RTX 6000 Ada Price, 48GB VRAM & Local LLMs
Price alert
We'll email you when the eBay (US) price reaches your target.
NVIDIA's RTX 6000 Ada (2022) packs 48GB of VRAM into a 300W dual-slot card, enough for 30B-class models at any precision. Don't confuse it with the older Quadro RTX 6000; look for “Ada” and 48GB in the listing.
eBay
RTX 6000 Ada price history
eBay
As an eBay Partner, VRAMGlass may be compensated if you buy through the eBay links on this page. The price you pay stays the same.
RTX 6000 Ada VRAM & full specs
- Memory48 GB · GDDR6 ECC
- Bandwidth960 GB/s
- FP16 / BF16364.3 TFLOPS
- FP8 / INT8728.5 / 728.5 TOPS
- Bus width
- 384-bit
- Architecture
- Ada Lovelace (AD102)
- Cores
- 18,176
- Tensor cores
- 568
- Ecosystem
- CUDA
- Interface
- PCIe 4.0 x16
- NVLink
- Not supported
- Power
- 300 W
- Power connector
- 16-pin 12VHPWR
- Slots
- 2
- Length
- 267 mm
- Type
- Workstation
- Launch date
- Sep 20, 2022
Best RTX 6000 Ada deals
Powered by VRAMGlass real-time price tracking — handpicking the best-value hardware deals across the web
As an eBay Partner, VRAMGlass may be compensated if you buy through the eBay links on this page. The price you pay stays the same.
As an Amazon Associate I earn from qualifying purchases. The price you pay stays the same.
What LLMs can a single RTX 6000 Ada run?
A single RTX 6000 Ada runs up to 35B at any precision on the card. 4-bit balances best, 27B near 40 tokens/s; 8-bit answers better, dense in the 20s.
| Model | Size | Runs? | Speed t/s | Max context |
|---|---|---|---|---|
 Gemma 4 12B12B | 4.7 GB | Runs | 123.5 | 256K |
| Gemma 4 26B A4B25.2B (3.8B active) | 10.5 GB | Runs | 140.8 | 256K |
 Qwen3.6 27B27.8B | 11.8 GB | Runs | 56.2 | 256K |
| Qwen3.8 27B27.8B | 9.8 GB | Runs | 66.3 | 256K |
 GLM 4.7 Flash30B (3B active) | 11.9 GB | Runs | 151.2 | 198K |
| Gemma 4 31B30.7B | 11.8 GB | Runs | 54.2 | 256K |
| Qwen3.6 35B A3B36B (3B active) | 12.3 GB | Runs | 161.1 | 256K |
| Model | Size | Runs? | Speed t/s | Max context |
|---|---|---|---|---|
| Gemma 4 12B12B | 7.1 GB | Runs | 88 | 256K |
| Gemma 4 26B A4B25.2B (3.8B active) | 16.9 GB | Runs | 118.8 | 256K |
| Qwen3.6 27B27.8B | 16.8 GB | Runs | 40.7 | 256K |
| Qwen3.8 27B27.8B | 16.5 GB | Runs | 41.4 | 256K |
| GLM 4.7 Flash30B (3B active) | 18.3 GB | Runs | 133.7 | 198K |
| Gemma 4 31B30.7B | 18.3 GB | Runs | 36.7 | 256K |
| Qwen3.6 35B A3B36B (3B active) | 22.1 GB | Runs | 136.7 | 256K |
| Model | Size | Runs? | Speed t/s | Max context |
|---|---|---|---|---|
| Gemma 4 12B12B | 12.7 GB | Runs | 52.7 | 256K |
| Gemma 4 26B A4B25.2B (3.8B active) | 26.9 GB | Runs | 95.6 | 256K |
| Qwen3.6 27B27.8B | 28.6 GB | Runs | 24.6 | 256K |
| Qwen3.8 27B27.8B | 29 GB | Runs | 24.3 | 256K |
| GLM 4.7 Flash30B (3B active) | 31.8 GB | Runs | 107.3 | 198K |
| Gemma 4 31B30.7B | 32.6 GB | Runs | 21.4 | 202K |
| Qwen3.6 35B A3B36B (3B active) | 36.9 GB | Runs | 111.2 | 256K |
RunsFits fully in VRAM; speed is estimated from VRAM bandwidth.
Needs RAM offloadMoE only: expert weights sit in system RAM, so speed is estimated from 70 GB/s RAM bandwidth.
Too bigWon’t fit on one card at this quantization, even with RAM offload.
Full calculation rules are on the Methodology page.
What LLMs can multiple RTX 6000 Ada cards run?
Two RTX 6000 Ada cards (96GB) run gpt-oss 120B at about 108 tokens/s; four run DeepSeek V4 Flash at 4-bit. No NVLink, but cards share data directly.
| Model | Size | Cards | Total price | Total VRAM | Speed t/s | Total power |
|---|---|---|---|---|---|---|
| Qwen3.8 Flash Next176B (6B active) | 78.9 GB | 2 cards | $15,554 | 96 GB | 112.5 | 600 W |
 DeepSeek V4 Flash284B (13B active) | 96.8 GB | 4 cards | $31,108 | 192 GB | 92.2 | 1,200 W |
 Hy3295B (21B active) | 107.2 GB | 4 cards | $31,108 | 192 GB | 45 | 1,200 W |
| GLM 5.3 Flash320B (18B active) | 108.7 GB | 4 cards | $31,108 | 192 GB | 75.4 | 1,200 W |
| GLM 5.3744B (40B active) | 253.9 GB | 8 cards | $62,216 | 384 GB | 39.7 | 2,400 W |
| Hy4 preview770B (49B active) | 307.2 GBest. | 8 cards | $62,216 | 384 GB | 30.1 | 2,400 W |
 Muse Spark Open816B (42B active) | 311.2 GB | 8 cards | $62,216 | 384 GB | 35.5 | 2,400 W |
 Kimi K2.7 Code1,000B (32B active) | 339.5 GB | 8 cards | $62,216 | 384 GB | 47.6 | 2,400 W |
| Kimi K2.61,000B (32B active) | 340.5 GB | 8 cards | $62,216 | 384 GB | 47.5 | 2,400 W |
| Model | Size | Cards | Total price | Total VRAM | Speed t/s | Total power |
|---|---|---|---|---|---|---|
| Qwen3.8 Flash Next176B (6B active) | 111.3 GB | 4 cards | $31,108 | 192 GB | 96.2 | 1,200 W |
| DeepSeek V4 Flash284B (13B active) | 155.1 GB | 4 cards | $31,108 | 192 GB | 69.1 | 1,200 W |
| Hy3295B (21B active) | 182.2 GB | 8 cards | $62,216 | 384 GB | 33.9 | 2,400 W |
| GLM 5.3 Flash320B (18B active) | 199.7 GB | 8 cards | $62,216 | 384 GB | 49.5 | 2,400 W |
| Model | Size | Cards | Total price | Total VRAM | Speed t/s | Total power |
|---|---|---|---|---|---|---|
| DeepSeek V4 Flash284B (13B active) | 161.9 GB | 4 cards | $31,108 | 192 GB | 67.2 | 1,200 W |
| Qwen3.8 Flash Next176B (6B active) | 188.2 GB | 8 cards | $62,216 | 384 GB | 71.7 | 2,400 W |
| Hy3295B (21B active) | 317.7 GB | 8 cards | $62,216 | 384 GB | 23.5 | 2,400 W |
| GLM 5.3 Flash320B (18B active) | 341 GB | 8 cards | $62,216 | 384 GB | 32.2 | 2,400 W |
Full calculation rules are on the Methodology page.
Who it suits
Good fit
- Running 30B-class models on one card with VRAM to spare
Qwen3.8 27B at 4-bit runs at about 41 tokens/s, and every model up to 35B can use its full context limit
- Running FP8 models in vLLM
This generation supports FP8 compute, and the previous-generation A6000 does not
- Packing several cards into one workstation
The dual-slot blower exhausts hot air out through the bracket, and each card draws 300W
- Tensor parallelism across multiple cards
Puget Systems measured direct transfers between two 6000 Adas, something NVIDIA has confirmed the 4090 can't do
Not a good fit
- Linking two cards with NVLink
This generation dropped it, though the A6000 still has it
- Using 4-bit floating-point formats like NVFP4
That takes a card from the 50-series generation
Before you buy
Get the model name right
The official name is RTX 6000 Ada Generation. Sellers often call it “A6000 Ada”, and the older Quadro RTX 6000 also goes by RTX 6000. In listings, look for both “Ada” and 48GB.
Plug in both cables on the 16-pin adapter
The card has a single 16-pin power connector and ships with a dual 8-pin adapter. Both 8-pin plugs must be connected: NVIDIA's power guide states that with only one plugged in, the adapter tells the card that just 150W is available, and a 300W card won't power on.
No need for an NVLink bridge
This card has no NVLink connector, so a bridge from an A6000 is of no use. Multi-GPU efficiency depends on the platform: Puget Systems measured direct card-to-card bandwidth of about 51 GB/s on Threadripper Pro and about 41 GB/s on Xeon W.
Size and cooling
It is about 267 mm long and dual-slot, and its blower fan pushes hot air out of the case through the bracket, so several cards can sit right next to each other.
Compared with similar GPUs
RTX 6000 Ada vs RTX A6000, L40S, and RTX PRO 5000 48GB: price, VRAM, bandwidth, compute and local LLM speed compared.
| Spec | RTX 6000 Ada | RTX A6000 | L40S | RTX PRO 5000 48GB |
|---|---|---|---|---|
| Memory | 48 GB | 48 GBSame | 48 GBSame | 48 GBSame |
| Bandwidth | 960 GB/s | 768 GB/s20% lower | 864 GB/s10% lower | 1,344 GB/s40% higher |
| BF16 / FP16 compute | 364 TFLOPS | 155 TFLOPS57% lower | 362 TFLOPSSame | 258 TFLOPS29% lower |
| INT8 compute | 729 TOPS | 310 TOPS57% lower | 733 TOPSSame | 516 TOPS29% lower |
| Qwen3.8 27B speed (2-bit) | 66.3 t/s | 54.1 t/s18% slower | 60.3 t/s9% slower | 89.5 t/s35% faster |
| Largest dense model (4-bit) | 79B | 79BSame | 79BSame | 79BSame |
| Power | 300 W | 300 WSame | 350 W17% higher | 300 WSame |
| GPU interconnect | Not supported | NVLink 112.5 GB/s | Not supported | Not supported |
| Interface | PCIe 4.0 x16 | PCIe 4.0 x16 | PCIe 4.0 x16 | PCIe 5.0 x16 |
| Released | 2022-09 | 2020-10 | 2023-08 | 2025-03 |
| eBay median |
As an eBay Partner, VRAMGlass may be compensated if you buy through the eBay links on this page. The price you pay stays the same.
RTX 6000 Ada rental price per hour
On-demand rental of the RTX 6000 Ada starts at $0.66 per hour (Vast.ai), about $161/month at 8 hours a day. Buying one (eBay median asking price $7,777) pays off in about 8.8 years, 4.4 years or 1.5 years at 4, 8 or 24 hours a day.
| Option | Price | 4 h/day | 8 h/day | 24 h/day | View |
|---|---|---|---|---|---|
 Vast.aiRentUpdated 29 min. ago | $0.66/hr | $81/mo | $161/mo | $483/mo | Rent |
 RunPodRentUpdated 29 min. ago | $0.74/hr | $90/mo | $180/mo | $540/mo | Rent |
 eBayUpdated 3 hr. ago | $7,777 | Payback ~8.8 yrsPower $7/mo | Payback ~4.4 yrsPower $13/mo | Payback ~1.5 yrsPower $39/mo | Buy |
Break-even uses the US residential average ($0.18/kWh) and 300 W rated power, excluding resale value and the rest of the build.
Rental links are referral links: we may receive a commission or credit if you sign up through them. As an eBay Partner, VRAMGlass may be compensated if you buy through the eBay links on this page. The price you pay stays the same.
Amazon.jp
Xianyu