NVIDIA GeForce GTX 680M vs NVIDIA GeForce GTX 880M Comparison

Head-to-Head Benchmarks

The recorded data for these two mobile GPUs shows a split decision across the two available benchmark tests. In Geekbench Metal, the NVIDIA GeForce GTX 880M posts a score of 10,458, while the NVIDIA GeForce GTX 680M scores 4,815. This gives the 880M a decisive advantage of 117.2% in that test, more than doubling the older card's result. The margin is substantial and suggests a major generational leap in the compute workloads that Metal accelerates.

However, the Geekbench OpenCL test tells a completely different story. Here, the GTX 680M scores 9,230, while the GTX 880M scores 5,622. The delta is 39.1% in favor of the 680M. This is a remarkable reversal, indicating that the 680M outperforms its successor by a wide margin in this specific OpenCL workload. The data shows a clear 1-1 split in the head-to-head comparison, with each card winning one test by a large margin.

The average benchmark score across both tests, however, favors the 880M. Its average is 8,040, compared to 7,023 for the 680M. This places the 880M in the 42nd percentile of all GPUs in the database, while the 680M sits in the 39th percentile. The average score reflects a combined view of the two tests, and it shows the 880M ahead by approximately 14.5% overall. Yet the individual test results are so divergent that a simple average masks the true nature of their performance profiles. One is not uniformly better; each has a clear domain of strength.

Where Each One Wins

The GTX 880M is the clear winner in Metal-based applications. Its score of 10,458 versus 4,815 represents a 117.2% advantage, so any workload that leverages Apple's Metal API will strongly favor the newer card. The 880M also wins in the overall average benchmark score, which is the database's aggregate metric. This makes it the more consistent performer when both tests are considered together.

The GTX 680M is the winner in OpenCL compute workloads. Its score of 9,230 versus 5,622 is a 39.1% advantage, meaning the older architecture is significantly faster in OpenCL. This is counterintuitive given the 880M's newer release date, but the measured data is unambiguous. For applications that rely on OpenCL, the 680M delivers more raw throughput.

Comparing the 880M to its nearest rivals in the database, its average score of 8,040 is essentially equal to the NVIDIA Quadro P5000 at 8,039 (0% delta), slightly below the GeForce GTX 650 Ti at 8,053 (0.2% slower), and roughly 0.5% behind the GTX 650 Ti Boost at 8,067. For the 680M, its average of 7,023 is 0.2% below the NVIDIA T600 at 7,035, 0.7% above the AMD Radeon R5 M240 at 6,975, and 1.1% above the GeForce GTX 675M at 6,946. It also trails the GeForce GTX 970 by 1.9%. These deltas show both cards sitting in a crowded mid-range field, with neither offering a dominant overall edge over its immediate peers.

Architecture Differences

Both GPUs are built on the same fundamental architecture: Kepler. Both use the GK104 chip, fabricated by TSMC on a 28 nm process node. The transistor counts are identical at 3,540 million, and the die size matches at 294 mm². This means the transistor density is exactly the same, 12.0M per mm², for both parts. There is no architectural shift between them; they are the same chip family, just configured differently.

The GTX 880M is the higher-tier configuration of the GK104 chip. It has 1,536 shading units, 128 texture mapping units (TMUs), and 32 raster output units (ROPs). The GTX 680M has fewer of each: 1,344 shading units, 112 TMUs, and the same 32 ROPs. The 880M therefore has 192 more shaders and 16 more TMUs, which directly contributes to its higher compute and texture throughput.

Clock speeds also differ significantly. The 880M runs at a base clock of 954 MHz and a boost clock of 993 MHz. The 680M runs at 719 MHz base and 758 MHz boost. This is a roughly 25% higher base clock for the 880M, and a similar margin at boost. The combination of more cores and higher clocks explains why the 880M has substantially higher theoretical throughput: its pixel rate is 31.78 GPixel/s versus 21.22 GPixel/s, and its texture rate is 127.1 GTexel/s versus 84.90 GTexel/s. Floating point performance (FP32) is 3.050 TFLOPS for the 880M versus 2.038 TFLOPS for the 680M.

The memory subsystem also differs. Both have a 256-bit bus, but the 880M uses 8 GB of GDDR5 at 1250 MHz (5 Gbps effective), yielding a bandwidth of 160.0 GB/s. The 680M uses 4 GB of GDDR5 at 900 MHz (3.6 Gbps effective), yielding 115.2 GB/s. The 880M therefore has both double the capacity and roughly 40% more bandwidth.

Specification Differences

The two cards differ in several key specification fields. The memory size is different: 8 GB for the 880M versus 4 GB for the 680M. The memory clock is different as well: the 880M runs at 1250 MHz (5 Gbps effective) while the 680M runs at 900 MHz (3.6 Gbps effective). This leads to the bandwidth difference of 160.0 GB/s versus 115.2 GB/s.

The shading unit count differs: 1,536 for the 880M versus 1,344 for the 680M. The TMU count also differs: 128 versus 112. The clock speeds differ substantially: the 880M has a base of 954 MHz and boost of 993 MHz, while the 680M has a base of 719 MHz and boost of 758 MHz. The pixel rate is different (31.78 GPixel/s versus 21.22 GPixel/s), as is the texture rate (127.1 GTexel/s versus 84.90 GTexel/s). The FP32 compute is different: 3.050 TFLOPS versus 2.038 TFLOPS. The TDP is also different: the 880M is rated at 122 W, while the 680M is rated at 100 W.

The release dates differ as well. The 880M was released on March 11, 2014, while the 680M was released on June 3, 2012. The predecessor and successor fields also differ: the 880M follows the GeForce 700M and precedes the GeForce 900M, while the 680M follows the GeForce 500M and precedes the GeForce 700M. The generation field confirms this: the 880M is from the GeForce 800M series, the 680M from the GeForce 600M series. The process node, foundry, die size, transistor count, architecture, chip, bus interface, display outputs, slot width, power connectors, and API support (DirectX 12 with 11_0 feature level, OpenGL 4.6, Vulkan 1.2.175) are all identical.

FAQ

Q: Which GPU is faster in the overall average benchmark score?

A: The GeForce GTX 880M has a higher average score of 8,040, compared to 7,023 for the GTX 680M. This places the 880M in the 42nd percentile of all GPUs, while the 680M sits in the 39th percentile.

Q: What is the biggest performance gap between the two cards?

A: In the Geekbench Metal test, the 880M outscores the 680M by 117.2%, with 10,458 points versus 4,815. This is the largest margin in either direction.

Q: Does the GTX 680M win any benchmark?

A: Yes, the 680M wins the Geekbench OpenCL test comfortably, scoring 9,230 versus 5,622 for the 880M, a 39.1% advantage.

Q: How do the memory capacities compare?

A: The 880M has 8 GB of GDDR5 memory, double the 4 GB found on the 680M. Both use a 256-bit bus, but the 880M's memory runs at a higher effective speed of 5 Gbps versus 3.6 Gbps.

Q: Are these two cards based on the same chip?

A: Yes, both are based on the GK104 chip with the Kepler architecture, produced on a 28 nm process at TSMC. They have the same 3,540 million transistors and the same 294 mm² die size.

Q: What are the TDP ratings for each?

A: The GTX 880M has a TDP of 122 W, while the GTX 680M has a TDP of 100 W. The 880M uses more power, corresponding to its higher clock speeds and more cores.

The Verdict

From the recorded data, the GTX 880M is the better choice for anyone prioritizing Metal performance or needing more memory capacity and bandwidth. Its shader core count and clock speed are both higher, which aligns with its dominant Metal score and its higher overall average benchmark. The 880M is also the only one of the two with 8 GB of memory, which is more future-proof for larger textures or datasets.

However, the GTX 680M is not merely an obsolete part. Its OpenCL score is far superior, and for any application that uses OpenCL, it will be the faster card. Its lower TDP of 100 W also makes it more power-efficient in raw terms, and its lower average score of 7,023 still places it within a few percentage points of its immediate rivals. The 680M's higher OpenCL performance is a niche advantage, but it is a clear one.

The overall average benchmark score favors the 880M, and the 880M is also the more modern part, released in 2014 versus 2012. Given that the 880M wins the aggregate metric and the Metal test decisively, it is the more versatile performer. The 680M should only be chosen if the workload is specifically known to be OpenCL-heavy, where it has a 39.1% lead. For general use and for Metal support, the GTX 880M is the stronger GPU. The data does not support the 680M as a better overall card; it only supports it as a specialist OpenCL performer.