Lenovo ThinkSystem vs Dell PowerEdge: SR630 vs R640

Both figures below come from the vendors' own documentation — Lenovo's published SR630 specifications and product guide, and Dell's R640 technical guide. Read the table looking for differences; there are fewer than the marketing on either side suggests.

This is the comparison that gets quoted most and understood least, so here are the actual numbers, from Lenovo's RAID 930 series product guide and Dell's H730P spec sheet. Both controllers are Broadcom MegaRAID silicon underneath — the vendors buy from the same supplier and rebadge.

What that actually buys you. Two things are real: the extra 2 GB of write cache, and a newer controller generation. On a write-heavy SAS array — a database doing checkpoints, a busy VM host, a backup target absorbing parallel streams — more write cache means fewer stalls when the cache floods, and it shortens the window where latency spikes during RAID 5/6 rebuilds. If your workload is write-bound on spinning disks or SAS SSDs, that is worth something.

What it doesn't buy you is anything to do with port count, and this is where the usual pitch falls apart. A 1U chassis holds ten drives. The 930-24i's twenty-four ports connect ten of them and leave fourteen idle. The H730P's eight ports reach all ten through the backplane's expander, which is how essentially every ten-bay 1U server has worked for a decade. Twenty-four ports matter in a 24-bay 2U chassis with direct-attach cabling. In this comparison they are a spec-sheet number with nowhere to go.

And both controllers sit behind the same PCIe 3.0 x8 link — roughly 7.9 GB/s of theoretical bandwidth, far more than ten SAS drives will ever ask for. Neither card is the bottleneck in a ten-bay build; the drives are.

One more practical note that outweighs both: if you are building all-flash or NVMe, look past both of these controllers. Neither the 930-24i nor the H730P is a tri-mode card, and hardware RAID in front of NVMe SSDs generally costs you more performance than it returns. For NVMe builds, choose the pass-through path — HBA330 on the Dell side, a 430-series HBA on the Lenovo — and do redundancy in software. We covered which drives belong in which tier in our SAS vs SATA vs NVMe guide.

Every used server listing mentions the controller. Very few mention the management license, and on these two platforms it's the difference between administering the machine from your desk and standing in front of it with a crash cart.

Dell iDRAC9 comes in tiers, and the one you want is Enterprise: it unlocks the HTML5 virtual console, virtual media (mounting an ISO over the network to install an OS), and the full alerting and automation set. Express gives you monitoring and IPMI but no remote console. Most refurbished R640s on the market ship with Enterprise already licensed — it was a common original-order option — but not all, and the license is tied to the service tag.

Lenovo XClarity Controller works the same way with different names: Standard, Advanced, Enterprise. Per Lenovo's documentation, remote console requires Advanced or Enterprise; Enterprise adds more simultaneous console sessions and deeper features. The tiering is arguably friendlier than Dell's — you get remote console one step down from the top — but the practical question is identical.

Familiarity counts too, and it's a legitimate tiebreaker. If your team already runs Dell servers, adding an R640 means the same iDRAC interface, the same Lifecycle Controller update flow, the same OpenManage tooling. Introducing a single Lenovo into a Dell shop means a second management stack for one machine. That's not a knock on XClarity — it's a good controller — it's an argument about your runbooks.

Here is where the comparison stops being theoretical. These are live prices and quantities from our own inventory on August 21, 2026 — the same day, the same grading standard, the same warranty on both sides.

Read that top pair carefully, because it cuts against the conventional advice. For $47 more than the Dell, the Lenovo arrives with two extra drive bays, all ten caddies fitted, and a 4 GB flash-backed RAID controller instead of a cacheless H330. Buy the H330 Dell and then add a PERC H730P and two caddies, and the Lenovo is the cheaper machine. That is a specific comparison on a specific day rather than a law of nature — but it is exactly why "just buy the Dell" is lazy advice. Price the actual configurations.

Where the Dell pulls ahead is depth and parts. Across our catalog today there are 47 R640 listings in stock against 34 for the SR630, and the gap on components is far wider than that: a Dell 2.5-inch caddy is $7.67 and we have over seven thousand of them on the shelf. Lenovo trays cost more and turn up in smaller quantities, across the whole secondary market and not just here. Over a five-year life, with drives failing and configurations changing, that difference compounds.

The same logic applies to the machines themselves. If a Dell dies out of warranty, replacing the chassis is a routine purchase at any hour. Some SR630 configurations take longer to source — particularly if you need a matching backplane or riser. Neither is exotic; the Dell is simply more of a commodity.

Most people comparing the SR630 and R640 are also weighing these three, so here's the short version of each.

R630 vs R640 is the comparison worth spending thirty seconds on, because the price gap is real and so is the difference. The R630 is a generation older: Xeon E5 v3/v4 instead of Xeon Scalable, DDR4-2400 instead of 2933, and a memory generation whose used prices are actually lower right now. If your workload is IO- or capacity-bound rather than CPU-bound, an R630 at $243 does the job of a $525 R640 and leaves budget for drives. If you're consolidating VMs, buy the newer cores.

R440 vs R640 comes down to one number: 16 memory slots against 24. For a lightly loaded application server that's irrelevant; for a virtualization host it caps you a third lower on RAM, which in this memory market is the expensive constraint. Buy the R440 only when the price difference is large.

Both machines share the constraints of the form factor, and they catch first-time buyers out more often than any spec difference between them.

• They are loud. A 1U chassis cools two 150 W processors with 40 mm fans spinning at high RPM. At idle in a cool room both are tolerable; under load, or in a warm room, both are genuinely unpleasant to sit near. If the server is going in an office rather than a rack room, a 2U machine with larger, slower fans — an R740 or SR650 — is dramatically quieter for the same compute. This is the single most common regret we hear about 1U purchases.
• Fan behavior reacts to what you install. Both vendors' firmware ramps fans up when it sees a card it can't identify or a drive without vendor firmware — a third-party NIC or HBA can leave a machine running noticeably louder than an identical one with certified parts. It's a real effect, not folklore, and it applies to both platforms.
• Power draw is modest. A dual-Silver build with a dozen DIMMs and a few SAS drives typically pulls 150–250 W under normal load. Dual 750 W supplies are ample; the 1100 W and 1600 W options exist for high-TDP processors and GPU configurations, not for ordinary builds. Buying bigger supplies than you need costs efficiency at low load.
• GPUs mostly don't fit. Neither of these is a GPU platform in any meaningful sense — a 1U takes low-profile cards only, with tight thermal limits. If you're planning inference or rendering, the answer is a 2U or a purpose-built GPU chassis; we sized those in our used GPU server guide.
• Rails and depth. Both are around 30 inches deep with cable management, which rules out shallow wall-mount racks. Confirm the rail kit is included — on the used market it frequently isn't, and buying rails separately can cost more than you would guess for a $500 server.

If none of that is a dealbreaker, a 1U is the most compute-per-rack-unit you can buy secondhand, which is exactly why both of these platforms sold in such volume.

By workload, with the reasoning rather than just the verdict:

Whichever way you go, the checks that save money are the same on both platforms: confirm remote console is licensed, confirm the drive caddies are included (ten missing caddies is a real cost), ask which controller is actually fitted rather than which the platform supports, and check whether the memory installed leaves you free slots in the right channels for later expansion.

Is Lenovo ThinkSystem better than Dell PowerEdge?

Neither is better as a platform — at this class they use the same Intel silicon, the same DIMM counts and the same Broadcom RAID chips, and both are built to the same standard. Dell wins on secondary-market depth: more machines, more spare parts, cheaper caddies, and iDRAC skills that most IT teams already have. Lenovo wins when a specific configuration is better value, which happens regularly. Compare the actual units in front of you rather than the brands.

Is the SR630 equivalent to the R640?

Yes, they are direct competitors: two-socket 1U, Xeon Scalable 1st/2nd generation, 24 DIMM slots, up to ten 2.5-inch bays, three PCIe slots. Dell qualified higher memory ceilings (128 GB LRDIMMs and Optane DCPMM), which matters only for very large memory builds. For most workloads the two are interchangeable.

Is the RAID 930-24i better than the PERC H730P?

It has twice the cache — 4 GB flash-backed versus 2 GB NV — and comes from a newer Broadcom MegaRAID generation, which helps on write-heavy arrays and shortens the latency impact of a RAID 5/6 rebuild. Its 24 internal ports are not an advantage in a ten-bay 1U server: only ten get used, and the H730P reaches all ten bays through the backplane expander. Both cards share the same PCIe 3.0 x8 host link, which neither comes close to saturating with ten SAS drives.

What is the difference between the R630 and the R640?

A generation. The R630 (13th gen) runs Xeon E5-2600 v3/v4 with DDR4-2400; the R640 (14th gen) runs Xeon Scalable with DDR4-2933, supports NVMe backplanes, and has more expansion. The R630 is meaningfully cheaper on the used market — from about $243 with 24 cores against $525 for an entry R640 — and its older memory is cheaper too. Choose the R630 for capacity- or IO-bound work, the R640 for virtualization and newer per-core performance.

What is the difference between the R440 and the R640?

The R440 is the cut-down 1U of the same 14th generation: 16 memory slots instead of 24, and less PCIe expansion. Both take the same Xeon Scalable processors. With DDR4 prices as high as they are in 2026, losing a third of the memory slots is the constraint that bites — buy the R440 only when it's substantially cheaper than an equivalent R640.

Do I need iDRAC Enterprise on a used R640?

If you want to administer the server remotely, yes. iDRAC9 Enterprise unlocks the HTML5 virtual console and virtual media — the ability to see the boot screen and mount an ISO from your desk. Express covers monitoring and IPMI but not the console. The Lenovo equivalent is XClarity Controller Advanced or Enterprise. Always ask a seller which tier a used machine carries; firmware downloads are free from both vendors, but the console entitlement is licensed.

Can I put non-Dell drives in an R640?

Yes, and the same goes for third-party drives in an SR630. Both platforms will flag drives that aren't vendor-certified in the management interface, and the arrays work regardless. What you lose is the vendor's firmware-update path for those drives and a tidy health dashboard. For a production array we'd stick with certified drives; for lab and secondary storage, third-party drives are common practice.

Which is cheaper to run over five years?

They're within noise of each other on power — same processors, same class of Platinum/Titanium supplies. The lifetime difference is parts: caddies, risers, backplanes and replacement chassis are cheaper and more plentiful on the Dell side. If you're buying one machine, that's a rounding error; if you're standing up a rack you'll keep for five years, it's the strongest argument in this whole comparison.