. Both have 2 GB of DDR3 memory, but the 820M can feed its shading units data roughly twice as fast.

Head-to-Head Benchmarks

The only direct comparison available is the Geekbench OpenCL test. Here, the 920M delivers a score of 3725, while the 820M scores 2784. The delta is 33.8%, which is a substantial gap for two GPUs with the same core count and memory type. This result indicates that the 920M’s higher clock speeds and newer architecture more than compensate for the 820M’s wider memory bus.

To put this in perspective, look at how each GPU fares against its own nearest rivals. The 920M’s average score of 3287 is within 0.9% of the GeForce GT 730M (3316) and 1.4% of the Intel HD Graphics 530 (3332). It is 2.4% faster than the GeForce GT 640 (3210) and 3% slower than the Intel HD Graphics P4600 (3389). This places the 920M in a tight cluster of mid-2010s integrated and entry-level discrete GPUs.

The 820M’s average score of 2784 is a different story. It is 0.4% slower than the GeForce RTX 3060 Ti GDDR6X (2795) and 0.6% slower than the AMD Radeon 550 (2801). It is 1.2% faster than the Intel UHD Graphics 610 (2752) and 3.5% faster than the GeForce RTX 5070 SUPER (2690). The presence of modern RTX cards in the 820M’s rival list is a quirk of the benchmark database, but it underscores how low the 820M’s absolute performance is; it is competing with iGPUs from the same era, not with dedicated gaming hardware.

Specification Differences

The two cards differ in several key specification fields. The chip is GK208B for the 920M and GK107 for the 820M. The architecture is Kepler 2.0 versus Kepler. The generation differs: GeForce 900M for the 920M, GeForce 800M for the 820M. The process node is the same 28 nm, but the die size and transistor count differ, as noted above.

Clock speeds are another differentiator. The 920M has a base and boost clock of 954 MHz. The 820M has no listed clocks, which means its operating frequency is unspecified in the data. The memory clock is effectively the same, 1800 Mbps vs 1802 Mbps effective, but the bus width differs (64-bit vs 128-bit), leading to a bandwidth difference of 14.40 GB/s vs 28.83 GB/s.

The render output units (ROPs) differ: 8 for the 920M, 16 for the 820M. This is notable, as the 820M’s higher ROP count should theoretically improve fill-rate-bound scenarios, yet its pixel rate is still lower (6.480 vs 7.632 GPixel/s) due to the clock deficit. The 920M has a TDP of 33 W and is an IGP (integrated graphics processor) slot width, while the 820M has a TDP of 45 W and uses an MXM Module slot width. The bus interface also differs: PCIe 3.0 x8 for the 920M versus PCIe 3.0 x16 for the 820M. Both are portable-device dependent for display outputs and share the same API support: DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175.

The shading units and TMUs are identical at 384 and 32, respectively. Both have 2 GB of DDR3 memory. Neither has RT cores or tensor cores, and neither has a listed launch MSRP. The 920M’s predecessor is the GeForce 800M and its successor is the GeForce 10 Mobile, while the 820M’s predecessor is the GeForce 700M and its successor is the GeForce 900M, which is the 920M’s own generation.

FAQ

Q: Which GPU is faster in OpenCL compute?

A: The NVIDIA GeForce 920M is significantly faster, scoring 3725 in Geekbench OpenCL compared to the 820M’s 2784, a 33.8% advantage.

Q: Do these GPUs support the same modern APIs?

A: Yes, both support DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.2.175, so there is no difference in API feature level.

Q: Does the 820M have any hardware advantage over the 920M?

A: Yes, the 820M has a 128-bit memory bus versus the 920M’s 64-bit, giving it 28.83 GB/s of bandwidth compared to 14.40 GB/s. It also has 16 ROPs versus 8.

Q: How does the 920M compare to its nearest rivals?

A: The 920M’s average score of 3287 is within 1.4% of the Intel HD Graphics 530 (3332) and within 0.9% of the GeForce GT 730M (3316). It is 2.4% faster than the GeForce GT 640 (3210).

Q: What is the performance class of the 820M based on its rivals?

A: The 820M’s average score of 2784 puts it within 1.2% of the Intel UHD Graphics 610 (2752) and within 0.4% of the GeForce RTX 3060 Ti GDDR6X (2795), indicating its performance is at the very bottom of the discrete GPU spectrum.

Q: Which GPU has a lower power draw?

A: The 920M has a TDP of 33 W, while the 820M has a TDP of 45 W, making the 920M the more power-efficient option.

The Verdict

The choice is straightforward if you are selecting between these two based on the available data. The NVIDIA GeForce 920M is the stronger performer in every measurable compute benchmark. It wins the only head-to-head test by 33.8%, has a higher clock speed (954 MHz vs unspecified), higher FP32 throughput (732.7 GFLOPS vs 622.1 GFLOPS), and higher pixel and texture rates. It also consumes less power (33 W vs 45 W) and is built on a newer architecture revision (Kepler 2.0). The 920M is the default pick for any workload that depends on shader compute.

The 820M is not a competitive alternative. Its only advantage is memory bandwidth (28.83 GB/s vs 14.40 GB/s) and ROP count (16 vs 8), which could theoretically help in memory-bound or fill-rate-bound scenarios. However, the data shows no benchmark where this translates into a win. Its lower average score (2784 vs 3287) and lower percentile ranking (18th vs 20th) confirm that it is the weaker card overall.

For a user with an 820M-equipped laptop, the data suggests that it is adequate only for basic tasks. It sits at the same performance level as integrated graphics like the Intel UHD Graphics 610, so expect similar frame rates in light gaming or general usage. For anyone with a choice, the 920M is the only rational option, offering a substantial compute lead without any penalty in power efficiency. The 820M should be considered a legacy part that is outperformed by its successor across the board.