8 replies to this topic

[Catamount]

So I've been playing with RAM a bit, trying to do some decent overclocking (half just for the hell of it, but some games are starting to see benefit) and in doing a bit of reading, since I haven't played with RAM much since the DDR1 days, I've been seeing some funny statements floating around a lot of forums.


I see a lot of people saying that VTT has to stay within 0.5v of the DIMM voltage, warning of dire consequences like a fried IMC if that isn't adhered to. Is there some actual source for that claim, or is it just some armchair rumor based on nothing? Stock VTT is almost always around 1.07-1.08 it looks like, so wouldn't that mean buying a 1.65v kit would basically automatically fry a CPU? How would a VTT/VDDQ difference even cause problems? I've also been likewise seeing people claim VCCSA should match the ever-increasing VTT, and that just seems like a cosmically, hilariously bad idea (AFAIK, that MIGHT well fry something).

Is there any validity, whatsoever, to these kinds of statements?

Edited by Catamount, 03 November 2014 - 06:42 AM.

[Goose]

Can't say I'm seeing this rule.

http://www.overclock...uide-at-the-end

http://www.overclock...us-motherboards

http://www.anandtech...-with-gskill/13

http://www.anandtech.com/show/2636

But you've seen some numbers move around?

[Catamount]

I haven't seen it show up in too many guides, but I've seen forum user after forum user say if the VTT/VDDQ separate by more than 0.5v the moon will crash into the Earth and the universe will catch on fire or something. I'd have to go back and look over where I was on my other machine (will do later just to show what I'm talking about), but I found it rather curious.

A friend of mine promises to go over it later and explain what the deal is (Australian, he's in bed now ).

I did manage to get DDR3-2000@10-11-10-30 stable at 1.58v in memtest, and the computer was stable in OCCT, but for some reason the machine was getting really flaking with POSTing correctly. It would turn on, power off, repeat, it would do its "failsafe" boot and the bios would complain about boot failure, then I could continue normally after telling it to just keep my OCed settings. You have any idea what that's about? It's a decent result for 1600mhz RAM, if I could just get it to behave. I'm going to try increasing VTT (and play around with RAM voltage) and see if that helps since most guides indicate it may help with stability in general, but if this is behavior anyone's seen before then not stabbing in the dark would be nice.

Edited by Catamount, 03 November 2014 - 09:29 AM.

[Goose]

Quote

'http://www.overclock.net/t/1247413/ivy-bridge-overclocking-guide-with-ln2-guide-at-the-end'[/url]]The voltages you should change for high memory overclocking on Z77 on air is the DDR Voltage, and if you like you can try increasing the VCCIO(VTT) and VCCSA(IMC) the VCCIO (VTT) can help with memory OC, however you will also need to increase VCCSA along with it on these GIGABYTE Z77 boards (except on the Sniper M3). If you want to increase VTT you need to increase IMC voltage to within 0.005v below it, so 1.1v VTT would be 1.095v IMC on these GIGABYTE boards. However I didn't really need to change it much at all.

Edited by Goose, 03 November 2014 - 10:45 AM.

[Catamount]

OTOH, this guide recommend never changing the VCCSA, and makes no mention of any need to keep VTT and VCCSA together

http://www.overclock...-asrock-edition

The Z77 Extreme 4 doesn't even have an option to increase VCCSA past a tiny increment. It's not like every other option where you can type in or select the voltage you want up to downright silly levels. It's a tiny list with four nearly - identical options, which is an implicit statement by Asrock to not screw with it. Literally every other voltage on the board can be increased to just about insta-fry levels so it's not like they don't provide general voltage headroom.

I also have never heard anyone tell me they messed with the system agent during OCing. My board's default values for VTT and VCCSA are way farther apart than 0.005 and the machine is perfectly stable with the 3770k at 4.4ghz. In fact my VTT and VCCSA are 30 times further apart than that at default (1.076 vs 0.925, 0.151v difference).

I'm starting to think all these claimed relationships are nonsense. Leaving a certain voltage too high or low can certainly damage things, and in VERY rare cases where one voltage in drawn from another (eg Haswell Vcore vs CPU input voltage) things can destabilize if there's too much difference (same with clocks), but I've never heard of a modest difference between two voltages causing damage, and numerous cases seem to violate all of these relationships people insist be followed without damage or instability.

I think for now I'm just going to increase VDDQ as I please, maybe up to 1.6 or so, increase VTT only if I hit instability, and leave VCCSA alone.

Edited by Catamount, 03 November 2014 - 11:41 AM.

[Goose]

http://www.overclock...0#post_23007496

found by http://www.overclock...0#post_23007496

[Catamount]

That pretty much confirms what I've come to figure by this point, so I'll go with that. Thanks. Now it's just a matter of seeing how far these chips will go. I'm hoping to get them to 2133, even if I have to loosen timings considerably.

[Odins Fist]

My friend runs his 2400 MHz RAM with no issues on an 1150 socket.

[Catamount]

Sadly, it's stable up to 2133 once the computer gets running, but the computer fails to POST correct above 1866. It fizzles 3 times (unless I press DEL), eventually boots with default settings, and then continues if I load my OC profile, too much work for just screwing around with RAM.

I appreciate the help guys, but sadly 1866@9-9-9-27 is as good as I'm getting out of my five year old 1600mhz G.Skill Ripjaw. I can't complain too bitterly there.