Intel Xeon E5-2600 / R2000GZ Server


Intel Grizzly Pass 2U Server Review

When I found out we were getting some of the newly released Intel Xeon E5-2600 series processors to review I was intrigued. Much of my life is spent installing and configuring servers of almost every conceivable role, so I was quite interested in seeing what Intel's best was capable of. Besides the usual round of benchmarks on this server we decided to test it as a web server. With co-location space being a rather expensive commodity, packing as much processing power per unit (U) is rather important. 

Stepping away from our typical desktop component reviews for a minute, we are going to take a look at another side of the computer world. Servers, without them we wouldn't have much of a network infrastructure. The Intel 'Grizzly Pass' server features some of the latest technology from Intel. Mainly it brings the latest LGA2011 processors to the server world. Recently we have been focused on the Intel 'Sandy Bridge' E processor and the Intel X79 chipset motherboards. Currently that combination of components is the top of the line for desktops. Recently though Intel announced the launch of the new Intel Xeon 'Sandy Bridge' EP processors in the LGA2011 package. The initial launch of the Intel Sandy Bridge architecture was just over a year ago for desktop users. The server variant however has had a lengthy delay as the system architecture is quite different from the desktop based LGA-1155 CPU's. While Intel released the LGA-1155 based Xeon E3 late last year, it was limited to a single socket and more designed to the high end workstation market. With the new Xeon E5-2600 series supporting 2 sockets and the upcoming 4600 series supporting 4 sockets the wait is finally over.

In order for a system like the Intel 'Grizzly Pass' to work, the processors need to be able to talk to each other. This is accomplished by each of the Intel Xeon E5-2660 processors have a pair of QPI (Quick Path Interconnects). So if you're hoping to drop in a pair of Intel Core i7 3960X processors or the more budget friendly Intel Core i7 3820 in this system or a desktop motherboard like the EVGA SRX dual LGA2011 Socket motherboard, you'll be out of luck. The Intel Core i7 processors series is left with only a single QPI. 


Sandy Bridge-EP die shot

The Intel Xeon E5-2600 series CPU contains up to 8 cores plus HyperThreading for a total of 16 threads per processor. The Intel 'Grizzly Pass' server we have on hand today has a pair of Intel Xeon E5-2660 Sandy Bridge EP processors, that gives us a total of 32 processing threads! The Intel Xeon E5-2600 series is built on 32nm technology and the overall die size is 416mm. Also on this huge 416mm die is 20MB of L3 cache adding up to a transistor count of around 2.263 billion transistors.

Romley EP Platform Block diagram

With 2 QPI links, 40 PCI Express 3.0 lanes per socket and Quad Channel LRDIMM (Load-Reduced DIMM) support the Xeon E5 appears to be ready for high performance workloads.

Intel Sandy Bridge EP E5-2660 CPUz

Each of the Intel Xeon E5-2660 processors runs at a default speed of 2.2GHz. When loading up the system with Prime95 the Intel Turbo Boost 2.0 kicks in and boost the processor speed up to 2.5GHz across all of the cores. When running lightly threaded applications the Intel Xeon E5-2660 has a maximum processor frequency of 3.0GHz.

Processor Support Support for one or two Intel Xeon processors E5-2600 product family with a Thermal Design Power (TDP) of up to 130 W.
Memory 24 DIMM slots – 3 DIMMs / Channel – 4 memory channels per processor
Unbuffered DDR3 (UDIMM), Registered DDR3 (RDIMM), and Load Reduced DDR3 (LRDIMM)
Memory DDR3 data transfer rates of 800, 1066, 1333 MT/s, and 1600 MT/s
DDR3 standard I/O voltage of 1.5V and DDR3 Low Voltage of 1.35V
Chipset Intel C600-A
External I/O connections Video – Back Panel + Front Panel
RJ-45 Serial-A Port
4x RJ-45 Network Interface Connectors supporting 10/100/1000Mb
USB 2.0 connectors -3 on back panel + 2 on front panel
Internal I/O connectors / headers One Type-A USB 2.0 connector One DH-10 Serial-B port connector
Additional I/O Modules Dual SFP+ port 10GbE module based on Intel® 82500 10 GbE controller – AXX10GBNIAIOM
System Fans Five managed system fans
One power supply fan for each installed power supply module
Riser Cards Support for two riser card slots. Each riser card slot has support for the following riser card options:
3-slot PCIe Riser Card: (Slots 1 & 2) – PCIe x16 slot, x8 lanes, (Slot 3) – PCIe x8 slot, x8 lanes
2-slot PCIe Riser Card: (Slot 1) – PCIe x16 slot, x16 lanes, (Slot 2) – PCIe x8 slot, x8 lanes
3-slot PCIx/PCIe Riser Card: (Slots 1 & 2) – PCIx 64-bit, (Slot 3) – PCIe x8 slot, x8 lanes
Video Integrated 2D Video Controller 16 MB DDR3 Memory
On-board storage controllers and options 1x eUSB 2x5 pin connector to support 2mm low-profile eUSB solid state devices
2x 7-pin single port AHCI SATA connectors capable of supporting up to 6 GB/sec
2x SCU 4-port mini-SAS connectors capable of supporting up to 3 GB/sec
Intel RAID C600 Upgrade Key support providing optional expanded SATA / SAS RAID capabilities (RKSAS8R5 'purple' key installed)
Intel Integrated RAID Module RMS25CB080 1GB
Server Management Integrated Baseboard Management Controller, IPMI 2.0 compliant
Support for Intel® Server Management Software
Intel Remote Management Module 4 Lite – Accessory Option
Intel Remote Management Module 4 Management NIC – Accessory Option
Power Supply 2x 750W Redundant, Hot Swap DPS-750XB (80+ Platinum)
Storage Bay 8x – 2.5” SATA/SAS Hot Swap Hard Drive Bays + Optical Drive support

 

Intel R2000GZ Server Specs


Intel R2000GZ Server - Front

The Intel R2000GZ is offered in several configurations and has many additional options that can be added to any of the SKU's. The model we received is an 8 drive, 2.5" bay configuration with a dual layer DVD-RW drive and front I/O module. This server line is also available in 16 and 24 bay configurations for the 2.5" form factor as well as 8 and 12 bay configurations for 3.5" drives.

Intel R2000GZ Server - Rear

Intel has included dual hot swap, redundant 750W power supplies in the unit as can be seen above. Also included was a 2-port 10GbE module which is connected internally via a proprietary on-board connector. With 4 GbE ports on top of the 10GbE module this server has plenty of network connectivity.

Intel R2000GZ Server - Hot Swap Drive Bays

Each of the drive cages in the Intel R2000GZ will carry a single 2.5" drive. For our testing today we will be playing with a handful of Intel 710 series SSD's in RAID 5

Intel SSD 710 Series

For testing this server was equipped with 5x Intel 710 Series 200GB SSD drives running in RAID-5. While not the most performance optimized setup, RAID-5 gives an excellent balance between speed and reliability.

Intel Integrated RAID Module RMS25CB080
Intel Integrated RAID Module RMS25CB080

Also included was an Intel Integrated RAID Module RMS25CB080 with 1GB of on-board memory. The RMS25CB080 is a 6Gb/s SATA/SAS module supporting RAID levels 0,1,5,6,10,50,6 and 60. This additional module is a hardware RAID card, meaning it uses dedicated hardware to control the array as opposed to the processing being done by the CPU via software. The biggest advantage of this type of card is performance, even with a CPU like the E5-2660 the performance hit for non simple (RAID 0/1) array is quite substantial. In addition to the performance boost you generally gain far better OS compatibility, especially when dealing with virtualization. This controller was configured in write-back mode for all testing unless otherwise noted.

Intel DPS-750XB Hot Swap Power Supplies

Each of the Intel DPS-750XB A power supplies (P/N: E98791-005)above is rated for 750 Watts maximum and is 80+ Platinum certified.

Intel R2000GZ 'Grizzly Pass' Interior

The server has a very cooling efficient design where the air is channeled from front to back helping ensure adequate airflow. Designs such as this have become increasingly common in rack-mount servers to aid in efficient cooling. Many data-centers are moving to rack-based cooling solutions to conserve energy and to help ensure uniform inlet air temperatures to equipment. Efficient server designs are one very important piece of this puzzle.

Intel R2000GZ 'Grizzly Pass' Memory

The plastic shroud over heatsinks and memory don't only serve as a mounting point for the Intel 710 series SSD. It also directs the air flow over the the memory and through the pair of heatsinks on the Intel Xeon E5-2660 processors. Keeping the air moving and the Intel R2000GZ server system cool is five Nidec UltraFlo V60E12BS1B5 60MM fans.

Intel R2000GZ 'Grizzly Pass' Drive Cage Backplane

The 2.5" drive hot-swap backplane supports up to 8 drives each. These are powered by a single 4 pin connector which in turn powers the connected drives. Two multi-port mini-SAS cables are connected each providing I/O for four SAS/SATA drives that are installed on the backplane. The thin connectors on the top corners of the backplane are for SMBus cables and are used as a management interface between the server board and the backplane(s).

Intel R2000GZ 'Grizzly Pass' PCIe Riser Cards

The Intel S2600GZ server board features two x24 super slots that each have a riser card installed. Each of the installed riser cards feature three PCIe  Gen 3 x8 slots with one as a half height slot.

Intel Grizly Pass S2600GZ Motherboard

Here is a better view of the Intel S2600GZ server board that powers this server. This server board has the following key features:

Intel S2600GZ and S2600GL boards are only available in rack-optimized server systems in an expansive family of 1U and 2U systems for the most demanding enterprise data center, cloud, or high performance computing (HPC) environments.

Intel Grizly Pass BIOS - Main Screen

Once we boot in to the Intel Grizzly Pass server BIOS the initial page gives us some basic information about the server. The Platform ID is also the motherboard model number, which happens to be the Intel S2600GZ. Below that we can see the BIOS version and build date. Just below that the total memory of 128GB is displayed. Below that we can enable or disable the Quiet Boot, POST Error Pause, and set the system date and time.

Intel Grizly Pass BIOS - Advanced Screen

The Advanced tab of the Intel S2600GZ BIOS displays the information about our Intel Xeon E5-2660 Sandy Bridge EP processors. Each of them run at 2.2GHz and have 20MB of Cache! Below the processor information we can enable and disable several features of the processor.

Intel Grizly Pass BIOS - Memory Configuration

While the Intel Server System R2000GZ can support up to 384GB of memory, we equipped ours with a more typical configuration of 128GB 1600MHz Registered DDR3.

Intel Sandy Bridge EP E5-2660 CPUz
Intel Sandy Bridge EP E5-2660 CPUz
Intel Sandy Bridge EP E5-2660 CPUz

SiSoftware Sandra 2012 SP2 Memory Bandwidth

Sisoftware Sandra 2012 SP2

The SiSoft Sandra 2012 SP2 benchmark utility just came out a few weeks ago and we have started to include it in our benchmarking. Sandra 2012 comes with support for Virtualisation (Virtual PC/Server, Hyper-V, VMware) and GPGPU (OpenCL, DirectX 11 DirectCompute), but today we will be using the program to look at memory performance!

Intel R2000GZ 'Grizzly Pass' Sandra 2012 SP2 Memory Benchmark Scores

Benchmark Results: The Intel Grizzly Pass server has an impressive 128Gb of memory in a quad channel configuration. That's 64Gb of RAM for each of the Intel Xeon E5-2660 CPUs to work with. Looking at the results above we can see that we topped out at 72.372GB/s on the Integer memory bandwidth test and 73.436GB/s on the Float Memory bandwidth test. In comparison our Intel X79 test bench is typically just above 40GB/s on both tests.

wPrime

wPrime

wPrime is a leading multithreaded benchmark for x86 processors that tests your processor performance by calculating square roots with a recursive call of Newton's method for estimating functions, with f(x)=x2-k, where k is the number we're sqrting, until Sgn(f(x)/f'(x)) does not equal that of the previous iteration, starting with an estimation of k/2. It then uses an iterative calling of the estimation method a set amount of times to increase the accuracy of the results. It then confirms that n(k)2=k to ensure the calculation was correct. It repeats this for all numbers from 1 to the requested maximum.

Intel Grizzly Pass wPrime Benchmark Results

Benchmark Results: The pair of Intel Xeon E5-2660 CPU's in the Intel Grizzly Pass server pretty much rocked everything I have tested in the past. With an average time of 4.097 seconds for the 32M tests it is significantly faster than the Intel Core i7 2600K which takes ~7.3 seconds to complete the test. We weren't to surprised by this since the Grizzly Pass server from Intel has 32 threads of processing power, four times that of the 2600K. We were a little bit surprised that it wasn't faster, but watching the task manager the 32M test was nearly done by the time that the Intel Xeon E5-2660 CPU's were ramped up to full throttle. Firing up the 1024M test the Intel Grizzly Pass server was able to complete the test in an average of 81.173 seconds! In comparison the Intel Core i7 2600K takes ~225 seconds and the Intel Core i7 3960X takes 146.111 seconds. The i7 3960X doesn't come close even when overclocked to 4.8GHz, at that speed it takes ~119 seconds!

CrystalDiskMark

CrystalDiskMark is a small benchmark utility for drives and enables rapid measurement of sequential and random read/write speeds. Note that CDM only supports Native Command Queuing (NCQ) with a queue depth of 32 (as noted) for the last listed benchmark score. This can skew some results in favor of controllers that also do not support NCQ.

Intel Grizzly Pass Server Configured in RAID 5 with Write Back

Intel Grizzly Pass with Intel 710 series SSD's CrystalDiskMark Benchmark Results

Intel Grizzly Pass Server Configured in RAID 5 with Write Through

Intel Grizzly Pass with Intel 710 series SSD's CrystalDiskMark Benchmark Results

Benchmark Results: We configured the RAID 5 array with two different modes for this test. The first mode that we ran was with Write Back enabled. Write back takes advantage of the LSI onboard memory to store the data being written as it writes to the RAID Array, this gives a boost to the performance since we are writing to the LSI controller and not directly to the Intel 710 series SSD's themselves. The sequential reads with write back hit an impressive 1346 MB/s and the Sequential Writes hit 871MB/s! The 4K read was 41.85MB/s while the 4K write was only 14.06MB/s (more on this result shortly). Setting the Intel 710 series SSD RAID array to write through mode or results dropped a rather significant amount. The Sequential read dropped from 1346MB/s previously to 992.2MB/s and the sequential write dropped from 871.0MB/s to 88.24MB/s. The 4K read which was 41.85MB/s fell to 20.01MB/s and the 4K write fell to 11.15MB/s. Small block writes in a RAID 5 array suffer greatly as you can see, this is due to the nature of RAID 5 and each small write ending up being two disk reads and two disk writes (4 I/O's). While typically this hit is helped via the controller's cache however in the write-back test we were sending more data than the cache could contain at once.


SQLIO

The SQLIO tool is a disk benchmark tool provided by Microsoft. The name is very misleading however, this tool has nothing to do with SQL Server, it is used to determine the I/O capacity of a storage configuration. This tool not only tests the throughput in MB/s like CrystalDiskMark, it also allows for testing Input/Output Operations Per second (IOPs). It also allows for user specified block sizes, test file size and if you want to run a random or sequential IO test.

Testing in SQLIO was done using a 3GB test file in several block sizes for 5 minutes per test. This was done so the test file would be larger then the on-board cache of the RAID controller and to ensure we were receiving accurate numbers.

Intel Grizzly Pass with Intel 710 series SSD's SQLIO Read IOPs Results

Intel Grizzly Pass with Intel 710 series SSD's SQLIO Write IOPs Results

Intel Grizzly Pass with Intel 710 series SSD's SQLIO Read MBs Results

Intel Grizzly Pass with Intel 710 series SSD's SQLIO Write MBs Results

Benchmark Results: The SQLIO tests were done with the array in Write Back mode. Again, Write back takes advantage of the LSI onboard memory to store the data being written as it writes to the RAID Array, this gives a boost to the performance since we are writing to the LSI controller and not the Intel 710 series SSD's directly. This test shows the relationship between I/O operations and MB/s rather clearly. As expected, sequential read speeds on the array were very good ranging from 934.77 MB/s to 1226.9 MB/s however the random reads ranged from a very impressive 1124.16 MB/s to an astonishing 1623.56 MB/s. RAID-5 really shines in random read performance and increases greatly the more drives that are added to the array.

POV-Ray 3.7 RC5

Processor Performance on Pov-Ray 3.7 RC5:

The Persistence of Vision Ray-Tracer was developed from DKBTrace 2.12 (written by David K. Buck and Aaron A. Collins) by a bunch of people (called the POV-Team) in their spare time. It is a high-quality, totally free tool for creating stunning three-dimensional graphics. It is available in official versions for Windows, Mac OS/Mac OS X and i86 Linux. The POV-Ray package includes detailed instructions on using the ray-tracer and creating scenes. Many stunning scenes are included with POV-Ray so you can start creating images immediately when you get the package. These scenes can be modified so you do not have to start from scratch. In addition to the pre-defined scenes, a large library of pre-defined shapes and materials is provided. You can include these shapes and materials in your own scenes by just including the library file name at the top of your scene file and by using the shape or material name in your scene. Since this is free software feel free to download this version and try it out on your own.

The most significant change from the end-user point of view between versions 3.6 and 3.7 is the addition of SMP (symmetric multiprocessing) support, which, in a nutshell, allows the renderer to run on as many CPU's as you have installed on your computer. This will be particularly useful for those users who intend on purchasing a dual-core CPU or who already have a two (or more) processor machine. On a two-CPU system the rendering speed in some scenes almost doubles. For our benchmarking we used version 3.7 RC5, which is the most recent version available.  The benchmark used all available cores to their fullest extent to complete the render.

Pov-Ray 3.7 RC3

Once rendering on the object we selected was completed, we took the elapsed time from the dialog box, which indicates the exact time it took for the benchmark to finish the benchmark. A lower time indicates faster system performance. This benchmark used all 12 threaded and loaded each one up at 100% load!

Pov-Ray 3.7 RC3

Benchmark Results: The pair of Intel XEON E5-2660 processors left everything in the dust! The LGA2011 desktop version Intel Core i7 3960X was 70.3% slower than the Intel Grizzly Pass server!

Cinebench R11.5

MAXON; CINEBENCH R11.5:

MAXON recently released CINEBENCH Release 11.5, an advanced hardware testing suite that assesses a computer's performance capabilities. CINEBENCH is based on the same powerful technology as MAXON's award-winning animation software CINEMA 4D, which is used extensively by studios and production houses worldwide for 3D content creation. The new version of CINEBENCH includes the ability to more accurately test the industry’s latest hardware, including systems with up to 64 processor threads, and the testing environment better reflects the expectations of today’s production demands. A more streamlined interface makes testing systems and reading results incredibly straightforward. Again, higher Frames/Second and point score equal better performance.

Cinebench R11.5

Cinebench R11.5 was able to put a 100% load across all the cores on all of the processors, which makes this a great benchmark to look at multi-core platforms.

Cinebench R11.5 Benchmark Results

Benchmark Results: When it comes to multi-core performance the Intel XEON E5-2660 processors scored 20.13 points. That is 75.7% faster than the Intel Core i7 3960X which scored 11.46 points.

Cinebench R11.5 Results

Benchmark Results: The single core performance of the Intel XEON E5-2660 processors wasn't as impressive. It was only able to squeeze out a score of 1.21 which is behind the Intel Core i3 2105. We have to keep in mind that the Intel Grizzly Pass server isn't designed for single threaded applications. Having a pair of the XEON E5-2660 chips it offers 16 cores and 32 threads of processing power, though they are running 900MHz slower than the Intel Core i7 3960X.

x264 HD Encoding

x264 HD Encoding Benchmark

Simply put, the x264 HD Benchmark is a reproducible measure of how fast your machine can encode a short HD-quality video clip into a high quality x264 video file. It's nice because everyone running it will use the same video clip and software. The video encoder (x264.exe) reports a fairly accurate internal benchmark (in frames per second) for each pass of the video encode and it also uses multi-core processors very efficiently. All these factors make this an ideal benchmark to compare different processors and systems to each other. We are using x264 HD v4.0 for this test.

x264 HD Encoding Benchmark

This application did fairly well when run on 12 threads, as you can see from the screen shot above. The first pass was not using all of the processing power available on the cores, but on the second pass all 12 threads were at ~95% load.

x264 HD Encoding Benchmark Results

Benchmark Results: The Dual XEON E5-2660 processors were able to run the second pass of the benchmark at an average of 71.1 frames per second. It is definitely lower than we expected though still quite strong. The lower processor frequency a contributing factor. The first pass which is dominated by the Intel Core i7 3960X with an average of 192.4 frames per second was average at only 114.9 frames per second, this pass isn't very heavily threaded and relies on processor frequency more than anything.

Windows IIS / DotNetNuke

DotNetNuke is one of the leading Web Content Management Platform (or CMS) for Microsoft ASP.NET. It is an open source CMS platform that also functions as a flexible development framework for creating content rich web applications. DotNetNuke is also designed to be expanded by the use of third party modules to enable almost any functionality a website could need.

DotNetNuke Sample Site

For the purpose of this benchmark, DotNetNuke was tested using its default configuration on Windows 2008 R2 Enterprise SP1 with a SQL Server 2008 Enterprise database back-end. Testing was done locally using a Dell PowerEdge 2900 as the client machine over a private network running on an HP 1810-8G switch.

To simulate user load we are using a software package called Apache JMeter. JMeter is a powerful Java based application with a wide range of very powerful features. It is used to test performance on both static and dynamic web resources by simulating different load types. For this set of testing we are using it to roughly determine the maximum load our server can handle while staying under a specified average response time. Simply, we are attempting to determine how many hits per second, on average, this hardware is capable of handling in an acceptable amount of time. 

One of the issues with running this type of test is simulating the exact load against different machines. While one set of hardware may respond at 1500ms under a load of 100 users, another server may only be able to handle 60 while responding within this time frame. This is the reason the response times will differ slightly between servers, different numbers of users had to be simulated to achieve comparable results.

Apache JMeter was run in this test against several pages, including logging into the system and displaying one of the back-end admin pages. An average server response time of under 1.5 seconds was the goal, however for this benchmark I did allow for the overall average to be slightly higher as the Admin page caused a fairly substantial increase on its own. Note that these response times are the overall response time and not the time to first byte.

DotNetNuke Average Response Time

These results are from runs of roughly 150,000 hits. One of the pages being generated was a back-end admin page that caused the average response time to increase fairly substantially. Removing this page from the average brings them down to around 1400ms, in line with our goal of < 1500ms. There are many factors that come into play for this type of test and all testing was done under no bandwidth constraints on a very low latency network, real world numbers will vary greatly.

DotNetNuke Throughput

Benchmark results:While these numbers were far lower than expected overall, they were roughly where expected in relation to each other with the E5-2660 being just under 4x faster. DotNetNuke is very resource intensive in its default configuration, however so we could reproduce the test on each machine and be able to accurately compare the results no changes were made. To give some idea what these numbers mean, per day the E5-2660 would be able to handle around 2.9 million requests compared to the roughly 740 thousand of the E5620, while keeping the average load time to under 1.5 seconds.

LAMP - WordPress & Joomla

WordPress is an open source blogging tool and content management system (CMS) that is based on PHP and MySQL. Its large feature list includes a plug-in architecture and a powerful template system. Wordpress is the most popular CMS in use on the internet, powering 14.7% of Alexa's 'top 1 million' websites as well as managing 22 out of every 100 new active domains in the US.

WordPress Sample Site

Joomla is an open source content management system (CMS) and MVC framework that is based on PHP and MySQL. Joomla's feature list includes news flashes, blogs and polls to name a few and can also be extended by third-party extensions.

Joomla Sample Site

For the purpose of this benchmark, WordPress 3.3.1 and Joomla 2.5.2 were tested using their default configurations on Ubuntu 12.04 LTS Beta with a MySQL back-end. Testing was done locally using a Dell PowerEdge 2900 as the client machine over a private network running on an HP 1810-8G switch.

To simulate user load we are using a software package called Apache JMeter. JMeter is a powerful Java based application with a wide range of very powerful features. It is used to test performance on both static and dynamic web resources by simulating different load types. For this set of testing we are using it to determine the maximum load our server can handle while staying under a specified average response time. Simply, we are attempting to determine how many hits per second, on average, this hardware is capable of handling in an acceptable amount of time.

One of the issues with running this type of test is simulating the exact load against different machines. While one set of hardware may respond at 1500ms under a load of 100 users, another server may only be able to handle 60 while responding within this time frame. This is the reason the response times will differ slightly between servers, different numbers of users had to be simulated to achieve comparable results.

Apache JMeter was run for these tests against several pages, including logging into the system and displaying one of the back-end admin pages. An average server response time of 1.5 seconds was the goal. Note that these response times are the overall response time and not the time to first byte.

LAMP Average Response Time

Getting the average response time to be similar over test runs of around 150,000 hits is a rather difficult thing to do. Our averages per test were close enough to at least get some rough throughput numbers. Remember there are many factors that come into play for this type of test and all testing was done under no bandwidth constraints on a very low latency network, real world numbers will vary greatly.

LAMP Throughput

Benchmark results: Considering the difference in core count between our two machines the numbers were right where we expected. Using the default install of WordPress the E5-2660 handles roughly 11.4 million hits per day while the E5620 was only able to process around 3 million. In Joomla the results were similar with the E5-2660 handling 12.4 million to the E5620's 3.3 million.

Final Thoughts & Conclusions

Intel R2000GZ - Grizzly Pass

At the heart of the Intel Grizzly Pass server are the Intel LGA2011 Sandy Bridge EP processors. In our particular server Intel sent it with a pair of the Intel Xeon E5-2660 processors. The Xeon E5-2660 alone retails for a measly $1347.28 after shipping. There are lower cost alternatives to the Xeon E5-2660, which of course will mean lower performance, but if you don't need all the performance offered by the E5-2660 you may as well save a few bucks. The Xeon E5 series range from the E5-2603 quad core processor which retails for $223.36, up to the Xeon E5-2690 which retails for $2078.10. Most likely Intel has a processor in the Xeon E5 family that will suite your needs of your server. Not surprisingly, Intel has done a fantastic job on the new Sandy Bridge-EP Xeon E5 processors. When it comes to sheer performance they simply blow away the previous generation Intel Xeon 5600. Possibly the most impressive part is that Intel was able to pack this much computing power into a 2U server, which is astonishing. We would guess it is only a matter of time until these amazing CPU's are available in a 1U server or as a blade.

The components that Intel used on the R2000GZ work fantastically together to provide simply amazing performance. The only minor gripe on the server the difficulty in routing cables and having the plastic airflow shield fit. It does all fit, but adding memory ends up being a fairly annoying task as replacing the shield properly requires getting several cables to stay exactly where they are required to be during the procedure. With this only needing to be done extremely rarely on servers this is a very minor issue. Outside of this minor headache however all the hardware is put together very well and is rather easy to access for maintenance even compared to other manufacturers models of a similar form factor.

Looking at the performance numbers, this server would be an excellent virtualization server as well, easily able to power multiple heavily loaded VM's. In fact with the broad range of I/O options that Intel has included or made available, I struggle to find a real-world workload we could reasonably throw at this machine that it would struggle with.

Intel Xeon E5-2660 Windows Task Manager

Since each of the Intel Xeon E5-2660 processors has eight cores plus hyper-threading, the Intel Grizzly Pass server has a total of 32 threads of processing power. When the Windows task manager was opened up for the first time, we couldn't help being impressed by the look of 32 threads. It's hard to imagine that just a few short years ago we were oohing and aahing over dual core processors! It's amazing how fast and far the technology has come in the last decade or so.

Legit Bottom Line: Overall the Intel R2000GZ 'Grizzly Pass' is a well built and amazingly fast server. The Intel Xeon E5-2660 processor is an immense powerhouse of a processor, sure to keep Intel as king of the server world for some time to come.