Anyone with Haswell E and LED vengeance DDR4 3000+ memory

[Brighttail]

GL to you.

[Korth]

Truly faulty sticks are indeed becoming more of a rarity. As in, sticks which either don't work at factory-rated specs or which fail to function at all - when installed in systems which are (otherwise) completely operated within factory specifications. You can't claim a memory stick is faulty just because it won't work in combination with other overclocks, you can claim it's faulty (and get it replaced) if it fails to operate at spec when the rest of the system runs lame stock/factory settings all across the board.

I second Brighttail's advice - test your memory kit, one stick at a time.

Some overclockers/reviewers used to religiously advise people personally "bin" every stick in each DDR4 kit. Test each stick, one stick at a time. Also test out the overclocking limits, maximum stable frequency and minimum timings available at each voltage. Exhaustive testing takes a while and involves many restarts. Carefully note every halt/crash/BSoD or software error, no matter how trivial it may seem, and track all the data on a graph of some kind - it's worth the effort if you really want to push hard limits.

There's almost always one (but almost never two) "weak" stick in every kit which just barely performs at the minimum spec for the whole kit (this is at least true with quad-channel DDR4 kits from Corsair, G.Skill, GeIL, HyperX, ADATA, Mushkin, and Crucial, lol). There's often one and sometimes even two "strong" sticks which can be pushed to speeds equivalent to the next performance bin, they got binned down because some obscure tertiary timings or guardband tolerance or something just won't play nice with all the faster higher-binned sticks. The "average" and "strong" sticks can often be overclocked even faster if more voltage is applied, unless their bin was already factory overclocked at a high voltage (1.35V+), while the "weak" stick holds the entire memory kit down.

My understanding is that "pro" overclockers basically toss out all the "weak" and "average" sticks, they prize the very best of the "strong" sticks during record overclock attempts. And my understanding is that Hynix silicon is generally preferred vs Samsung or Micron for subzero overclock attempts, though I could be wrong and it hardly matters to most of us "amateur" overclockers anyhow, lol.

[Brighttail]

Truly faulty sticks are indeed becoming more of a rarity. As in, sticks which either don't work at factory-rated specs or which fail to function at all - when installed in systems which are (otherwise) completely operated within factory specifications. You can't claim a memory stick is faulty just because it won't work in combination with other overclocks, you can claim it's faulty (and get it replaced) if it fails to operate at spec when the rest of the system runs lame stock/factory settings all across the board.

I second Brighttail's advice - test your memory kit, one stick at a time.

Some overclockers/reviewers used to religiously advise people to "bin" each stick in a DDR4 kit. Test each stick, one stick at a time. Also test out the overclocking limits, maximum stable frequency and minimum timings available at each voltage. There's almost always one "weak" stick in every kit which just barely performs at the minimum spec for the whole kit (this is at least true with quad-channel DDR4 kits from Corsair, G.Skill, GeIL, HyperX, ADATA, Mushkin, and Crucial, lol). There's often one and sometimes even two "strong" sticks which can be pushed to speeds equivalent to the next performance bin, they got binned down because some obscure tertiary timings or guardband tolerance or something just didn't play nice with all the faster sticks. The "average" and "strong" sticks can often be overclocked even faster if more voltage is applied, unless their bin was already factory overclocked at a high voltage (1.35V+). My understanding is that "pro" overclockers basically toss out all the "weak" and "average" sticks, they prize the best of the "strong" ones during record overclock attempts. And my understanding is that Hynix silicon is generally preferred vs Samsung or Micron for subzero overclock attempts, though I could be wrong and it hardly matters to most of us "amateur" overclockers anyhow, lol.

When you are using Corsair Dominator Platinum Special Edition kits, where there are only 500, kinda hard to bin them with other kits of the same time to get the 'best' of them all, especially when these are supposed to be the best of the bestest best bests.

[Brighttail]

I'm really just curious how far I can push this kit but I don't know enough about the purpose to each timing. I've looked around to see if anyone has attempted to push it and I haven't seen anyone do a write up of it.

[Korth]

I still think you could push one or two sticks much faster than the others, there's always a "weakling" holding things back. Running just the single "strongest" stick would be great for achieving top benchmarks, but not so great when you need a realistic amount of RAM for real computing/gaming.

It would be interesting to see how much variance exists in these four sticks, especially when these are supposed to be the best of the bestest best bests, lol.

[Brighttail]

I still think you could push one or two sticks much faster than the others, there's always a "weakling" holding things back. Running just the single "strongest" stick would be great for achieving top benchmarks, but not so great when you need a realistic amount of RAM for real computing/gaming.

It would be interesting to see how much variance exists in these four sticks, especially when these are supposed to be the best of the bestest best bests, lol.

I would love to take the time to test it but the methodology shown earlier in this thread seems a little dated.

[Korth]

Can you suggest a faster method?

[Brighttail]

A faster method? me ? Lol.

[adamr]

An update.*

Supplier tested memory and found it to be faulty. (An entire week of testing). Got my replacement kit and so far it appears good (will do more testing as this is early days )*

[CMDRZOD]

And both had faulty sticks. The first set one stick failed after a day. Nothing I did could get that stick to work again. Returned it for another kit. That kit had one stick that would not work right out of the box. Brought it back and I'm using my old Gskill 2400Mhz. Thats the first time I've ever seen failed RAM sticks.