RTX 3090 workstation (Ryzen 9 7950X, 64 GB)

A 24 GB Ampere card that remains common in local-inference builds. Its 936 GB/s GDDR6X bandwidth is close to an RTX 4090's, while used pricing makes it an important baseline for 20B-30B quantized models.

NVIDIAcomponentAvailable13 models fit1 measured
Memory
24 GB
Official spec GDDR6X
Memory bandwidth
936 GB/s
Official spec ~75% achieved in practice
Specifications
CPU
AMD Ryzen 9 7950X Official spec
GPU
GeForce RTX 3090 24 GB Official spec
Memory
24 GB GDDR6X Official spec
Memory bandwidth
936 GB/s Official spec source
Achieved bandwidth
~75% (702 GB/s) Measured
FP16 compute
~142 TFLOPS Official spec
System RAM
64 GB Official spec
Architecture
Discrete GPU with dedicated VRAM Official spec
Nominal power
600 W Official spec
Measured load power
Assumption
Idle power
Measured
Released
24 Sept 2020 Official spec
Exact Ryzen 9 7950X, RTX 3090 24 GB and 64 GB DDR5 configuration. Hardware specifications are based on AMD and NVIDIA product documentation. The EUR 2,199 used-market estimate is anchored to an exact-match completed listing at USD 1,999 and checked against an active exact-match European listing; asking prices vary with component quality and condition.
Model performance on RTX 3090 workstation (Ryzen 9 7950X, 64 GB)1 measured, 12 estimated
13 of 13 rows
ModelQuantMemoryDecodePrefillContextFitConfidence
Nemotron 3.5 Lightning 30B-A3B
NVIDIA · 31.578B (3B active)
Q4_K_M
recommended
20.0 GB~213–307 t/s~1520–3150 t/s192KBorderlineEstimated
gpt-oss-20b
OpenAI · 20.915B (3.6B active)
MXFP4
recommended
13.0 GB~205–295 t/s~1700–3530 t/s128KComfortableEstimated
Qwen3-Coder 30B-A3B
Alibaba Qwen · 30.532B (3.3B active)
Q4_K_M
recommended
20.0 GB~197–284 t/s~1470–3060 t/s16KBorderlineEstimated
Qwen3 30B-A3B
Alibaba Qwen · 30.532B (3.3B active)
Q4_K_M
recommended
20.0 GB~197–284 t/s~1470–3060 t/s16KBorderlineEstimated
Gemma 4 26B-A4B
Google DeepMind · 25.806B (3.8B active)
Q4_K_M
recommended
18.3 GB~176–253 t/s~1490–3100 t/s16KFitsEstimated
Qwen3 8B
Alibaba Qwen · 8.191B
Q8_0
recommended
10.6 GB~62.1–89.4 t/s~1800–3740 t/s40KComfortableEstimated
Phi-4 14B
Microsoft · 14.66B
Q5_K_M
recommended
12.9 GB~52.4–75.4 t/s~1010–2090 t/s16KComfortableEstimated
Devstral Small 24B
Mistral AI · 23.572B
Q4_K_M
recommended
16.4 GB~38.8–55.8 t/s~627–1300 t/s32KComfortableEstimated
Mistral Small 3.2 24B
Mistral AI · 24.011B
Q4_K_M
recommended
16.6 GB~38.1–54.8 t/s~615–1280 t/s32KComfortableEstimated
Gemma 3 27B
Google DeepMind · 27.432B
Q4_K_M
recommended
21.3 GB~33.5–48.2 t/s~538–1120 t/s8KBorderlineEstimated
Qwen3.8 27B
Alibaba Qwen · 27.781B
Q4_K_M
recommended
measured: Q5_K_M
18.1 GB~33.1–47.6 t/s
Measured 74.88 t/s
~532–1100 t/s32K
evidence 1K
FitsEstimated
Measured · llama.cpp
Qwen3.6 27B
Alibaba Qwen · 27.781B
Q4_K_M
recommended
18.1 GB~33.1–47.6 t/s~532–1100 t/s32KFitsEstimated
Granite 4.2 30B
IBM · 29.277B
Q4_K_M
recommended
20.5 GB~31.4–45.2 t/s~504–1050 t/s8KBorderlineEstimated
Rows in grey are estimates from our bandwidth model, not measurements — they are shown as a range and never as a precise figure. Use the “measured only” filter to see just the 1 pairing on this machine that a real benchmark backs.
Submit a benchmarkContributions are reviewed before publication.
Ownership economicsUnited States (federal) · C corporation · 8h/day
Monthly economic cost
$26
Calculated after tax
Codex/Claude Code
$100/mo
≈ $100/mo · local is $74 less
Net cash at purchase
$1,981
Calculated VAT not reclaimable
Total over 5 years
$1,578
Calculated after tax, after resale
Cost per USD/1M tokens
$0.80
Calculated 32.9M tokens/month

The $100 comparison uses the Codex Pro 5x / Claude Max 5x plans. A fixed planning conversion is used for USD. This compares monthly spend only: subscriptions have usage limits, local hardware has different capabilities and constraints, and taxes or regional pricing may change the charged amount. Prices checked 7 September 2026.

Purchase-price input: $1,981 in United States (federal), tax/VAT excluded · estimated · Used market exact-match reference (converted catalogue estimate). This is the localized purchase input; the after-tax economic cost is calculated separately below.

Monthly breakdown

Depreciation
1,545 over 5 years, straight-line to a 436 residual
$25.75
Electricity
31.3 kWh/month at 0.140/kWh
$4.38
Cost of capital
4.0%/yr on 1,208 average capital employed
$4.03
Monthly cost before tax$34.16
Electricity tax shield
Running costs are deductible business expenses
−$0.92
First-year expensing
§179 (100.0%)
−$6.93
Monthly economic cost after tax$26.30

Three different numbers, deliberately

Cash cost
$1,981
Money that leaves the bank account on day one, net of reclaimable VAT.
Accounting depreciation
$25.75/month
$1,545 written down over 5 years to a $436 residual.
After-tax economic cost
$26.30/month
Depreciation plus running costs plus cost of capital, less the tax those deductions save. This is the figure to compare between machines.
Assumptions and sources (verified 2026-09-07)
VAT rate
0% Official spec
VAT recoverable
0% Official spec
Federal C-corporation estimate at the flat 21% rate. Pass-through entities should use the sole-proprietor estimate as a rougher proxy.
Effective deduction rate
21.00% Calculated
Headline marginal rate 21.00%.
Depreciation
5 years, straight-line Assumption
Residual value
$436 (22%) Assumption
Extrapolated.
Electricity
$0.140/kWh Assumption
Average power draw
178 W Calculated
Load 600 W for 20% of powered hours, idle 72 W for the rest — a machine that is on is not generating tokens the whole time.
Investment allowances
§179 100.0% Official spec
Worth $416 in total — first-year expensing that replaces later tax depreciation.
Cost of capital
$4.03/month Assumption
Assumes the machine generates tokens 20% of its 176 powered hours per month. Cost per token scales inversely with this number — halve the utilisation and the cost per token doubles.
Federal planning estimate only, not tax advice. State and local income tax, sales/use tax and incentives are excluded. It assumes 100% business use, a Section 179 election, enough business income to use it, and that the full annual limit remains available.