New (used) vmware hosts for home

farscapesg1

2[H]4U
Joined
Aug 4, 2004
Messages
2,648
We decided to retire some older servers that are out of warranty at work so I grabbed a couple to play with. My home servers are getting a little constrained due to memory (32GB with a Xeon X3440 and only 24GB with a Xeon E5640 in the second host).

I grabbed a couple Dell R710's with dual Xeon X5650 procs and 96GB RAM (4 onboard nics, extra 2 port PCIe nic, dual single port 4GB fiber cards, 8x 2.5 drive bays, 2 146GB 15k SAS drives). Hey.. for free can't beat it right?

The problem comes down to worrying about power consumption since they will be at home running 24/7. I use my existing servers for lab/training (really need to get more certs) as well as some home 24/7 uses (untangle, SageTV/Media Center PVR, 2012 Server Essentials for PC backup, torrent/usenet, plex server, vcenter).

My current E5640 system runs at about 80W idle and I've gotten the R710 down to around 125W idle by dropping it to 64GB RAM, taking out the fiber cards, and replacing the two SAS drives with SSD drives. I still need to tinker with it, but I'm trying to decide if replacing the X5650 procs with L5640 ones from ebay would be worth it for energy savings vs the decreased performance, since they will be on 24/7 but the CPUs won't see more than 10% usage (if that) for 95% of the day. At 60W TDP versus 95W TDP, at full load that would only be about 70W difference, or about $6.50 a month per server at 12 cents/kwh, so the procs would pay for themselves (about $65 each on Ebay) in approximately 2 years... if running at full power 24/7. However, what I can't seem to find is a good comparison of idle power usage between the two procs to determine if my low load requirements would make a difference in power consumption.

TLDR;
Anyone know the idle power differences between dual L5640's vs dual X5650's?
 
Thanks! So @20 watts difference at idle... doesn't seem worth replacing them then as the X5650's will get the work done and get back to an idle state faster than the L5640's.

Gotta work on some cabling and get one of the R710's added to my vcenter so i can move all my VMs over and watch the power usage for a day or so (and noise... though at idle it isn't any louder than my 1st gen Norco 4220 beast, won't know for every day usage until I get some VMs running on it).
 
Just to give you some idea' again based on x5670's coz that what I have.
These were folding rigs so all cores were maxed out

Dual L5640's, 12Gb Ram, 80Gb hard disk was 217w from the wall on a Seasonic x400 PSU
Dual X5670's 6Gb Ram, 80 Gb hard disk was 297w from the wall on a corsair AX 650w PSU

Currently gaming with the dual x5670's, 12gb ram, GTX 670, 2 x 240Gb SSD and 2 320Gb HDD runs 250w from the wall on an old HX 1000 PSU
 
Thanks for that info. I'm still tinkering with them to see if there is anything else I can tweak to bring the power down a little more, but I noticed the one I'm testing with 3 VMs was idling at about 120W. Only 40W more than my single cpu E5640 host. Running Handbrake on a DVD maxxing out 8 vCPU brought it up to around 250W, but the conversion finished in 20 minutes vs 40-50 minutes on the old system (which could only use 4 vcpu and maxxed out at about 200W).

Gotta plug my X3440 system in and get some readings on it at idle and load. That one is going to a bigger difference I think, but if I can get both R710's to hover around 150W 80% of the time it will probably be worth the extra electricity cost for the compute power and flexibility to bring up more test VMs for short periods of time to self train (MCSE, Citrix, VCAP, etc.).
 
Back
Top