Which is faster 2xXeon 2.0ghz or Pentium D 805 as a server?

[praeses]

I have a peculiar situation here with two servers, one that was a mediocre server when originally purchased a few years ago, and a relatively recent absolute budget server. Soon I plan on rolling out windows 2003 on both servers primarily for shadow copy to handle the small "whoopsies" when I am not around. I would like the faster server to be paired with the more demanding office.

I realize they are not quite apples to apples comparisons since one server is running hardware RAID 1 while the other is SATA with software RAID 1 and the lacks of ECC capable RAM.. The budget server has proven to be quite reliable to date however.

Server 1:
Dual Xeon 2.0Ghz 1MB L2
1GB DDR
2x32GB GB Ultra 160

Server 2:
Pentium D 805
1GB DDR
2x80gig SATA

Overall currently Server 1 averages 4% CPU utlilization during work hours, spiking much higher during the backup procedures scheduled in the evening, while Server 2 averages 1% but also has much less of a load.

Although Server2 seems to fair much better in most benchmarks I have found it rather difficult to find a benchmark that simulates the environment that they currently run in(Mostly very small files, email, authentication, etc). Since both offices are in differing cities, and the future OS installation for both of them will probably be the last during their lifetime (2010) I am looking for other people's expertise before potentially wasting a day or two's work.

Benchmark suggestions would be wonderful as well.

 

[InlineFive]

It seems that in your case CPU utilization is irrelevant. So with that in mind I would place Server 1 at the main office because it appears to have more fault tolerant and faster hardware.

 

[The-Noid]

Server 2 is definetly faster hardware. I would use that. I don't believe the Xeon 2.0GHZ is DDR ram. It is more than likely that it uses RDRAM, which is horribly ineffiecent with the 100 mhz bus. Server 2 will run laps around the first. The only thing the first has going is the SCSI drives, but unless they are 15000 rpm you will nto see a huge difference.

Cheers

 

[praeses]

I don't believe the Xeon 2.0GHZ is DDR ram. It is more than likely that it uses RDRAM, which is horribly ineffiecent with the 100 mhz bus.

The Xeon 2.0 is indeed using DDR at a measely 266mhz over the default 533mhz
bus compared to the dual channel DDR2 at 533 over the 800mhz bus.

However, you did get me to double check the specs I posted, the Xeons are 512k L2(no L3), versus the 1MB/core on the 805. Granted, there are architecture differences such as pipeline stages etc, similar to the old northwood vs prescott debate as well as some chipset differences.

In the end, the drives in Server 2 are much faster (both 7200rpm however HDTach and Sisandra show faster throughput despite running in software RAID 1), also have more cache (2meg vs 512k) but conceivably the cpu utilization may be substantially higher with numerous small files which they typically deal with.

Edit: originally mistyped Raid0, meant RAID1.

Personally I would much sooner replace both of them at the end of 2008/start of 2009 but I had to make some choices in the budget and there were some more pressing matters to attend to(ie 256megs ram on desktop running Windows Vista in 2010? I think not...)

 

[Ayah]

If they're both 7200rpms, of course the SATA drives will be MUCH faster. SCSI doesn't shine unless it's 15K with many, many users accessing at the same time.

 

[praeses]

Rougly 30devices are attached to the server, but only half of those are actively used at any given time.

 

[XBoxLPU]

Um first post said Raid 1 for both and than your second post stated that Server # 2 is using Raid 0 (which of course IS going to be faster in benchmarks) ..

 

[praeses]

Typo, should be read RAID 1 for both.

Edit: corrected now.

 

[TerryMathews]

Architecturally speaking, the Pentium D series chips work the same as Intel Xeon DP chips. The Pentium D is two fully-functional P4 processor cores tied on the GTL+ bus with the proper ID bits set for each core.

If you look at a bus diagram, they're wired the same.

Now as to overall performance, I'm assuming the dual Xeons are on a 'proper' Xeon motherboard, and not a workstation-class board (look below for explanation). That would mean that it requires registered RAM, which has both pros and cons. The dual Xeon (probably) has a lot more DIMM sockets and the capability to add a lot more RAM, which could offset any performance differences between the two systems. Registered RAM is slower than unbuffered (standard) RAM, by one CAS latency, so for raw performance it's at a disadvantage even if it's running at the same clock frequency, which it's likely not.

The Pentium D is probably faster overall, although you could make the dual Xeon into a better server.

Now as to the explanation, two or three years ago towards the end of the Socket 604 Xeon's life, Intel was feeling the squeeze of the AMD Opteron line and didn't have a server-class chipset that supported 533 or 800MHz bus. They created the specification for a workstation-class board that a few companies like ASUS implemented - it was basically an i875-based Xeon board. Just like we all used to overclock Northwood P4s, we could now use to overclock Prestonia (basically Northwood) Xeons.

At the same time, some 1.6GHz LV Xeons appeared on eBay that overclocked 100% to 3.2GHz. The rest, as they say, is history.

 

[VooDooAddict]

There are also drastic differances in storage size.

The systems are similar enough that I would put them into service based on projected user storage needs.

That said, if reliablity and uptime are critical to one location and not another, use the Xeon/ECC system at said location. Nothing is a garontee, but you are reducing the chances of a problem at the critical location that way.