iXBT Labs - Computer Hardware in Detail






Dual Intel Xeon Configuration

Any good for a desktop?

December 17, 2009

<< Previous page

     Next page >>

There was a time when multi-way configurations were popular not only in servers, but also in higher-end workstations and even custom PCs. The populatiry was achieved, thanks to the SMP support added by Intel to Pentium-based machines, which allowed building dual-way rigs with no architectural tricks involved. The support was further improved in Pentium Pro, allowing up to four-way configurations to be built. Desktop Pentium II processors stepped down to dual-way, but they were cheaper. Since even the lowest-end Celerons could be used in such PCs after slight tweaking, the corresponding motherboards quickly became very widespread. Then AMD joined the race by offering its dual-way workstation solutions. You could either purchase an Athlon MP processor or make it yourself from mainstream Athlon XP CPUs. This quickly attracted users to dual-socket motherboards based on the 760MP and 760MPX chipsets.

Later, manufacturers realized that the simplicity of assembling SMP machines from inexpensive parts had quite a negative effect on their revenues, so mainstream single-socket and single-CPU solutions were separated from high-performance ones up to complete incompatibility on some occasions. However, that was done to earn more, not to finish that market off. As a result, some enthusiasts lost interest in dual-socket configuratins, but the most enthusiastic ones, having real need for high performance, continued to actively purchase dual-way workstations. The pinnacle of that were the first Xeon and Opteron processors. Then it ceased quickly.

The reason was the rollout of dual-core, and then quad-core, desktop processors. Essentially, CMP and SMP didn't differ much in most tasks, but that way users could build machines from regular PC parts, and spend less at that. Of course, two sockets allowed for twice as many cores than one socket, but there were only a few who needed that many. Users quickly found out that two cores were almost always better than one, even in case of single-thread applications. Such PCs were more responsive, thanks to the necessary resources provided by the second core. Four cores, however, could only be used by quite special software for a long time. This hasn't changed much since then actually. But even if four cores are not enough for you, there is still no need to purchase a dual-socket board, as both Intel and AMD are going to rollout six-core desktop processors in the near future. Then you will be able to install a couple of those as well. But what are you going to do with twelve processing cores? In other words, increasing the number of cores was only critical in the transition from one to more than one. SMP configurations were quite necessary when there were no other ways of getting more than one core. However, their attractiveness quickly ceased as CMP evolved. Today, of all successful desktop SMP systems, only Apple Mac Pro comes to mind, but about six month ago the company put the lid on it by releasing a modification with a single "one-socket" Xeon 3500 processor.

However, we all know there are applications that can be greatly sped up by increasing the number of cores in any way possible. It's just that certain tasks are very parallelizable -- some on the level of algorithms, some by running similar independent parts of code at the same time. Although such tasks are quite rare among usual desktop software. Or are they? We thought it would be interesting to check this out. Our processor test method is largely oriented to desktop CPUs anyway. Besides, it shows decent performance boosts when the number of threads is increased (by means of both CMP and, to a lesser extent, Intel's SMT). What if we activate all of the existing performance boosting technologies: SMT (more threads per core, the idea behind Hyper-Threading), CMP (more cores per device), and SMP (more processors per system)? Let's find out.


CPU Core i7-860 Core i7-975 Extreme Edition Xeon L5520 Xeon X5570
Core Name Lynnfield Bloomfield Bloomfield Bloomfield
Process Technology, nm 45 45 45 45
Core Clock (Std/Max), GHz 2.8/3.47 3.33/3.6 2.26/2.53 2.93/3.33
Bus/Core Ratio 21 25 17 22
Turbo Boost 5-4-1-1 2-1-1-1 2-1-1-1 3-3-2-2
Cores/Threads 4/8 4/8 4/8 4/8
L1 Cache, I/D, KB 32/32 32/32 32/32 32/32
L2 Cache, KB 4 x 256 4 x 256 4 x 256 4 x 256
L3 Cache, KB 8192 8192 8192 8192
Uncore Clock 2.4 2.66 2.13 2.66
Supported RAM 2 x DDR3-1333 3 x DDR3-1066 3 x DDR3-1066 3 x DDR3-1333
QPI, GT/s 4.8 6.4 5.86 6.4
Socket LGA1156 LGA1366 LGA1366 LGA1366
TDP, W 95 130 60 95

Let us explain the selection of processors. Core i7-860 can be considered a reasonable baseline. Those actually interested in high performance won't buy anything cheaper, while higher-end processors are considerably more expensive. Core i7 Extreme 975 is the current top limit for desktop systems. It sure is expensive (though it's priced similar to one Xeon 5500 CPU), but there's nothing faster yet. The two Xeons, tested separately and together, are the main characters of today's review. Why these models exactly? X5570 is the fastest of "non-extreme" Xeons, with 95W TDP. Two of these require 190W which is quite a lot for a desktop, but faster ones (W5580 and W5590) have the TDP of 260W! That's why X5570 remains the limit for many systems. As for L5520, it cannot boast of extreme performance, but it only requires 60W. In other words, two of these are more power-efficient than a single Extreme 975, for example. Late this summer, Intel rolled out L5530, a faster modification with lower power consumption. However, increasing the clock rate by 133MHz cannot provide a radical performance boost, so L5520 is fine too.

  Motherboard RAM
LGA1156 Intel DP55WG (P55) 4 x 2 GB (1333; 9-9-9-24)
LGA1366 Intel DX58SO (X58) 3 x 2 GB (1333; 9-9-9-24 for 975 EE/X5570, 1066; 8-8-8-19 for L5520)
Dual LGA1366 Intel S5500HCV (5500) 6 x 1 GB (1333; 9-9-9-24 for X5570, 1066; 8-8-8-19 for L5520)

We publish the results of Core i7-860 with 8GB of RAM. Having only 4GB of memory available greatly reduces performance in some applications participating in our test method. But there's no such performance difference between 6GB and 8GB of RAM. Machines featuring single and dual LGA1366, formally, have identical memory characterisics, but the actual modules differ: 3 x 2GB or 6 x 1GB, divided equally between the sockets. Perhaps, such a configuration will result in performance drop, but clogging a dual-way machine with twice as much RAM would be even wronger.

We also need to say a few words about setting up server platforms from the angle of power consumption. The matter is that the latter has many more capabilities than desktops. For example, it sometimes allows to limit total power consumption, so the machine will try to meet the requirement despite anything. This might be a problem at times. In particular, simply enabling Turbo Boost in BIOS is not enough to turn it on. This matter requires a few other settings to be adjusted. We mentioned this just in case. Obviously, a user assembling such a rig knows what he's doing. Still, remember about that, if your server or dual-way workstation performs worse than expected.

Write a comment below. No registration needed!

Next page >>

Article navigation:

Page 1: Introduction, testbeds

Page 2: Tests

Page 3: More tests, total score

blog comments powered by Disqus

  Most Popular Reviews More    RSS  

AMD Phenom II X4 955, Phenom II X4 960T, Phenom II X6 1075T, and Intel Pentium G2120, Core i3-3220, Core i5-3330 Processors

Comparing old, cheap solutions from AMD with new, budget offerings from Intel.
February 1, 2013 · Processor Roundups

Inno3D GeForce GTX 670 iChill, Inno3D GeForce GTX 660 Ti Graphics Cards

A couple of mid-range adapters with original cooling systems.
January 30, 2013 · Video cards: NVIDIA GPUs

Creative Sound Blaster X-Fi Surround 5.1

An external X-Fi solution in tests.
September 9, 2008 · Sound Cards

AMD FX-8350 Processor

The first worthwhile Piledriver CPU.
September 11, 2012 · Processors: AMD

Consumed Power, Energy Consumption: Ivy Bridge vs. Sandy Bridge

Trying out the new method.
September 18, 2012 · Processors: Intel
  Latest Reviews More    RSS  

i3DSpeed, September 2013

Retested all graphics cards with the new drivers.
Oct 18, 2013 · 3Digests

i3DSpeed, August 2013

Added new benchmarks: BioShock Infinite and Metro: Last Light.
Sep 06, 2013 · 3Digests

i3DSpeed, July 2013

Added the test results of NVIDIA GeForce GTX 760 and AMD Radeon HD 7730.
Aug 05, 2013 · 3Digests

Gainward GeForce GTX 650 Ti BOOST 2GB Golden Sample Graphics Card

An excellent hybrid of GeForce GTX 650 Ti and GeForce GTX 660.
Jun 24, 2013 · Video cards: NVIDIA GPUs

i3DSpeed, May 2013

Added the test results of NVIDIA GeForce GTX 770/780.
Jun 03, 2013 · 3Digests
  Latest News More    RSS  

Platform  ·  Video  ·  Multimedia  ·  Mobile  ·  Other  ||  About us & Privacy policy  ·  Twitter  ·  Facebook

Copyright © Byrds Research & Publishing, Ltd., 1997–2011. All rights reserved.