Westmere EP Die: 6 cores (12 threads) and 12mb L3 cache

Xeon 5600 Series "Westmere EP" Processor

Intel has recently been releasing some new 1P desktop processors based on the new 32nm "Gulftown" core, but today is the day that the 2P server and workstation space gets its shot in the arm. As you have undoubtedly seen in reveiws of its desktop counterpart, the Westmere core brings improved performance to the table via both process improvments and architecture enhancements and additions. Let's face it... 50% more cores and 50% more L3 cache are never a bad thing either.

One of the major enhancements on the Westmere core is the addition of AES-NI. AES-NI adds six new instructions that are designed to aid in encryption. Intel has stated that things like whole drive encryption that have been too slow and cumbersome to implement on the end user side are within reach since those functions are now supported on the processor itself.

You may remember that the Nehalem core supported three channels of DDR3 memory, but you could only acheive maximum speed (1333mhz) with a single DDR3 DIMM per channel. With Westmere, Intel has tweaked the memory controller for performance, so now you can still get to that 1333mhz number with two DIMMs installed per channel. An end user can now buy cheaper, lower density DIMMs without fear of performance degradation. The new and improved memory controller also supports new low-voltage DDR3 memory modules.

Since the push is still on for more performance using even less power, Intel has added a power gate to the Westmere core. The new power gate allows the processor to completely turn off parts of the "uncore" when the processor is idle, using less net power.

Other highlights of the Westmere core include improved virtualization performance, PCLMULQDQ instructions, an Always Running APIC timer, support for larger (1gb) pages and PCID (Processor Context ID). I'm still not sure I understand what all the improvments are or what all they do, but suffice it to say that Intel has poked, prodded and tweaked Nehalem's core, shrunk it and added more cores and cache and arrived at Westmere.

There are a bevy of new SKUs for Westmere EP in LV, 4 core (Yes. I said 4 core), and 6 core configurations. Rather than trying to explain all the differences and how the new SKUs match-up with the "old" Nehalem SKUs, I'll post a nice chart provided by Intel that breaks it all down graphically.

Test System

12 cores (24 threads), 24mb L3 cache and 24gb of 1333mhz DDR3

One of the main benefits of Westmere EP is pin compatibility with existing Nehalem EP motherboards. As long as your current motherboard was built to Intel's specs (for voltage, etc), Westmere EP is a drop-in replacement for Nehalem EP. Because of this the only change to the test system configuration throughout the entire review process were the actual processors themselves. Everything else remained consistent for all tests.

The system I used was the same Supermicro workstation used in my previous Nehalem EP review. I was required to change out the motherboard (X8DAI) for a newer revision (rev 1.31) that supports Westmere EP. Apparently Supermicro was NOT one of the companies that followed the guidelines on the original Nehalem motherboards, but both pairs of processors were tested in the new motherboard with the exact same auxiliary hardware.

For the sake of comparison I used the current fastest Nehalem EP based Xeon (W5590) which matches-up quite nicely with the fastest (6 core) Westmere EP based Xeon (X5680). Both the W5590 and the X5680 run at 3.33ghz with a 6.4GT/s QPI speed.

Other key features of the test system include:

- Windows Server 2008 Enterprise x64 (configured as a workstation)
- Quadro FX3400 with latest NVIDIA reference drivers
- 6x 4gb Nanya 1333mhz DDR3
- Ablecom 805W PSU

Again, for the purposes of this review, all graphs will show the processors by their official designation; X5680 for Westmere EP and W5590 for Nehalem EP.

Workloads

I used the following versions of our test applications:

- SiSoftware SANDRA 2010c Lite
- Black & Scholes Kernel v5.0
- SunGard Adaptiv Credit Risk Analysis v4.0
- CINEBENCH R11.5 x64
- POV-Ray v3.7 beta 35a x64
- SPECjbb2005 v1.07
- Euler3d CFD Benchmark v2.2
- Matrix Multiply v3.5
- FlamMap (FSPRO)
- MyriMatch
- x264 Benchmark HD v3.0

Noticably absent from this review is an old-time favorite, 3ds Max. I did attempt to run our custom 3ds Max benchmark on both the 2009 and 2010 versions of the software, but the application would simply not load on the Westmere box with hyper-threading enabled. Evidently Autodesk didn't plan far enough ahead to write their software for more than 16 threads. Once there is an update that addresses this issue, I will happily add 3ds Max back into the benchmarking mix.

All benchmark numbers presented were an average of three runs. Desktop resolution was set at 1920x1200, 32bit color with a 60Hz refresh rate. "Optimized Defaults" were used in the BIOS on all machines.

Advanced power management settings were enabled on both systems. Both sets of processors were tested with SMT and Turbo set to ON for all tests. For the SPECjbb tests, Windows Server's "lock pages in memory" setting was enabled for the benchmark user and in the BIOS, "hardware prefetch" and "adjacent cache line prefetch" options were disabled.

All power measurements throughout this review were taken with an Extech 380803 power meter. Readings were for the computer itself, and all other devices (monitor, speakers, etc) were plugged-in separately and not included in the measurements.