TR Forums

Can't seem to work out what the deal is with "OC" ram settings, with regard to what's "safe" for the CPU. Wonder if the trusty Techreport forums can enlighten me

I just ordered a 5960x off ebay - apparently a brand-new warranty-replacement from Intel. The seller happened to warn me that Intel specifically advised he shouldn't have been running 3000MHz RAM, as they reckoned that was what caused the IMC, and thus the chip, to fail.

That's gotten me concerned, as yes, I was intending to run Trident Z 3200MHz RAM with the thing.

I've read various bits and pieces regarding the "CPU strap" setting, and how certain settings can be better or worse for performance/stability, and thus that certain XMP profile speeds tend to be preferable to others... with suggestions that for various reasons, 2800 and 3000 ram isn't so good, but that 3200 works better.

Can anyone dispel the voodoo on this topic for me? The concensus seems to be that running OC ram can't damage the CPU, but Intel very clearly thinks differently (and whether or not they're feeling forgiving, officially it seems running OC ram invalidates their warranty).
Is it the raw speed itself that can stress the IMC, or is it stressed less running at certain CPU strap settings? If so, which is better... keeping it to 100MHz, or is that the *more* stressing option, and it'd be better set to 133MHz or something?

Boosting RAM voltage definitely stresses the DRAM controller. Not sure how much of an effect RAM clock has.

Ah right, so it might be okay to run an OC, up to the limit of what 1.2V can manage? Looks like 2800 is the upper bound most RAM is specced to still run on 1.2V.

The speed of the memory will not damage the chip, but voltage will.

Anything under 1.4v is considered the max safe for normal usage 24/7 on intel's DDR4 memory controller.

Interesting. Thanks guys.

So, seeing as Intel specifically only state support for RAM operating at 1.2V, is there a chance that 1.35V could cause damage over time? Or is that unlikely?

The guy I'm buying off reckoned that he'd read suggestions that Haswell-E generation chips possibly didn't have as strong IMCs as Broadwell ended up having, as in the early days of DDR4 they didn't anticipate the speeds of RAM people would end up using, and so Haswell-E should be kept within spec. Does that sound plausible?

A lot of people run 1.35V RAM without issue, but then a lot of people run 1.35V Vcore without issue, and I had a CPU (4690K) fail under 1.27V Vcore. It's all about probabilities of failure, and what tradeoffs between that and speed you want to make. Most people would say 1.35V RAM is no problem at all and you should go for it, but if reliability is a high priority for you, 1.2V will be a touch safer. I've seen 1.25V RAM around too, which might be good.

As far as clocks, I can't imagine stuff in the 2800-3200 ballpark being a problem if the voltage isn't. Higher clocks probably leave less headroom in case something degrades, but if it could actually cause damage, I'd think the 3733 and higher kits would have picked up a worse reputation than they've got by this point.

1.35v is safe.

I am running this memory on a 5.1ghz 7700k (1.375v), XMP enabled (1.35v memory, 3600mhz) and I expect no issues..
https://www.newegg.com/Product/Product. ... 6820232194

I always try to find the fastest memory at the default voltage. I keep my machines for a very long time and this has worked well for me.

Your Haswell-E is "safe" to 1.5v.

http://www.legitreviews.com/what-is-the ... ing_150115

1.2V or lower = Best for DDR4
1.35V = okay voltage for overclocking kits
1.5V =absolute max voltage allowed for Intel XMP 2.0 profiles and max suggested voltage

http://www.legitreviews.com/intel-x99-m ... own_150008

The ‘high’ DDR4 voltage (1.35 to 1.5V) that we were using on the board shouldn’t have caused any issues since the memory controller in Haswell-E can actually support DDR3 and DDR4, although only DDR4 is being implemented. DDR3 memory kits run at 1.5V, so that almost eliminates that from being the issue

Also found Skylake/Kabylake's IMC supports DDR3L, which is 1.35v, that is the highest I would go on that platform, and to be honest if I had an X99 setup, 1.35v is the highest I would want to go. IMO intel silicon does best 1.4v-ish or lower long term.

Thanks all for the input. Much appreciated!

ColeLT1 wrote:

The ‘high’ DDR4 voltage (1.35 to 1.5V) that we were using on the board shouldn’t have caused any issues since the memory controller in Haswell-E can actually support DDR3 and DDR4, although only DDR4 is being implemented. DDR3 memory kits run at 1.5V, so that almost eliminates that from being the issue

Well that's an intriguing nugget... after searching around, it seems Asrock even released a couple of DDR3-supporting Haswell-EP boards, and there are a select 3 different Xeon E5 v3 models that have stated DDR3 support. Seems there's mileage to that assertion.

Well, I'll give it a bit more thought while the rest of my build turns up, but maybe I'll give it a shot then, and hope for the best!