NVIDIA GeForce RTX 3060 Ti vs NVIDIA GeForce RTX 3070

Every spec side by side, from each maker's published figures.

Only show differences

Better figure, where one is clearly better

Share comparison

GeForce RTX 3060 Ti compared with GeForce RTX 3070

	NVIDIAGeForce RTX 3060 Ti	NVIDIAGeForce RTX 3070
Overview
Released
Released	December 2020	October 2020
Performance
Chip maker
Chip maker	NVIDIA	NVIDIA
Architecture
Architecture	Ampere	Ampere
Shader cores
Shader cores	4,864	5,888
Boost clock
Boost clock	1,670 MHz	1,730 MHz (better)
Architecture
Architecture	NVIDIA reference specifications; photo shows Gainward's original GDDR6 Ghost partner card, whose cooler and dimensions differ from the Founders Edition. NVIDIA announced worldwide partner availability of the GDDR6X revision from November 2022.	GA104; 46 Streaming Multiprocessors. NVIDIA reference specifications; photo shows PNY's stock-clock UPRISING Dual Fan LHR revision, whose cooler and dimensions differ from the Founders Edition.
Manufacturing process
Manufacturing process	–	Samsung 8 nm 8N NVIDIA custom process; 17.4 billion transistors; 392.5 mm² die
Ray tracing
Ray tracing	Second-generation RT Cores	46 second-generation RT Cores
AI cores
AI cores	Third-generation Tensor Cores	184 third-generation Tensor Cores
Upscaling and frame generation
Upscaling and frame generation	NVIDIA DLSS upscaling; no DLSS Frame Generation	NVIDIA DLSS upscaling; no DLSS Frame Generation
Base clock
Base clock	1.41 GHz; NVIDIA rounds boost to 1.67 GHz. Gainward's original Ghost and PNY's GDDR6X VERTO both specify 1,665 MHz boost.	1.50 GHz; the NVIDIA architecture whitepaper specifies 1,725 MHz boost, rounded to 1.73 GHz on the current specification page
Cache
Cache	–	4,096 KB L2
Memory
Video memory
Video memory	8 GB GDDR6 or 8 GB GDDR6X	8 GB GDDR6
Memory bus
Memory bus	256-bit	256-bit
Memory bandwidth
Memory bandwidth	448 GB/s or 608 GB/s	448 GB/s
Memory detail
Memory detail	Original GDDR6 is 14 Gbps and 448 GB/s (Gainward Ghost); later GDDR6X is 19 Gbps and 608 GB/s (PNY stock-clock VERTO). Both have 8 GB on a 256-bit bus.	14 Gbps GDDR6
Connections and power
Board power
Board power	200 W or –	220 W
Interface
Interface	PCIe 4.0	PCIe 4.0
Display outputs
Display outputs	3 × DisplayPort 1.4a, 1 × HDMI 2.1; up to 4 displays; HDCP 2.3	3 × DisplayPort 1.4a, 1 × HDMI 2.1; up to 4 displays; HDCP 2.3
Maximum resolution
Maximum resolution	7680 × 4320; HDMI supports 4K 120 Hz HDR or 8K 60 Hz HDR	7680 × 4320; HDMI supports 4K 120 Hz HDR or 8K 60 Hz HDR
Power connector
Power connector	1 × PCIe 8-pin cable through the included adapter to 1 × 12-pin (original GDDR6 Founders Edition); check the maker for GDDR6X partner cards	1 × PCIe 8-pin cable through the included adapter to 1 × 12-pin (Founders Edition)
Recommended power supply
Recommended power supply	600 W for the original Founders Edition. NVIDIA's current table lists 200 W card power without separating memory revisions; PNY publishes 225 W and a 650 W system recommendation for its GDDR6X VERTO. Its board power is not treated as a verified universal GDDR6X reference figure.	650 W, based on NVIDIA's Intel Core i9-10900K test system
Slot width
Slot width	2-slot (original GDDR6 Founders Edition); GDDR6X partner cards vary	2-slot (Founders Edition)
Size (reference card)
Size (reference card)	242 × 112 mm, 2-slot (original GDDR6 Founders Edition); no separate GDDR6X reference geometry verified	242 × 112 mm, 2-slot (Founders Edition)
Video encode and decode
Video encode and decode	Seventh-generation NVENC, fifth-generation NVDEC; hardware AV1 decode, no AV1 encode	Seventh-generation NVENC, fifth-generation NVDEC; hardware AV1 decode, no AV1 encode