Server refresh is a recurring question in every SMB: the hypervisor has run for five years, will soon drop out of the warranty contract, RAM consumption of the VMs has grown, the SATA SSDs are not full but have slowed down in random IO. The question now is: upgrade or new purchase?
This decision is rarely made data-driven in practice. Often there is either the sales argument “it is older than five years, replace everything” or the cost-saving argument “just add another 256 GB of RAM and keep the box running”. Both are often wrong. This article delivers the criteria for a well-founded decision.
The Six Decision Factors
Six factors determine whether an upgrade or a new purchase makes more sense:
- Platform age (socket, chipset, generation)
- Remaining warranty and support
- Energy efficiency of current vs. new CPU generation
- Spare part availability over the planned lifetime
- Workload trend (is load growing or stable?)
- Investment budget (CAPEX available, or OPEX/leasing?)
Each of these factors alone can force a new purchase. We go through them in order.
Factor 1: Platform Age
A server is not “old” because it has run for five years, but because its platform has expired. Platform here means CPU socket + chipset + BIOS generation.
| Platform | Example | Status 2026 |
|---|---|---|
| Intel LGA3647 (Skylake/Cascade Lake) | Dell R740, R740xd | EOL, replacement CPUs increasingly expensive |
| Intel LGA4189 (Ice Lake / Cooper Lake) | Dell R750, HPE DL380 Gen10+ | medium — still supported, EOL approaching |
| Intel LGA4677 (Sapphire / Emerald Rapids) | Dell R760, R770, HPE Gen11 | current, several years of support |
| AMD SP3 (EPYC 7002/7003) | Dell R7515/R7525 | EOL, but stable CPU replacement market |
| AMD SP5 (EPYC 9004/9005 Genoa/Turin) | Dell R7625, HPE Gen11/12 | current, long roadmap |
Rule of thumb: Once the platform has been replaced by a successor in the market, the upgrade window narrows — spare parts become more expensive and harder to find. Two generations behind current is usually the tipping point toward replacement.
Factor 2: Remaining Warranty and Support
Servers typically come with 3-5 years of workshop warranty from purchase, extendable to 7-8 years (Dell ProSupport Plus, HPE Foundation Care). What happens after expiry?
- Hardware support keeps costing — extensions are possible but priced as a percentage of list (can get expensive)
- Firmware updates may stop — security patches for older iDRAC/iLO versions eventually end
- Service levels degrade — next-business-day becomes rare, parts-only is standard
If the existing server runs out of support and is critical, the service contract must be evaluated. A server without an SLA in production is overconfidence. Sometimes new purchase is cheaper than renewal — sometimes not.
Factor 3: Energy Efficiency
CPU generations become more energy-efficient. A server that drew 250-300 W idle six years ago can today deliver the same performance at 150-180 W. At German electricity prices that adds up to four- to five-figure amounts over five years — depending on server count.
Example calculation (simplified, no PUE):
- Old server: 280 W * 24h * 365d * 5 years = 12.3 MWh
- New server: 170 W * 24h * 365d * 5 years = 7.4 MWh
- Delta: 4.9 MWh
At 0.30 EUR/kWh that is 1,470 EUR additional cost per server over 5 years — without cooling overhead. Across ten servers it becomes a five-figure number.
Important: The numbers above are illustrative, not benchmark figures. Anyone calculating seriously should use the actual consumption of current servers (via PDU or iDRAC) and the spec values of new models.
Factor 4: Spare Part Availability
Even if a server is formally still supported, spare parts get tighter over time. In particular:
- Specific DIMM generations (DDR4 RDIMM for 14th/15th-gen Dell)
- Backplane boards and riser cards
- Proprietary NIC mezzanine cards
- Power supplies with specific connector shape and wattage
Well-maintained vendors like Dell and HPE often guarantee 5-7 years of spare parts supply beyond end-of-sale — after that it becomes a lottery. To run a critical server for 8+ years, buy spare hardware up front (second NIC, spare PSU).
Factor 5: Workload Trend
The most important question: is the load growing or stable?
- Stable / declining (same VM count, same headcount, no new system planned): an upgrade can last a long time
- Growing (more VMs, new apps, AI workload, more staff): the upgrade is a band-aid, new hardware is coming in 1-2 years anyway
SMBs often underestimate the workload trend. “We are not building anything new” — and then a new ERP, BI solution or video surveillance with recording arrives. Anyone buying servers should plan for growth.
Factor 6: Investment Budget
Last but not least, the budget.
| Model | Upside | Downside |
|---|---|---|
| CAPEX new purchase | clear ownership, depreciable | high one-off cost |
| Hardware leasing | OPEX, predictable | more expensive overall |
| Hardware-as-a-Service | service included, automatic refresh | lock-in, hard to swap |
| Refurbished enterprise | very cheap, with remaining warranty | platform usually already EOL |
Example Calculation: Dell PowerEdge R740 in 2026
A typical SMB setup: Dell PowerEdge R740, purchased 2020, with 2x Xeon Silver 4216, 256 GB RAM, 6x 1.92 TB SATA SSD in RAID 10. Support contract ends end of 2026.
Option A: Upgrade (RAM + NVMe)
- Add 256 GB RAM: approx. 1,500-2,500 EUR
- 4x 3.84 TB NVMe add-on in bays: approx. 5,000-7,000 EUR (if backplane supports NVMe)
- 2-year support extension: approx. 2,000-3,500 EUR
- Total: 8,500-13,000 EUR
Option B: New Dell PowerEdge R760
- 2x Xeon Gold 5418Y, 512 GB RAM, 8x 3.84 TB NVMe U.2, NBD 5 years
- approx. 15,000-22,000 EUR depending on configuration
Option C: New Wortmann TERRA server
- Comparable configuration as TERRA server: often noticeably cheaper
- 3 years of warranty in the German service network included
- 2024/2025 platform generation, similar performance
- Wortmann TERRA Server overview
Decision guidance:
- Stable workload + platform replacement already planned in 2 years -> Option A (upgrade) makes sense
- Growing workload + platform suitable for the next 5 years -> Option B or C
- Tight budget + tolerance for “not the premium brand” -> Option C as affordable alternative
Wortmann as an Alternative
Wortmann TERRA servers are a typical SMB sweet spot: German manufacturer with its own service network, off-the-shelf components (no vendor lock-in for DIMMs or NICs), readily available spare parts, transparent configuration. For classic hypervisor hosts with Proxmox VE or TrueNAS storage backends they are a pragmatic choice — as long as specific Dell/HPE platform features (ProLiant integration in OneView etc.) are not mandatory.
What Not to Forget on a Refresh
Whether upgrade or new purchase, these items belong in the plan:
- Migration of existing VMs/data — plan downtime, back up before refresh
- License review — Microsoft, Windows Server, possibly core fees
- Network compatibility — do switch Twinax/optical modules match the new NIC?
- Power in the rack — new servers may have different connector shape (C13 vs. C19)
- Climate in the server room — newer generations can be denser and thus warmer per RU
- Training — new iDRAC/iLO generations have different UIs
Rules of Thumb for a Quick Check
| Situation | Recommendation |
|---|---|
| Server <3 years, platform current, stable load | Upgrade (RAM/NVMe) |
| Server 3-5 years, platform still supported | Upgrade plus support extension |
| Server >5 years, platform EOL | Plan new purchase |
| Workload growing strongly | New purchase, not upgrade |
| Energy cost critical | New purchase often pays back via power |
| Compliance demands current platform | New purchase |
| SMB with tight budget | Evaluate Wortmann or refurb |
Conclusion
Server refresh is an investment decision, not an emotional one. Working through the six factors — platform age, support, energy, spares, workload, budget — leads to a well-founded answer. Often the honest answer is: the upgrade carries you another year or two, then the new purchase is due anyway. Communicating that openly is planning instead of improvising.
DATAZONE advises on these decisions vendor-neutral — we sell both Wortmann and Dell and support existing HPE and Lenovo environments. If you are facing a refresh decision and want a second opinion on the configuration, reach out — we break the six factors down to your concrete environment.
Related Articles
More articles
Server Room Energy Costs: 5 Levers
Save energy in the server room without new hardware. Five concrete levers: hot-aisle / cold-aisle separation, server power profiles, idle VM consolidation, UPS efficiency, and right-sized cooling. Plus: when ARM / EPYC hardware pays off.
Caddy: Reverse Proxy With Automatic HTTPS
Caddy is a modern reverse proxy with built-in Let's Encrypt automation. Ten lines of Caddyfile instead of fifty lines of Nginx — and no certbot. We show setup, examples for three self-hosted services behind one public IP, and the limits compared with Nginx.
Systemd Security: Hardening and Securing Linux Services
Systemd security hardening: unit hardening with ProtectSystem, PrivateTmp, NoNewPrivileges, CapabilityBoundingSet, systemd-analyze security, sandboxing, resource limits, and creating custom timers.