GPU server support: how to reach our engineers

GPU server support at FullGPU is available 24/7 and handled by engineers. You reach us through the contact form, or through the channel provided with your server. We help with incidents, with drivers, CUDA, NCCL and networking questions, and with sizing before you order.

How to reach us

• Contact form: for incidents, technical questions, billing, plan changes and cancellations.
• The channel provided with your server: for reporting incidents, as an alternative to the contact form.

Never send passwords or private SSH keys in a request. We do not access the data on your servers except when necessary to operate or repair the service at your request, or when required by law.

What to include in your request

A complete first message saves a round trip. Include:

• the server's public IP address;
• when the problem started, with your time zone (UTC is simplest);
• what you were running and what you expected to happen;
• the output of nvidia-smi, or of amd-smi on AMD Instinct servers;
• any Xid lines from the kernel log. Xid messages are error reports the NVIDIA driver writes to the kernel log, and they help tell a hardware fault from a software one.

nvidia-smi
nvidia-smi -q > nvidia-smi.txt
sudo dmesg -T | grep -i xid

Our GPU monitoring guide explains these tools in more detail.

Incidents and the SLA

Every server is covered by a 99.9% monthly availability commitment for power, network and hardware. Downtime is counted from when we detect an incident or you report it, so report problems as soon as you see them. Failed GPUs and components are replaced as a priority, and when a repair cannot be completed quickly, we offer an equivalent server where available.

If availability falls below 99.9%, request a service credit through support within 30 days after the end of the affected month. See SLA credits and our service level agreement.

Help before you order

You do not need a running server to ask for help. Our engineers can help you size a configuration: how much GPU memory your model needs at a given precision, whether one GPU is enough or you need an 8-GPU server, and which OS image to start from. Describe your model, precision, context length and workload in the contact form. For a first estimate, try the VRAM calculator.

Reporting abuse

To report abuse coming from one of our IP addresses, such as attacks, spam or phishing, use the contact form and include as much detail as possible: the IP address, the time and any evidence. We investigate reports of abuse and, depending on severity, may ask the customer to stop, suspend the server or terminate the service. The rules are in our acceptable use policy.