Capordia3 days
USD 887.571 lot · 2 GPU
180 GB per GPU
Capordia · 2 × NVIDIA B200 SXM rental
B200: two GPUs, a single lot plan.
Capordia brings together two B200s of 180 GB each in one rental lot. Build a capacity for two autonomous jobs or one application explicitly distributed across GPUs.
- 2 GPUs of 180 GB each per lot
- The price already includes both cards
- Three periods to organize your jobs
Direct crypto payment, no required top-up. Capordia accepts crypto payment without requiring an ID document or a KYC process in this ordering flow. An account with first name, last name, email and password is required; this is not an anonymous purchase. In particular, you can choose BTC on Bitcoin or USDT on Tron TRC-20.
View all assets and networksA complete batch, a chosen period
Three packages for 2 × NVIDIA B200 SXM.
7 days
USD 2,071.001 lot · 2 GPU
180 GB per GPU
30 days
USD 7,391.001 lot · 2 GPU
180 GB per GPU
Amounts in USD for the entire batch. Multiply by the number of batches; no automatic renewal.
The right model for your work
Two autonomous jobs or one distributed application
The B200 lot can suit two independent tasks, each sized for its card's 180 GB. For a single distributed application, define instead the data placement, the exchanges, and the recovery criteria. These two organizations are prepared differently.
The two memories do not automatically form a single 360 GB space. Your software must support the distribution, and the topology actually provided must meet its needs. If a single card is enough, compare the H100 or H200 lots before choosing the two B200s.
A budget for your team
A B200 lot for seven days of team work
A single lot already includes 2 GPUs of 180 GB each. Select 1 lot for seven days; the total is not to be multiplied by two again.
USD 2,071.001 lot(s) × USD 2,071.00 · 7 days
2 GPUs in total · 180 GB per GPU
Two options to compare
Choose based on memory and how the work is organized.
NVIDIA H200 SXM 141GB
Choose a 141 GB GPU if your work fits on one card and does not call for distribution. The lot then includes only one GPU.
USD 640.00 · 7 days
1 lot · 1 GPU · 141 GB per GPU
NVIDIA H100 SXM 80GB
Compare an 80 GB card for a task that fits within that capacity. Two autonomous tasks may justify two H100 lots; their budget then equals twice the price of one lot.
USD 475.00 · 7 days
1 lot · 1 GPU · 80 GB per GPU
Prepare your choice before ordering.
Before committing to the plan, have the CPU resources, host RAM, storage, network, topology, region and schedule that your project depends on specified. The displayed quantity does not reserve a future slot. Your team prepares and validates its software; the configuration is not a ready-to-use application. External fees, taxes and commercial documentation remain for you to verify for your purchase.
Your model, your period
Build your rental.
The selector uses the duration from the chosen link. Adjust the number of batches and check the total before continuing.
Two cards, an explicit work plan
Each pack includes two B200 GPUs with 180 GB HBM3e each, on the Blackwell architecture. The combined 360 GB of the pack does not automatically form a single memory space accessible to a program. For distributed processing, the application must handle data placement and inter-GPU exchanges; the topology actually accessible must also be verified.
If your tasks are independent, assign each one a GPU, an identifier and an output location. For a single distributed computation, instead document the model partitioning, the communications and the recovery after interruption. These two modes call for different acceptance criteria.
Validate the Blackwell software stack
Prepare a version of your environment compatible with Blackwell, including compiled extensions for PyTorch or your CUDA application. A successful launch is not enough: verify a representative result, the maximum memory and a restart from a checkpoint. Keep the versions used for this validation.
When to prefer an H200
If the entire workload fits within 141 GB and stays on a single GPU, the H200 SXM is worth comparing. It avoids introducing a distribution scheme solely to use a second device. Conversely, two genuinely autonomous projects can justify the B200 pack even without a distributed application. In that case, compare completed deliverables, not just the reported utilization rate.
Book the duration around your milestones
A 3-day window can cover qualification of both GPUs; 7 days let you plan a first campaign with an intermediate check; 30 days suit a schedule of repeated processing. In each case, allow time for retrieving results.
Select the duration and the number of packs, then the environment and the order details. Check the total number of GPUs in the summary. After payment, track its verification and the preparation of the rental separately before starting your execution schedule.
Choose 2 × NVIDIA B200 SXM
The answers before your rental.
Does the B200 price already include two GPUs?
Yes. A Capordia lot contains 2 B200s with 180 GB each. 3 days: USD 887.57 ; 7 days: USD 2,071.00 ; 30 days: USD 7,391.00. These USD amounts cover both cards in the lot. Only the number of lots multiplies the rate.
Can I use 360 GB as the memory of a single card?
No, not automatically. Each B200 has 180 GB. An application must explicitly handle placement and inter-GPU exchange to distribute its workload; the available topology must also be suitable.
Do I need a distributed application to use a B200 bundle?
Not necessarily. Two standalone jobs can each use one card if your organization and software allow it. A single distributed application, however, requires software support and suitable communication.
When should I compare B200, H100 and H200 for my team?
Start with the memory each task needs and the number of independent jobs. H100 provides 80 GB on one GPU, H200 141 GB on one GPU, and the B200 bundle two GPUs of 180 GB each. Compare the full budget and compatibility; the number of cards alone does not predict performance.