Tesla T10 Price, 16GB VRAM & Local LLMs
Price alert
We'll email you when the eBay (US) price reaches your target.
NVIDIA's Tesla T10 (2020) is a single-slot cloud gaming GPU with 16GB of VRAM that runs Gemma 4 12B at 4-bit at about 40 tokens/s. It was never sold at retail and has no fan, so you'll need to add your own cooling.
eBay
Tesla T10 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.
Tesla T10 VRAM & full specs
- Memory16 GB · GDDR6
- Bandwidth403 GB/s
- FP16 / BF1680 TFLOPS
- INT8160 TOPS
- Bus width
- 256-bit
- Architecture
- Turing (TU102)
- Cores
- 3,584
- Tensor cores
- 448
- Ecosystem
- CUDA
- Interface
- PCIe 3.0 x16
- NVLink
- Not supported
- Power
- 150 W
- Power connector
- 1× 8-pin
- Slots
- 1
- Length
- 267 mm
- Type
- Data center
- Launch date
- Feb 4, 2020
Best Tesla T10 deals on eBay
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.
UsedNVIDIA TESLA T10 16GB GDDR6 PCIe x16 GPU MODEL PG150$369+ shippingBuy on eBay
UsedNVIDIA 699-2G150-0215-505 TESLA T10 16GB GDDR6 GPUTop Rated$495Buy on eBay
Used699-2G150-0215-505 NVIDIA TESLA T10 16GB GDDR6 GPUTop Rated$495Buy on eBay
UsedNVIDIA Tesla T10 16GB GDDR6 Turing GPU Accelerator PCIe x16 699-2G150-0215-505$581Buy on eBay
UsedNVIDIA TESLA T10 16GB Tensor Core GDDR6 Tesla T10 PCIe 3.0 x16 Accelerator GPU$583Buy on eBay
What LLMs can a single Tesla T10 run?
A single T10 suits 9B to 12B at 4-bit. 30B-class MoE fits only at 2-bit, where Qwen3.6 35B A3B runs about 121 tokens/s, its fastest big model.
| Model | Size | Runs? | Speed t/s | Max context |
|---|---|---|---|---|
 Gemma 4 E4B8B | 3.8 GB | Runs | 71.7 | 128K |
 Qwen3.5 9B9.7B | 4.1 GB | Runs | 66.3 | 256K |
| Gemma 4 12B12B | 4.7 GB | Runs | 57.6 | 256K |
| Gemma 4 26B A4B25.2B (3.8B active) | 10.5 GB | Runs | 96 | 256K |
| Qwen3.6 27B27.8B | 11.8 GB | Runs | 24.7 | 64K |
| Qwen3.8 27B27.8B | 9.8 GB | Runs | 29.4 | 93K |
 GLM 4.7 Flash30B (3B active) | 11.9 GB | Runs | 108.1 | 75K |
| Gemma 4 31B30.7B | 11.8 GB | Runs | 23.8 | 41K |
| Qwen3.6 35B A3B36B (3B active) | 12.3 GB | Runs | 120.7 | 181K |
| Model | Size | Runs? | Speed t/s | Max context |
|---|---|---|---|---|
| Gemma 4 E4B8B | 5 GB | Runs | 56.1 | 128K |
| Qwen3.5 9B9.7B | 5.7 GB | Runs | 49.3 | 256K |
| Gemma 4 12B12B | 7.1 GB | Runs | 39.8 | 256K |
| Gemma 4 26B A4B25.2B (3.8B active) | 16.9 GB | Needs RAM offload | 34.4 | 256K |
| Qwen3.6 27B27.8B | 16.8 GB | Too big | — | — |
| Qwen3.8 27B27.8B | 16.5 GB | Too big | — | — |
| GLM 4.7 Flash30B (3B active) | 18.3 GB | Needs RAM offload | 34.4 | 198K |
| Gemma 4 31B30.7B | 18.3 GB | Too big | — | — |
| Qwen3.6 35B A3B36B (3B active) | 22.1 GB | Needs RAM offload | 46.5 | 256K |
| Model | Size | Runs? | Speed t/s | Max context |
|---|---|---|---|---|
| Gemma 4 E4B8B | 8.2 GB | Runs | 35.5 | 128K |
| Qwen3.5 9B9.7B | 9.5 GB | Runs | 30.7 | 196K |
| Gemma 4 12B12B | 12.7 GB | Runs | 23.1 | 183K |
| Gemma 4 26B A4B25.2B (3.8B active) | 26.9 GB | Needs RAM offload | 23.2 | 256K |
| Qwen3.6 27B27.8B | 28.6 GB | Too big | — | — |
| Qwen3.8 27B27.8B | 29 GB | Too big | — | — |
| GLM 4.7 Flash30B (3B active) | 31.8 GB | Needs RAM offload | 21.5 | 198K |
| Gemma 4 31B30.7B | 32.6 GB | Too big | — | — |
| Qwen3.6 35B A3B36B (3B active) | 36.9 GB | Needs RAM offload | 30.7 | 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 Tesla T10 cards run?
One T10 is too small; eight (128GB) run Qwen3.8 Flash Next at 4-bit at about 55 tokens/s, and four suit gpt-oss 120B at 2-bit. No NVLink.
| Model | Size | Cards | Total price | Total VRAM | Speed t/s | Total power |
|---|---|---|---|---|---|---|
| Qwen3.8 Flash Next176B (6B active) | 78.9 GB | 8 cards | $3,960 | 128 GB | 68.1 | 1,200 W |
 DeepSeek V4 Flash284B (13B active) | 96.8 GB | 8 cards | $3,960 | 128 GB | 51.7 | 1,200 W |
| GLM 5.3 Flash320B (18B active) | 108.7 GB | 8 cards | $3,960 | 128 GB | 39.9 | 1,200 W |
| Model | Size | Cards | Total price | Total VRAM | Speed t/s | Total power |
|---|---|---|---|---|---|---|
| Qwen3.8 Flash Next176B (6B active) | 111.3 GB | 8 cards | $3,960 | 128 GB | 54.8 | 1,200 W |
At this quantization, even 8 × Tesla T10 cannot hold any model over 50B.
Full calculation rules are on the Methodology page.
Who it suits
Good fit
- Running only small 9B to 12B models
Gemma 4 12B at 4-bit runs at about 40 tokens/s, and Qwen3.5 9B at about 49
- Running Qwen3.6 35B A3B at 2-bit
At 12.3 GB it fits entirely in VRAM, at about 121 tokens/s with 181K of context
- Packing several into a server
Single slot and 150W, so 8 cards draw about 1,200W in total
- Using CUDA on Linux
This generation is still supported by CUDA 13 and vLLM
Not a good fit
- Running 27B and up at 4-bit
The weights of Qwen3.8 27B alone take 16.5 GB
- Using it directly on a Windows desktop
The standard driver doesn't recognize it, so you need a cloud gaming driver or a manually modified driver file
- A regular case with no cooling mods
It is passively cooled and quickly passes 85°C without added fans
Before you buy
There is no official documentation for this card
The T10 is a custom part for GeForce NOW servers. NVIDIA has published no datasheet, so the specs come from GPU-Z readings of the card itself; its PCI device ID is 1E37. Long ago there was also a “Tesla T10 Processor” with the same name, so include 1E37 when searching for drivers and information to avoid turning up that old card.
Verify capacity on listings titled “T10 8G”
In the PCI ID database, 1E37 is listed as Tesla T10 16GB, with subsystems registered as GRID RTX T10-4 and T10-8; there is no 8GB Tesla T10. A listing titled “8G” has most likely mistaken one of these names for the capacity, so check with nvidia-smi when the card arrives.
Drivers: Linux works directly, Windows takes effort
The official Linux driver recognizes it (the 550 driver's README lists the Tesla T10), and once it is installed, CUDA works. The standard Windows driver doesn't recognize the device: one review used AWS's cloud gaming driver, and others have used the GRID driver with NVCleanstall to add 1E37 manually.
Cooling: single-slot passive, dual-slot once modified
The stock card is single-slot, about 27 cm long and passively cooled; without a cooling mod it hits 85°C as soon as it is under load, and above 86°C it throttles. The common desktop fix is to fit an RTX 2080 Ti blower cooler shroud, with the fan connected to the motherboard and its speed controlled in software; after the mod it takes up two slots.
Compared with similar GPUs
Tesla T10 vs Tesla V100 16GB, Tesla P40, and GeForce RTX 5060 Ti 16GB: price, VRAM, bandwidth, compute and local LLM speed compared.
| Spec | Tesla T10 | Tesla V100 16GB | Tesla P40 | GeForce RTX 5060 Ti 16GB |
|---|---|---|---|---|
| Memory | 16 GB | 16 GBSame | 24 GB50% more | 16 GBSame |
| Bandwidth | 403 GB/s | 900 GB/s123% higher | 347 GB/s14% lower | 448 GB/s11% higher |
| BF16 / FP16 compute | 80 TFLOPS | 112 TFLOPS40% higher | — | 95 TFLOPS19% higher |
| INT8 compute | 160 TOPS | 56 TOPS65% lower | 47 TOPS71% lower | 190 TOPS19% higher |
| Qwen3.8 27B speed (2-bit) | 29.4 t/s | 62.6 t/s113% faster | 25.5 t/s13% slower | 32.6 t/s11% faster |
| Largest dense model (4-bit) | 23B | 23BSame | 37B61% more | 23BSame |
| Power | 150 W | 250 W67% higher | 250 W67% higher | 180 W20% higher |
| Interface | PCIe 3.0 x16 | PCIe 3.0 x16 | PCIe 3.0 x16 | PCIe 5.0 x8 |
| Released | 2020-02 | 2017-06 | 2016-09 | 2025-04 |
| 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.
Amazon.jp
Xianyu