News
June 18, 2026
Selecting the right server power supply is a balancing act: enough headroom for CPU and GPU TDP, enough efficiency to control operating cost, but not so much over-provisioning that you waste CapEx and operate the PSU in its least efficient load range. The APEX Titanium family offers two models — 1300W and 2000W — with identical feature sets. The question is which one fits your deployment.
| Parameter | APEX 1300W Titanium | APEX 2000W Titanium |
|---|---|---|
| Max Output | 1300W | 2000W |
| Main Rail Current | 108 A (+12V) | 166 A (+12V) |
| Standby Rail | +12Vsb / 3 A | +12Vsb / 3 A |
| Efficiency @ 50% | 96% | 96% |
| Efficiency @ 20% | 94% | 94% |
| 80Plus Rating | Titanium | Titanium |
| Form Factor | 1U hot-plug | 1U hot-plug |
| Input | 90–264 VAC / 180–320 VDC | 90–264 VAC / 180–320 VDC |
| Max Input Current | 6.5 A @ 100VAC | 10 A @ 100VAC |
| Input Connector | IEC 320 C-20 | IEC 320 C-20 |
| PMBus | 1.2 + PEC + SMBAlert# | 1.2 + PEC + SMBAlert# |
| Cold Redundancy | Yes (CR_BUS) | Yes (CR_BUS) |
| In-System FW Update | Yes (dual-bank) | Yes (dual-bank) |
| Black Box Recording | Yes | Yes |
| FRU EEPROM | IPMI 1.0 | IPMI 1.0 |
The most straightforward metric. A typical dual-socket Xeon server with 8× DIMMs, 4× NVMe SSDs, and 2× network adapters draws approximately 450–600W at full load. Add a single GPU (300–450W) and you're at 750–1050W. Add dual GPUs and you're pushing 1200–1500W.
Both PSUs share the same Titanium efficiency curve. The key insight: a PSU operated well below its rated power wastes CapEx and may run at lower efficiency.
A 600W server running on a 1300W PSU operates at 46% load — near the peak efficiency point (50%). The same server on a 2000W PSU operates at 30% load — still in the 94%+ range, but you've paid for 700W of headroom you're not using. Over a 500-server deployment, that's 350 kW of unused power supply capacity. At roughly $0.30/W for Titanium PSUs, that's $105,000 in overspend.
Both PSUs support N+1 and cold redundancy. The configuration affects how much usable power you get per rack slot:
| Configuration | 1300W: Usable Power | 2000W: Usable Power |
|---|---|---|
| 1+1 (2 PSUs, 1 active) | 1300W | 2000W |
| 2+1 (3 PSUs, 2 active) | 2600W | 4000W |
| 3+1 (4 PSUs, 3 active) | 3900W | 6000W |
A 2U4N server with four independent nodes, each drawing 500W, needs 2000W in an N+1 configuration. With 1300W PSUs, you'd need 3+1 (3900W usable — overkill). With 2000W PSUs, 2+1 gives you 4000W usable with one fewer PSU slot consumed.
Consider the rack power budget, not just the server. A 42U rack with two 30A 208V circuits provides approximately 8.6 kW of usable power (derated to 80%).
| Scenario | Recommended Model | Rationale |
|---|---|---|
| General-purpose enterprise servers (dual Xeon, no GPU) | 1300W | Adequate headroom, lower CapEx, optimal efficiency range |
| Storage servers (high drive count) | 1300W | Drive spin-up peaks are transient; 1300W handles with peak power margin |
| Single-GPU AI inference servers | 1300W | Typical draw 750–1000W. 1300W provides 30%+ headroom |
| Dual-GPU training / HPC nodes | 2000W | Dual A100/H100 draw exceeds 1300W headroom. 2000W required |
| High-density blade enclosures | 2000W | Per-sled power aggregation demands 2000W per PSU bay |
| Edge / Telco (single PSU, no redundancy) | 1300W | Cost-optimized, adequate for edge server TDP |
| Hyperscale (3+1 cold redundant) | Depends on per-node TDP | Calculate total N-node power ÷ N active PSUs. Match to model. |
Both the APEX 1300W and 2000W Titanium PSUs share the same digital management platform — PMBus, cold redundancy, firmware update, black box recording. The decision is purely about the power envelope: under 1000W of sustained server draw, the 1300W saves CapEx and operates at peak efficiency. Above 1000W, the 2000W is the right call — and essential above 1300W.
Contact [email protected] for volume pricing, evaluation units, and technical documentation for both models.
