GPU Comparison

The recorded data paints a lopsided picture: the desktop GeForce GTX 950 wins every head-to-head benchmark against the mobile-focused GeForce MX350, and the gaps range from comfortable to enormous. Both are end-of-life NVIDIA parts separated by roughly a generation, and both carry the same 2 GB of GDDR5, but that is where the similarity ends. The GTX 950 sits at the 53rd percentile of all GPUs in the database with an average benchmark score of 13189, while the MX350 lands at the 49th percentile with an average of 10883. Here is how the numbers break down.

Head-to-Head Benchmarks

Only two tests exist for both cards, and the GTX 950 takes both.

The OpenCL result is the headline: the GTX 950 scores 15662 against the MX350's 8689, an 80.3% victory. That is not a close race, it is a different weight class. Any OpenCL-accelerated workload, from compute kernels to GPU-assisted rendering, will run dramatically faster on the desktop card.

The Vulkan test is tighter but still decisive: 16734 for the GTX 950 versus 13077 for the MX350, a 28% gap in the desktop card's favor. Interestingly, the MX350's Vulkan score is its strongest result by far, and it nearly matches the GTX 950's average benchmark score of 13189. But "nearly matching an average" is not the same as winning a direct comparison.

The database also records a Geekbench Metal score of 7172 for the GTX 950; there is no corresponding MX350 Metal entry, so no head-to-head can be drawn there. Final tally: 2 wins for the GTX 950, 0 for the MX350.

For context against the wider field, the GTX 950's rivals include the NVIDIA GTX 480 (average score 13300, 0.8% ahead of it), the Tesla M2090 (13075, 0.9% behind), the AMD Radeon Pro 555X (13321, 1% ahead), and the AMD FirePro M6100 (13354, 1.2% ahead). The MX350 clusters with the AMD Radeon Pro 450 (10804, 0.7% behind it), the NVIDIA Quadro K2200 (10761, 1.1% behind), the GTX 1650 SUPER (11047, 1.5% ahead), and the AMD Radeon RX 550 (11075, 1.7% ahead). In short, each card competes in its own neighborhood, and the GTX 950's neighborhood is measurably faster.

FAQ

Q: Which GPU is faster overall?

A: The GTX 950. It averages 13189 across recorded benchmarks versus 10883 for the MX350, and it won both shared tests: OpenCL by 80.3% and Vulkan by 28%.

Q: How much faster is the GTX 950 in compute workloads?

A: In Geekbench OpenCL the margin is 80.3%: 15662 versus 8689. This is the single largest gap in the head-to-head data.

Q: Do the two cards have the same amount of memory?

A: Yes. Both ship with 2 GB of GDDR5. However, the GTX 950 uses a 128-bit bus delivering 105.8 GB/s, while the MX350 uses a 64-bit bus delivering 56.06 GB/s, less than half the bandwidth.

Q: Which card uses less power?

A: The MX350, by a wide margin. Its TDP is 20 W and it needs no power connector. The GTX 950 has a 90 W TDP, requires a 6-pin PCIe power connector, and carries a 250 W suggested PSU rating.

Q: Which architecture is newer?

A: The MX350. It uses the Pascal-based GP107S chip on a 14 nm Samsung process, while the GTX 950 uses the Maxwell 2.0 GM206 chip on a 28 nm TSMC process. The MX350 was released in 2020, the GTX 950 in 2015.

Q: Do they support the same APIs?

A: Yes. Both support DirectX 12 (feature level 12_1), OpenGL 4.6, and Vulkan 1.4. Neither has RT cores or tensor cores.

Where Each One Wins

The GTX 950 wins every performance category in the recorded data, so the split is really about performance versus efficiency.

Pick the GTX 950 when throughput matters. It delivers 1.825 TFLOPS of FP32 compute on paper versus the MX350's 1.879 TFLOPS, but the benchmark results show the desktop card converting its wider hardware into real wins: 38.02 GPixel/s of pixel throughput versus 23.49, 57.02 GTexel/s of texture throughput versus 46.98, and double the memory bandwidth. With 768 shading units, 48 TMUs, and 32 ROPs against the MX350's 640, 32, and 16, the GTX 950 has more rasterization muscle everywhere it counts. Its 1.4 times larger framebuffer bandwidth pool is what feeds those units.

Pick the MX350 when power is the constraint. At 20 W with no power connector, it exists in a thermal and electrical envelope the GTX 950 simply cannot touch. The desktop card draws 90 W and demands a 6-pin connector plus a 250 W suggested PSU. In a portable device, where the MX350's display outputs are described only as "Portable Device Dependent," the choice is made for you. The MX350 also punches reasonably hard for its envelope: its Vulkan score of 13077 puts it within 1.7% of cards like the Radeon RX 550 in its rival cluster.

Specification Differences

The two cards differ on nearly every measurable spec:

• Bus interface: GTX 950 uses PCIe 3.0 x16; MX350 uses PCIe 3.0 x4, a quarter of the host bandwidth.
• Memory bus and bandwidth: 128 bit and 105.8 GB/s for the GTX 950 versus 64 bit and 56.06 GB/s for the MX350. Same 2 GB GDDR5 capacity on both.
• Clocks: The MX350 actually clocks higher, 1354 MHz base and 1468 MHz boost, against 1024 MHz base and 1188 MHz boost for the GTX 950. Its memory also runs faster at 1752 MHz (7 Gbps effective) versus 1653 MHz (6.6 Gbps effective). Higher clocks cannot compensate for the narrower bus.
• Render hardware: 768 shading units, 48 TMUs, 32 ROPs for the GTX 950 versus 640, 32, and 16 for the MX350.
• FP16 support: The MX350 records 29.36 GFLOPS of FP16 at a punitive 1:64 rate; the GTX 950 records no FP16 figure at all.
• Power: 90 W TDP, 1x 6-pin connector, 250 W suggested PSU for the GTX 950; 20 W TDP, no connectors for the MX350.
• Physical form: The GTX 950 is a dual-slot card measuring 202 mm (8 inches) long with fixed outputs (1x DVI, 1x HDMI 2.0, 3x DisplayPort 1.2). The MX350's slot width and dimensions are not specified, consistent with an integrated mobile part.
• Launch MSRP: The GTX 950 launched at 159 USD. No launch MSRP is recorded for the MX350.

Architecture Differences

These are two different design philosophies from different eras. The GTX 950 is the GM206 chip, Maxwell 2.0 architecture, part of the GeForce 900 generation, built on TSMC's 28 nm process. It packs 2,940 million transistors onto a 228 mm² die, a density of 12.9M per mm². Its lineage runs from the GeForce 700 series to the GeForce 10 series.

The MX350 is the GP107S chip, Pascal architecture, from the GeForce MX (3xx) generation, built on Samsung's 14 nm process. It fits 3,300 million transistors into just 132 mm², a density of 25.0M per mm², nearly double the GTX 950's density on a die less than 60% of the size. That smaller die and modern node are what make the 20 W power envelope possible.

Feature support is otherwise identical: both handle DirectX 12 (12_1), OpenGL 4.6, and Vulkan 1.4, and neither includes RT cores or tensor cores. Neither card benefits from hardware ray tracing or machine-learning accelerators of any kind.

The Verdict

The benchmark data leaves no ambiguity: the GTX 950 is the faster card. It wins both shared tests, by 80.3% in OpenCL and 28% in Vulkan, sits four percentile points higher against the full GPU database (53rd versus 49th), and out-averages the MX350 13189 to 10883. Anyone choosing between them for gaming, compute, or any throughput-driven task should take the GTX 950, provided the system can supply a 6-pin connector and accommodate a dual-slot, 202 mm card.

The MX350's case rests entirely on its electrical profile. At 20 W with no external power, it delivers performance that keeps it within striking distance of cards like the Radeon Pro 450 and Quadro K2200, and its Vulkan result of 13077 shows it is no slouch in modern graphics APIs. But it loses every direct comparison here. Treat it as an efficiency play for systems where the GTX 950's 90 W draw is physically impossible, and treat the GTX 950 as the performance pick everywhere else.