Honestly, I have used this as my Linux box, but once I had review HP EX470, which has only single core CPU, I decided to put WHS on Shuttle K45 with dual core Intel CPU and see the differences it may have. Although I have only 160GB on K45 box, it should give an idea whether or not CPU can make any significant difference.
First thing first, my K45 box specification.
|CPU||Intel Celeron dual-core E1200 @1.6GHz 512kB L2 Cache 800MHz|
|Operating System||Windows Home Server (based on Windows Server 2003)|
|Memory||2*1GB DDR2-800 RAM (2 slots in total)|
|Graphic card||Don’t Care|
|Chipset||Intel i945 + Intel ICH7|
|Hard Drive||1*Seagate ST3160811AS 160GB|
|Networking||Marvell Yukon 88E8056 10/100/1000 (jumbo frame supported)|
|Extra||2*Internal HDD bay
5*USB (4 on the back), 1* Ethernet, 1*PCI slot
|Dimension (LWH)||11” x 7.5” x 6.5”|
As you see, I tried to have closet specification as HP EX470 as much as I can although one biggest factor here is the hard disk drive. It’s only 160GB while on HP EX470 I have 2 of 1TB from Hitachi and Seagate. Well, we’ll see if this hurts a lot, but please bear with me because I couldn’t find any spare bigger one at the moment too.
With demigrator hard-working, you could expect the main HDD performance on HP EX470 drop quite a bit, but as you see both 1TB still outperform 160GB easily. That was testing in working condition with 80% space used while there is basically none K45 box.
Dual-core vs Single-core CPU
What we want to find out here is whether dual core could help anything =) By comparing with HP EX470 which has AMD Sempron 3400+@2GHz 256kB L2 cache, Intel Celeron dual-core E1200@1.6GHz 512kB L2 cache is such a suitable opponent—not too different when core-by-core is concerned. Ok, wPrime can show exactly how different :-P
When only thread is concerned, surprisingly AMD Sempron 3400+ outpaces Intel Celeron E120 by 12 seconds—that’s about 10%. When 2 threads is running simultaneously, Dual-core CPU shows its potential clearly. It cuts processing time by half as expected while single-core CPU have to run 1 second more to complete. What about 4 threads then? the same situation applied; dual-core needs only about half of single-core processing time to do the same job.
What does it mean to us? It means if you use WHS, demigrator.exe—the most CPU time killer process in WHS—will take less CPU time to do things and CPU will have more time available to other tasks.If you are using SageTV, antivirus, etc, on your WHS box, dual-core CPU will definitely helps, but if not, I don’t know if that will help us about network throughput or not though.
System 1 configuration is Thinkpad X61T;
|CPU||Intel Core 2 Duo L7700 @ 1.8GHz|
|Operating System||Vista Business 32-bit|
|Chipset||Intel GM965 + ICH8-M|
|Memory||3GB PC2-5300 DDR2 SDRAM|
|Network||Intel 82566DM onboard gigabit|
|Hard drive||120GB 5400rpm 2.5" Hitachi|
System 2 configuration is white box desktop
|CPU||Intel Pentium E2180 @ 2.00GHz|
|Operating System||Vista Ultimate 32-bit|
|Chipset||Geforce 7150 + nForce 630i|
|Memory||3GB PC2-5300 DDR2 SDRAM|
|Network||nVidia nForce 10/100/1000|
|Hard drive||500GB Seagate ST3500630AS 7200rpm 16MB Buffer|
First we tested with iozone;
iozone -Rab .\k45-1000.xls -i 0 -i 1 -+u -f z:\x.tmp -y 64k -q 64k -n 32M -g 2G –z
Yeah, testbeds are the same as HP EX470 review, but I expect to see something different in term of throughput.
Comparing to HP EX470 test, trend is awfully similar, but that’s expectable since we use the same OS. The value, which is lower than HP EX470, could come from 2 factors: hard drive & NIC chipset. Oh I wish I had spare 1TB HDD to test on this.
The same applied for robocopy, there is no different at all. I don’t think 1TB could help get significant boost though. I might be wrong though.
Write performance dropped a lot here. It has to be hard drive factor indeed since this 160GB has such a bad seek time comparing much higher plate density.
After all these result, I couldn’t stop curiosity of how WHS use CPU to process things when only transferring file was concerned. So I had a look of how busy CPU was when there was only 1 io-zone test was involved.
That was only one core job!! another core just sat doing nothing here. What if there was many transfer file involved then? Unfortunately, CPU usage was still similar. Its pattern was not the same as what you saw in HP EX470 because there was no demigrator working in the background—1 drive w/o any duplicate.
This was a quick test just after I was wonder if dual-core CPU could help anything performance or not, but very different HDDs in both test might effect more than I thought. However, what I can I see so far is dual-core PC won’t help you NAS produce any magnificent performance over single-core CPU, but for things like WHS with many hard drives, it might help demigrator process a little bit more smooth =)