AMD Radeon HD 8950M vs NVIDIA GeForce GTX 1660 Comparison

AMD
RADEON

AMD Radeon HD 8950M

CORE STATE Saturn
VRAM 2 GB
CLOCK SPEED —
TDP 100 W
BUS WIDTH 128 bit
ARCHITECTURE GCN 2.0
nm
PROCESS 28 nm
LAUNCH DATE 2013
VS
NVIDIA
GEFORCE

GeForce GTX 1660

CORE STATE TU116
VRAM 6 GB
CLOCK SPEED 1785 MHz
TDP 120 W
BUS WIDTH 192 bit
ARCHITECTURE Turing
nm
PROCESS 12 nm
LAUNCH DATE 2019

PERFORMANCE BENCHMARKS

geekbench_opencl
13,772
47,850
geekbench_vulkan
12,979
50,137
3dmark_3dmark_steel_nomad_dx12
N/A
1,065
passmark_directx_10
N/A
61
passmark_directx_11
N/A
79
passmark_directx_12
N/A
49
passmark_directx_9
N/A
177
passmark_g2d
N/A
776
passmark_g3d
N/A
11,646
passmark_gpu_compute
N/A
4,963

Analysis: AMD Radeon HD 8950M vs NVIDIA GeForce GTX 1660

Head-to-Head Benchmarks

The recorded data includes two direct comparison tests between the AMD Radeon HD 8950M and the NVIDIA GeForce GTX 1660. The NVIDIA card wins both, and the margins are substantial. In Geekbench OpenCL, the GTX 1660 scores 47,850 against the HD 8950M's 13,772. That is a 71.2% advantage for NVIDIA. In Geekbench Vulkan, the gap is even wider: the GTX 1660 posts 50,137, while the AMD part manages 12,979, a 74.1% deficit. These are not close contests. The NVIDIA GPU more than triples the AMD GPU's compute output in both APIs.

The average benchmark score tells a similar story, though with a smaller relative gap. The HD 8950M averages 13,376 across its recorded tests, while the GTX 1660 averages 11,680. The HD 8950M actually holds a higher average because its two scores are both near 13,000, whereas the GTX 1660 includes several lower-scoring DirectX and compute tests in its suite. The percentile ranking is close as well: the AMD card sits at the 54th percentile of all GPUs, and the NVIDIA card at the 51st. But the head-to-head results are unambiguous. When both cards run the same OpenCL and Vulkan workloads, the GTX 1660 wins by roughly three to four times the performance.

Looking at the nearest rivals in the database provides context. The HD 8950M's closest competitors are the AMD Radeon RX 5500M, the AMD FirePro M6100, and the AMD Radeon Pro 555 and 555X. Their average scores range from 13,321 to 13,407, and the deltas are tiny: the RX 5500M is 0.2% ahead, the FirePro M6100 is 0.2% ahead, and the Radeon Pro 555 is 0.2% behind. The HD 8950M is essentially performance-equivalent to those parts. The GTX 1660's nearest rivals include the AMD Radeon RX 7800 XT, the AMD Radeon Pro 5500M, the AMD Radeon RX 6500 XT, and the NVIDIA Tesla K20c. Their average scores span 11,479 to 11,842, with deltas of 0.5% to 1.8% relative to the GTX 1660. So the GTX 1660 sits in a slightly different performance tier, near mid-range desktop cards, while the HD 8950M is aligned with older mobile workstation parts.

Architecture Differences

The two GPUs come from different architectural generations and target different market segments. The AMD Radeon HD 8950M uses the Saturn chip, built on GCN 2.0 architecture, and belongs to the Solar System generation of the HD 8900M series. It is fabricated on a 28 nm process at TSMC. The NVIDIA GeForce GTX 1660 uses the TU116 chip, built on Turing architecture, and belongs to the GeForce 16 series. It is fabricated on a 12 nm process, also at TSMC. The process node difference is significant: 12 nm versus 28 nm, which directly affects transistor density and power characteristics.

The transistor counts are very different. The HD 8950M packs 2,080 million transistors on a 160 mm² die, for a density of 13.0 million transistors per square millimeter. The GTX 1660 contains 6,600 million transistors on a 284 mm² die, for a density of 23.2 million per square millimeter. The GTX 1660 has more than three times the transistor count, on a die that is less than twice the size. That density advantage comes from the newer process.

The memory subsystems also differ sharply. The HD 8950M has 2 GB of GDDR5 on a 128-bit bus, with memory clocked at 1375 MHz (5.5 Gbps effective) and a bandwidth of 88.00 GB/s. The GTX 1660 has 6 GB of GDDR5 on a 192-bit bus, with memory clocked at 2001 MHz (8 Gbps effective) and a bandwidth of 192.1 GB/s. The NVIDIA card has three times the memory capacity and roughly 2.2 times the bandwidth. That is a large practical difference for texture-heavy workloads and larger data sets.

Compute resources follow the same pattern. The HD 8950M has 768 shading units, 48 texture mapping units, and 16 ROPs. Its pixel rate is 17.20 GPixel/s, and its texture rate is 51.60 GTexel/s. The GTX 1660 has 1,408 shading units, 88 TMUs, and 48 ROPs. Its pixel rate is 85.68 GPixel/s, and its texture rate is 157.1 GTexel/s. The GTX 1660 has roughly 1.8 times the shader count, 1.8 times the texture units, and three times the ROPs. Its pixel rate is nearly five times higher, and its texture rate is about three times higher.

The FP32 throughput is also lopsided. The HD 8950M delivers 1.651 TFLOPS. The GTX 1660 delivers 5.027 TFLOPS, and it also has a listed FP16 rate of 10.05 TFLOPS with a 2:1 ratio. The HD 8950M has no listed FP16 capability. The GTX 1660 also shows a higher memory clock and a higher effective data rate, and its base clock is 1530 MHz with a boost of 1785 MHz, while the HD 8950M has no recorded base or boost clock.

The TDP figures are closer than the performance gap might suggest. The HD 8950M is rated at 100 W, and the GTX 1660 at 120 W. The NVIDIA card uses only 20 W more, while delivering roughly three times the FP32 throughput. The GTX 1660 is a dual-slot card with a 229 mm length, a 111 mm height, and a 35 mm width, and it uses a single 8-pin power connector. The AMD part is an MXM module, designed for portable devices, with display outputs described as portable-device dependent. The GTX 1660 has fixed display outputs: one DVI, one HDMI 2.0, and one DisplayPort 1.4a.

The API support differs as well. The HD 8950M supports DirectX 12 (feature level 12_0), OpenGL 4.6, and Vulkan 1.2.170. The GTX 1660 supports DirectX 12 (feature level 12_1), OpenGL 4.6, and Vulkan 1.4. The NVIDIA card has a higher DirectX feature level and a newer Vulkan version.

The Verdict

The data points to a clear performance hierarchy. The NVIDIA GeForce GTX 1660 is decisively faster than the AMD Radeon HD 8950M in every direct head-to-head test. The OpenCL score is 71.2% higher, and the Vulkan score is 74.1% higher. The GTX 1660 also has more than three times the memory capacity, roughly 2.2 times the memory bandwidth, and nearly three times the FP32 throughput. Its architecture is newer, its process node is smaller, and its transistor density is almost double.

The HD 8950M is not without merit, but its strengths are contextual. It is an MXM module, which means it is intended for laptops and portable workstations. Its 100 W TDP is lower than the GTX 1660's 120 W, and its 160 mm² die is much smaller. It also has a higher average benchmark score in the database, 13,376 versus 11,680, but that is a function of the test suite composition, not a sign of overall superiority. The direct comparisons are the reliable indicators, and they favor NVIDIA overwhelmingly.

For a user choosing between these two, the decision depends on the platform. If the requirement is a replaceable MXM module in a portable system, the HD 8950M is the only option of the two. If the requirement is a desktop GPU, the GTX 1660 is the clear choice. The GTX 1660 also has a longer production life: it was released in March 2019, while the HD 8950M came out in May 2013. Both are now end-of-life, but the GTX 1660 is a much more recent design.

The percentile rankings are close, 54 versus 51, but that is misleading in context. Those percentiles reflect all GPUs in the database, including many older and slower parts. The HD 8950M's ranking is inflated by its small test set. The GTX 1660's ranking is dragged down by its inclusion of PassMark DirectX 9, 10, 11, and 12 tests, where it scores 177, 61, 79, and 49 respectively. Those are legacy API tests that do not reflect modern workload performance. The GTX 1660's Geekbench scores, 47,850 and 50,137, are far above the HD 8950M's 13,772 and 12,979.

FAQ

Q: Which GPU has the higher average benchmark score in the database?

A: The AMD Radeon HD 8950M has an average benchmark score of 13,376, while the NVIDIA GeForce GTX 1660 averages 11,680.

Q: What is the margin of victory for the GTX 1660 in Geekbench Vulkan?

A: The GTX 1660 scores 50,137 in Geekbench Vulkan, which is 74.1% higher than the HD 8950M's 12,979.

Q: How much memory bandwidth does each card have?

A: The HD 8950M has a bandwidth of 88.00 GB/s over a 128-bit bus with 2 GB of GDDR5. The GTX 1660 has 192.1 GB/s over a 192-bit bus with 6 GB of GDDR5.

Q: What is the transistor density difference between the two GPUs?

A: The HD 8950M has a density of 13.0 million transistors per square millimeter on a 28 nm process. The GTX 1660 has 23.2 million per square millimeter on a 12 nm process.

Q: Which GPU supports a newer DirectX feature level?

A: The GTX 1660 supports DirectX 12 with feature level 12_1, while the HD 8950M supports DirectX 12 with feature level 12_0.

Q: What are the TDP ratings for each card?

A: The HD 8950M is rated at 100 W, and the GTX 1660 is rated at 120 W.

Where Each One Wins

The NVIDIA GeForce GTX 1660 wins in every measured compute benchmark. It is 71.2% ahead in Geekbench OpenCL and 74.1% ahead in Geekbench Vulkan. Its FP32 throughput is 5.027 TFLOPS versus 1.651 TFLOPS, its texture rate is 157.1 GTexel/s versus 51.60 GTexel/s, and its pixel rate is 85.68 GPixel/s versus 17.20 GPixel/s. It also has the memory advantage: 6 GB versus 2 GB, and 192.1 GB/s versus 88.00 GB/s. For any workload that stresses compute shaders, rasterization, or large textures, the GTX 1660 is the better part.

The AMD Radeon HD 8950M has a narrower set of advantages, but they are real. It draws 100 W compared to the GTX 1660's 120 W, making it the lower-power option. Its die is much smaller, 160 mm² versus 284 mm², which is relevant for compact mobile chassis. It is an MXM module, meaning it is designed for socketed replacement in laptops and portable workstations, while the GTX 1660 is a dual-slot desktop card. The HD 8950M also has a higher average benchmark score in the database, 13,376 versus 11,680, though that is due to the different test sets.

In practical terms, the HD 8950M is the correct choice only for systems that require an MXM form factor and have a 100 W thermal envelope. The GTX 1660 is the correct choice for any desktop build where space and power allow. The performance gap is too large to ignore. The GTX 1660's nearest rivals in the database are the RX 7800 XT, Radeon Pro 5500M, RX 6500 XT, and Tesla K20c, all within 1.8% of its average score. The HD 8950M's nearest rivals are the RX 5500M, FirePro M6100, and Radeon Pro 555 series, all within 0.4%. These are different performance tiers, and the GTX 1660 sits clearly above the HD 8950M in the database's hierarchy.

DETAILED SPECIFICATIONS

SPECIFICATION
HD 8950M
GTX 1660
Core Specs
Shading Units
768
1,408 +83.3%
Shaders
768
1,408 +83.3%
TMUs
48
88 +83.3%
ROPs
16
48 +200.0%
Compute Units
12
—
SM Count
—
22
Clocks
Base Clock
—
1530 MHz
Boost Clock
—
1785 MHz
GPU Clock
1075 MHz
—
Memory Clock
1375 MHz 5.5 Gbps effective
2001 MHz 8 Gbps effective
Memory
Memory Size
2 GB
6 GB
VRAM (MB)
2,048
6,144 +200.0%
Memory Type
GDDR5
GDDR5
Memory Bus
128 bit
192 bit
Bandwidth
88.00 GB/s
192.1 GB/s
Cache
L1 Cache
16 KB (per CU)
64 KB (per SM)
L2 Cache
256 KB
1536 KB
Performance
Pixel Rate
17.20 GPixel/s
85.68 GPixel/s
Texture Rate
51.60 GTexel/s
157.1 GTexel/s
FP32 (TFLOPS)
1.651 TFLOPS
5.027 TFLOPS
FP64 (TFLOPS)
103.2 GFLOPS (1:16)
157.1 GFLOPS (1:32)
FP16 (TFLOPS)
—
10.05 TFLOPS (2:1)
Power
TDP
100 W
120 W
TDP (W)
100
120 +20.0%
Suggested PSU
—
300 W
Power Connectors
—
1x 8-pin
Architecture
Architecture
GCN 2.0
Turing
GPU Name
Saturn
TU116
Generation
Solar System (HD 8900M)
GeForce 16
Process Size
28 nm
12 nm
Transistors
2,080 million
6,600 million
Die Size
160 mm²
284 mm²
Foundry
TSMC
TSMC
Density
13.0M / mm²
23.2M / mm²
API Support
DirectX
12 (12_0)
12 (12_1)
OpenGL
4.6
4.6
Vulkan
1.2.170
1.4
OpenCL
2.1
3.0
CUDA
—
7.5
Shader Model
6.5
6.8
Physical
Slot Width
MXM Module
Dual-slot
Length
—
229 mm 9 inches
Height
—
111 mm 4.4 inches
Outputs
Portable Device Dependent
1x DVI1x HDMI 2.01x DisplayPort 1.4a
Bus Interface
PCIe 3.0 x16
PCIe 3.0 x16
Other
Launch Price
—
219 USD
Production
End-of-life
End-of-life
Predecessor
London
GeForce 10
Successor
Gem System
GeForce 20
View Radeon HD 8950M Details View GeForce GTX 1660 Details