Getting the Titanium PSU 30 Up and Running
The Titanium PSU 30 is a mean machine, but the manual glosses over some stuff that matters. I got two of these into rack servers last week after a long week of debugging why one unit would not stay stable at full load, and I am going to share what actually works here. These instructions assume you already have a basic understanding of server hardware. If you do not, start somewhere else. Start by powering down the server completely. This means unplugging the PDU if you have one, not just hitting the power button on the front panel. I learned this the hard way when a hot-swap unit arced across the backplane because the other PSU was still feeding power through the shared rail. The Titanium 30 will sit in either slot A or slot B of a dual-PSU chassis, but only if both slots are rated for 30 amps. Check your chassis manual for per-slot limits before you do anything. Most modern 1U and 2U servers handle it, but a few older Dell and HP models cap slot B at 20 amps and you will get an error code on boot if you ignore that. Slide the unit into the bay until the latches click. Do not force it. There is a metal tab on each side that needs to align with a notch in the chassis frame, and if you misalign that tab by even a millimeter, the connector pins will bend. I bent three pins on my first attempt because I was rushing. It took me twenty minutes to straighten them with a pair of tweezers and a magnifying lamp. Worth noting: the connector is not keyed perfectly, so you can absolutely insert it backwards if you are careless. Look for the small triangular marking on the PSU housing and match it to the notch on the chassis rail.
Once it is seated, connect the power cables. The Titanium 30 uses C13 outlets on the back, so standard server-grade power cords work. If you are running dual feeds from separate PDUs, make sure each PSU is on a different circuit. Redundancy means nothing if both cables terminate at the same PDU that loses power. I wasted an entire afternoon troubleshooting a phantom brownout because someone had plugged both supply lines into the same UPS branch circuit. After the physical install, the firmware handshake is where most people run into trouble. Power on the server and check the BMC or IPMI interface. The Titanium 30 reports its status through the management controller, and you should see a reading of approximately 3000 watts nominal, plus the temperature and fan speed. If the unit shows as "not present" or "faulty," do not immediately assume the hardware is bad. In my experience, about half of those false faults come from a loose ribbon cable connecting the PSU to the system board. On the second unit I installed, the ribbon cable connector on the motherboard side was slightly seated because the retention clip had loosened during shipping. I reseated it, and the unit reported correctly on the next boot. Tighten that clip with a #2 Phillips if it feels floppy. One thing the documentation does not mention clearly is the initial inrush current. When you first apply power, the capacitors on the Titanium 30 will draw a spike that can trip sensitive overcurrent protection on the PDU. This spike lasts about 80 to 120 milliseconds, but it can reach 45 amps. If you are running these off a PDU with a 30-amp breaker per outlet, you will need to sequence the power application. Plug in one PSU first, let it stabilize for about ten seconds, then plug in the second. This usually avoids the tripped breaker scenario entirely.
There are also configuration nuances through the management software. The Titanium 30 supports both active-active and active-passive redundancy modes, and the default is often active-passive, which means only one unit provides power at a time until the other fails. For most deployments, switching to active-active is the better move because it spreads the thermal load and gives you actual redundancy instead of just backup readiness. The setting is under the PSU configuration menu in the BMC web interface. Look for the redundancy mode dropdown and set it to active-active, then save and apply. The change takes effect within thirty seconds and you should see both units reporting load percentages instead of one sitting at zero percent. A word of caution on the cooling. These units run hotter than you might expect under full load, especially in confined 1U spaces. I measured the exhaust air coming off the Titanium 30 at roughly 55 degrees Celsius when both units were under 90 percent load in a densely packed rack. That is within spec, but if your rack has poor airflow management, that heat will stack. Make sure you have blanking panels in every empty slot behind the PSU bays. Uncovered openings let hot exhaust air recirculate back into the intake path, and that will cause the unit to throttle down its output to protect itself. I saw this happen on a unit that would only deliver 2200 watts instead of its rated 3000 watts because the exhaust was being pulled right back in through an open slot. Adding a simple blanking panel brought it back to full output immediately. If you are installing more than one Titanium 30 in a single server, the load balancing between them is managed automatically by the firmware, but it is not perfect. You will typically see a 3 to 5 percent variance in power delivery between the two units under steady load, which is normal. If the variance exceeds 10 percent, check the input voltage on each leg. A 120-volt circuit that is brownouting will cause one PSU to pull significantly more current than the other, and the load sharing will appear unbalanced. Run a voltage check at the PDU outlet with a multimeter while the server is under load. If one leg is reading below 114 volts, you have a wiring issue upstream, not a PSU issue.
Get the Full Details
The warranty process is straightforward but slow. Keep your serial number and the date of installation handy. Sapiens AI support typically responds within two business days, but replacement units can take up to ten business days to ship depending on your region. There is no expedited option unless you have a contractual SLA with them. If your unit fails within the first 48 hours, try a firmware update before requesting a replacement. I had a unit that exhibited random fault codes on the BMC, and a firmware update from version 2.14 to 2.17 resolved it completely. The changelog for 2.17 specifically mentions fixes for false fault reporting during high-load transitions, so it is worth checking the current firmware version against what Sapiens AI has posted before you start the RMA dance. For the actual download of the installation guide, head to the Sapiens AI support portal and search for "Titanium PSU 30." The PDF is about 14 megabytes and includes diagrams for the latch mechanism, the ribbon cable routing, and the BMC configuration screens. It is more useful than the quick-start sheet they ship in the box, which skips the ribbon cable detail entirely and assumes you already know how to access the management interface.
