NVIDIA GeForce RTX 3090 vs NVIDIA GeForce RTX 4090 vs NVIDIA GeForce RTX 3060

NVIDIA GeForce RTX 3090

The Nvidia GeForce RTX 3090 is high-end desktop graphics card based on the Ampere generation. It uses the big GA102 chip and offers 10,496 shaders and 24 GB GDDR6X graphics memory. Compared to the slower RTX 3080, the 3090 offers 10 - 20% higher gaming performance, but only above 4k. Still it has not enough power for 8k gaming at maximum detail settings. The RTX3090 is rated at 350 Watt but the power efficiency is still quite good.

NVIDIA GeForce RTX 4090

The Nvidia GeForce RTX 4090 is a high-end desktop graphics card based on the AD102 chip using the Ada Lovelace architecture. The 4090 offers 16,384 of the 18,432 cores and 24 GB GDDR6X graphics memory with a 384-bit memory bus, and is clocked at 21 Gbits. The card offers an excellent ray tracing performance thanks to the 128 dedicated ray tracing cores. For AI calculations (and DLSS 3 in games), the RTX 4090 has 512 Tensor cores.

Performance

The gaming performance is very well suited for 4k gaming in maximum detail settings. Even ray tracing effects can be used without DLSS in most games. Compared to the competition, the RTX 4090 is the fastest graphics card in 2022, with the AMD Radeon RX 7900 XTX and the RTX 4080 in second place.

Power consumption

The AD102 chip is manufactured in the modern 5nm process at TSMC (4N with EUV) and specified with a TGP (Total Graphics Power) of 450 watts. Nvidia is recommending an 850-watt power supply for the RTX 4090.

NVIDIA GeForce RTX 3060

The Nvidia GeForce RTX 3060 is a fast desktop graphics card. It is based on the GA106 Ampere chip and offers 12 GB GDDR6 graphics memory with a 192 Bit memory bus. It offers 3,584 cores and supports PCIe 4.0.

The performance is similar to the older GeForce RTX 2070 and therefore best suited for 1080p gaming of modern and demanding games in high settings (in 2021).

The GA106 chip offers 3,840 FP32 ALUs of which half can also execute INT32 instructions (i.e. 1,920 INT32 ALUs). With Turing all shaders could still execute FP32 or INT32 instructions. The raytracing and tensor cores on the chip were also improved according to Nvidia. The Ampere chips also include an improved 5th generation video encoder (NVENC for H.264 and H.265) and a 7th generation decoder (for various formats now including AV1).

The GA106 chip is manufactured by Samsung in 8nm (8N), which is not quite able to keep up with the 7nm node at TSMC (e.g. used by AMD and also for the professional GA100 Ampere chip). The TDP of the RTX 3060 is specified at 170 Watt and therefore slightly higher than the previous RTX 2060 (160 Watt).