MatPool vs TensorDock (USA): which is better?

Price, cashback, and pros/cons comparison across two platforms · Data updated: 2026-10-26 · Always refer to the order page for final details.

GPU model	MatPool	TensorDock (USA)
RTX 3090	¥1.85/hourPay-as-you-go	¥1.44/hourLowerPay-as-you-go
RTX 4090	¥2.20/hourLowerPay-as-you-go	¥2.52/hourPay-as-you-go
RTX 5090	¥2.99/hourLowerPay-as-you-go	¥4.10/hourPay-as-you-go
T4 16GB	¥1.35/hourLowerPay-as-you-go	—
A6000 48GB	¥2.85/hourLowerPay-as-you-go	¥3.24/hourPay-as-you-go
Cashback	Voucher Tiered rebates (5%–20% on referral spending, valid for 180 days, issued as Compute Beans)	Voucher No public referral program available
	Rent on MatPool	Rent on TensorDock (USA)

How to Choose:MatPool (Competitive List Pricing)— RTX 4090 listed prices are competitive; billing starts only when the instance is running; flexible daily/weekly/monthly plans. Note: documentation is average, and multi-GPU support is limited.
TensorDock (USA) — Global Low-Cost Option—Officially lists the lowest starting price globally (H100 SXM at $2.25/hour), offering 45 GPU models across 20+ countries and 100+ data centers, with $5 minimum top-up and per-second billing. Note: Popular low-cost GPUs frequently run out of stock—expect wait times; uses third-party hosted infrastructure; foreign currency payment required.
Prices change hourly—don’t rely on a single screenshot:Real-time price comparison table on homepageAlways Sorted by Price — Also Supports Subscribe to Price Drop Alerts— Automatic Email Alerts on Price Drops.

More comparisons:AutoDL vs Gpushare · AutoDL vs MatPool · AutoDL vs Compshare · AutoDL vs Suanjiayun · AutoDL vs Paratera · AutoDL vs Vast.ai (USA) · Side-by-side comparison of four major platforms · Real-time price comparison across all 26+ platforms