• 🎉 Big News: Circle City Servers is backing our CloudFest Americas coverage. Read more

Which CPUs are you running in your entry-level dedicated servers lately?

MladenVM

MladenVM

Member
Just curious what the go-to CPUs are these days for low to mid-range dedi boxes?

Are folks still sticking with older Intel stuff like 6152/6262? Or has everyone moved on to EPYC chips like 7713, 9274F, etc.?

For those of you reselling or deploying for smaller hosting clients, are you prioritizing:
  • Core count?
  • Single-thread performance?
  • Power efficiency?
  • Availability / cost?
I’ve been seeing a mix of older Intel dual CPU setups and newer EPYC builds with crazy RAM configs (512GB+). Curious to hear what’s working best for you.

Also, are people still asking for dual CPU setups, or has that faded?
 
I can say the following - Intel processors of the old generation are enough for Low-medium servers. What are the customers looking for? As a rule, those who are just starting work (although it often happens that large projects) are looking for cheap servers in order to invest less and earn more, but if you still have a fairly heavy project, then you need more cores and their efficiency.
 
For entry-level dedicated servers, I’ve seen a good balance in using single-socket AMD EPYC 7402P / 7443P for efficiency and modern architecture, especially when clients prioritize high core counts with decent single-thread performance.

Older Intel Xeon setups like 6152 still have a place in budget-friendly deployments, mainly when availability and cost are the main factors, but power efficiency and performance-per-watt are pushing more users toward EPYC.

For smaller hosting clients, I usually prioritize core-to-price ratio + stability over pure single-thread speed. Dual CPU requests have dropped a lot unless the workload specifically needs massive parallelism or huge RAM configs.
 
Single-socket CPUs with relatively high clock speeds matter a lot in these entry tiers. Higher boost clock (4-5 GHz) shows up frequently. Core counts are moderate: generally 6-8 cores (12-16 threads) unless you're going for a more performance-dense “lower mid tier” server. Providers seem to strike a balance: enough cores for parallel tasks but keeping latency/single-threaded performance respectable.
 
Most folks I see are leaning into EPYC now - especially 7313P or 7453 for mid-range. Older Intel like 6262V3 still pops up in budget dual-socket boxes, but new deployments? Rare. Power efficiency's a nice bonus, but cost and availability drive decisions more than specs these days.
 
These days most deployments have shifted toward EPYC-based servers rather than older dual Xeon setups like 6152/6262, mainly because a single modern EPYC CPU offers higher core density, better memory bandwidth, more PCIe lanes, and improved power efficiency, which is especially important for VPS hosting and multi-tenant environments.
 
For low to mid-range dedicated boxes, many still rely on older generation processors for cost and availability, especially where workloads are not overly demanding on single-thread performance. However, newer server platforms offer significantly better power efficiency and higher core counts for tasks that can utilize them. It largely comes down to the specific client's application needs and balancing initial hardware cost with operational expenses. Dual CPU setups are much less common now unless there's a very specific legacy application or extreme I/O requirement.
 
Older dual Xeons like Gold 6152/6262 are absolutely still around, particularly where the chassis/RAM/CPU inventory is already depreciated. Leaseweb, for example, still has plenty of older dual-Xeon systems alongside current EPYC hardware.
 
One dimension that has not come up: the answer changes by market, and not
because of chip prices.

In several of the markets we operate in - Jakarta, Manila, and parts of
Latin America - the binding constraint is not cost per core, it is
kilowatts per rack. Plenty of facilities there are still provisioned
around 3 to 5 kW a rack, where Frankfurt or Ashburn will sell you three
times that without blinking. Once you are power capped rather than space
capped, a 400W part stops looking efficient no matter how good its
performance per watt is on paper, because the number you actually care
about is usable cores per rack, and that is set by the power envelope.

That flips the Intel versus EPYC arithmetic depending on where the box
lands. The same SKU comparison can come out differently in two of your
own sites, which is not something you can reason about from a spec sheet.

On dual socket fading - agreeing with voxfor and derek_carter, with one
addition from the VPS side. NUMA makes per-vCPU performance depend on
where a guest happens to land, which is awkward when the thing you are
selling is a predictable vCPU. Single socket is easier to promise, not
just cheaper to buy.

Provider side here, so read that as our experience rather than a general
rule.
 
For a hosting business, I’d pay less attention to the CPU name alone and compare cores/threads, sustained performance, RAM, NVMe storage, network speed, and provider-side limits. For instance: AMD Ryzen 5 / Ryzen 7 are popular for budget servers where strong single-core performance matters; AMD Ryzen 9 is often found in affordable performance-oriented dedicated servers; as for Intel Xeon E-series they are still common in traditional entry-level server configurations.
 
You must log in or register to reply here.

Sponsor

Supporters

Back
Top