Skip to main contentFrequently Asked Questions
Find answers to common questions about Visent Telemetry, Atlas, and Forge. Get help with setup, troubleshooting, pricing, and best practices for GPU monitoring and optimization.
Can’t find what you’re looking for? Contact our support team at [email protected].
General Questions
What is Visent?
Visent is an intelligent GPU observability suite consisting of three products: Telemetry for real-time monitoring, Atlas for pricing intelligence, and Forge for performance benchmarking. It helps organizations optimize GPU infrastructure costs and performance.
Which GPU types does Visent support?
Visent supports NVIDIA GPUs (Tesla, GeForce RTX, Quadro series), AMD Instinct GPUs, and Intel discrete GPUs. We’re continuously adding support for new GPU architectures as they become available.
Do I need to install software on my GPU nodes?
Yes, Visent Telemetry requires a lightweight agent installed on each GPU node to collect metrics. The agent has minimal performance impact and supports Linux, Windows, and containerized environments.
Is Visent suitable for both cloud and on-premises infrastructure?
Absolutely. Visent works with GPU infrastructure anywhere - cloud providers (AWS, GCP, Azure), on-premises data centers, edge deployments, and hybrid environments.
Telemetry Questions
How much overhead does the Telemetry agent add?
The Telemetry agent typically uses less than 1% CPU and 50MB RAM, with negligible impact on GPU performance. Metrics collection runs in the background without interfering with your workloads.
Can I customize which metrics are collected?
Yes, you can configure the agent to collect specific metrics, adjust collection intervals, and define custom metrics for your applications. See our configuration guide for details.
How long is monitoring data retained?
Standard retention is 30 days for high-resolution metrics and 1 year for aggregated data. Enterprise plans offer extended retention options and data export capabilities.
Can I integrate with existing monitoring systems?
Yes, Visent Telemetry integrates with Prometheus, Grafana, Datadog, and other monitoring platforms through our exporters and APIs. Check our integrations guide for setup instructions.
Atlas Questions
How accurate is the pricing data?
Atlas pricing data is updated in real-time and sourced directly from cloud provider APIs and official pricing pages. We validate data accuracy continuously and maintain 99.9% accuracy for current pricing.
Which cloud providers does Atlas cover?
Currently, Atlas covers AWS, Google Cloud Platform, and Microsoft Azure GPU pricing. We’re expanding coverage to include additional cloud providers and specialized GPU platforms.
Can I get historical pricing trends?
Yes, Atlas provides historical pricing data going back 90 days for trend analysis and forecasting. Enterprise plans include extended historical data and advanced analytics.
Does Atlas include spot pricing?
Yes, Atlas tracks both on-demand and spot pricing across all supported providers, including availability zones and interruption rates for spot instances.
Forge Questions
What types of benchmarks can I run?
Forge supports machine learning training/inference, scientific computing (HPC), graphics rendering, and custom benchmarks. You can also create custom benchmark profiles for specific workloads.
How long do benchmarks typically take?
Benchmark duration varies by type and configuration. Simple GPU tests take 5-15 minutes, while comprehensive ML training benchmarks may take 30-60 minutes. You can configure timeout limits.
Can I compare results across different GPUs?
Yes, Forge provides detailed comparison reports showing performance differences across GPU models, configurations, and optimization settings. Results include statistical analysis and recommendations.
Is there an API for automated benchmarking?
Yes, the Forge API allows you to trigger benchmarks, monitor progress, and retrieve results programmatically. This enables integration with CI/CD pipelines and automated testing workflows.
Technical Questions
What are the system requirements?
Telemetry agent requires Linux (Ubuntu 18.04+, CentOS 7+) or Windows Server 2019+, with NVIDIA drivers 470+ or AMD drivers 21.30+. See our installation guide for detailed requirements.
How do I secure API access?
Visent uses API key authentication with configurable permissions and rate limiting. We recommend storing keys securely, rotating them regularly, and using environment variables. Review our security guide for best practices.
Can I run Visent in air-gapped environments?
Yes, Visent supports air-gapped deployments with local data storage and offline reporting. Contact our enterprise team for deployment assistance and licensing options.
What happens if my internet connection is interrupted?
The Telemetry agent buffers data locally during outages and uploads automatically when connectivity is restored. No monitoring data is lost during temporary network interruptions.
Pricing and Plans
Is there a free tier?
Yes, Visent offers a free tier with basic monitoring for up to 5 GPU nodes, 7-day data retention, and standard API limits. Perfect for small deployments and evaluation.
How is usage calculated?
Usage is typically based on the number of monitored GPU nodes and API requests. Atlas pricing queries and Forge benchmark execution have separate usage limits. Check our pricing page for current rates.
Do you offer enterprise discounts?
Yes, we offer volume discounts and custom enterprise plans for large deployments. Contact our sales team at [email protected] for pricing discussions.
Can I change plans at any time?
Yes, you can upgrade or downgrade your plan at any time. Changes take effect immediately, and billing is prorated for the current period.
Troubleshooting
The Telemetry agent won’t start
Check that GPU drivers are properly installed, the agent has necessary permissions, and firewall rules allow outbound connections. See our troubleshooting guide for detailed steps.
I’m not seeing GPU metrics
Verify that GPUs are detected by running nvidia-smi or equivalent command, check agent logs, and ensure the node is properly registered. GPU detection issues are often driver-related.
API requests are failing
Check your API key validity, rate limits, and request format. API errors include detailed error codes and messages. Review our API documentation for troubleshooting steps.
Benchmark jobs are timing out
Adjust timeout settings, check GPU availability, and verify that the benchmark configuration is valid. Large benchmarks may require extended timeout limits.
Support
- Email: [email protected]
- Community forum: community.visent.ai
- Documentation: docs.visent.ai
- Status page: status.visent.ai
What’s included in support?
All plans include email support and access to documentation. Premium plans include priority support, phone/video calls, and dedicated customer success managers.
Do you provide implementation services?
Yes, we offer professional services for large deployments, custom integrations, and migration assistance. Contact our team to discuss your specific requirements.
Next Steps
