0

I keep hearing all these terms - EN, EP and EX level servers. What do these terms mean? What are they classifications of? How are they demarcated from one another? How are they different from the thin client-thick client server classification used in most literature?

Arpith
  • 274
  • 6
  • 17

2 Answers2

2

It depends on the CPU families;

  • between the 56xx (Nehalem) series CPUs and the Sandy-Bridge Xeons the EP CPUs were designed for two-socket servers, EX CPUs are designed for four-or-more-socket servers. There were also single-socket Xeons sometimes labelled as EP but most were the W range.

  • from Sandy Bridge onwards they're dropping 'EX' and moving to a different naming convention with 'EN' meaning CPUs for one-or-two socket server and 'EP' meaning (confusingly) two-or-four-socket servers.

Four-way 'EP' CPUs can still be used to make >4-socket servers but they need 'glue' chips to manage that, i.e. they can't just do it out of the box.

As for thin/thick-client, there's no direct link - any can CPU can do any role.

Chopper3
  • 101,299
  • 9
  • 108
  • 239
  • What about the various phrases such as entry level, enterprise level and data center level servers? Do these classifications relate to the EN, EP and EX classifications in any way? – Arpith Aug 07 '12 at 07:26
  • This *is* some loose correlation between these CPUs and phrases, certainly nobody would have classed an EX server as 'entry' but those phrases relate more to the server design than the CPUs. Features such as hot-swappable PSUs/fans/memory/CPUs/adapter-cards, a wide range of sensors and the physical quality of the components used is more likely to define whether a server is 'enterprise' class or not, with the opposite being true - a server without these features is very much an 'entry' one. – Chopper3 Aug 07 '12 at 07:30
  • Great answer. Helped a lot. Thanks for clarifying. :-) – Arpith Aug 07 '12 at 07:34
0

[I was going to expand on the answer and comments, especially since EX didn't end with Sandy Bridge. I spent a lot of time figuring out the deal with Intel's product lines and I've tried to condense it down to what may be useful to you, reader from my future.]

It's worth noting this old page explaining their numbering scheme. While not reflective of their current apparent scheme, there are some similarities. Namely the first number is the maximum number of processors. Core count is not easily identifiable from the number, however. The version number is starting from Sandy Bridge (and Westmere-EX, offsetting the lack of a Sandy Bridge-EX), this puts their version one behind the consumer Core generation (except for Westmere-EX).
It's also worth noting that the suffixes EP, EN, and EX are sometimes collectively called E (eg Haswell-E includes Haswell-EN, Haswell EP, and Haswell-EX)

Xeon E3s (previous Xeon 3xxx line) [suffixless/WS] E3-12xx chips use the same socket as their consumer counterparts, but support ECC RAM. there's also the Gladden E3-11xx v1 & v2 lines, and the Skylake-H E3-15xx v5 line which are SoC (BGA 1440, FCBGA 1667 sockets). It's also worth mentioning Broadwell-DE here which are the Xeon D-15xx line (also SoC with FCBGA 1667 socket). There are also a couple SoC E3-12xx series chips (E3-1284L v3, E3-1258L v4, E3-1278L v4) [BGA 1364].

EN series include E5-14xx and E5-24xx series. These are the entry level server socket CPUs.

EP series include E5-16xx, E5-26xx, and E5-46xx series.

EX Series include E7-28xx v1 & v2, E7-48xx series, and E7-88xx series