DFI unveiled the P55 MI-T36 to the public a couple of months ago. It's an exciting little board based upon the P55 chipset, cut back to the most desirable essentials in order to facilitate a shrink down to the mini-ITX form factor.

What DFI are attempting to do is bring near enthusiast level performance to a form factor that is associated largely with web-browsing and HTPC use. Teamed up with a suitable GPU, the prospects of having a compact multi-purpose PC capable of delivering a high-end gaming experience is rather compelling.

Unfortunately, there's not exactly an abundance of suitable PC cases on the market catering for such builds. Right now, you're limited to a couple of ITX offerings from Silverstone, namely the SG-05 and SG-06 models, while Lian-Li have the PC Q07. Obvious limitations in going such a route are related to PSU requirements. Once you team a board like this up with a hefty GPU, cooling and power demands throw minimum case dimensions towards M-ATX sizing, which defeats the attraction of going for mini-ITX in the first place.

Nevertheless, products like these create demands for innovation and we're sure the MI-T36 has given PC case and PSU vendors something to think about for the future. So although the MI-T36 seems like a bit of an outcast in certain scenarios at this stage of proceedings, it's still interesting to see what DFI have managed to cram into the mini-ITX form factor.

We've spent the past week or so running various benchmarks on the MI-T36 to see how it fares against regular ATX sized P55 offerings. Of course, boards from DFI usually feature overclocking centric BIOSes, so we've taken the time to explore what's possible on that side of things too...


Our board arrived with the P55MIB02 BIOS, which worked well within the context of stock processor speed testing. Onboard peripherals work as they should for the most part, although we have seen reports of the board failing to work properly with USB wireless adaptors. Our issues/niggles during testing were primarily related to overclocking.

First of all, some of the memory sub-timings don't respond to manual changes, so you have to work with what you get. Although, we should add that the default settings are more than adequate in most situations, unless you decide to push the board harder than you should.

One of the other niggles is that you don't get any report of the applied operating frequency in the BIOS other than BCLK. Failed OC recovery tends to be hit and miss too; our board frequently froze on the post screen after OC recovery had kicked in, leaving us needing to reset CMOS to get back into the BIOS. We also found that the board failed to recognize our hard drives on occasion when loading saved BIOS profiles or when the board recovered from a failed overclock. Lastly, S3 resume seems to be limited somewhere between 150-160BCLK on the current BIOS. Any higher than this and the board gets caught in a reboot loop.

All of the above issues have been reported back to DFI who are working on fixes as we speak. The new BIOS will also cap over-current protection to around 90 amps with a maximum VID of 1.40V (currently set to 110 amps). The reason is that onboard power delivery for CPU VCore does not respond kindly to being pushed under heavy loads. We've seen one reported case of a reviewer pushing the board over 3.8GHz with a HT enabled 4 core CPU running stress tests like OCCT and burning out a couple of FETs. Just a heads up to those who are thinking of pushing the MI-T36: it's not designed for it. We'd say you're better off teaming a board like this up with a 750 CPU, and keeping operating speeds below 4GHz. Cooling the PWM FET's actively may help a little, but given the location of some of them, we don't think it is going to be practical or easy. It's something that DFI should have addressed prior to release.

The alternative is to wait and see what the upcoming H55/H57 platform and processors bring to the table. The IGP makes it a more ideal candidate than P55 for mini-sizing, while the reduced core count of Clarksdale processors should allow for reasonable overclocking headroom without being too demanding of power delivery. We haven't heard of any upcoming M-ITX boards from other vendors yet, although we're sure it's a work in progress now that DFI have laid down the gauntlet with the MI-T36. A beefier VRM from the likes of Volterra would see a board like this upstaging everything priced above it for overclocking. FET's delivering up to 45 amps a piece are readily available, and can deliver buckets of current within a small footprint so long as cooling is taken care of properly. Don't be surprised if you see someone running with the concept.

What DFI have done is to pave the way for future products by giving vendors a glimpse of what needs to be addressed and catered for to make the mini-ITX form factor attractive for such use. As it stands right now on the current shipping BIOS, the MI-T36 has a bit of an identity crisis. You've got overclocking features on tap that are usually reserved for boards costing twice as much, while onboard power delivery imposes limitations before you'll really make use of them. Being able to deliver both would have been nice, but we guess compromises were made to work within a reasonable target price. We applaud DFI for pushing the boat out regardless. The execution may not suit everyone, but we think the MI-T36 deserves consideration if you're looking for a little work horse. Just be prepared to work within its limitations.

Performance Summary


View All Comments

  • DigitalFreak - Monday, January 04, 2010 - link

    If DFI provides the BIOS update, you can use the new Clarkdale Cire i3 & i5 CPUs with this board. You obviously can't use the onboard video, but the CPU should work fine.

    I would expect to see H57 / H55 mini-itx boards in the next couple of months.
  • ScavengerLX - Monday, January 04, 2010 - link

    I'm using this board with a 5770 and a 860 in a Silverstone SG-06 case. I love it! I overclocked the bclock to 150 (w/ HT on) while leaving all of the voltages on Auto and its perfectly stable. Pushing beyond 150 bclock causes system instability. With 8 threads running in prime95 Vcore is at 1.15V (3.3GHz on 4 cores, 3.7GHz on 2 cores). Idle temps are in the upper 20s with all of the power saving features on. I'm using a Coolermaster GeminII S with a 120x20mm Yate Loon and it barely fits.

    I'm running all of this on the stock 300w PSU!
  • stefi - Wednesday, January 06, 2010 - link

    Great, I was thinking of using SG-06 with HD 5750 and i5-750 :)) Do you have some pics to post? What brand of 5770 do you have? Is it quiet enough when idle? Do you have any aditional fans on the case? Reply
  • Mr Perfect - Monday, January 04, 2010 - link

    Nice build! That's about what I'm eyeing myself, assuming nothing new pops up at CES. :)

    It's impressive to see someone push the 300watt supply too. 1KW+ PSUs have been the norm for so long, that everyone automatically assumes you can't do jack with anything lower then 600 or so. That said, hopefully the 450(http://www.sffclub.com/index.php?option=com_conten...">http://www.sffclub.com/index.php?option...amp;cati... replaces the 300 in Silverstone's cases.
  • fr500 - Monday, January 04, 2010 - link


    I mean, the GeminII does not have mounting options for 1156 chips

    I'm waiting for the 450w psu for the SG05, anyway my Zotac 9300-D-E with an e8500@3.7Ghz and a GTS250 overclocked is enough for most console ports :D and HTPC duty

  • fr500 - Monday, January 04, 2010 - link

    Edit: Duh I just realized you said GeminII S. My bad Reply
  • GeorgeH - Monday, January 04, 2010 - link

    On the board features page, it lists 6x SATA ports from the chipset and 4x from a JMB322. If there are only 4 ports on the board, what's the point of the JMB322? And isn't the JMB322 a 1 or 2 port part?

    Also, did DFI say who the target market was? Without an IGP and with limited SATA ports I'm having trouble thinking of any good usage scenarios for this board.
  • Rajinder Gill - Monday, January 04, 2010 - link

    Hi George,

    That was my mistake on the SATA ports. I used a template XLS and forgot to delete some of the cell data. It's fixed now.


  • tomoyo - Monday, January 04, 2010 - link

    I was really hopeful for some good clarkdale mini-itx mobos, with only one chip in addition to the processor, it makes for better board layouts. I hope we'll start getting more case choices also, especially a few in server config. I'd love to make a mini-itx Raid box that can outdo the pre-built raid machines out there. Reply
  • zer0sum - Wednesday, January 06, 2010 - link

    That is already possible using ion 330 based boards like the POV.
    4 x sata and 1 x esata, 1Gb NIC supports jumbo frames with 7200 MTU under linux

    Then just add a Chenbro ES34069 chassis with 4 x hot swap sata bays.

    I can't seem to get Freenas working with jumbo frames as I suspect the NIC driver is not quiet up to speed under FreeBSD.
    Ubuntu works perfectly though and I can easily get 80-100Mb/s off a 2 disk mirror.
    Just waiting on more disks to see how raid 5 speed works out.

    It is an expensive little chassis but it is also damn cool



Log in

Don't have an account? Sign up now