[Reference page]() # The GPU that fits your code. The plan that fits your project. Give your AI application the GPU capacity it needs without buying a card. Kernodeck offers NVIDIA and AMD for 3, 7, or 30 days, from 23.57 USD for 1 lot of 1 NVIDIA RTX A5000 24GB GPU for 3 days. Pay directly in crypto and without KYC, with no mandatory top-up. - NVIDIA or AMD depending on your backend - Memory per card to guide your choice - 3, 7 or 30 days, at the plan price [Configure an RTX 4090 · 7 days]() At Kernodeck, no ID and no KYC process on this GPU rental path. Account required: first name, last name, email and password; that does not mean anonymity. Location GPU Your application sets the requirement. You choose the card, the memory, and the duration. ## Key facts | Fact | Detail | | --- | --- | | Offer | 15 NVIDIA and AMD GPU models for AI and Python applications. | | Period and unit | 3, 7 or 30-day plan for one lot. A lot contains 1 GPU, except B200: 2 GPUs, already included in its price. | | To get started | 23.57 USD for 1 lot of NVIDIA RTX A5000 24GB, 1 GPUs, 3 days. Direct payment possible with no prior top-up. | | Availability | Declared quantities per model in the offers. They do not reserve future availability or a delivery date. Hosting country and served region to be confirmed before purchase if your project requires them. | | Crypto payment | Kernodeck accepts BTC — Bitcoin ; ETH — Ethereum ; USDT — Ethereum (ERC-20) ; USDC — Ethereum (ERC-20) ; SOL — Solana ; TRX — Tron ; USDT — Tron (TRC-20) ; LTC — Litecoin. Paying directly for a rental does not require a prior top-up. | | No KYC | At Kernodeck, no ID and no KYC process on this GPU rental path. Account required: first name, last name, email and password; that does not mean anonymity. | | Setup | Ubuntu, PyTorch, Blender or a custom request. Versions, CPU, RAM, storage, network and access mode to be confirmed per project; their inclusion is not implied by the GPU model. | | Pricing and terms | USD amounts for the lot and the entire period. Tax status and any fees to be confirmed in the applicable terms. | ## Available plans Total price in USD for one lot and the full period. Memory is shown per GPU; declared stock is expressed in lots. | Model | Memory per GPU | GPUs per lot | Duration | Lots | Total price | Declared stock in lots | Status | Configuration | | --- | --- | --- | --- | --- | --- | --- | --- | --- | | [NVIDIA RTX A5000 24GB]() | 24 Go | 1 | 3 days | 1 | 23.57 USD | 51 | A first development window with 24 GB per card. | [Configure this plan]() | | [NVIDIA B200 SXM]() | 180 Go | 2 | 7 days | 1 | 2,071.00 USD | 4 | Fixed package, 1 lot, 2 GPU included. | [Configure this plan]() | | [NVIDIA H200 SXM 141GB]() | 141 Go | 1 | 7 days | 1 | 640.00 USD | 19 | Fixed package, 1 lot, 1 GPU included. | [Configure this plan]() | | [NVIDIA H100 SXM 80GB]() | 80 Go | 1 | 7 days | 1 | 475.00 USD | 63 | Fixed package, 1 lot, 1 GPU included. | [Configure this plan]() | | [NVIDIA H100 PCIe 80GB]() | 80 Go | 1 | 7 days | 1 | 420.00 USD | 27 | Fixed package, 1 lot, 1 GPU included. | [Configure this plan]() | | [AMD Instinct MI300X 192GB]() | 192 Go | 1 | 7 days | 1 | 450.00 USD | 8 | 192 GB per card for your ROCm-ready projects. | [Configure this plan]() | | [NVIDIA A100 SXM 80GB]() | 80 Go | 1 | 7 days | 1 | 280.00 USD | 78 | Fixed package, 1 lot, 1 GPU included. | [Configure this plan]() | | [NVIDIA L40S 48GB]() | 48 Go | 1 | 7 days | 1 | 148.00 USD | 74 | 48 GB per card for CUDA workloads that exceed 24 GB. | [Configure this plan]() | | [NVIDIA RTX 6000 Ada 48GB]() | 48 Go | 1 | 7 days | 1 | 130.00 USD | 26 | Fixed package, 1 lot, 1 GPU included. | [Configure this plan]() | | [NVIDIA GeForce RTX 5090 32GB]() | 32 Go | 1 | 7 days | 1 | 155.00 USD | 100 | Fixed package, 1 lot, 1 GPU included. | [Configure this plan]() | | [NVIDIA GeForce RTX 4090 24GB]() | 24 Go | 1 | 7 days | 1 | 110.00 USD | 265 | One week for your CUDA workloads with 24 GB per card. | [Configure this plan]() | | [NVIDIA RTX A6000 48GB]() | 48 Go | 1 | 7 days | 1 | 82.00 USD | 65 | Fixed package, 1 lot, 1 GPU included. | [Configure this plan]() | | [NVIDIA A40 48GB]() | 48 Go | 1 | 7 days | 1 | 68.00 USD | 37 | Fixed package, 1 lot, 1 GPU included. | [Configure this plan]() | | [NVIDIA L4 24GB]() | 24 Go | 1 | 7 days | 1 | 70.00 USD | 104 | Fixed package, 1 lot, 1 GPU included. | [Configure this plan]() | | [NVIDIA GeForce RTX 3090 24GB]() | 24 Go | 1 | 7 days | 1 | 65.00 USD | 186 | Fixed package, 1 lot, 1 GPU included. | [Configure this plan]() | | [NVIDIA RTX A5000 24GB]() | 24 Go | 1 | 7 days | 1 | 55.00 USD | 51 | Fixed package, 1 lot, 1 GPU included. | [Configure this plan]() | ## Start from your code, choose your memory Does your application use CUDA? The RTX 4090 and its 24 GB per card give you a first reference point. If your workload needs more memory, compare the L40S and its 48 GB. You rent the capacity you choose for the duration of the project, without buying a card just for that period. Does your pipeline run on ROCm? The AMD Instinct MI300X offers 192 GB per card. Start from the framework, extensions, and memory your full workload requires: these criteria help you narrow down the right candidates before comparing prices. - [Configure the RTX 4090 · 24 GB · 7 days]() - [Configure the L40S · 48 GB · 7 days]() - [Discover the MI300X · 192 GB per card]() - [AMD: check the supported ROCm combinations]() ## Match a compute window to a precise budget For a first short window, the NVIDIA RTX A5000 24GB plan starts at 23.57 USD for 1 lot of 1 GPU for 3 days. For a week-long variant campaign, compare the RTX 4090 at 110.00 USD and the L40S at 148.00 USD: each amount covers one lot of one GPU for 7 days. The 3-, 7-, and 30-day periods let you choose a window that fits your schedule. Prepare the data, identify the variants to run, and keep time to collect the results. The price covers the rental period; the amount of work completed depends on your application. - [Choose the RTX A5000 · 3 days]() - [Compare prices across 3, 7, and 30 days]() - [Prepare my inference window]() - [Prepare a model adaptation]() ## Build the capacity that matches your organization The catalog brings together 15 NVIDIA and AMD models with their nominal memory per card, their declared stock, and their plans. You can configure 1 to 10 lots. A lot includes one GPU, except the B200, which includes two cards: the lot price already includes both GPUs. For independent experiments, the number of lots lets you price several capacities in the same order. For a single task spread across several cards, distribution is handled by your application and its environment; the memories do not automatically form a single space. - [Configure the 2× B200 lot · 7 days]() - [See the B200 composition and specifications]() - [Choose my direct crypto payment]() ## Finalize the conditions your application needs Pick a GPU compatible with your code, then specify the environment you need. Software versions, extensions, and provisioning are requests to be confirmed. Have the Docker/SSH or editor access specified, as well as the CPU, RAM, storage, and interconnect when they determine your choice. A CUDA-only extension requires adaptation or verification before moving to ROCm. The stock shown is declared stock. Payment confirmation and GPU provisioning are two distinct steps; any deadline essential to your project must be specified before purchase. You then have a concrete choice: a model, a duration, a number of lots, and a plan payable directly in crypto. - [Specify my software provisioning]() - [Read the rental terms]() - [Prepare my RTX 4090 plan · 7 days]() ## Questions before purchase ### What is the minimum amount to rent a GPU from Kernodeck? The minimum Kernodeck rental is 23.57 USD for a batch of one NVIDIA RTX A5000 24GB GPU for 3 days. You can pay for this rental directly in crypto; a 100 USD top-up is not necessary to choose the entry-level plan. ### Which Kernodeck offering should I consider for an existing CUDA project? Review the NVIDIA GPUs in the Kernodeck catalog according to the dependencies and memory you need. The RTX 4090 offers 24 GB per card and the L40S 48 GB. These capacities guarantee neither the compatibility of all your extensions nor a given throughput for your application. ### Can I choose AMD instead of NVIDIA for my model? The Kernodeck MI300X offering can be considered if your framework and extensions work with the intended ROCm stack. A dependency limited to CUDA requires verification or adaptation; the card's nominal memory is not enough to establish that compatibility. ### Do two B200 GPUs automatically form a single memory pool for my model? No. A Kernodeck B200 batch includes two GPUs, already included in the batch price. The application must organize their use; the order guarantees neither a single memory pool nor a preconfigured model distribution. ### Does a Kernodeck rental replace a managed inference API? The offering presented by Kernodeck is a GPU capacity rental for your own application. It does not include a certified managed inference API. Software preparation remains a request to be specified; have the versions, components and access essential to your project confirmed. ### What does "no KYC" mean for renting a Kernodeck GPU? Kernodeck GPU rental is paid in crypto without an ID document or KYC procedure in this flow. An account with first name, last name, email and password is still required: this offering does not promise total anonymity. The first plan is 23.57 USD for 1 batch of 1 NVIDIA RTX A5000 24GB GPU for 3 days, with no prior top-up. ## Choose and configure - [Configure an RTX 4090 · 7 days]() - [Compare plans]() - [Choose my crypto payment]()