Service status & availability Crypto-only billing · No-KYC signup

EPYC bare metal

EPYC 7513 dedicated server

A 32-core EPYC host with 256 GB listed ECC memory and four NVMe devices, from $263 per thirty-day term.

ProcessorEPYC 7513
Compute32c / 64t
Memory256 GB ECC
Storage4×3.84 TB NVMe
Network10 Gbps unmetered
AddressingIPv4 /29 · IPv6 /64

What this configuration is suited to

EPYC 7513 balances dedicated compute with local NVMe. Choose it when predictable CPU scheduling, fixed transfer cost and direct control of the operating system matter more than minute-by-minute elasticity.

Included with the server

  • Unmetered network port
  • IPv4 /29 block
  • IPv6 /64 allocation
  • Full IPMI/KVM access
  • Baseline DDoS protection

Operating systems

Ubuntu, Debian, Rocky Linux, AlmaLinux, Windows Server, Proxmox VE, VMware ESXi, TrueNAS, pfSense, Custom ISO. Commercial software licenses are not included unless stated on an order.

Available locations

Run it in Dallas, Ashburn or Portland.

Provisioning estimates can vary by hardware availability. A final region check happens before the ledger is debited.

Before you choose

Where this configuration makes sense.

3 min guide

More devices create a design choice, not an automatic layout

The EPYC 7513 configuration provides four NVMe devices with 15.36 TB combined raw capacity. Compared with a paired-device host, that creates more ways to allocate storage roles or implement redundancy, subject to the operating system and controller configuration.

Decide whether the application needs one pool, separate workloads or a particular recovery arrangement before installation. More devices do not establish independent failure domains for the whole server, and an unconfirmed layout should not be treated as delivered usable capacity.

Check whether additional cores need additional memory

The 7513 increases core count over the 7443P while retaining the same listed memory amount. That can suit workers or guests whose extra parallelism does not require proportionally more memory. It can also leave memory as the first limiting resource for a different workload.

List peak demand per guest or worker and include the host, storage services and maintenance. A larger CPU allocation is not a license to add guests until the machine becomes difficult to recover. Preserve a realistic margin for failure investigation and restoration.

Include background I/O in the acceptance test

Evaluate the application while normal backup, logging and maintenance work is running. A storage design that appears sufficient in a quiet test may behave differently when it also writes backup data or rebuilds working files. Use the real job mix rather than a peak device specification.

The 10 Gbps port can help make network capacity part of the shortlist, but achieved transfer rates require measurement across the complete path. Keep external backups and verify that the restore destination can receive data at the rate your recovery plan assumes.

Choose between storage layout and more memory

Compare EPYC 7443P when its smaller core and device allocation already meets the requirement. Compare EPYC 9354P when more memory is the priority: it changes the memory platform and uses fewer, larger NVMe devices despite the same raw-capacity total.

That trade-off is more informative than treating the processors as a simple ascending ladder. Choose the memory and disk arrangement that supports your operational design, confirm the required software and controller behavior, and include migration effort when replacing an existing host.

Direct answers

Questions before you order

Does four NVMe mean RAID 10 is already configured?

No. It describes the available devices. The installed storage layout must be agreed and confirmed, and its usable capacity calculated separately.

Is this always a better virtualization host than the 7443P?

No. It adds cores and devices but not a larger listed memory allocation. Choose it when the intended guests benefit from those changes and still leave sufficient host headroom.