RTX 3090 Specifications
Last consumer card with NVLink. Cheap 24 GB on the used market for hobbyist fine-tuning.
Buy it for
The cheapest credible 24 GB for local inference and small-scale fine-tuning, and it retains NVLink — the last GeForce to do so, which lets a pair present 48 GB to frameworks that support it.
Look elsewhere if
You need reliability or ECC. These are heavily represented among ex-mining cards, and GDDR6X runs hot enough that memory-junction temperature is a real failure mode.
Will it fit your system?
- Form factor
- PCIe triple-slot
- Cooling
- active
- Power connectors
- 2 × 8-pin PCIe (Founders Edition uses a 12-pin adapter)
- Board power
- 350 W
- Suggested PSU
- 850 W or better
- Interface
- PCIe Gen 4
PSU figure assumes one card plus a mainstream CPU and drives, with headroom for transient spikes — modern cards can draw well above their rated TDP for milliseconds, which is what trips an undersized supply.
Supported formats & features
What the RTX 3090 does and does not support in hardware.
Blackwell-generation feature — B200 has it
Predates hardware FP8; use INT8 or BF16 instead — H100 has it
Wider exponent than FP16 — fewer overflow issues when training
Drop-in FP32 speedup with no code changes
Quantised inference supported
INT4 tensor operations supported
Precision must be managed manually
112 GB/s GPU-to-GPU
Hardware ray tracing for rendering workloads
No ECC, so silent corruption is possible on multi-day runs
What fits in memory
Largest model that fits on one RTX 3090, assuming ~20% of the 24 GB is reserved for KV cache and activations.
| Model | FP16 / BF16 | INT4 / FP4 |
|---|---|---|
| Llama 8B | Fits | Fits |
| Llama 70B | 8× GPUs | 2× GPUs |
| Mixtral 8x22B | 15× GPUs | 4× GPUs |
| Llama 405B | 43× GPUs | 11× GPUs |
Weights only. Long contexts, large batch sizes and optimiser state during training all need substantially more — full fine-tuning typically needs 3–4× the weight memory.
Workload character
Plenty of bandwidth per unit of compute. Memory-bound work — single-stream token decode above all — runs close to its ceiling, and compute is usually what runs out first.
Cost per unit of compute
Breakeven counts hardware cost against rental only. Power, cooling, hosting and depreciation push the real crossover substantially later — see colocation pricing for what hosting actually costs.
Cost to run
Electricity for one card at 100% duty, including cooling and PSU overhead (4,292 kWh/year).
At the US average rate, three years of continuous operation costs $1,545 — roughly 103% of the card's $1,499 launch price, before any cooling or hosting fees.
Availability & market
Best value per gigabyte on the used market. Buy from a seller who will confirm the card was not mined on, and re-paste the memory pads.
Discontinued. Expect secondary-market stock only, with no manufacturer warranty and limited driver support ahead.
Memory
- Capacity
- 24 GB
- Type
- GDDR6X
- Bandwidth
- 936 GB/s
- Bus width
- 384-bit
- GB per TFLOPMemory relative to compute — higher favours large models
- 0.338
Compute
- FP64
- —
- FP32
- 35.6 TFLOPS
- TF32 dense
- —
- FP16 dense
- 71 TFLOPS
- FP16 sparseRequires structured sparsity; rarely reached in practice
- 142 TFLOPS
- FP8 dense
- —
- INT8 dense
- —
Silicon
- Codename
- GA102
- Process node
- Samsung 8N
- Transistors
- 28.3B
- Die size
- 628 mm²
- CUDA cores
- 10,496
- Tensor cores
- 328
- SM count
- 82
- Boost clock
- 1,695 MHz
Platform & power
- TDP
- 350 W
- TFLOPS per wattDense FP16 ÷ TDP — drives your power bill at scale
- 0.20
- Form factor
- PCIe triple-slot
- Cooling
- active
- Interconnect
- NVLink
- Interconnect speed
- 112 GB/s
- PCIe generation
- Gen 4
- MIG
- No
- Launch
- 2020-09
- MSRP at launch
- $1,499
RTX 3090 — common questions
Answered from the specifications above. Questions this card's data cannot answer with confidence are left out rather than guessed.
- How much VRAM does the RTX 3090 have?
- The RTX 3090 has 24 GB of GDDR6X memory with 936 GB/s of bandwidth. In practice plan on roughly 19 GB for weights: KV cache, activations and fragmentation take the rest, and long contexts or large batches take considerably more.
- What size model can the RTX 3090 run?
- On a single RTX 3090, about 9B parameters at FP16, or roughly 38B quantised to FP4. Anything larger has to be split across several GPUs. These are weight-only figures — a long context window can consume tens of gigabytes on its own.
- Can you put the RTX 3090 in a desktop PC?
- Yes. The RTX 3090 is a self-cooled PCIe triple-slot card and works in a normal desktop, provided the case has clearance and the power supply is rated around 850 W or better.
- How much power does the RTX 3090 use?
- The RTX 3090 is rated at 350 W board power. It is fed through 2 × 8-pin PCIe (Founders Edition uses a 12-pin adapter). For a single-card system, a 850 W supply or better leaves headroom for the CPU, drives and the transient spikes that trip undersized units.
- How much does it cost to run the RTX 3090?
- At the US commercial average of $0.12/kWh, running one RTX 3090 continuously costs about $515 a year in electricity. That figure includes cooling and power-supply overhead at a factor of 1.4 over board power, and over three years comes to roughly 103% of what the card cost at launch.
- Does the RTX 3090 support FP8?
- No. The RTX 3090 has no hardware FP8 path. FP8-quantised checkpoints will still load, but the values are upcast before they reach the tensor cores, so you get the memory saving without the speedup.
- Does the RTX 3090 support FP4?
- No. The RTX 3090 predates hardware FP4. You can run 4-bit quantised weights on it and the memory saving is real, but they are dequantised before compute, so there is no arithmetic speedup.
- Does the RTX 3090 support INT8?
- Yes. The RTX 3090 supports INT8 acceleration. INT8 remains the most widely supported quantised inference format across frameworks.
- Does the RTX 3090 support BF16?
- Yes. The RTX 3090 supports bfloat16. Its wider exponent range makes training far less prone to the overflow that plagues FP16, which is why most current training recipes assume it.
- Does the RTX 3090 have NVLink?
- Yes. The RTX 3090 supports NVLink (NVLink) at 112 GB/s, so GPU-to-GPU traffic bypasses PCIe. This matters for tensor-parallel serving, where every token crosses the link.
- Does the RTX 3090 have RT cores?
- Yes. The RTX 3090 includes dedicated ray-tracing cores, so it is usable for rendering and visualisation as well as compute.
- Does the RTX 3090 have ECC memory?
- No. The RTX 3090 has no ECC. For short interactive work this rarely matters; for multi-week training runs it is a real risk, and it is one reason datacentre parts cost more than consumer ones.
- Is the RTX 3090 worth buying?
- Buy it for: The cheapest credible 24 GB for local inference and small-scale fine-tuning, and it retains NVLink — the last GeForce to do so, which lets a pair present 48 GB to frameworks that support it. Look elsewhere if: You need reliability or ECC. These are heavily represented among ex-mining cards, and GDDR6X runs hot enough that memory-junction temperature is a real failure mode.