Security and integration.
The health layer sits beside designs already in production. It is small, it stays out of your logic, it runs on your site, and it changes nothing without your approval.
Where it sits
Beside your design, not inside it.
A small, fixed region of the device, next to your design. Your logic, routing and I/O stay as built.
Telemetry flows out to your fleet tools, inside your network. Actions flow back only after your approval, through a handshake.
What moves
Two directions, two rules.
On the card
Measurements are taken at speed while the design computes. Nothing is written into your logic; no debug port is opened.
Out to your fleet
Health data leaves the card in one vendor-neutral schema and lands in your existing monitoring. The platform is deployed on your site, so nothing leaves your network. Telemetry, not a control channel.
Back only with a handshake
Voltage, frequency and repair actions are opt-in, scheduled in your windows, and applied only after an explicit handshake.
The contract
Six guarantees.
Your design stays intact
Logic, routing and I/O stay as built. The health layer occupies a small region reserved beside them. Repair, if you enable it, moves only a path whose element is degrading.
No JTAG exposure
No debug port has to be opened on production systems for monitoring to work.
Zero downtime
The design keeps computing for the whole measurement. A full sweep takes about a second.
Small, on any design
Under 1% of LUTs and 5% of flip-flops on a production shell, under 2% added delay.
Deterministic, scheduled changes
Opt-in, scheduled, handshake-gated, and coordinated per device, per server or fleet-wide.
On your site
The platform runs on your premises. The same health metric, telemetry schema, thresholds and aging model on AMD and Altera parts, all inside your network.
Control
Read-only is the default.
Each further step is one you turn on.
- Read-only by default. Monitoring and aging models never change an operating point.
- Opt-in actions. Voltage scaling, frequency scaling and repair are enabled separately, by policy.
- Your windows. Changes run only in maintenance windows you schedule.
- Your thresholds. You set the alert and action thresholds for your fleet.
- A handshake on every action. Nothing is applied to a running card without an explicit handshake.
FAQ
Common questions.
Do we need to change our design?
Your RTL is not changed. Integration needs a small region reserved beside your design, sized and placed with your team during evaluation.
Does monitoring need JTAG access?
No. No JTAG port needs to be exposed on production systems.
Can we keep it read-only?
Yes. Monitoring and modeling are read-only. Voltage scaling, frequency scaling and repair are separate steps that you choose to enable.
Where does the telemetry go?
Into your existing monitoring, in one vendor-neutral schema. The platform is deployed on your site; nothing leaves your network.
How does the sensing work?
We share the sensor architecture, deployment mechanics and calibration methods with evaluation partners under NDA. Ask for a technical session.
Which devices are supported?
AMD 7-series, UltraScale and UltraScale+ are supported today, with select Altera Agilex 7 devices. See device support for the full list and status.
Measure a device at speed, live.
A 30-minute session: in-system slack measurements on a supported AMD or Altera device, and what an evaluation on your boards involves.