NVIDIA GeForce GTX 670M vs NVIDIA Quadro K620 Comparison

# Head-to-Head Benchmarks

The single direct comparison available in the database places the NVIDIA Quadro K620 ahead of the NVIDIA GeForce GTX 670M in Geekbench OpenCL compute performance. The K620 scores 6693 points against the GTX 670M’s 6513 points, a margin of 2.7% in favor of the workstation card. While the percentage gap is modest, it is consistent across the compute workload tested, indicating that the Maxwell-based K620 maintains a measurable edge in raw OpenCL throughput over the older Fermi-based GTX 670M.

Contextualizing these results against their nearest rivals clarifies the competitive positioning. The GTX 670M’s score of 6513 places it within a tight cluster of similarly performing mobile and integrated solutions. It sits a mere 0.3% above the NVIDIA GeForce GT 555M (6493 points) and 0.5% above the NVIDIA Quadro M5000M (6481 points). The AMD Radeon Vega 10 Mobile trails by 0.6% at 6476 points, while the Intel UHD Graphics P750 leads the GTX 670M by 0.6% at 6554 points. This grouping suggests that the GTX 670M’s compute performance is essentially at parity with a broad swath of mid-range mobile GPUs from multiple generations and vendors.

The K620’s 6693-point OpenCL result, by contrast, places it in a different neighborhood. Its nearest rivals include the NVIDIA GeForce RTX 5070 Ti SUPER and NVIDIA GeForce RTX 4070 Ti SUPER AD102, both averaging 6270 points — meaning the K620 sits 0.2% above those modern enthusiast-class cards in this particular benchmark. On the AMD side, the Radeon R7 M350 (6327 points) trails by 0.7%, and the Radeon Pro WX 4100 (6330 points) trails by 0.8%. It is important to note that these rival scores come from a single average benchmark figure per GPU, and the K620’s OpenCL score of 6693 is its stronger result; its Vulkan score of 5870 is considerably lower.

The deltaPct values reveal an interesting asymmetry. The GTX 670M’s rivals are all within a 1.2-percentage-point band, underscoring how tightly packed the mid-range mobile landscape was at its release. The K620’s nearest rivals span a slightly wider range, from -0.8% to +0.2%, but the K620 still emerges as the leader of its immediate cluster. In terms of overall percentile ranking, the GTX 670M holds a 38th percentile position across all GPUs, while the K620 sits at the 36th percentile — a slight reversal of the head-to-head outcome, suggesting that the GTX 670M’s single score is more representative of its overall standing relative to the full GPU population.

FAQ

Q: Which GPU has the higher Geekbench OpenCL score?

A: The NVIDIA Quadro K620 scores 6693 points in Geekbench OpenCL, while the NVIDIA GeForce GTX 670M scores 6513 points. The K620 leads by 2.7% in this head-to-head benchmark.

Q: How does the GTX 670M compare to its nearest rival, the GeForce GT 555M?

A: The GTX 670M scores 6513 points, which is 0.3% higher than the GeForce GT 555M’s average score of 6493 points. This places the two GPUs at near parity in compute performance.

Q: What is the K620’s strongest benchmark result?

A: The K620’s strongest recorded benchmark is its Geekbench OpenCL score of 6693 points. Its Geekbench Vulkan score of 5870 points is noticeably lower, indicating that compute performance varies significantly depending on the API used.

Q: Which GPU has the better overall percentile ranking?

A: The GeForce GTX 670M holds a 38th percentile ranking across all GPUs, while the Quadro K620 sits at the 36th percentile. Despite the K620 winning the head-to-head OpenCL test, the GTX 670M ranks slightly higher in the overall distribution.

Q: How does the K620 perform against modern RTX-class GPUs?

A: In the average benchmark score used for rival comparisons, the K620 at 6282 points sits 0.2% above both the RTX 5070 Ti SUPER and the RTX 4070 Ti SUPER AD102, each averaging 6270 points. This is a narrow margin but still places the older workstation card ahead in the aggregate.

Q: What are the approximate margins between the GTX 670M and its closest competitors?

A: The GTX 670M’s nearest rivals are extremely close: it leads the GT 555M by 0.3%, the Quadro M5000M by 0.5%, and the Radeon Vega 10 Mobile by 0.6%, while trailing the Intel UHD Graphics P750 by 0.6%. All five GPUs fall within a 1.2-point-percentage range.

Where Each One Wins

The NVIDIA GeForce GTX 670M wins in the context of overall GPU population ranking. Its 38th percentile placement surpasses the K620’s 36th percentile, indicating that the GTX 670M’s single benchmark score of 6513 positions it better relative to the full historical GPU landscape. This is notable given that the GTX 670M is an older, mobile-oriented part from the GeForce 600M generation, yet it still holds a slight edge in distribution-wide standing.

The NVIDIA Quadro K620 wins the direct compute benchmark comparison. Its Geekbench OpenCL score of 6693 outperforms the GTX 670M’s 6513 by 2.7%, making it the clear choice for workloads that rely on OpenCL compute acceleration. The K620 also demonstrates broader API support, with a recorded Vulkan score of 5870, whereas the GTX 670M has no Vulkan benchmark result in the database. For users targeting Vulkan-based compute tasks, the K620 is the only one of the two with verified performance data.

In terms of nearest-rival dominance, the K620 also holds a stronger position. Its average score of 6282 leads its closest competitors by margins of 0.2% (RTX 5070 Ti SUPER and RTX 4070 Ti SUPER AD102) and sits well above the Radeon R7 M350 (-0.7%) and Radeon Pro WX 4100 (-0.8%). The GTX 670M, meanwhile, is effectively tied with its nearest rivals, with all deltas under 1% and a mix of wins and losses. The K620’s rival cluster shows it as the clear leader of its group, while the GTX 670M is merely one member of a tightly bunched pack.

For compute density, the K620’s higher FP32 throughput of 863.2 GFLOPS versus the GTX 670M’s 803.7 GFLOPS reinforces its advantage in raw arithmetic performance. This 59.5 GFLOPS difference, combined with the higher pixel rate of 17.98 GPixel/s versus 8.372 GPixel/s, suggests the K620 is better suited to tasks that stress rasterization throughput and shader math simultaneously.

Specification Differences

The two GPUs diverge significantly across nearly every major specification category. The GTX 670M uses a 192-bit memory bus paired with 1536 MB of GDDR5 memory, delivering 72.00 GB/s of bandwidth. The K620 uses a narrower 128-bit bus with 2 GB of DDR3 memory, resulting in just 28.80 GB/s of bandwidth — less than half the GTX 670M’s figure. This is a substantial difference that favors the GTX 670M in memory-bound workloads, despite the K620 having a larger capacity.

Clock speeds tell the opposite story. The GTX 670M’s memory runs at 750 MHz (3 Gbps effective), while the K620’s memory operates at 900 MHz (1800 Mbps effective). The K620 also has explicit base and boost clocks of 1058 MHz and 1124 MHz, respectively, whereas the GTX 670M has no listed base or boost clocks, only its memory clock. This means the K620’s core operating frequencies are documented and verifiable, while the GTX 670M’s are not specified in the data.

The compute machinery differs in configuration as well. The GTX 670M packs 336 shading units, 56 texture mapping units, and 24 raster output units. The K620 has more shading units at 384 but fewer TMUs (24) and ROPs (16). This yields a higher texture rate for the GTX 670M at 33.49 GTexel/s versus the K620’s 26.98 GTexel/s, but a much higher pixel rate for the K620 at 17.98 GPixel/s versus 8.372 GPixel/s. The FP32 throughput slightly favors the K620 at 863.2 GFLOPS versus 803.7 GFLOPS.

Physical and interface characteristics also differ. The GTX 670M is an MXM Module with an MXM-B (3.0) bus interface, while the K620 is a single-slot card with a PCIe 2.0 x16 interface. The K620 has defined dimensions of 160 mm in length and 69 mm in height, while the GTX 670M has no listed dimensions. The K620 also specifies a suggested PSU of 200 W, whereas the GTX 670M lists no suggested PSU. Display outputs are portable-device-dependent for the GTX 670M, while the K620 offers 1x DVI and 1x DisplayPort 1.2.

Architecture Differences

The architectural divide between these two GPUs is generational and profound. The GTX 670M is built on NVIDIA’s Fermi 2.0 architecture using the GF114 chip, fabricated on a 40 nm process at TSMC. The K620 uses the Maxwell architecture with the GM107 chip, manufactured on a 28 nm process, also at TSMC. This process shrink from 40 nm to 28 nm is the primary driver of the K620’s improved efficiency and higher clock potential.

The chip layouts reflect the process difference. The GF114 die measures 332 mm² and contains 1,950 million transistors, yielding a transistor density of 5.9 million per square millimeter. The GM107 die is dramatically smaller at 148 mm² but packs a similar 1,870 million transistors, achieving a density of 12.6 million per square millimeter — more than double the GTX 670M’s density. This density advantage allows the K620 to deliver comparable transistor counts in a much smaller footprint, which is reflected in its lower 45 W TDP versus the GTX 670M’s 75 W.

The generation labels further contextualize the gap. The GTX 670M belongs to the GeForce 600M series, with the GeForce 500M as its predecessor and the GeForce 700M as its successor. The K620 is categorized under the Quadro Kepler (Kx200) generation, despite using Maxwell architecture — an apparent naming anomaly in the database. Its predecessor is listed as Quadro Fermi and its successor as Quadro Maxwell, which suggests the K620 sits at a transitional point in NVIDIA’s workstation lineup.

API support shows both similarities and differences. Both GPUs support DirectX 12 (11_0) and OpenGL 4.6. However, the K620 adds Vulkan 1.4 support, while the GTX 670M has no Vulkan version listed. This is a meaningful distinction for modern compute and graphics applications that leverage Vulkan, as the K620 has verified compatibility while the GTX 670M’s Vulkan status is unconfirmed. The K620’s Vulkan benchmark score of 5870 further validates its functional Vulkan implementation.

Memory technology differences are rooted in the architectures’ design priorities. The GTX 670M pairs GDDR5 with a 192-bit bus for high bandwidth, whereas the K620 uses DDR3 on a 128-bit bus, trading bandwidth for capacity and lower power. The GTX 670M’s 72.00 GB/s bandwidth is 2.5 times the K620’s 28.80 GB/s, a direct consequence of the wider bus and faster memory type. The K620’s 2 GB capacity, however, exceeds the GTX 670M’s 1536 MB, offering more headroom for large datasets that fit within slower memory.

The TDP difference of 30 W (75 W for GTX 670M versus 45 W for K620) is notable for deployment scenarios. The GTX 670M’s MXM form factor and portable-device-dependent outputs indicate a mobile-first design, while the K620’s single-slot, PCIe 2.0 x16 form factor and fixed display outputs point to a desktop workstation orientation. These architectural and physical differences define two distinct use cases despite their comparable compute scores.