Sadly enough, the UD5H challenged me and Massman during the hours we tested the board intensively. We already tested a sample provided by the Benelux office and that one was categorised as 'faulty sample' because it was giving massive issues under cold. Anyway, the second board did better than the previous sample (it actually worked below -35°C), but the final outcome was still a mixed bag for extreme overclocking. Here's a list of our findings with the retail F4, F5 and S4 bios. Plus the replies from a Gigabyte Tech rep on our findings in italic.

Hardware used:

• i7 3770K (retail)
• UD5H (bios: S4)
• G.Skill Flare (F3-16000CL7D-4GBFLS) 2x2GB
• G.Skill RipjawsX (F3-17000CL8D-4GBXMD) 2x2GB
• G.Skill RipjawsZ (F3-19200CL9Q-16GBZMD) 2x4GB
• G.Skill TridentX ("2666 CL11 kit") 2x4GB
• MSI GTX580 TwinFrozr PE/OC

- coldbootbug at -149°C (this cpu had no coldbootbug issue on competitors boards)
- coldbug at -167°C (similar remark about the coldbug issue as before ) This time the board shuts down during PI-32M when going below this temperature. Witnessed on other setups too. Very annoying bug.

There are two CBBs from my observation. One is when system still remembers your bclock which tends to normally be at -150 (some chips don’t have any) and second one is when you clear CMOS and some chips can have a CBB as bad as -75C. Please use S2 bios and see if this behaviour changes!

- PSC, BBSE and TridentX kit didn't show very good OC compatibility over 2400Mhz. The PSC/BBSE had to run AUTO timings to boot at DDR3-2600 and failed with tightening up the CL to 8. Fyi, the PSC kit was tested on the Gene V board and could run DDR3-2666CL8-12-8 stable. The TridentX did not pass DDR3-2800 even though we're sure the CPU can handle that memory frequency (previously tested on GeneV)

In regards to RAM clocking. Please note that GIGABYTE bios tightens secondary subtimings a lot more than Asus and one of the main reasons why you will reach the frequency but be very inefficient until you manually tighten up the right timings. If you loosen up secondary subtimings it will clock higher but logically it performs less. Try slots 1 and 3 on UD3H and UD5H as it clocks a bit higher than slots 2 and 4. BBSE sticks were tougher for me to tune and I think we still need some bios work on all boards.

EDIT : set tRRD to 7 and tRRSR to 5, this will allow most kit's to boot at 2600mhz.

- Regarding timings, the bios displays incorrect timings set by the user or the XMP profile. This might lead to confusion from where to start.

- The CPU ratio is locked in x39 after a hard crash. I'm setting x56 in the BIOS, but in OS it shows x39. I can, however, increase the CPU ratio with the GTL software to x56 no problem. Small issue, but very strange nonetheless.

In terms of CPU ratio observation this happens when you select only 1 core on CPU! If you select 2 cores you will get proper CPU ratio otherwise it will be at x39 which can be adjusted in GTL as you’ve found out.

- The board fails at 110MHz BCLK (in fact, it fails at 107MHz at -50°C too). Massman tested all 10 retail Ivy CPUs on a another board before and all could run 110MHz BCLK easily. Fyi, the UD3H has no issues with that BCLK.

Bios S4 issue , plz flash back to S2

- Debug led no longer functions after a series of crashes. When this happens, you can only get the board up and running after CMOS clear, but any profile you load will result in no boot/no debug again. I had to warm up the system to -107°C before everything worked normal again. We've seen many Gigabyte users required doing this at the EOS event. Sometimes switching to Bios 2 got the board alive again.

- Once in a while, after an OC failure, there are no options visible in the BIOS. It shows the background image, but none of the three OC fail options (load default and reboot, load default and boot, enter bios). We had that very often on the previous board, only a couple of times with this sample.

- Using the GTL software, the system freezes in OS when changing the Command Rate.
- Memory frequency is reported incorrectly after failed OC. When booting up with the DDR3-2666, the post fails at '51', then recovers and shows 1680MHz in BIOS (DDR3-3360).

Known bugs and being worked on

So is the UD5H board bad or what ? Nope, seems we have been using the wrong bench bios during our final testing. The S biosses are far more forgiving and high performance in certain benchmarks, than the early F4 and F5 bios we used with our first board. Small recap that the efficiency is great on these Gigabyte boards. But the bios engineers are slowly working their way through some annoying bugs ( RAM compatibility and the CB - CBB probs ) Those should be corrected as soon as possible. The Z77X-UD3H was far more cooperative during our LN2 sessions. Therefore in our book it's more recommended for Extreme benching then this UD5H.

For air overclocking both the Z77-D3H and the Z77X-UD5H had no issues to be prime95 test stable at 4700mhz with our 3770K ES CPU.