If anyone's interested, just updated my Gigabyte Q57/i5-650 ESXi box to the following:

MSi B75MA-P45 (B75 (Chipset)
Intel Core i5 3550
24GB Ram

First impressions are very, very favorable. With the system running 11 VMs, the performance increase is impressive and the power consumption is about 30% lower than the dual core:

~28 watts at idle
~34 - 38 watts at light load

Measured with kill-a-watt @220v. Figure aren't comparable to Windows mind, there's an ~10w ESXi idle tax even with a custom power profile.

The supplied Intel fan is LOUD though, its right next to me as I type and its seriously irritating even with the case closed. Once Id updated to ESXi 5.0U1 (didn't do this before swapping the board, not smart) all SATA controllers were detected and VT-d/device passthrough was available, despite the option not being present in BIOS.

Full system specs are now:

MSi B75MA-P45 (B75 (Chipset)
Intel Core i5 3550
24GB Ram
Pico PSU 90w 12V
Trust (yeah I know) 120w brick
Intel Desktop CT GB LAN card
Onboard Realtek GB LAN for MGMT traffic enabled (crazy, but for ESXi Realtek is an advantage...)
1 x WD 128 GB SSD
1x OCZ 120 GB SSD
1x 1TB Samsung 2.5"
1x 640GB Samsung 2.5"

Despite some of the negative press Ivy Bridge has received, I couldn't be more chuffed.

how comes you are going to put 2 ssd drives in ur system?

They are used as datastores for storing the Virtual Machines, SSDs are a really cheap & power efficient method for obtaining a lot of IOPS. Unless they wear out in 6 months .

Just a quick update on the motherboard - VT-d is listed in the BIOS, finding my way around the UEFI bios is taking some time to get used to.

The fan control is working well and in combination with a tower cooler has made the system quiet enough for my needs. Some guys here, would not agree of course.

While searching for an allround esx server at home I came accross your post regarding an intel cpu and msi mobo.
I've recently bought an HP Microserver N40L but found the performance too bad for my use.
I'm interested in getting about the same setup you have and I have a question about power usage.
The wattages you mention are these the ones you measured (actual usage) or the number of watts it uses less than your previous setup?
Because the Microserver in my config uses about 50Watt at idle with 3 disks (older satas) and 2 VMs (Freenas and XP).

The OP's build has no desktop drives, a MSI board and a pico so the 28W idle figure isn't surprising.
Note that this build is not equivalent to a Microserver from a reliability perspective (but maybe you don't care).

It's possible your Microserver has a crappy power supply (HP has been changing its specs across variants).
But some desktop drives have a high power consumption and running three of them could easily explain your 50W figure.

If you want low idle power consumption, you're going to have to pay attention to the power supply as well as the drives.
Using a proper server board or a more powerful CPU shouldn't make much difference as long as you avoid inefficient boards made by the likes of Asus.

The power use I quoted is for the whole system, the cpu is underclocked to 2.5GHz and the RAM undervolted but apart from that theres nothing special about the config and you should easily be able to achieve the same results with similar components.

Id be suprised if the cpu performance of the Microserver isnt enough for a couple of VMs and a small hard drive pool. You dont say how much RAM you have in there, is it the origional 2GB? IF your using ZFS I believe you need more like 4-8GB to get decent performance, perhaps thats the issue? You dont say which virtualisation platform your using, however Im not aware of any that dont give some fairly detailed performance logging which should be able to point you iin the right direction.

As for the power consumption, I understand the Microserver normally idles at about 20w. Im guessing those hard drives are old 7200 models? They could easily use 10w a piece and Freenas may not be spinning them down properly?

Well I tested the performance and even with 2VMs (Freenas 4GB and XP 2GB Ram) the system is not too quick, I've done serveral tests and although the test with iperf showed around 500mbit internal and to my linux box the actual throughput when copying files is slow. From Freenas to linux box = fast 40MB/s copying to it is abnormally slow with around 4-5 MB/s. I've got the microserver configured with 16GB RAM.
Also even the XP system lags and performs way below expectations.

I'm using ESXi 5.0 U1 as hypervisor and done some performance checking but I don't really get the strange figures when doing network tests.

I've migrated my freenas from an old athlon system with only 1GB ram (I use the same disks) and there the performance was consistent and got much higher throughput and also overall system speed was quicker (booting etc).

The harddrives (2x 320GB sata) have been migrated and coupled to the VM via RDM.

What I do notice is the latency on the disks (2x sata on RDM) goes above the 100 milliseconds when performing I/O. I've also configured 3 vmdks (20GB each) as raid z and get a latency of +300 milliseconds. So for some reason I/O has high latencies and I don't know the reason.

I've also ordered an IBM M1015 and my goal is to use pci passthrough so Freenas gets full access to the controller (which can't be done on the microserver).

I dont think the Microserver supports VT-d so I wouldnt plan on using passthrough, although you may be able to use it as a RAID card in ESXi if you have the BBWU/cache.

Just a thought, are you using 1vCPU per VM? If you are you probably shouldnt be. Whats your CPU usage during a write action? I have tried Freenas but quickly abandoned it so Im no expert, perhaps the Freenas forums could be of more help?

Ofcourse using 1vCPU per VM, it's my day to day job running vSphere farms so luckily I knew that one
CPU Usage was quite ok even memory, but the disk latencies skyrocketed like I have never seen on another system.
Just as a test I moved ESXi and the Freenas VM + disks to my desktop system (quad-core amd) and performance more than doubled especially writes.
Also disk latencies were around 20-30 milliseconds, so much better than on the N40L.

I know it doesn't support VT-d, a colleague of mine owns it now, for him it's good enough .

I've bought the Gigabyte B75M-D3H and an i5-3550. Going to try some underclocking/volting and hope too bring it down enough to have a low power consumption and be stable.
Don't know the difference in power usage on motherboards from Gigabyte and MSI.

Finally a decent ESXi Box so I can reinstall my home labo.

What type of memory do you have installed and what voltages are you running at? Just to give me an indication of what I can aim for.

Did you plug your drives in the four main slots? Did you by any chance neglect to enable the drives' cache? It's disabled by default (at least on some Microservers).

Glad to hear its working out for you. If you ever try and successfully implement GPU passthrough please let me know how you did it, Im getting nowhere...

I used 2x4GB Kingston 1333mhz Value Ram and 2x8GB Corsair 1333mhz (again, the cheap stuff), both clocked at 1066mhz and running at 1.35v

Well I installed the system last weekend and am pleasantly surprised.
Current setup:
Gigabyte B75M-D3H
i5-3550
16GB Ram
1 x 250GB 2,5" HDD (VMFS Datastore)
1 x WD 1TB Caviar (non Green) (RDM for freenas)
450Watt PSU (somewhere between 8-10 years old, so not efficient at all)
Without underclocking/undervolting I get 50Watt idle and 60Watt under load.
Currently only 2 VMs since I'm just getting started.

In regards to the GPU passthrough, are you trying with a pci-e card or the onboard GPU?
I've tried it on my other system, which only had a pci-e graphics card and the system hung.
If you plug in an extra graphics card and force the onboard GPU as primary (I think you can give it priority), then it should be possible to perform passthrough.
As soon as I find the time I will try if it works that way, but by my guess it should work.

Sounds like a new PSU would pay for itself in no time. Have you changed the power profile yet?

As for the GPU passthrough, I tried it with a Radeon 5450 using the latest drivers which have been shown to work. However from what Ive read, GPU passthrough has only been successfuly on Supermicro or other true server/workstation class boards. Even then, its a pain to get up and running. Shame, I was hoping to get down to one box for everything (router/WLAN/firewall/HTPC/VM's), but for that it look like Ill need to move to Xen.

A new PSU is one of the next things I will be buying. Searched for the efficiency of this old PSU (from 2004/2005) and found this:
http://www.legitreviews.com/article/151/1/
63% Efficiency which means I'll be around 30 Watts Idle and 40 Watts under light load.
And the efficiency will probably be lower after so many years.
Was thinking of either getting an Seasonic fanless or picopsu.

Also the current 2 drives are going to be swapped.
Probably 3x3TB instead of 4x2TB (lower total power consumption) and a 256GB SSD as datastore.
Also another 16GB ram will be installed because 16 is not sufficient.

I was trying to accomplish about the same, 1 box for all but soon gave up the idea.
Running a dedicated GPU (which is always on) will overtime probably consume more power then powering on my xbmc box when needed (asus eeebox thingy).

Was this powered with a picoPSU? I'm seriously considering one, but am thinking about what wattage to take.

Final system will have max 4 3TB green drives and probably later on an ssd for VMFS Datastore (currently using 2,5" HDD)