GPU comparison

GeForce RTX 3090 vs GeForce RTX 4090

Compared on the specs that decide a purchase: memory, bandwidth, power, and which workloads each card actually fits — gaming, creative, or local models when that is the job.

Specifications verified October 2, 2026. Launch MSRP is not today's price — in September 2026 the memory shortage pushed street prices 1.2×–3× above it; the notes row below carries a dated lowest UK/US price where we have one.

NVIDIA

GeForce RTX 3090

24GB VRAM · 936 GB/s

NVIDIA

GeForce RTX 4090

24GB VRAM · 1008 GB/s

Amazon UK search links are provided for current availability; no affiliate programme is currently configured, and no live price is claimed here. Affiliate disclosure.

GPU specification, memory, bandwidth, and price comparison
Specification	GeForce RTX 3090	GeForce RTX 4090
Vendor
Vendor	NVIDIA	NVIDIA
Memory
Memory	24GB	24GB
Usable for a model
Usable for a model	24GB	24GB
Bandwidth
Bandwidth	936 GB/s	Winner: 1008 GB/s
Segment
Segment	consumer	consumer
Released
Released	September 24, 2020	October 12, 2022
Launch MSRP
Launch MSRP	~$1,499~£1,110	~$1,599~£1,184
Notes
Notes	24GB GDDR6X and 936GB/s theoretical bandwidth. NVIDIA's September 2020 announcement lists September 24 availability at $1,499. Current production status, new stock and used-card pricing or condition are unverified. Local-model fit is a planning estimate; model files, context, cache, backend and driver support determine actual memory use and speed.	24GB GDDR6X and 1008GB/s theoretical bandwidth. NVIDIA's launch notice lists October 12, 2022 availability starting at $1,599. Current production status, retailer stock and used prices are unverified here. Local-model fit is a planning estimate; model files, context, cache, backend and driver support determine actual memory use and speed.
UK street price
UK street price	Search Amazon (opens in new tab)	Search Amazon (opens in new tab)

Local models

Estimated local-model fit

Planning estimates from published VRAM and a weights-plus-overhead heuristic. Context length, KV cache, quantization and runtime can change actual memory needs.

Local model fit comparison for both GPUs
Model	GeForce RTX 3090 (est.)	GeForce RTX 4090 (est.)
Qwen 3 4B4B parameters
Qwen 3 4B4B parameters	Est. fit	Est. fit
Qwen 3.8 27B27B parameters
Qwen 3.8 27B27B parameters	Est. fit	Est. fit
Gemma 4 31B30.7B parameters
Gemma 4 31B30.7B parameters	Est. tight	Est. tight
Llama 3.1 8B Instruct8B parameters
Llama 3.1 8B Instruct8B parameters	Est. fit	Est. fit
Qwen 3 14B14.8B parameters
Qwen 3 14B14.8B parameters	Est. fit	Est. fit
Gemma 3 27B27B parameters
Gemma 3 27B27B parameters	Est. fit	Est. fit
Qwen 3 30B-A3B30.5B parameters
Qwen 3 30B-A3B30.5B parameters	Est. tight	Est. tight
Qwen 3 32B32.8B parameters
Qwen 3 32B32.8B parameters	Est. tight	Est. tight
Llama 3.3 70B70B parameters
Llama 3.3 70B70B parameters	Est. no fit	Est. no fit
Muse Glimmer 30B29.6B parameters
Muse Glimmer 30B29.6B parameters	Est. tight	Est. tight
Mixtral 8x22B141B parameters
Mixtral 8x22B141B parameters	Est. no fit	Est. no fit
Qwen 3 235B-A22B235B parameters
Qwen 3 235B-A22B235B parameters	Est. no fit	Est. no fit
Qwen 3 Coder 480B-A35B480B parameters
Qwen 3 Coder 480B-A35B480B parameters	Est. no fit	Est. no fit
DeepSeek V4-Flash284B parameters
DeepSeek V4-Flash284B parameters	Est. no fit	Est. no fit

Common questions

GeForce RTX 3090 vs GeForce RTX 4090

Answered from the verified figures on this page rather than general guidance.

Which costs less at launch MSRP, GeForce RTX 3090 or GeForce RTX 4090?

GeForce RTX 3090 launched lower at ~$1499 against ~$1599 for GeForce RTX 4090. Launch MSRP is an anchor, not a live price.

Which is better for high-refresh gaming, GeForce RTX 3090 or GeForce RTX 4090?

GeForce RTX 4090 has more memory bandwidth — 1008 GB/s against 936 GB/s. That usually helps high-refresh 1440p and 4K once VRAM is sufficient for your settings. Still verify with dated independent tests for the games you play.

For local inference only — which holds larger models, GeForce RTX 3090 or GeForce RTX 4090?

Neither — both have 24GB usable and hold the same models at Q4. Decide on bandwidth and price instead.

For local inference only — is GeForce RTX 3090 or GeForce RTX 4090 faster at generating tokens?

GeForce RTX 4090, at 1008 GB/s against 936 GB/s — roughly 1.1× the memory bandwidth. Bandwidth is the practical ceiling on token throughput once a model fits, but it only matters if the model fits in the first place.

For local inference only — can either run a 70B model?

Neither. A 70B model needs about 48GB at Q4 including overhead, above both cards. The largest that fits is Qwen 3 32B on GeForce RTX 3090 and Qwen 3 32B on GeForce RTX 4090.

If one card holds a model the other cannot, that capacity difference usually outweighs a bandwidth advantage. A model that spills to system memory can be substantially slower, but the actual impact depends on the backend, transfer path, context, and workload.

All GPU comparisons Compare hosted models instead