AMD EPYC 7543 32-Core 2.8GHz Socket SP3 Processor: Milan, 256MB L3, 225-240W Configurable TDP
Enterasource is proud to present the AMD EPYC 7543, a 32-core third-generation EPYC processor on socket SP3 with 64 threads via SMT, a 2.8 GHz base clock, 256MB of L3 cache and eight DDR4-3200 memory channels. The specification most often stated wrongly for this part is its power: AMD documents a configurable TDP of 225 to 240 W, not a fixed 225 W, and the chassis thermal solution has to be rated for the ceiling rather than the default.
Configurable TDP Is 225 to 240 W — Size the Cooling to the Ceiling
AMD publishes the configurable TDP range for this processor as 225-240 W. The server vendor sets the operating point within that range, typically in firmware, depending on the platform’s thermal design.
That means a chassis whose heatsink and airflow are specified for 225 W may not carry this processor at its upper configured point. If you are matching a heatsink, planning airflow, or sizing a thermal budget, use 240 W. This is the single most useful non-attribute fact about the part and it is AMD’s own figure.
Core, Thread and Cache Configuration
| Cores | 32 |
|---|---|
| Threads | 64 — simultaneous multithreading (SMT) |
| Base clock | 2.8 GHz |
| L3 cache | 256 MB |
| Socket | SP3 |
| Socket count | 1P and 2P |
| Series | AMD EPYC 7003 (Milan) |
A terminology note that matters for anyone cross-shopping against Intel parts: AMD’s two-threads-per-core technology is SMT — simultaneous multithreading. "Hyper-Threading" is Intel branding for the equivalent feature and AMD does not use it. The 32-core / 64-thread pairing is AMD’s own.
Memory Subsystem
| Memory channels | 8 |
|---|---|
| Memory type | DDR4-3200 |
| Peak memory bandwidth | 204.8 GB/s per socket |
Eight channels per socket is the defining architectural feature of this EPYC generation against its contemporaries, and it is why populating all eight channels matters more here than on a six-channel platform — a partially populated EPYC gives up bandwidth in proportion to the channels left empty.
AMD publishes no maximum memory capacity figure and no ECC field anywhere in its server-processor dataset, so neither appears on this page. Your platform’s documentation is the authority on both.
PCIe
PCIe 4.0, with AMD’s published lane count for this processor being x128.
We state AMD’s figure and stop there. There is a widely repeated gloss about how lanes are apportioned between expansion and inter-socket fabric in a two-socket configuration; whatever its merits, it is not in AMD’s dataset and we do not publish it as a specification.
Platform and Socket — and the One Caveat We Are Not Turning Into a Claim
AMD documents the platform as SP3, supporting one-socket and two-socket configurations. That is the fitment claim the manufacturer supports and the one we make.
A caveat worth being aware of as a buyer, stated as context rather than as a specification: socket SP3 also carried the earlier EPYC 7001 and 7002 generations, so socket compatibility and platform support are not the same question on this socket. AMD publishes nothing about firmware requirements, so we make no claim about them — but check your server vendor’s processor support list for your specific board and firmware level rather than relying on the socket alone.
The Two Order Part Numbers, and Which One This Is
AMD publishes two ordering part numbers for this processor: a tray OPN and a boxed OPN. This unit matches the tray OPN.
We state which one it matches and stop there. AMD does not describe packaging contents in this dataset, so we make no statement about what is or is not supplied with a tray processor — ask us and we will tell you what is physically in stock.
Condition Disclosure: This Is a Dell-Locked Processor
This unit is a Dell-locked tray processor.
We disclose it because it is decisive: a vendor-locked processor is intended for that vendor’s platforms, and buying one for a different vendor’s board is the most consequential mistake available on this listing. If you are building on a Dell PowerEdge platform, this is a normal part. If you are not, this is not the part for you.
This is our condition disclosure about the unit, not a manufacturer specification — AMD publishes nothing about OEM processor locking, and we are not going to explain a mechanism we cannot source.
Workloads AMD Names for This Processor
From AMD’s own workload affinity field: analytics; CAE, CFD and FEA; ERM, SCM and CRM applications; media streaming; medium-capacity data management including non-relational and relational databases; VDI; and virtual-machine density.
The 32-core, 256MB-L3, eight-channel shape is a general-purpose virtualisation and database profile rather than a frequency-optimised one — per-core licensing cost and per-socket density are the two axes it is designed against.
Supported Technologies
AMD Infinity Guard and AMD Infinity Architecture.
Important Notice: this is a Dell-locked tray processor for socket SP3. It is intended for Dell PowerEdge platforms and is not suitable for other vendors’ boards. Configurable TDP reaches 240 W, so the thermal solution must be rated accordingly. Every processor is tested before it ships and is covered by the Enterasource warranty shown on this page.
