NVIDIA GeForce GTX 675MX vs NVIDIA GeForce GTX 970 Comparison

FAQ

Q: How does the NVIDIA GeForce GTX 675MX compare to the NVIDIA GeForce GTX 970 in raw compute performance?

A: The GTX 970 is significantly ahead. In Geekbench OpenCL, the GTX 970 scores 29,982 versus 10,723 for the GTX 675MX, a delta of -64.2% for the older mobile chip. The GTX 970 also leads in Geekbench Metal with 13,595 against 6,131, a -54.9% difference.

Q: Which GPU has the higher transistor count and larger die?

A: The GTX 970 uses the GM204 chip with 5,200 million transistors on a 398 mm² die. The GTX 675MX uses the GK104 chip with 3,540 million transistors on a 294 mm² die. Both are built on TSMC's 28 nm process, but the GTX 970 has a higher transistor density at 13.1M / mm² versus 12.0M / mm².

Q: What are the memory specifications of each card?

A: The GTX 675MX has 2 GB of GDDR5 on a 256-bit bus with 115.2 GB/s bandwidth and 3.6 Gbps effective memory speed. The GTX 970 has 4 GB of GDDR5 on a 256-bit bus with 224.4 GB/s bandwidth and 7 Gbps effective memory speed.

Q: Which GPU supports newer DirectX features?

A: The GTX 970 supports DirectX 12 (12_1), while the GTX 675MX supports DirectX 12 (11_0). The GTX 970 also has a newer Vulkan version (1.4) compared to Vulkan 1.2.175 on the GTX 675MX.

Q: How do the cards differ in physical design and power requirements?

A: The GTX 675MX is an MXM Module with no power connectors and a 100 W TDP, designed for portable devices. The GTX 970 is a dual-slot card requiring 2x 6-pin power connectors, a 300 W suggested PSU, and has a 148 W TDP. Its dimensions are 267 mm length, 111 mm height, and 40 mm width.

Q: What is the average benchmark score for each GPU?

A: The GTX 675MX has an average benchmark score of 8,427, placing it in the 43rd percentile of all GPUs. The GTX 970 has an average score of 7,157, placing it in the 39th percentile. Despite the GTX 970 winning the head-to-head tests, the GTX 675MX has a higher aggregate average across all recorded benchmarks.

Where Each One Wins

The GTX 970 dominates both recorded head-to-head benchmark categories. In Geekbench Metal, it scores 13,595 versus 6,131 for the GTX 675MX, a 54.9% advantage. This indicates a substantial lead in Apple's Metal API compute workloads, which often translate to real-world graphics and compute tasks in macOS environments.

In Geekbench OpenCL, the gap widens further. The GTX 970 scores 29,982 while the GTX 675MX manages only 10,723, a 64.2% deficit for the older chip. OpenCL is a cross-platform compute framework, and this large margin suggests the GTX 970's Maxwell 2.0 architecture handles parallel compute workloads far more efficiently than the Kepler-based GTX 675MX.

The GTX 675MX does not win any of the recorded head-to-head benchmarks. However, its aggregate average benchmark score of 8,427 is higher than the GTX 970's 7,157. This discrepancy comes from the different test suites recorded for each GPU. The GTX 675MX has only two benchmark entries, both Geekbench tests, while the GTX 970 has eleven entries spanning 3DMark, PassMark, and Geekbench tests. The GTX 970's PassMark G2D score of 764 and DirectX 9 score of 143 pull its average down, while the GTX 675MX's two strong Geekbench scores keep its average higher.

For users seeking maximum compute throughput in OpenCL or Metal workloads, the GTX 970 is the clear winner. For those looking at a broader benchmark profile that includes 2D graphics and legacy DirectX tests, the GTX 675MX shows a higher average score, but this is a function of the limited test set recorded for that card rather than an indication of superior overall performance.

Architecture Differences

The GTX 675MX is built on the Kepler architecture, specifically the GK104 chip. The GTX 970 uses the Maxwell 2.0 architecture with the GM204 chip. Both are fabricated by TSMC on a 28 nm process, but the architectural generations are distinct. Kepler was designed for the GeForce 600M series, while Maxwell 2.0 powers the GeForce 900 series.

The transistor budgets differ significantly. The GK104 packs 3,540 million transistors into a 294 mm² die, yielding a density of 12.0M / mm². The GM204 contains 5,200 million transistors on a 398 mm² die, achieving 13.1M / mm² density. This higher density reflects the architectural improvements in Maxwell 2.0, which emphasizes efficiency per clock cycle.

The shading engine configuration also differs. The GTX 675MX has 960 shading units, 80 texture mapping units, and 32 raster output units. The GTX 970 increases these to 1,664 shading units, 104 TMUs, and 56 ROPs. This is a substantial increase in parallel processing resources, explaining the large compute performance gap.

Clock behavior differs as well. The GTX 675MX has no recorded base or boost clock, while the GTX 970 has a base clock of 1,050 MHz and a boost clock of 1,178 MHz. The memory clocks also differ: the GTX 675MX runs at 900 MHz (3.6 Gbps effective), while the GTX 970 runs at 1,753 MHz (7 Gbps effective). The GTX 970's memory clock is nearly double that of the GTX 675MX.

The API support reflects the newer architecture. The GTX 970 supports DirectX 12 (12_1) and Vulkan 1.4, while the GTX 675MX supports DirectX 12 (11_0) and Vulkan 1.2.175. Both support OpenGL 4.6. The GTX 970's higher DirectX feature level and newer Vulkan version enable modern rendering techniques that the older Kepler chip cannot handle.

Both GPUs have no ray tracing or tensor cores, as these features did not exist in their respective generations. The GTX 675MX's interface is MXM-B (3.0), designed for laptop modules, while the GTX 970 uses PCIe 3.0 x16 for desktop motherboards.

Specification Differences

The two GPUs differ in nearly every measurable specification. The GTX 675MX has 2 GB of GDDR5 memory, while the GTX 970 has 4 GB. Both use a 256-bit memory bus, but the GTX 970's bandwidth of 224.4 GB/s is nearly double the GTX 675MX's 115.2 GB/s. This comes from the GTX 970's 7 Gbps effective memory speed versus 3.6 Gbps on the GTX 675MX.

The compute resources differ substantially. The GTX 970 has 1,664 shading units, 104 TMUs, and 56 ROPs, while the GTX 675MX has 960 shading units, 80 TMUs, and 32 ROPs. The GTX 970's pixel rate of 65.97 GPixel/s is over five times the GTX 675MX's 13.08 GPixel/s. Its texture rate of 122.5 GTexel/s is more than double the GTX 675MX's 52.32 GTexel/s. FP32 performance is 3.920 TFLOPS for the GTX 970 versus 1,255.7 GFLOPS for the GTX 675MX.

The power profiles differ as well. The GTX 675MX has a 100 W TDP and no power connectors, consistent with its mobile MXM form factor. The GTX 970 has a 148 W TDP, requires 2x 6-pin power connectors, and recommends a 300 W PSU. The GTX 970 is a dual-slot card measuring 267 mm by 111 mm by 40 mm, while the GTX 675MX uses the MXM Module slot width.

Display outputs differ: the GTX 675MX is "Portable Device Dependent," while the GTX 970 offers 1x DVI, 1x HDMI 2.0, and 3x DisplayPort 1.2. The GTX 970's launch MSRP was 329 USD. The GTX 675MX has no recorded launch MSRP.

Release dates show a two-year gap: the GTX 675MX launched on 2012-09-30, and the GTX 970 on 2014-09-18. The GTX 675MX succeeds the GeForce 500M and precedes the GeForce 700M, while the GTX 970 succeeds the GeForce 700 and precedes the GeForce 10 series.

Head-to-Head Benchmarks

The recorded head-to-head data contains two benchmark comparisons, both won by the GTX 970.

In Geekbench Metal, the GTX 970 scores 13,595 against 6,131 for the GTX 675MX. The delta is -54.9%, meaning the GTX 675MX trails by more than half. This is a decisive margin. Metal is Apple's low-level graphics and compute API, and the result shows the GTX 970's Maxwell architecture executing Metal workloads at more than double the rate of the Kepler-based GTX 675MX. The GTX 970's advantage likely stems from its higher shading unit count, faster memory, and more efficient scheduler design.

In Geekbench OpenCL, the gap is even larger. The GTX 970 scores 29,982, while the GTX 675MX scores 10,723, a delta of -64.2%. OpenCL is a general-purpose compute API used across multiple platforms, and this result indicates that the GTX 970 handles compute shaders, physics simulations, and data-parallel workloads far more effectively. The GTX 970's FP32 throughput of 3.920 TFLOPS compared to 1,255.7 GFLOPS for the GTX 675MX provides a theoretical basis for this 3.1x advantage in raw floating-point compute.

The GTX 970 also has a significant memory bandwidth advantage that contributes to these results. With 224.4 GB/s versus 115.2 GB/s, the GTX 970 can feed its compute units with data nearly twice as fast. This is particularly important for OpenCL workloads that often involve large data arrays.

The GTX 675MX's nearest rivals in the database include the AMD Radeon 880M (-0.1%), NVIDIA GeForce MX330 (-0.4%), and AMD Radeon HD 8870M (-0.4%), all with average scores very close to its 8,427. The GTX 970's nearest rivals include the Intel Iris Pro Graphics P580 (-0.2%), NVIDIA GeForce GTX 560 SE (-0.2%), and AMD Radeon Vega 8 Mobile (-0.6%), all near its 7,157 average.

The overall benchmark data shows the GTX 970 as the superior performer in compute-intensive tasks, with the GTX 675MX's higher aggregate average score reflecting a smaller benchmark suite rather than actual performance parity. In every direct comparison, the GTX 970 wins by a wide margin, making it the clear choice for compute-heavy applications.