COMPARE / GTX 1660 SUPER VS RTX 3050

GTX 1660 SUPER vs RTX 3050

Specs from the chip makers' datasheets; prices from the live Compute Cafe panel, refreshed daily. Every rate is an on-demand published price with a source receipt.

LOADING LIVE DATA...

GTX 1660 SUPERRTX 3050
MAKERNVIDIANVIDIA
ARCHITECTURETuringAmpere
LAUNCHED20192022
MEMORY6 GB GDDR68 GB GDDR6
MEMORY BANDWIDTH336 GB/s224 GB/s
TENSOR COMPUTE5 TFLOPS FP32; no tensor cores9.1 TFLOPS FP32; 3rd-gen tensor cores
POWER DRAW125 W130 W
INTERCONNECTPCIe 3.0 x16PCIe 4.0 x8

HOW TO CHOOSE

GTX 1660 SUPER vs RTX 3050
The 3050 adds tensor cores and 2 GB more memory - the minimum for any ML work.
RTX 3050 vs GTX 1660 SUPER
Same price floor, but the 3050's tensor cores make it the only one of the two usable for ML. For transcoding they are equivalent.
Workload fit
GTX 1660 SUPER: video transcoding, game streaming and vdi, ml inference. RTX 3050: learning and dev, 7b inference, serving or training.

GTX 1660 SUPER VS RTX 3050 - FAQ

GTX 1660 SUPER or RTX 3050 - which should I rent?
GTX 1660 SUPER: The 1660 Super is a gaming card from before tensor cores reached this price - it has none. RTX 3050: The 3050 is the cheapest Ampere card with tensor cores - the minimum hardware for modern ML frameworks' fast paths. The side-by-side spec table and live price rows above give the full picture.
Where can I rent GTX 1660 SUPER and RTX 3050?
Each chip's page on Compute Cafe lists every provider renting it with links to the provider's own published rate.

Full pricing: GTX 1660 SUPER · RTX 3050 · all comparisons