Awhile back, I posted somewhere in the Hot Deal subforum that Newegg had an open-box Asus Maximus IX Apex on sale for extremely cheap. I, unfortunately, was slow and Newegg ran out of stock, so I bought the next best thing; an Asus Maximus VIII Impact on sale. Since open-box items lack a manufacter's warranty, I had 30 days to test the board. So I bought a cheap Celeron and some good ram to test the board, and decided to use the non-k oc features of the board. Sorry in advanced for a rather long-winded and rambling filled post.
Specs of the system
CPU: Celeron G3920 (2.9 GHz)
Cooler: Ye' Old Hyper 212 Evo
Motherboard: Asus Maximus VIII Impact (Z170)
RAM: G.Skill Ripjaws V 2x4GB 4000 MHz 19-21-21-41 (Samsung E-Die ICs)
GPU: Sapphire Nitro OC+ R9 Fury
After building the system, I went ahead and downloaded the non-k oc bios from here.
I attempted to downgrade from the BIOS on it to the non-k oc BIOS using the Ez-Flash utility built into the BIOS. The Ez-Flash utility refused to recognize the BIOS as being valid. It turns out that Asus locked out BIOS downgrades after BIOS version 3*** to prevent people from using the non-k BIOS. Luckily for me, the Maximus VIII Impact has a USB BIOS Flashback feature which can circumvent this downgrade restriction. If I was using a different Asus board without BIOS flashback, I would have been SOL. After using the feature, I booted right up into the the non-k oc BIOS.
I immediately set the Vcore to 1.35V, LLC to Level 4 (Level 5 overshoots at idle), and started to raise the BCLK is steps of 20. While doing this, I had to lower the memory straps from the XMP value of 1:40. I aimed to keep the ratio as close as possible to resulting in a 4000 MHz memory clock. I had no issues up until 180 BCLK, at which point it couldn't boot for obvious reasons. I tried lowering the base clock to 170 MHz, but that wasn't stable. I ended up achieving a BCLK of 165(x 29 = 4784 MHz) at 1.35V for my 24/7 overclock with a memory clock of 3960 MHz.
To put this to the test, I ran some benchmarks. The results of which can be seen here.
(Do note that the memory timings are tighter than stock [but not 100% stable] on most of the benchmark runs and some of them were run with higher core overclocks.) Overclocking definitely helped boost performance quite a bit on the 2c/2t CPU and was worthwhile.
Now, onto the obligatory complaints with the build section.
1) My first problem with the build is the motherboard setting the System Agent and VCCIO voltage to dangerous levels when loading the XMP profile before overclocking. For whatever reason, the M8I (Maximus VIII Impact) thought setting these voltages to 1.325V on auto was safe. The XMP profile may have a high frequency and Asus may set these voltages high enough to get every IMC to boot XMP, but jesus. That's enough voltage to degrade the IMC relatively fast. It's better for the board to fail to post than to degrade the IMC of the CPU.
2) The whole blocking BIOS downgrades thing. Asus shouldn't have done that.
3) The IMC on the Celeron is pretty bad. It is unable to boot 1T Command Rate at frequencies above ~3466 MHz even with enough memory, System Agent, and VCCIO voltage. It is known that smaller die Skylake/Kabylake (i3s, Pentiums, and Celerons) have significantly weaker IMCs then big die Sky/Kabylake.
4) For whatever reason, the BIOS refuses to cover the full display. This is just an annoying bug specific to my board though as other M8Is do not have this issue.
As a whole, I'm happy with this build, and would do it again. I meant for this post to talk about Skylake non-k oc exclusively, but I started rambling off I guess.
Celeron G3920 @ 4.7 GHz, Maximus VIII Impact, 8 GB of 4000 MHz 19-21-21-41 Samsung E-Die