F3-12800-4GBECO - Voltages and OC on an X48 board

[ph34rb0t]

Greetings!

First of all, I'm a new user of the F3-12800-4GBECO kit and I find it a truly awesome piece of memory (especially considering the price)!

On my Gigabyte X48T-DQ6 (Socket 775) with an Intel C2Q 9450, it runs well with the XMP profile at 1.53V (the board overvolts a bit by default and can't provide anthing less than that voltage).

During searching just for that perfect combination between high CPU frequency, memory throughput, low latency and acceptable temperatures I've experienced that the Ram needs a bit of voltage to get it going in terms of timings above FSB1600. Could be a mainboard-RAM combination thing (this Ram was designed for i7s/i5s after all...) or so.

Anyways, after some trials I've found a truly awesome frequency/timing combination. FSB 6-8-6-24 with T(read) at 6, however at the cost of a VDIMM of 1.68V (1.66V) under load. The PC won't boot with anything less. That has me a bit worried. To measure Ram temperature at those 1.68V, I've stuck a temperature sensor between the PCB and the heat spreader of one of the two Ram sticks. During a three hour OCCT Run and one hour of Linpack, that sensor never indicated more than 44?C (111.2?F); mostly temps in the 42?C range . Upon touching them, the heatspreader felt only lukewarm.


Is that temperature and voltage okay for 24/7 use?

It's voltage that kills components, after all.

(I've been reading a bit about the ECOs across the WWW, but I couldn't find a definitive remark about it's capabilities in terms of voltage and I'd kind of hate to throw away those nice timings, hence me asking here.)


Looking forward to a reply.


Greetings from Germany!



P.S: North bridge voltage has already been raised by 0.1V.

[GSKILL TECH]

No problem at all, as long as the memory is using the voltage, it is fine. Only when you over volt it for no reason, does it become a problem and possibly damage the hardware.

Thank you
GSKILL TECH

[ph34rb0t]

Thanks for your reply!

No problem at all, as long as the memory is using the voltage, it is fine. Only when you over volt it for no reason, does it become a problem and possibly damage the hardware.

Does that mean that 24/7 use of fairly high voltages is okay with low timings and a high FSB, but dangerous at high timings and a low FSB?



P.S:
I think I can scrape 0.05V off the Rams by increasing VMCH of the mainboard from 1.35V to 1.45-1.475V. Would that be worth it?

[GSKILL TECH]

Yes, too high of voltage on too high of timings and low FSB can damage the hardware. For example, if you have DDR3-1600 1.50V, and you shoot it up to 1.70V because you can't stabilize the memory, that will probably cause damage to not only the memory, but the memory controller on the CPU as well since that is trying to support your memory voltage.

If you are familliar with Intel, the new motherboards state 1.65V memory max or it may damage the CPU. Well, not exactly. The reason why they do this is if you set memory very high and leave QPI Voltage (memory controller) at the default 1.10V, it will damage the CPU. They will get many RMAs because people are wreckless with upping memory voltage. This is actually a very good tactic by Intel.

But as serious overclockers know, 1.65V+ VDIMM is not a problem, as long as you raise the QPI Voltage with it.

Same with our memory, people always ask if it is safe to raise voltage. Yes, it sure is, as long as it is done correctly. For example, our low voltage memory is 1.35V, but people have found that it is still effective at 1.70V. As a result, the ECO series is a very good overclocking kit since it is so flexible. At the price, not many memory can beat its performance.

Thank you
GSKILL TECH

[ph34rb0t]

Wow, you guys are quick!


QPI is the FSB replacement and integrated into the new i7 CPUs, right?

So if you say raising QPI voltages along with DIMM voltages is the best way to get more out of you Ram, the same principle would also apply to the old system with a North Bridge, correct?

Keeping this in mind and getting back to my example, tightening the gap between NB voltage and DIMM voltage would actually be more beneficial than keeping the NB voltage low and the DIMM voltage high, wouldn't it?



Excuse the ton of (seemingly stupid) questions, but I want to be absolutely positively sure on this matter.




And yes, the cost-benefit ratio of the ECOs is nothing short of awesome!

[GSKILL TECH]

Yes, QPI is basically the FSB or Northbridge replacement, so basic principles will similarly apply.

Yes, tightening the gap is ideally what you want to do, but if you have to raise NB by +0.1V just to shave 0.05V off VDIMM, it may not be worth it as higher NB temps aren't worth it. DDR3 has on board voltage regulators, so over voltage is not necessarily a concern.

Perfect example is AMD users raising VDIMM to 1.60V for 1.50V memory, thinking it is a memory problem, but in actuality, they should just raise the CPU-NB (memory controller) instead. VDIMM 1.60V forces the CPU-NB to use higher voltages too, so it is like a natural overvolt.

Questions are great. If you have one, don't hesistate to ask. Chances are more people, if not already, have the same question in mind. So if we document it, hoepfully they will search and find it. That is the reason why I may over explain at times, but what I'm actually hoping for is that people will search for this information. More information for them, and more key words.

But not everyone searches, like the AMD problem, the same question is still posted every day.

Thank you
GSKILL TECH

[ph34rb0t]

Yes, tightening the gap is ideally what you want to do, but if you have to raise NB by +0.1V just to shave 0.05V off VDIMM, it may not be worth it as higher NB temps aren't worth it. DDR3 has on board voltage regulators, so over voltage is not necessarily a concern.

Cool!

A german G.Skill techie in a german hardware forum also said that 1.68V isn't too much for 24/7 use, so since double info is twice as good, I'll stick to my OC settings.

Questions are great. If you have one, don't hesistate to ask. Chances are more people, if not already, have the same question in mind. So if we document it, hoepfully they will search and find it. That is the reason why I may over explain at times, but what I'm actually hoping for is that people will search for this information. More information for them, and more key words.

Exactly my point of view!
I *did* a forum search before registering here and found a similar topic from an AMD user about running the ECOs at 1.6V. Yet, I hesitated to dig that thread up since my "problem" was more specific.
Hopefully we can really help others with our little discussion.

But not everyone searches, like the AMD problem, the same question is still posted every day.

Ouch, I feel your pain. What about a thread about that very subject pinned to the top of the forum?



Anyways, thanks for the helpful info.
You delivered one of the, if not *the* best support I've ever got concerning PC hardware!