AMD Radeon RX 550 vs NVIDIA GeForce GTX 670 Comparison

AMD
RADEON

AMD Radeon RX 550

CORE STATE Lexa
VRAM 2 GB
CLOCK SPEED 1183 MHz
TDP 50 W
BUS WIDTH 128 bit
ARCHITECTURE GCN 4.0
nm
PROCESS 14 nm
LAUNCH DATE 2017
VS
NVIDIA
GEFORCE

GeForce GTX 670

CORE STATE GK104
VRAM 2 GB
CLOCK SPEED 980 MHz
TDP 170 W
BUS WIDTH 256 bit
ARCHITECTURE Kepler
nm
PROCESS 28 nm
LAUNCH DATE 2012

PERFORMANCE BENCHMARKS

3dmark_3dmark_steel_nomad_dx12
127
N/A
geekbench_metal
20,838
7,424
geekbench_opencl
11,063
15,341
geekbench_vulkan
12,270
15,553

Analysis: AMD Radeon RX 550 vs NVIDIA GeForce GTX 670

Head-to-Head Benchmarks

The benchmark data reveals a split decision between these two cards, with the older NVIDIA GeForce GTX 670 taking two of three head-to-head tests, while the AMD Radeon RX 550 secures one decisive victory. In Geekbench Metal, the RX 550 posts a score of 20838, which is 64.4% higher than the GTX 670's 7424. That is a massive margin, reflecting the RX 550's much stronger performance in Apple's Metal API workloads. For users prioritizing Metal compute, the RX 550 is the clear choice.

However, the GTX 670 strikes back in the other two API tests. In Geekbench OpenCL, the GTX 670 scores 15341 against the RX 550's 11063, giving NVIDIA a 38.7% lead. In Geekbench Vulkan, the GTX 670 scores 15553 versus 12270, a 26.8% advantage. So while the RX 550 dominates Metal, the GTX 670 is substantially faster in both OpenCL and Vulkan. The data shows a 2-1 win split in favor of the GTX 670 across these three workloads.

The average benchmark scores confirm the GTX 670's overall edge. The GTX 670 averages 12773 across all recorded tests, while the RX 550 averages 11075. That is a 15.3% difference in favor of NVIDIA. The GTX 670's percentile ranking is 52, slightly above the RX 550's 50th percentile, meaning the GTX 670 sits just above the median GPU in the database while the RX 550 sits exactly at the median.

Looking at the nearest rivals in the database, the GTX 670's average score of 12773 places it very close to a cluster of GPUs: the NVIDIA GeForce GTX 590 (12830, only 0.4% higher), the AMD Radeon Pro 455 (12831, 0.4% higher), the AMD Radeon RX 7600M XT (12710, 0.5% lower), and the AMD FirePro W5100 (12847, 0.6% higher). The RX 550's average of 11075 sits near the NVIDIA RTX PRO 6000D Blackwell Max-Q (11088, 0.1% higher), the NVIDIA RTX PRO 6000 Blackwell Max-Q (11088, 0.1% higher), the NVIDIA GeForce GTX 1650 SUPER (11047, 0.3% lower), and the AMD FirePro W4300 (11225, 1.3% higher). This shows both cards are tightly packed with their performance peers.

FAQ

Q: Which GPU wins in Geekbench Metal?

A: The AMD Radeon RX 550 wins decisively, scoring 20838 compared to the NVIDIA GeForce GTX 670's 7424. This is a 64.4% higher score for the RX 550.

Q: How does the GTX 670 perform in OpenCL?

A: It outperforms the RX 550 by 38.7% in Geekbench OpenCL, with scores of 15341 for the GTX 670 and 11063 for the RX 550.

Q: What about Vulkan performance?

A: The NVIDIA GeForce GTX 670 leads in Geekbench Vulkan, scoring 15553 against the RX 550's 12270, which is a 26.8% advantage for NVIDIA.

Q: What are the average benchmark scores for each card?

A: The GTX 670 has an average benchmark score of 12773 across all tests, while the RX 550 averages 11075. The GTX 670 is 15.3% ahead on average.

Q: How many test wins does each card have?

A: The GTX 670 wins 2 out of 3 head-to-head tests (OpenCL and Vulkan), while the RX 550 wins 1 (Metal).

Q: Where does each card rank percent-wise against all GPUs in the database?

A: The GTX 670 ranks at the 53rd percentile, while the RX 550 ranks at the 50th percentile. Both are near the middle of the performance distribution.

Architecture Differences

The underlying architectures are fundamentally different, which explains the benchmark split. The NVIDIA GeForce GTX 670 uses the GK104 chip based on the Kepler architecture, manufactured on a 28 nm process by TSMC. The AMD Radeon RX 550 uses the Lexa chip based on GCN 4.0, manufactured on a 14 nm process by GlobalFoundries. This process difference gives the RX 550 a much higher transistor density: 21.4 million transistors per square millimeter versus 12.0 million for the GTX 670. The RX 550 packs 2,200 million transistors into a 103 mm² die, while the GTX 670 has 3,540 million transistors spread over a 294 mm² die.

The compute configurations are drastically different. The GTX 670 has 1344 shading units, 112 texture mapping units, and 32 render output units. The RX 550 has only 512 shading units, 32 TMUs, and 16 ROPs. Despite the GTX 670's larger compute configuration, the RX 550 achieves a higher base clock of 1100 MHz versus 915 MHz, and a higher boost clock of 1183 MHz versus 980 MHz. The GTX 670 compensates with raw unit counts, achieving 2.634 TFLOPS of FP32 performance, while the RX 550 produces 1,211.4 GFLOPS. The RX 550 also supports FP16 at a 1:1 ratio (1,211.4 GFLOPS), while the GTX 670 has no FP16 performance listed.

Memory architecture also differs. Both cards have 2 GB of GDDR5, but the GTX 670 uses a 256-bit bus with a bandwidth of 192.3 GB/s, while the RX 550 uses a 128-bit bus with a bandwidth of 112.0 GB/s. The GTX 670's memory runs at 1502 MHz (6 Gbps effective), while the RX 550 runs at 1750 MHz (7 Gbps effective), but the narrower bus limits the RX 550's total bandwidth. The GTX 670 also has higher pixel and texture rates: 27.44 GPixel/s and 109.8 GTexel/s, respectively, versus 18.93 GPixel/s and 37.86 GTexel/s for the RX 550.

The API support differs. The GTX 670 supports DirectX 12 (11_0), while the RX 550 supports DirectX 12 (12_0). Both support OpenGL 4.6, but the GTX 670 supports Vulkan 1.2.175, while the RX 550 supports Vulkan 1.3. Display outputs also differ: the GTX 670 has 2x DVI, 1x HDMI 1.4a, and 1x DisplayPort 1.2, while the RX 550 has 1x DVI, 1x HDMI 2.0b, and 1x DisplayPort 1.4a.

Specification Differences

The two cards differ across nearly every specification field. The process node: 28 nm for the GTX 670 versus 14 nm for the RX 550. Transistor count: 3,540 million versus 2,200 million. Die size: 294 mm² versus 103 mm². Transistor density: 12.0 million per mm² versus 21.4 million per mm².

Clock speeds: base 915 MHz versus 1100 MHz, boost 980 MHz versus 1183 MHz, memory 1502 MHz (6 Gbps effective) versus 1750 MHz (7 Gbps effective). Memory bus: 256-bit versus 128-bit. Bandwidth: 192.3 GB/s versus 112.0 GB/s.

Compute units: 1344 shading units versus 512, 112 TMUs versus 32, 32 ROPs versus 16. Pixel rate: 27.44 GPixel/s versus 18.93 GPixel/s. Texture rate: 109.8 GTexel/s versus 37.86 GTexel/s. FP32: 2.634 TFLOPS versus 1,211.4 GFLOPS. FP16: not available versus 1,211.4 GFLOPS (1:1). TDP: 170 W versus 50 W. Power connectors: 2x 6-pin versus none. Suggested PSU: 450 W versus 250 W. Bus interface: PCIe 3.0 x16 versus PCIe 3.0 x8.

Dimensions: the GTX 670 is 241 mm long, 111 mm high, and 38 mm wide (9.5 x 4.4 x 1.5 inches). The RX 550 is 145 mm long (5.7 inches) with no recorded height or width. Release dates: the GTX 670 launched on 2012-05-09, the RX 550 on 2017-04-19. The GTX 670 sits between the GeForce 500 and GeForce 700 generations, while the RX 550 sits between Arctic Islands and Vega. The GTX 670's launch MSRP was 399 USD, the RX 550's launch MSRP was 79 USD.

The Verdict

The data points to the NVIDIA GeForce GTX 670 as the more capable overall GPU. It wins the two most common compute APIs (OpenCL and Vulkan), has a higher average benchmark score (12773 versus 11075), and ranks higher in percentile (52 versus 50). Its memory bandwidth advantage (192.3 GB/s versus 112.0 GB/s) and far higher shading unit count (1344 versus 512) translate into a 38.7% lead in OpenCL and a 26.8% lead in Vulkan. The GTX 670 is also the older card by nearly five years, but its high-end Kepler architecture still outperforms the entry-level Polaris chip in most tests.

However, the AMD Radeon RX 550 is not without merit. It wins Geekbench Metal by a massive 64.4% margin, which indicates that for any workload leveraging Apple's Metal API, the RX 550 is the clear choice. Its newer GCN 4.0 architecture, smaller die, and lower power draw (50 W versus 170 W) make it a more efficient component. The RX 550 also supports newer display outputs (DisplayPort 1.4a and HDMI 2.0b) and a newer Vulkan version (1.3 versus 1.2.175).

The verdict is situation-dependent. If the workload is primarily Metal-based, the RX 550 is the pick. For any other compute scenario, the GTX 670 is faster, and its higher bandwidth and raw compute strength make it the better all-around performer.

Where Each One Wins

NVIDIA GeForce GTX 670 wins in:

  • OpenCL performance: 38.7% ahead of the RX 550.
  • Vulkan performance: 26.8% ahead of the RX 550.
  • Average benchmark score: 12773 versus 11075.
  • Higher memory bandwidth (192.3 GB/s) and a wider 256-bit bus.
  • More shading units (1344), TMUs (112), and ROPs (32).
  • Higher pixel rate (27.44 GPixel/s) and texture rate (109.8 GTexel/s).

AMD Radeon RX 550 wins in:

  • Geekbench Metal: 64.4% ahead of the GTX 670.
  • Lower power draw: 50 W versus 170 W, with no power connectors required.
  • Newer display outputs: HDMI 2.0b and DisplayPort 1.4a.
  • Newer Vulkan support: 1.3 versus 1.2.175.
  • Smaller die (103 mm²) and higher transistor density (21.4M/mm²) due to the 14 nm process.
  • Higher native FP16 performance (1,211.4 GFLOPS).

For users who run Metal-based applications, the RX 550 is the only logical choice. For users working with OpenCL or Vulkan, the GTX 670 provides substantially better performance. The data also shows the GTX 670's nearest rivals are all within 0.6% of its average score, indicating it is a borderline mid-range card, while the RX 550's rivals are similarly clustered. Both cards are end-of-life, but the GTX 670 remains the stronger legacy performer for general compute tasks.

DETAILED SPECIFICATIONS

SPECIFICATION
RX 550
GTX 670
Core Specs
Shading Units
512
1,344 +162.5%
Shaders
512
1,344 +162.5%
TMUs
32
112 +250.0%
ROPs
16
32 +100.0%
Compute Units
8
Clocks
Base Clock
1100 MHz
915 MHz
Boost Clock
1183 MHz
980 MHz
Memory Clock
1750 MHz 7 Gbps effective
1502 MHz 6 Gbps effective
Memory
Memory Size
2 GB
2 GB
VRAM (MB)
2,048
2,048 0.0%
Memory Type
GDDR5
GDDR5
Memory Bus
128 bit
256 bit
Bandwidth
112.0 GB/s
192.3 GB/s
Cache
L1 Cache
16 KB (per CU)
16 KB (per SMX)
L2 Cache
512 KB
512 KB
Performance
Pixel Rate
18.93 GPixel/s
27.44 GPixel/s
Texture Rate
37.86 GTexel/s
109.8 GTexel/s
FP32 (TFLOPS)
1,211.4 GFLOPS
2.634 TFLOPS
FP64 (TFLOPS)
75.71 GFLOPS (1:16)
109.8 GFLOPS (1:24)
FP16 (TFLOPS)
1,211.4 GFLOPS (1:1)
Power
TDP
50 W
170 W
TDP (W)
50
170 +240.0%
Suggested PSU
250 W
450 W
Power Connectors
None
2x 6-pin
Architecture
Architecture
GCN 4.0
Kepler
GPU Name
Lexa
GK104
Generation
Polaris (RX 500)
GeForce 600
Process Size
14 nm
28 nm
Transistors
2,200 million
3,540 million
Die Size
103 mm²
294 mm²
Foundry
GlobalFoundries
TSMC
Density
21.4M / mm²
12.0M / mm²
API Support
DirectX
12 (12_0)
12 (11_0)
OpenGL
4.6
4.6
Vulkan
1.3
1.2.175
OpenCL
2.1
3.0
CUDA
3.0
Shader Model
6.7
6.5 (5.1)
Physical
Slot Width
Dual-slot
Dual-slot
Length
145 mm 5.7 inches
241 mm 9.5 inches
Height
111 mm 4.4 inches
Outputs
1x DVI1x HDMI 2.0b1x DisplayPort 1.4a
2x DVI1x HDMI 1.4a1x DisplayPort 1.2
Bus Interface
PCIe 3.0 x8
PCIe 3.0 x16
Other
Launch Price
79 USD
399 USD
Production
End-of-life
End-of-life
Predecessor
Arctic Islands
GeForce 500
Successor
Vega
GeForce 700
View Radeon RX 550 Details View GeForce GTX 670 Details