Test Configuration

Software Configuration

Windows 2003 was configured with /3GB and /PAE switches in the boot.ini to support the 8GB of memory used for our tests. SQL Server Enterprise was set to use AWE extensions, and a maximum memory limit was set at 6144MB.

Intel Bensley 3.46 Pre-Production System
Dual 3.46GHz Dual-Core Dempsey Processors
Pre-production Blackford based Intel Motherboard
8GB FBDIMM DDR-2 533Mhz
Windows 2003 Enterprise Server (32 Bit) SP1
8 x 36GB 15,000RPM Ultra320 SCSI drives in RAID-0
LSI Logic 320-2 SCSI Raid Controller

Opteron 280 System
Tyan S2882 K8S Motherboard
Dual Opteron 280 (Dual-Core) Processors
8GB Corsair PC3200 DDR
Windows 2003 Enterprise Server (32 Bit) SP1
8 x 36GB 15,000RPM Ultra320 SCSI drives in RAID-0
LSI Logic 320-2 SCSI Raid Controller

Measuring Power

To measure power consumption of each system, we used an EXTECH Instruments Power Analyzer Model 380803. This power analyzer allows us to view current power consumption, and log the consumption at various intervals during a test to a text file. For this test, we used the same Power Supply for both systems, although we recorded the difference between a 750W power supply and a 550W power supply, and it was less than 3 Watts. We should note that the Raid Array was powered by a separate power supply that was not plugged into our analyzer, so we were measuring strictly bare system power consumption. If you’re curious, the Raid Array used about 98 Watts spun up, and averaged 110 Watts during the database tests.

Idle – To measure a system at idle, we booted each system into Windows and let it stabilize by watching the task manager in Windows and the Wattage readings. Once we were at a stable reading, we began recording for 100 iterations of our data logger (which logs every 2 seconds). We then took those numbers and averaged them to get the idle power reading.

50% – We used our database benchmark to measure a loaded system, by adjusting the thread count for the test to a level that produced a half loaded system. Then, we would run our database test for its duration while recording to the data logger. Finally, we averaged those results.

100% – To produce a fully loaded system, we used the same technique as above, except increasing the number of threads until we achieved a fully loaded system.

The future is performance per Watt. Database Benchmark Results
POST A COMMENT

67 Comments

View All Comments

  • Furen - Friday, December 16, 2005 - link

    I must say that performance is very good on these (seriously). The cost may be a bit prohibitive (then again, decent servers are always expensive as hell) since it introduces FB-DIMMs (and 4 channels for this performance). Also, I would like to see someone test these at a 667 FSB just to see how much of a choke point it becomes, since every Dempsey besides the top end 3.46GHz one will use this (I think). Reply
  • Beenthere - Friday, December 16, 2005 - link

    Intel holds a Dog & Pony Show for some hand picked journalists it feels will be "Intel friendly" as a result of getting the "scoop" over the mainstream PC media on Intel products as far as three years off. Then Intel proceeds to provide prototype CPUs for testing months if not years before they will actually be available. What a manipulation of the media and public opinion.

    This is damage control in action folks. Intel is desperate to save face and as many customers as it can while it hopelessly tries to deliver some competitive product in a year or two. The problem is AMD is so far ahead in technology, they can just release better CPUs any time they desire and Intel has nothing to counter AMD's superior products. Even the Intel fanboys and "media friendly journalists" have had to admit that purchasing an Intel product now or in the foreseeable future would be a very poor investment.

    The bad situation Intel is in couldn't happen to a nicer, more deserving company IMNHO.
    Reply
  • ElJefe - Friday, December 16, 2005 - link

    Bensley. thats a gay name.

    The chip does nicely though.

    i hope opteron makes something nicer. i mean, just increase the speed .2ghz on the chip and it will most likely blow it away and still not be power hungry.

    hm.

    1st
    Reply
  • Brian23 - Friday, December 16, 2005 - link

    The cost per month is wrong.

    Example:

    The intel chip pulls 479W at 100% load.
    In 24 hours, that's 24*479 = 11496W/d
    Assume 31 days in a month, thats 11496*31 = 356,376W/m = 356kWh/month
    Assume 14 cents per kWh 356*0.14 = $49.89 dollars per month
    Reply
  • coldpower27 - Friday, December 16, 2005 - link

    Actually I think I agree with you.

    1 kWH = 3,600,000J

    Worse Case Scenario
    479W = 479J in 1 Second, 1,724,400J in 1 Hour, 41,385,600J in 1 Day, 1,282,953,600J in 31 Days.

    Divide by 1 kWH = 3,600,000
    = 356.376kWH

    Multiply by 0.14/kWH
    = $49.89 Per month, the above poster is correct.
    Reply
  • coldpower27 - Friday, December 16, 2005 - link


    The actual difference between running 40 Opteron Systems & 40 Bensley Systems for 1 Year @ 40-60% Load comes to a difference of $5890.4 ~ 1/10 the amount Anandtech reports.
    Reply
  • coldpower27 - Friday, December 16, 2005 - link

    Disregard, used money figures of 172 and 292 for Wattage Oops :P. Reply
  • coldpower27 - Friday, December 16, 2005 - link

    In regard to $5000 ish figure. Reply
  • Furen - Friday, December 16, 2005 - link

    The difference is around $8140 for 40-60% load (which is realistic) and around $10,500 ($23,500 compared to around $13,500) for full load.

    The problem, however, is that the system's power consumption is not the only thing a data center deals with. The more power the system uses, the more heat it throw off. Energy consumption for cooling can match the system's power consumption. Another thing to take into account is the AC-DC and DC-AC power conversion inefficiencies (this is before even hitting the system's power supply, which will lead to even more inefficiency) which will probably add another 20-30% to the real power consumption. So instead of having a difference of $8140 you end up with a difference of $19,536, and that's assuming that you don't need to purchase any new equipment aside from the 40 servers themselves. Another VERY important thing is power density. You could conceivably throw 64 1U systems onto a single rack using Opterons, with a ~17KW peak power draw, but 64 1U Benseley systems would require a peak power draw of ~31KW, not to mention that it's probably very stupid to stick 2 Dempseys into a 1U system (but hey, I'd say the same thing about sticking 4 dual-core Opterons onto a 1U system but people still do it).

    That is not to say that Anandtech's data is right, 'cause it isn't. I just wanted to point out that though measuring power consumption in itself is important, trying to draw conclusions from the power consumption BY ITSELF is not very useful, since it ignored all other related costs and limitations.
    Reply
  • coldpower27 - Friday, December 16, 2005 - link

    How does a difference of $8140 increase to $19,536 which is an increae of over 130%? Reply

Log in

Don't have an account? Sign up now