Compare GPU, CPU and AI accelerator specs

Check the best NVIDIA and AMD prices on Amazon, Newegg, Walmart and Best Buy

Every row above links out to live listings so you can check and compare before you buy — no guesswork, no stale numbers.

Why this site does not list prices

Hardware prices move daily and vary by store, seller and condition. A stale number is worse than no number: it sends you to a checkout expecting one price and finding another. Rather than publish figures we cannot keep current, every row links to the live listing on Amazon so the price you see is the price Amazon is charging right now. The Launch MSRP column is different — it's a fixed historical figure from the manufacturer, shown only as a rough sense of tier, not a claim about what the part costs today.

Once this site is approved for Amazon's Product Advertising API, price-per-unit columns — price per GB of VRAM, per CPU core, per TOPS — will return, sourced directly from Amazon and refreshed on the schedule their terms require.

How to read the specs

VRAM

VRAM is the hard limit on what you can run locally: a model that does not fit in video memory either runs far slower or does not run at all, and no amount of raw compute compensates. For running large language models locally, VRAM capacity usually matters more than the generation of the card.

Cores and threads

Cores and threads matter for data preparation, compilation and classical machine learning rather than neural network training, which is dominated by the GPU.

TOPS

TOPS measures trillions of operations per second on a dedicated inference accelerator. These devices target edge and embedded inference — they are not substitutes for a GPU when training.

TDP

TDP is a rough guide to the power supply and cooling a part demands, and is worth checking before buying a card that will not fit your system's power budget.