Used Dell PowerEdge R640/R750 Buying Guide 2026

Last month a client in Austin called after discovering their R640 had a failed power supply that wasn't caught during intake. They'd paid $4,200 for the unit thinking it was fully functional. The bad PSU was a Dell 750W platinum (part number 0TCJN9, roughly $800-1,100 to replace), and worse, the server had already been powered on repeatedly during their deployment testing. We worked out a partial refund and arranged a swap, but that conversation stuck with me. In this market, knowing what to inspect before you deploy makes the difference between a $4K decision and a $5,200 hole.

I've handled over 200 R640 and R750 transactions since 2024. These servers are still workhorses—you see them everywhere from small hosting shops to enterprise hybrid clouds. They're also incredibly common in the secondary market, which means pricing is predictable but competition is fierce. Let me walk you through what actually matters when you're evaluating one.

The R640 vs. R750 Choice

The R640 is a 1U, 2-socket 3rd gen Intel Xeon (Ice Lake era). It handles up to 48 cores per socket, maxes out at 768GB RAM, and draws power like a machine from 2019 because, well, it is. The R750 is the 4th gen refresh—newer BIOS, slightly better power efficiency, supports more memory DIMM slots, and costs $800-1,500 more for similar compute.

Here's my honest take: if you're buying used, the R640 is the sweet spot. Depreciation hit it hard already (2019 launch), and the used market is flooded, driving prices down to $4,200-$6,800 for decently specced models. The R750 (2021 launch) is still shedding value—you're looking at $5,800-$8,200—and new hardware is close enough in price that the math doesn't always work unless you have a specific workload that needs the newer architecture. Most of my R640 sales are data centers consolidating, not end-of-life failures. That's a good sign.

If you're running dense virtualization or need maximum memory, the R750's ability to hold 24 DIMMs instead of 12 matters. But if you're running four to eight VMs per box? The R640 does the job at $2,000 less.

Three Failure Modes You'll Actually Encounter

PSU and Power Distribution Board Failures

This is the single most common issue I see. The R640 uses one of four PSU models depending on configuration: the 495W, 750W, or 1100W variants. Look for part numbers ending in 0TCJN9 (750W), 0YK6F7 (1100W), or 0NPT0N (495W). These fail in a specific way—you won't get total blackout. Instead, the server boots fine, runs for 20-30 minutes, then shuts down unexpectedly. It looks like thermal throttling until you dig into event logs.

When inspecting: power the server on and let it sit under no load for 45 minutes. If it survives, stress-test the CPU with Prime95 or Linpack for another 30 minutes. If the PSU is failing, it'll drop somewhere in that window. The replacement cost runs $800-$1,100 for an OEM unit (Dell will sell you one new). Aftermarket replacements exist at $450-$650, but they're hit-or-miss for warranty and compatibility. I've seen bad ones before.

RAID Controller Battery Degradation

The H740p controller uses an intelligent battery backup (part number 0G3JFT or 0G3K6F depending on SAS revision). The battery holds a charge for five years, not longer. After that, the controller disables write-back caching as a safety measure, and your I/O tanks by 40-60%. You'll notice random latency spikes and slower disk operations even on fast NVMe. If the seller hasn't replaced the battery, factor in $200-$320 for a new one.

Checking is simple: run omreport from Dell's OpenManage tools (free download) and look at the battery health status. "Good" is green. Anything else? Budget for replacement.

Memory DIMM Errors and Latent CPU Faults

I've seen exactly four R640 CPUs that were physically defective and passed initial tests. They'd run fine for weeks, then suddenly throw uncorrectable memory errors or core hangs. The machines are stable at 4-hour stress tests but fail at 72-hour production loads.

The real inspection: get the machine for 48 hours if possible. Run Memtest86+ on all DIMMs (available free, boot from USB) for at least one complete cycle. Then run Cinebench R23 or Geekbench multiple times. Watch the system event logs for microcode exceptions. If you see even one, walk away or negotiate a $1,000-$1,500 discount and plan for a CPU swap (replacement Xeon Platinum 8380 = $2,800-$3,600 retail, or $1,200-$1,800 used).

2026 Pricing Reality

Let me give you numbers from actual transactions in my pipeline:

R640, Dual Xeon Platinum 8280 (28c/s), 256GB RAM, 2x1.92TB SSD: $5,400-$6,200

R640, Dual Xeon Gold 6242 (16c/s), 128GB RAM, Mixed SAS drives: $4,200-$4,900

R640, Single Socket (yes, they exist), Xeon Gold 5220, 64GB RAM: $2,800-$3,400

R750, Dual Xeon Platinum 8380 (28c/s), 384GB RAM, 4x3.84TB NVMe: $7,800-$8,900

R750, Dual Xeon Gold 6348 (16c/s), 192GB RAM, SAS drives: $6,100-$6,900

These prices reflect machines in "Grade B+" condition—cosmetic wear, fully functional, tested, but not pristine. Rack handles might be bent. A few scuffs on the bezel. Everything works.

Grade A machines (nearly new cosmetics) run $600-$1,200 higher. Grade C machines (visible rack damage, scratches on chassis) go $400-$800 lower but you're taking more risk on hidden issues.

The Practical Path Forward

Buy from someone who ran the omreport diagnostics, burned in the CPUs, and tested every DIMM individually. If the seller can't provide that documentation, the price needs to drop $500-$1,000 to justify your own testing load. Don't buy "as-is" used servers at enterprise prices.

The R640 is your value play right now. It's old enough to be cheap, new enough to run modern workloads, and common enough that parts are abundant. The R750 makes sense if you specifically need newer silicon or the extra memory capacity and you're comparing against new hardware.

Contact Caladan Semi at caladansemi.com for current pricing on Dell PowerEdge R640 and R750 servers, power supplies, and replacement components.