NVIDIA GeForce GTX 670MX vs NVIDIA Quadro K620M Comparison

# Head-to-Head Benchmarks

The only direct comparison available between these two mobile GPUs is the Geekbench OpenCL test, and the results are remarkably close. The NVIDIA GeForce GTX 670MX scores 6125, while the NVIDIA Quadro K620M trails slightly at 5957, putting the GeForce card ahead by 2.7%. While that delta is small in absolute terms, it is consistent with the broader pattern of the GTX 670MX being the stronger compute performer. The Quadro K620M's OpenCL score of 5957 places it within 0.2% of the AMD Radeon HD 8750M (5970) and just 0.4% behind the NVIDIA Quadro K4000 (5982). Meanwhile, the GTX 670MX's 6125 result puts it 2.1% ahead of the NVIDIA Quadro M500M (5604) and 0.5% above the AMD Radeon HD 8790M (5691).

It is worth remembering the GTX 670MX also has a Vulkan benchmark score of 5316, which the Quadro K620M lacks entirely. This suggests the GeForce card offers broader API coverage for modern workloads, though the absence of a comparable Vulkan score for the K620M makes direct head-to-head comparison impossible in that test.

The overall percentile rankings tell a similar story. The GTX 670MX sits at the 33rd percentile of all GPUs, while the K620M ranks at the 34th percentile. That one-percentile difference is essentially negligible, indicating both cards occupy the same performance tier relative to the entire GPU landscape. However, when averaging across all available benchmarks, the GTX 670MX pulls ahead more decisively: its average benchmark score of 5721 is dragged down by the Vulkan result, but its OpenCL score remains the stronger of the two.

# FAQ

Q: Which GPU has the higher raw compute throughput?

A: The GTX 670MX delivers 1,153.9 GFLOPS of FP32 performance versus 863.2 GFLOPS for the K620M, a gap of roughly 34%. This aligns with its 2.7% OpenCL benchmark advantage, though the compute advantage is far larger than the benchmark delta suggests.

Q: Are these cards from the same architecture generation?

A: No. The K620M uses the Maxwell architecture (chip GM108S, 28 nm process), while the GTX 670MX uses the older Kepler architecture (chip GK104, also 28 nm). The K620M is listed under the "Quadro Kepler-M (Kx200M)" generation label despite its Maxwell core, while the GTX 670MX belongs to the GeForce 600M generation.

Q: How do the memory subsystems compare?

A: The GTX 670MX has 3 GB of GDDR5 memory on a 192-bit bus, delivering 67.20 GB/s of bandwidth. The K620M has 2 GB of DDR3 memory on a 64-bit bus, yielding just 16.02 GB/s. That is a 4.2x bandwidth advantage for the GTX 670MX.

Q: Which card has more shading units and texture units?

A: The GTX 670MX packs 960 shading units, 80 TMUs, and 24 ROPs. The K620M has 384 shading units, 16 TMUs, and 8 ROPs. The GeForce card has 2.5x the shaders, 5x the TMUs, and 3x the ROPs.

Q: What is the power consumption difference?

A: The K620M has a 30 W TDP, while the GTX 670MX is rated at 75 W. That is a 2.5x power draw for the GeForce card, which makes sense given its much larger silicon and memory interface.

Q: Which card supports Vulkan?

A: Both support Vulkan, but with different versions. The K620M supports Vulkan 1.4, while the GTX 670MX supports Vulkan 1.2.175. Both support DirectX 12 (11_0) and OpenGL 4.6.

# Where Each One Wins

The GTX 670MX wins in almost every performance-related category that matters for gaming and general compute. Its OpenCL score is 2.7% higher, its FP32 throughput is 33.7% higher, and its memory bandwidth is over four times greater. The 192-bit bus and GDDR5 memory make it the clear choice for bandwidth-hungry workloads like high-resolution textures or large compute buffers. Its 960 shading units and 80 TMUs also give it substantial advantages in pixel-heavy and texture-heavy rendering tasks. The Vulkan score of 5316 further indicates that modern graphics APIs run well on this card, which is relevant for games that leverage Vulkan for lower overhead.

The Quadro K620M's wins are more modest but still meaningful. Its 28 nm process node and TSMC foundry are identical to the GTX 670MX, but it achieves higher clock speeds: 1029 MHz base and 1124 MHz boost versus 601 MHz for both base and boost on the GeForce card. That higher clock rate partially compensates for its smaller core, but not enough to overcome the GeForce's massive resource advantage. The K620M also has a higher transistor density at 13.2M per mm² versus 12.0M per mm², reflecting the efficiency gains of Maxwell over Kepler. For users who prioritize power efficiency, the K620M's 30 W TDP makes it far more suitable for thin-and-light laptops where thermal headroom is scarce.

# Specification Differences

The two cards differ substantially across nearly every specification. The K620M uses the GM108S chip with 1,020 million transistors on a 77 mm² die, while the GTX 670MX uses the GK104 chip with 3,540 million transistors on a 294 mm² die. That is over three times the transistor count and nearly four times the die area for the GeForce card. Transistor density slightly favors the K620M at 13.2M/mm² versus 12.0M/mm².

Clock speeds are the one area where the K620M leads decisively. It runs at 1029 MHz base and 1124 MHz boost, while the GTX 670MX is locked at 601 MHz for both. Memory clocks also differ: the K620M's memory runs at 1001 MHz (2 Gbps effective) versus 700 MHz (2.8 Gbps effective) for the GTX 670MX.

Memory configuration is another major differentiator. The K620M has 2 GB of DDR3 on a 64-bit bus, while the GTX 670MX has 3 GB of GDDR5 on a 192-bit bus. Bandwidth is 16.02 GB/s versus 67.20 GB/s, respectively. The bus interface also differs: the K620M uses MXM-A (3.0), while the GTX 670MX uses PCIe 3.0 x16. The slot width for the K620M is listed as MXM Module, while the GTX 670MX has no listed slot width.

Both cards have no power connectors and portable-device-dependent display outputs. The K620M's TDP is 30 W versus 75 W for the GTX 670MX. Release dates also differ significantly: the K620M launched on 2015-02-28, while the GTX 670MX launched on 2012-09-30, nearly two and a half years earlier.

# Architecture Differences

The architectural gap between these two GPUs is substantial. The K620M is built on Maxwell (chip GM108S), which is a newer architecture than the Kepler-based GK104 in the GTX 670MX. Despite the K620M's generation label reading "Quadro Kepler-M (Kx200M)," the underlying silicon is Maxwell. This explains its higher transistor density and superior clock scaling at lower power.

The GTX 670MX uses the GK104 chip, which is a large, power-hungry Kepler design. It has 3,540 million transistors spread across 294 mm², giving it a dense array of shading units, TMUs, and ROPs. Kepler was designed for maximum throughput at the cost of efficiency, which is why the GTX 670MX consumes 2.5x the power of the K620M despite being on the same 28 nm process.

Cache hierarchies, ray tracing cores, and tensor cores are not listed for either card, so no comparison can be made there. However, the FP16 performance is also absent for both, meaning their half-precision compute capabilities are unknown from this data. The K620M supports Vulkan 1.4, while the GTX 670MX supports Vulkan 1.2.175, suggesting the newer Maxwell architecture has a more mature Vulkan driver stack. Both support DirectX 12 (11_0) and OpenGL 4.6, so API compatibility is broadly similar for older workloads.

The K620M's predecessor is the Quadro Fermi-M and its successor is the Quadro Maxwell-M, indicating a direct lineage in the professional mobile space. The GTX 670MX's predecessor is the GeForce 500M and successor is the GeForce 700M, placing it in the consumer mobile gaming line.

# The Verdict

The data paints a clear picture: the NVIDIA GeForce GTX 670MX is the more powerful GPU in almost every measurable way. Its 2.7% OpenCL benchmark lead, 33.7% higher FP32 throughput, and 4.2x memory bandwidth make it the obvious choice for gaming, rendering, or any compute-heavy task where performance matters more than power draw. The 960 shading units versus 384, 80 TMUs versus 16, and 24 ROPs versus 8 are overwhelming advantages that no amount of clock speed can overcome. The 3 GB GDDR5 frame buffer with a 192-bit bus is also far more future-proof than 2 GB of DDR3 on a 64-bit bus.

The Quadro K620M, however, is not without its merits. Its 30 W TDP makes it suitable for laptops where battery life and thermal output are critical concerns. Its higher boost clock of 1124 MHz and newer Maxwell architecture give it better transistor density and Vulkan 1.4 support, which could matter for specific professional applications that leverage newer driver features. But the benchmark data shows it is essentially matched by the AMD Radeon HD 8730M (5955, 0% delta) and only slightly ahead of the Intel UHD Graphics 730 (5929, 0.5% delta), placing it firmly in entry-level territory.

For a buyer choosing between these two end-of-life mobile GPUs, the decision comes down to workload and power constraints. If the laptop will be used for gaming, 3D modeling, or GPU-accelerated compute, the GTX 670MX is the only rational choice. Its Vulkan support, though older version, still enables modern gaming APIs, and its massive memory bandwidth advantage is decisive. If the laptop must run cool and quiet on battery power, and the workload is limited to light productivity or legacy OpenGL applications, the K620M's efficiency and higher clocks make it a defensible pick. But for anyone who values raw performance, the GTX 670MX wins outright — the data does not leave much room for debate.