NVIDIA GeForce 930A vs NVIDIA Quadro K620 Comparison

The NVIDIA Quadro K620 beats the NVIDIA GeForce 930A in every recorded head-to-head measurement, and the margin is not subtle. In the one benchmark where the database holds results for both cards, Geekbench OpenCL, the Quadro K620 scores 6693 against 5317 for the GeForce 930A, a 25.9 percent advantage. Both are end-of-life Maxwell parts built by NVIDIA on TSMC's 28 nm process, both carry 2 GB of DDR3 memory, and both went on to occupy the lower stretches of the historical performance tables, with the K620 sitting at the 36th percentile against all GPUs in the database and the 930A at the 31st. The story of this matchup is straightforward: identical shading hardware on paper, but the K620 pairs it with a wider memory subsystem, more render outputs, and higher clocks, and the results follow.

Head-to-Head Benchmarks

The only direct comparison in the recorded data is Geekbench OpenCL. The Quadro K620 posts 6693; the GeForce 930A posts 5317. That is a gap of 25.9 percent, and it belongs entirely to the K620, which also wins the aggregate comparison one benchmark to zero.

The size of that gap is worth unpacking, because the raw shader counts suggest the two cards should be closer. Both chips expose 384 shading units and 24 texture mapping units. The divergence comes from everything around that core configuration. The K620 runs at a base clock of 1058 MHz with a boost of 1124 MHz, while the 930A runs at 928 MHz base and 941 MHz boost. The K620 feeds its shaders through a 128-bit memory bus delivering 28.80 GB/s of bandwidth; the 930A makes do with a 64-bit bus and 16.02 GB/s. Compute throughput reflects the combination: 863.2 GFLOPS FP32 for the K620 versus 722.7 GFLOPS for the 930A, roughly a 19 percent edge in theoretical shading performance.

Yet the measured OpenCL gap exceeds the theoretical compute gap by a clear margin. That points to memory bandwidth as the deciding factor in real workloads. With 28.80 GB/s against 16.02 GB/s, the K620 holds nearly 80 percent more bandwidth, and OpenCL workloads that stream data through the GPU will feel that difference far more than they feel a clock-speed delta measured in tens of megahertz. The recorded 25.9 percent victory sits comfortably between the compute advantage and the bandwidth advantage, which is exactly the pattern expected when both resources constrain performance.

The K620 also holds a second recorded benchmark the 930A lacks entirely: a Geekbench Vulkan score of 5870. There is no Vulkan result for the 930A in the database, so no direct comparison is possible there, but it is notable that the K620's Vulkan result lands within striking distance of its OpenCL result, indicating consistent performance across the two compute interfaces rather than a result skewed by one favorable API.

Context from each card's neighborhood reinforces the verdict. The K620's average benchmark score of 6282 places it alongside the AMD Radeon R7 M350 at 6327 and the AMD Radeon Pro WX 4100 at 6330, both marginally ahead, within a percentage point. The 930A's average of 5317 puts it in a cluster with the NVIDIA GeForce 840M at 5322, the NVIDIA GeForce GTX 980M at 5308, and the NVIDIA GeForce 940M at 5284, all effectively even. These are two different performance tiers in the database's historical rankings, roughly a thousand points of average benchmark score apart.

Where Each One Wins

On pure benchmark evidence, the Quadro K620 wins everywhere the database has measurements. It is the only card of the pair with a recorded head-to-head victory, it holds the higher average benchmark score (6282 versus 5317), and it sits five percentile points higher against the full GPU population (36 versus 31). For any workload represented by these Geekbench compute tests, general-purpose GPU compute, OpenCL-accelerated applications, and by extension any graphics task that scales with pixel throughput, the K620 is the stronger card.

The pixel pipeline difference is dramatic. The K620's 16 render outputs and 1124 MHz boost clock produce a pixel fill rate of 17.98 GPixel/s. The 930A, with 8 ROPs and a 941 MHz boost, manages 7.528 GPixel/s. That is close to a factor of 2.4 in favor of the K620, which matters for any workload bounded by pixel output rather than shading math. Texture throughput is closer, 26.98 GTexel/s for the K620 against 22.58 GTexel/s for the 930A, since both share the same 24 TMUs, but the K620 still leads by a comfortable margin on the strength of its clocks.

The 930A's case rests on platform traits rather than measured performance. It is an integrated-solution part with an IGP form factor, a 33 W thermal envelope, and no power connectors, against the K620's 45 W single-slot discrete card. Its display output is listed as portable-device dependent, which marks it as a mobile-oriented solution rather than a workstation card. Neither of these traits translates into a benchmark win, and the database records none for the 930A in this pairing, but they define where the 930A physically fits: inside machines where a discrete single-slot card is not the design target.

Architecture Differences

Both GPUs belong to NVIDIA's Maxwell architecture and both are fabricated by TSMC on a 28 nm process, so the architectural generation is a wash. The differences are in implementation scale.

The K620 uses the GM107 chip with 1870 million transistors on a 148 mm² die, yielding a density of 12.6 million transistors per square millimeter. The 930A uses the smaller GM108 chip with 1020 million transistors on a 77 mm² die, at a slightly higher density of 13.2 million transistors per square millimeter. The K620's die is nearly twice the area and carries roughly 83 percent more transistors, spending that budget on the doubled memory interface, the doubled ROP count, and the wider memory controller that its 128-bit bus requires.

The K620's 2 GB of DDR3 runs at an effective 1800 Mbps across 128 bits for its 28.80 GB/s of bandwidth. The 930A's 2 GB of DDR3 runs at a faster effective 2 Gbps, but across only 64 bits, which caps bandwidth at 16.02 GB/s. Faster memory per pin cannot compensate for half the pins.

Bus interface is another split: the K620 connects over PCIe 2.0 x16 while the 930A uses PCIe 3.0 x8. Feature support is otherwise identical, with DirectX 12 (11_0 feature level), OpenGL 4.6, and Vulkan 1.4 listed for both. Neither card has RT cores or tensor cores, as expected for Maxwell-era silicon. The K620's professional orientation shows in its fixed display outputs, one DVI and one DisplayPort 1.2, versus the 930A's device-dependent output. The K620 also carries a suggested power supply figure of 200 W in the database, while the 930A has none listed, consistent with its integrated classification. On the generational tree, the K620 belongs to the Quadro Kx200 line, successor to Quadro Fermi and predecessor of Quadro Maxwell, released July 2014; the 930A belongs to the GeForce 900A series, successor to the GeForce 800A, released March 2015.

FAQ

Q: Which GPU is faster in benchmarks?

A: The Quadro K620. It wins Geekbench OpenCL 6693 to 5317, a 25.9 percent margin, and holds the higher average score, 6282 versus 5317.

Q: Both cards have 384 shading units, so why is the K620 faster?

A: Clocks and memory. The K620 boosts to 1124 MHz versus 941 MHz, and its 128-bit bus delivers 28.80 GB/s against the 930A's 16.02 GB/s on a 64-bit bus. It also has 16 ROPs versus 8.

Q: Do they support the same graphics APIs?

A: Yes. Both list DirectX 12 (11_0), OpenGL 4.6, and Vulkan 1.4.

Q: How much power does each draw?

A: The K620 has a 45 W TDP; the 930A has a 33 W TDP. Neither requires power connectors.

Q: How do they rank against all GPUs in the database?

A: The K620 sits at the 36th percentile; the 930A sits at the 31st.

Q: Are these cards still in production?

A: No. Both are listed as end-of-life.

Specification Differences

• Chip: GM107 (K620) versus GM108 (930A)
• Transistors: 1870 million versus 1020 million
• Die size: 148 mm² versus 77 mm²; density 12.6M/mm² versus 13.2M/mm²
• Base / boost clock: 1058 / 1124 MHz versus 928 / 941 MHz
• Memory: 1800 Mbps effective on a 128-bit bus (28.80 GB/s) versus 2 Gbps effective on a 64-bit bus (16.02 GB/s); both 2 GB DDR3
• ROPs: 16 versus 8
• Pixel rate: 17.98 GPixel/s versus 7.528 GPixel/s
• Texture rate: 26.98 GTexel/s versus 22.58 GTexel/s
• FP32 compute: 863.2 GFLOPS versus 722.7 GFLOPS
• TDP: 45 W versus 33 W
• Form factor: single-slot discrete card versus IGP
• Bus: PCIe 2.0 x16 versus PCIe 3.0 x8
• Outputs: 1x DVI, 1x DisplayPort 1.2 versus portable-device dependent
• Dimensions: 160 mm long, 69 mm high (K620) versus unspecified (930A)
• Release: July 2014 versus March 2015

The Verdict

For anyone choosing strictly on the recorded data, the Quadro K620 is the pick. It wins the only shared benchmark by 25.9 percent, posts higher average scores, ranks five percentile points higher against the full database, and doubles the 930A's render output and memory bandwidth. Its fixed DisplayPort 1.2 and DVI outputs and its single-slot 45 W envelope describe a compact workstation card that fits machines the IGP-class 930A was never intended to challenge on throughput.

The 930A's appeal exists only outside the benchmark table: a smaller GM108 die, a 33 W envelope, an integrated form factor, and a device-dependent display path that ties it to portable platforms. Where the K620 is a discrete card you install, the 930A is a solution that ships inside a machine. If the decision is purely about GPU capability, the data gives the K620 a clean sweep, and the size of the OpenCL gap, driven by bandwidth and pixel throughput rather than shader count, suggests that lead would hold across most graphics and compute workloads these chips can still run today.