WE HAVE talked a lot about the 'green' energy saving focus that the computer industry, both the vendor and user sides, is starting to accentuate more over the past year or two. It's not just all the gigaflops and gigabytes or TPC per Watt figures for big machines, but also how much power for the given performance is taken by your corporate client or home theatre PC. Here, of course, the lower the better.
The recent crop of high end workstation, server and PC processors have both high performance and good power/performance options. In the case of the Intel Xeon, you'll recognise the X-series as the top speed, high powered chips with TDP declared as 130W, followed by the E-series mainstream 95W and 80W units with lower speeds, as well as power-optimised L-series with speeds similar to the low end of the E-series but with much power lower consumption. The L-series TDP ratings will vary between 40W and 60W, depending on the model.
In real life, most Intel processors will consume far less power than what the TDP states. If coupled with low voltage memory and power saving peripherals, you might end up with a very efficient system while still keeping quite high performance. In fact, the recent Nehalem and upcoming Westmere chips push the power saving capabilities to the limits at both the CPU processor and system chipset overall levels.
Here we have two quad-core platforms from Gigabyte. The well-known major Taiwanese brand is very aggressive in the high-end space now, with a range of server and workstation boards that have its own unique designs. The Gigabyte GA-5YASV is using the old X3360 LGA775 socket Xeon based on the Core 2 platform, while the latest one, the Gigabyte GA-6FASV1, uses a Nehalem-based X3460 Xeon chip in the LGA1156 socket. Both are microATX sized compact servers, with dual-channel memory and similar clock frequencies - 2.83GHz and 2.80GHz respectively. Gigabyte's two mainboards are fairly resource rich, each with a nice combination of nest generation I/O interfaces as well as the old serial and other ports that are still useful for some peripherals.
Also, both boards used the standard reference, 8GB Kingston server memory at default speeds, DDR2-667 and DDR3-1333 respectively, in four 2GB modules each. We also kept the default BIOS power state settings for the CPUs. The newer Xeon's HyperThreading and Turbo features were enabled as well, since that's the default setting most users will stick with. How do they compare in terms of actual power consumption and other parameters when running?
We start with the power consumption observations at boot time in BIOS, using our old trusty MGE Magnum power supply with its LCD power usage display: when running alone using the integrated graphics and no storage attached, the older X3360 based Xeon board took on average 76W, by itself not a bad result for the system base. However, the recent X3460 based Nehalem Xeon setup ran at just 59W, showing an immediate quarter power savings.
When adding an Intel X25-M 160GB SSD drive and booting 64-bit Windows 7 Ultimate, and running Sandra 2010 CPU bench, the older Xeon consumption went up to 126W, while the new Xeon stood at 102W, pretty much keeping with the expected power saving scale. And that was keeping in mind the around 35 per cent performance boost in the CPU benchmark by the newer Nehalem based CPU.
At the same time, the CPU temperature comparison was also showing the improvement on the newer CPU, but not by that much. In the BIOS, using the Xigmatek heat sink fan, the default temperature for the older Xeon 3360 was 39 degrees C, while the newer one stood at 37 degrees C, at an ambient temperature in Singapore of 32 degrees C. While running the Sandra CPU tests, and using the Sandra Environment Monitor for observation, the temperature on the old Xeon jumped up to 73 deg C, while the newer one went up to 68 deg C, again keeping in mind that the Nehalem Xeon 3460 did provide the extra performance benefit.
Actually, both of these systems are fairly low power consuming. If you've got the one with the older Xeon 3360, I'm not exactly recommending you to rush out and upgrade it to the 3460 unless there are other compelling reasons that would justify it. However, if you have a generation before that or even earlier, then the combined performance and power advantages - for instance, if you have a dual-core Core 2 Xeon based UP server, you'd get three times the performance at the same power envelope by moving to the Xeon 3460 solution - are more than enough to consider the upgrade or a whole new system.
Price wise, at the present, the street price of the newer solution might be around $1,350 versus $1,080 for the older version - that is, the configurations as tested with 8GB Kingston RAM and with a Xigmatek or equivalent heat pipe heat sink fan combo. The difference at this time reflects well the performance benefit of the newer platform. However, when the power saving advantages are included, the overall preference might go to the newer Nehalem Xeon system.
The two Gigabyte platforms provided an interesting base for a close, one generation gap quick performance and power improvement test, and we can see first hand how much benefit is there from immediate upgrades from platforms divided by just over a year between launch dates. The Nehalem Xeon platform - basically the server bin of the Lynnfield desktop quad-core Nehalem in an LGA1156 socket - easily wins overall, and the power consumption, for that matter, was even lower than what I expected. You want it even better in this format and price? Well, then you'll have to wait exactly another year for the Sandy Bridge based system. µ
Both platforms have good power, performance and price index, Nehalem Xeon wins overall.
The older Xeon platform had no PCIe X16 slots for GPUs, so no 3D graphics test comparisons were possible.
There's nothing ugly here, just different tehnology for different purposes.
Bartender's Score for the old one
Bartender's Score for the new one
Sign up for INQbot – a weekly roundup of the best from the INQ