← All machines

Multi-GPU · DIY build

2x RTX 5090

64 GB of VRAM with the highest memory bandwidth of any consumer pair. Power is the hard part: at ~1.4 kW it outgrows a normal 120 V household circuit.

Memory for models

62.5 GB

VRAM across all cards

Memory bandwidth

1,792 GB/s

per card; 2 cards take turns on each token

Parts

~$4,109 + unpriced parts

GPU price not included

Largest at 4-bit

Llama 4 Scout 17B-16E Instruct

62.4 GB at 8,192 tokens of context

gpt-oss-20b, 4-bit

~224–327 tok/sestimate

no measurement on this hardware yet

Parts

PartQtyEach
GPUNVIDIA GeForce RTX 50902not priced
CPUAMD Ryzen 9 9950X (16-core, 170 W)1~$649
MotherboardASUS ProArt X870E-CREATOR WIFI (2x PCIe 5.0 x16 slots, x16 or x8/x8)1~$550
RAMCorsair Vengeance 64GB (2x32GB) DDR5-6000 CL301~$1,275
PSUCorsair WS3000 (3000 W, 220-240 V input only)1~$630
CaseCorsair 9000D RGB AIRFLOW super full tower (580 mm GPU clearance)1~$520
StorageCorsair MP600 CORE XT 2TB PCIe 4.0 NVMe1~$485
Priced parts only$4,109GPU price not included

New parts are priced at the maker’s own store or list price. Used cards have no price here yet, and a total with a missing GPU would understate the build, so none is shown.

What it runs at 4-bit

Every model in the catalog at its 4-bit file with 8,192 tokens of context. 24 of 41 fit. Speeds are the median of measured runs on the same GPUs where any exist; otherwise an estimate from memory bandwidth. Try other quants and context lengths.

ModelNeedsFitGeneration speed
Kimi K32.8T, 104B active · UD-Q4_K_XL1406.8 GBcontext memory estimatedDoes not fit
DeepSeek V4 Pro1.6T, 49B active · Q4_K_M890.5 GBcontext memory not countedDoes not fit
Kimi K2.61T, 32B active · UD-Q4_K_XL545.7 GBcontext memory estimatedDoes not fit
DeepSeek V4.1 Flash552B, 16B active · Q4 (DwarfStar)484.5 GBDoes not fit
GLM-5.3753.3B · UD-Q4_K_XL437.4 GBcontext memory estimatedDoes not fit
DeepSeek V3.2671B, 37B active · Q4_K_M379.6 GBcontext memory estimatedDoes not fit
MiniMax M3428B, 23B active · Q4_K_M244.8 GBcontext memory not countedDoes not fit
Qwen3.5 397B-A17B397B, 17B active · Q4_K_M229.1 GBcontext memory estimatedDoes not fit
GLM-5.3 Flash320B, 18B active · UD-Q4_K_XL187.6 GBcontext memory estimatedDoes not fit
Hy3295B, 21B active · Q4_K_M173.6 GBDoes not fit
MiniMax M2.7228.7B · Q4_K_M132.7 GBDoes not fit
Qwen3.8 Flash-Next125B, 6B active · Q4_K_M113.1 GBcontext memory estimatedDoes not fit
Nemotron 3 Super 120B-A12B120B, 12B active · Q4_K_M82.6 GBcontext memory estimatedDoes not fit
Qwen3.5 122B-A10B122B, 10B active · Q4_K_M73.0 GBcontext memory estimatedDoes not fit
Mistral Medium 3.5 128B128B · Q4_K_M74.0 GBDoes not fit
GLM-4.5-Air106B, 12B active · Q4_K_M70.9 GBDoes not fit
Mistral Small 4 119B (2603)119B, 6.5B active · Q4_K_M69.3 GBcontext memory estimatedDoes not fit
Llama 4 Scout 17B-16E Instruct109B, 17B active · Q4_K_M62.4 GBcontext memory not countedFits~44–64 tok/sestimate
gpt-oss-120b117B, 5.1B active · MXFP4 (native)60.2 GBcontext memory estimatedFits~163–238 tok/sestimate
Qwen3 Coder Next80B, 3B active · Q4_K_M46.9 GBcontext memory estimatedFits~245–358 tok/sestimate
Llama 3.3 70B Instruct70B · Q4_K_M43.6 GBFits~21–32 tok/sestimate
Nemotron 3.5 Lightning 30B-A3B30B, 3B active · Q4_K_M25.3 GBcontext memory estimatedFits~175–256 tok/sestimate
Qwen3.6 35B-A3B35B, 3B active · Q4_K_M22.4 GBcontext memory estimatedFits~233–341 tok/sestimate
Ornith 1.5 35B-A3B35B, 3B active · Q4_K_M21.9 GBcontext memory estimatedFits~240–350 tok/sestimate
Qwen3 Coder 30B-A3B Instruct30.5B, 3.3B active · Q4_K_M19.5 GBFits~222–324 tok/sestimate
Gemma 4 31B IT30.7B · Q4_K_M18.9 GBcontext memory estimatedFits~48–75 tok/sestimate
Granite 4.2 30B29.3B · Q4_K_M20.3 GBFits~49–77 tok/sestimate
Qwen3.8 27B27B · Q4_K_M18.2 GBcontext memory estimatedFits~50–79 tok/sestimate
Gemma 4 26B-A4B IT25.2B, 3.8B active · Q4_K_M17.4 GBcontext memory estimatedFits~174–253 tok/sestimate
Qwen3.6 27B27B · Q4_K_M17.7 GBcontext memory estimatedFits~52–82 tok/sestimate
Devstral Small 2 24B (2512)24B · Q4_K_M16.1 GBFits~61–96 tok/sestimate
gpt-oss-20b21B, 3.6B active · MXFP4 (native)12.5 GBcontext memory estimatedFits~224–327 tok/sestimate
Phi-414.7B · Q4_K_M11.3 GBFits~99–155 tok/sestimate
Ministral 3 14B Reasoning (2512)13.9B · Q4_K_M10.4 GBFits~107–167 tok/sestimate
Gemma 4 12B IT11.95B · Q4_K_M8.2 GBcontext memory estimatedFits~123–194 tok/sestimate
Qwen3.5 9B9B · Q4_K_M7.0 GBcontext memory estimatedFits~155–243 tok/sestimate
Granite 4.2 8B8.8B · Q4_K_M7.9 GBFits~159–249 tok/sestimate
Gemma 4 E4B IT8B · Q4_K_M6.5 GBcontext memory not countedFits~289–454 tok/sestimate
Qwen3 8B8.2B · Q4_K_M7.3 GBFits~175–274 tok/sestimate
Llama 3.1 8B Instruct8B · Q4_K_M7.1 GBFits~179–280 tok/sestimate
Qwen3.5 4B4B · Q4_K_M4.3 GBcontext memory estimatedFits~321–503 tok/sestimate

Measured speeds

Nobody has published a run on this GPU combination with a stated model, quant and backend that we could find. The speeds above are estimates.

With speculative decoding

A small draft model or extra prediction heads guess several tokens ahead. Generation is faster on text the guesses get right, so these are kept out of the medians above.

Runs with speculative decoding
ModelQuantBackendGenerationPromptContextMeasured onSource
Qwen3.8 27BAutoRound INT4 (fp8 KV)vLLM v0.27.1 (club-3090 dual-superfast compose)192 tok/s5,650 tok/s10,000-token prompt0this machineindependent benchmark2026-08
Qwen3.8 27BAutoRound INT4 (fp8 KV)vLLM v0.27.1 (club-3090 dual-superfast compose)94 tok/s7,400 tok/s10,000-token prompt0this machineindependent benchmark2026-08

Specifications

GPUs
2 × NVIDIA GeForce RTX 5090
VRAM
64 GB
System RAM
64 GB
RAM bandwidth
2 channels x 6000 MT/s x 8 bytes = 96.0 GB/s (dual-channel AM5)
Power supply
3000 W
Estimated load
2x RTX 5090 575 W + Ryzen 9 9950X 170 W + 100 W = 1,420 W
Software
CUDA, Vulkan

Before you build it

  • Corsair's WS3000 requires 220-240 V high-line input; on 120 V the next step down, the HX1500i, delivers only 1,200 W, below this build's ~1,420 W.
  • A US 15 A / 120 V circuit is 1,800 W on paper, but the NEC 80% rule for continuous loads (3 h+) limits it to 12 A, about 1,440 W, shared with everything else on that circuit.
  • On the ProArt X870E-CREATOR both CPU-attached slots share 16 lanes: x16 with one card, x8/x8 with two.
  • Open-air cards in adjacent slots blow exhaust straight into the next card's intake; check slot spacing (and NVLink bridge spacing) before buying the board.
  • RTX 5090 has no NVLink; the two cards talk over PCIe 5.0 x8.

Sources

Compared with

Similar machines