GeForce GT 640 vs GeForce GTX 660: specs and benchmark comparison
Short verdict
Not enough data for a performance conclusion: the database has no comparable benchmarks for this pair. Below are the key differences in official specifications.
Compare with AIAdvantages
Only measured and rated values: a noticeable benchmark gap and specs where more is better.
Where GeForce GT 640 is better2
Board power 75 W lower65 W (TDP) vs 140 W (TDP)
Launch price $130 lower$99 vs $229
Where GeForce GTX 660 is better3
Shader units 576 more960 (CUDA cores) vs 384 (CUDA cores)
Memory bus 64 bit more192 bit vs 128 bit
Memory bandwidth 115.7 GB/s more144.2 GB/s vs 28.5 GB/s
Benchmarks
Results are compared within one benchmark, one version and the same conditions.
Blender Benchmark 2.93.1
GPU, bmw27 (CUDA) · s, lower is better
GeForce GT 6401,298.5 sGeForce GTX 660357.5 sNo advantage conclusion: small sample in one of the results.
About the measurement and source
- A · NVIDIA GeForce GT 640
1,298.5 s · community measurement, sample 1 · November 14, 2026
Blender Open Data (Blender Foundation)
Median of 1 submissions (0 outliers excluded), interquartile range 1298.45–1298.45; submissions 2022-01-07 to 2022-01-07; CUDA backend. Snapshot 2026-11-14.
- B · NVIDIA GeForce GTX 660
357.5 s · community measurement, sample 5 · November 14, 2026
Blender Open Data (Blender Foundation)
Median of 5 submissions (0 outliers excluded), interquartile range 343.11–358.52; submissions 2021-10-01 to 2022-02-11; CUDA backend. Snapshot 2026-11-14.
difference = |A − B| ÷ the smaller value × 100%. The 5% noticeable threshold is a UX heuristic, not statistical significance.
Render time of a single Cycles scene without synchronization. Median of submissions after filtering.
Reference results (27)
| Benchmark | A | B | Difference |
|---|---|---|---|
| Blender Benchmark 2.90GPU, bmw27 (CUDA) | 1,209.9 | 361.6 | reference |
| Blender Benchmark 2.92GPU, bmw27 (CUDA) | 1,241.9 | — | no pair |
| Blender Benchmark 2.82GPU, classroom (CUDA) | 6,499.8 | 1,246.2 | reference |
| Blender Benchmark 2.90GPU, fishy_cat (CUDA) | 3,537.9 | 689.4 | reference |
| Blender Benchmark 3.5.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | 36.7 | 122.6 | reference |
| Blender Benchmark 3.6.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | 31.9 | 124.1 | reference |
| Blender Benchmark 2.81GPU, bmw27 (CUDA) | — | 342.5 | no pair |
| Blender Benchmark 2.82GPU, bmw27 (CUDA) | — | 352.6 | no pair |
| Blender Benchmark 2.83GPU, bmw27 (CUDA) | — | 362.7 | no pair |
| Blender Benchmark 2.81GPU, classroom (CUDA) | — | 1,168.2 | no pair |
| Blender Benchmark 2.83GPU, classroom (CUDA) | — | 1,301.3 | no pair |
| Blender Benchmark 2.90GPU, classroom (CUDA) | — | 1,289.5 | no pair |
| Blender Benchmark 2.93.1GPU, classroom (CUDA) | — | 1,286.5 | no pair |
| Blender Benchmark 2.92GPU, fishy_cat (CUDA) | — | 690.1 | no pair |
| Blender Benchmark 2.83GPU, koro (CUDA) | — | 1,369.4 | no pair |
| Blender Benchmark 2.90GPU, koro (CUDA) | — | 1,357.1 | no pair |
| Blender Benchmark 2.90GPU, pavillon_barcelona (CUDA) | — | 2,055 | no pair |
| Blender Benchmark 3.0.1GPU, sum of scenes monster + junkshop + classroom (CUDA) | — | 151.2 | no pair |
| Blender Benchmark 3.1.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | — | 121.2 | no pair |
| Blender Benchmark 3.2.1GPU, sum of scenes monster + junkshop + classroom (CUDA) | — | 127.7 | no pair |
| Blender Benchmark 3.3.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | — | 124.2 | no pair |
| Blender Benchmark 3.4.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | — | 128.8 | no pair |
| Blender Benchmark 4.0.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | — | 114.6 | no pair |
| Blender Benchmark 4.1.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | — | 113.2 | no pair |
| Blender Benchmark 4.2.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | — | 106.4 | no pair |
| Blender Benchmark 4.3.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | — | 115.5 | no pair |
| Blender Benchmark 4.5.0GPU, sum of scenes monster + junkshop + classroom (CUDA) | — | 115.7 | no pair |
Each benchmark version has its own scale: numbers are compared only within a row.
Specifications
| Specification | A · GeForce GT 640 | B · GeForce GTX 660 | Difference |
|---|---|---|---|
| Chip | |||
| GPU chip | GK107-300-A2 | GK106-400-A1 | differs |
| Architecture | Kepler | Kepler | equal |
| Shader unitsShader unit counts are comparable only within one architecture. | 384 (CUDA cores) | 960 (CUDA cores) | B: 576 more |
| Clocks | |||
| Base clock | 902 MHz | 980 MHz | B: 78 MHz more |
| Boost clockClocks of different architectures do not reflect performance directly. For laptop chips, the clock depends on the laptop’s power limit. | no data | 1,032 MHz | no data for A |
| Memory | |||
| Video memory | 2 GB | 2 GB | equal |
| Memory type | DDR3 | GDDR5 | differs |
| Memory bus | 128 bit | 192 bit | B: 64 bit more |
| Memory bandwidthExcludes cache (Infinity Cache / L2), which affects effective bandwidth differently. | 28.5 GB/s | 144.2 GB/s | B: 115.7 GB/s more |
| Power and dimensions | |||
| Board powerTGP (NVIDIA) and TBP (AMD/Intel) are calculated differently; higher is not better. For laptops, the chip maker sets the range and the laptop maker picks the value, so the same chip behaves differently in different laptops. | 65 W (TDP) | 140 W (TDP) | B: 75 W higher |
| Interface | PCIe 3.0 x16 | PCIe 3.0 x16 | equal |
Chip
GPU chip
AGK107-300-A2
BGK106-400-A1
differsArchitecture
AKepler
BKepler
equalShader units
A384 (CUDA cores)
B960 (CUDA cores)
B: 576 moreShader unit counts are comparable only within one architecture.
Clocks
Base clock
A902 MHz
B980 MHz
B: 78 MHz moreBoost clock
Ano data
B1,032 MHz
no data for AClocks of different architectures do not reflect performance directly. For laptop chips, the clock depends on the laptop’s power limit.
Memory
Video memory
A2 GB
B2 GB
equalMemory type
ADDR3
BGDDR5
differsMemory bus
A128 bit
B192 bit
B: 64 bit moreMemory bandwidth
A28.5 GB/s
B144.2 GB/s
B: 115.7 GB/s moreExcludes cache (Infinity Cache / L2), which affects effective bandwidth differently.
Power and dimensions
Board power
A65 W (TDP)
B140 W (TDP)
B: 75 W higherTGP (NVIDIA) and TBP (AMD/Intel) are calculated differently; higher is not better. For laptops, the chip maker sets the range and the laptop maker picks the value, so the same chip behaves differently in different laptops.
Interface
APCIe 3.0 x16
BPCIe 3.0 x16
equal
“More” does not always mean “better”: higher power draw or size is a system requirement, not an advantage (such bars are dimmed).
Compare with AI
An AI model reviews the specs and benchmarks of this pair and gives a plain-language verdict on who each model suits.
Questions and answers
How much video memory do NVIDIA GeForce GT 640 and NVIDIA GeForce GTX 660 have?
NVIDIA GeForce GT 640: 2 GB DDR3, 128-bit bus. NVIDIA GeForce GTX 660: 2 GB GDDR5, 192-bit bus.
Which PCIe interface do the cards use?
NVIDIA GeForce GT 640: PCIe 3.0 x16. NVIDIA GeForce GTX 660: PCIe 3.0 x16.
See also
Data sources (2)
- Wikipedia (English) — specifications of NVIDIA GeForce GT 640, specifications of NVIDIA GeForce GTX 660, release date and launch pricechecked November 14, 2026
- Blender Open Data (Blender Foundation) — benchmark Blender Benchmark 2.90, benchmark Blender Benchmark 2.92, benchmark Blender Benchmark 2.93.1, benchmark Blender Benchmark 2.81, benchmark Blender Benchmark 2.82, benchmark Blender Benchmark 2.83, benchmark Blender Benchmark 3.5.0, benchmark Blender Benchmark 3.6.0, benchmark Blender Benchmark 3.0.1, benchmark Blender Benchmark 3.1.0, benchmark Blender Benchmark 3.2.1, benchmark Blender Benchmark 3.3.0, benchmark Blender Benchmark 3.4.0, benchmark Blender Benchmark 4.0.0, benchmark Blender Benchmark 4.1.0, benchmark Blender Benchmark 4.2.0, benchmark Blender Benchmark 4.3.0, benchmark Blender Benchmark 4.5.0checked November 14, 2026
How we compare: methodology. Where the data comes from: sources.