©ASUSTEK COMPUTER INC. ALL RIGHTS RESERVED.

avsquare · ‎05-21-2015

Hi all,

Finally took the plunge into the X99 chipset with the R5E mobo.

Noticed that the Haswell-E is a little tad tricky to tune compared to my i7-4790K on the Maximus VII Hero mobo combo (which I achieved 4.8GHz on 1.246V stable).

I'm currently clocking my i7-5820K on 1.220V. Ran 2hrs of Prime95 (ver.26.6, not that later versions that would cause trouble to Haswells) and AIDA64 each without any issues.

I tried running the RealBench, set on 2hrs, but wasn't able to complete it. BSoD'd twice.

I set the overclocking AI on XMP using the 2666MHz profie of my RAM, strap at 125, multiplier of 34 for all cores at 1.220V, cache multiplier of 30 with cache voltage on auto. CPU input voltage on auto as well which is about 1.8XX shown on BIOS (BIOS already on latest). Everything else on auto.

I'm not sure if I want to pass this as stable. Increasing VCORE further might push temperatures beyond 80c during the bench (right now it is at 77-79c, Singapore is a damn hot place with ambient easily at 30c)

Any suggestions on what other tweaks I can do other than increasing VCORE?

Korth · ‎05-21-2015

125 strap is a lot more finicky than 100 strap. I think your iMC might need more voltage because 4x4GB DDR4-2666 (1333MHz) might be pretty heavy for a weakling processor (and, let's face it, the i7-5820K is already a lower-binned part).

Chino's DDR4-3000 Setup Guide For Rampage V Extreme has a lot of awesome gritty specifics which should help with any X99/DDR4 setup (or overclock).

"All opinions are not equal. Some are a very great deal more robust, sophisticated and well supported in logic and argument than others." - Douglas Adams

[/Korth]

avsquare · ‎05-23-2015

Thanks for the reply.

I referred to the RAM guide link you've referred to and did the 400% coverage test on my RAM's 2800MHz XMP profile, no errors.

I read up on Praz's comments about the link between VCCIN and VCORE on the other sticky thread, tried tweak around with that and set LLC to Level 9. Not exactly sure what I did wrong or right but maybe I was too aggressive with lowering the cache speed and voltages. I've managed to get a stable config of 4.3GHz CPU speed @ 1.255V with Cache at 3.8GHz @ 1.29V. CPU input voltage at 1.95V, no system agent offset applied. Passed hours of stress tests on AIDA64 and RealBench at around 78c which I find it comfortable. Some of you would probably push for 4.4-4.5GHz but that would push the temperatures to 8Xc range which I guess I would avoid for now 🙂

Whew, OC'ing Haswell-E is sure a different business with Haswell/Devil's Canyon itself! 100MHz base strap was much easier to manage with I feel haha.

Korth · ‎05-23-2015

lol I think what it all boils down to is that the lowest DDR4 spec - 2133MTps (1066MHz) - already exceeds all but the most extreme DDR3 speeds. And 4x4GB is a mountain of RAM, while 4x8GB or 8x8GB is an even mightier mountain. We see all the time that DDR4 is available at 2400, 2866, 3000, 3333 and beyond ... so it's easy to forget that even slow DDR4 is already amazingly fast stuff. And it's easy to be disappointed that our overclocked Haswell-E parts can't as easily run DDR4 at those insane speeds and capacities.

"All opinions are not equal. Some are a very great deal more robust, sophisticated and well supported in logic and argument than others." - Douglas Adams

[/Korth]

avsquare · ‎05-25-2015

Just some concerns here.

I'm not sure if this is a fault (that warrants a RMA of the chip) or the fact that I've been unlucky this time and gotten a really weak chip, but it seems that this chip requires a significant 1.255V @ 4328MHz (127.3 x 34) to be considered stable. Running 1.240V doesn't causes any BSoD crashes but both AIDA64 and OCCT stress test will stop either immediately or very soon due to "hardware failure detected" or "core #0 or core #2 error detected".

No matter what voltage tweaks I've applied, it will never be stable with the 35th multiplier (4455MHz).

Any idea what could be the reason? one or two of the cores could not keep up with the overclocking? (since I use sync all cores)

Praz · ‎05-25-2015

Hello

Seems to be an average overclock. I'm not sure why an overclock not meeting expectations would qualify for a RMA.

avsquare · ‎05-25-2015

Hi, thanks for the reply - not exactly meaning that but just want to check with the experienced members here to confirm the details.

During initial OC steps I've already have an idea that this is a sub-par to average chip. But usually when my OC steps fail it results in a crash rather than "hardware error detected" messages, so I just want to make sure it isn't a physical fault (which this does warrant a RMA I guess).

Hope you get what I mean 🙂

EDIT: Would also like to know what does such "hardware error detected" could mean, for knowledge purposes.

Korth · ‎05-25-2015

lol, agreed, an unspectacular overclock is certainly not a valid reason to RMA a part.

But there is always the Intel Performance Tuning Protection Plan ... which basically gives you one more ticket in the silicon lottery for 35 bucks.

"All opinions are not equal. Some are a very great deal more robust, sophisticated and well supported in logic and argument than others." - Douglas Adams

[/Korth]

avsquare · ‎05-25-2015

Actually I could live with it, but as I've mentioned I'm keen in what those "hardware failure detected" messages mean. Living with an unspectualr OC is okay, but living with a chip with fault isn't, I hope you see what I meant here.

In any case, I read the FAQ of the Tuning Plan, it covers for processor failures when you operate it outside of Intel's technical specifications. So I understand it as you fried/dmg the chip during OC, but it doesn't simply allow you to swap it for another chip just because you wanted to right?

MacClipper · ‎05-25-2015

avsquare wrote:
Actually I could live with it, but as I've mentioned I'm keen in what those "hardware failure detected" messages mean. Living with an unspectualr OC is okay, but living with a chip with fault isn't, I hope you see what I meant here...

Means the hardware failed at the o'ced speed you are pushing, it does not necessarily mean the hardware is faulty as sold ie. run at stock. In other words, run it at stock and see if you still get these messages.

As for all the spectacular HW-E o'cs that you may see, take them with a pinch of salt as others may have a different tolerance of acceptable stability. Some are pure gamers who can game all day with impressive looking high clocks but fail spectacularly in alternative full CPU loads eg. encoding but they may never know cos they only game with their rigs. :cool:

A Mind Forever Voyaging... Across The Universe

Need some tips with Haswell-E OC