Problems running 4 sticks of RAM?

[BlueWeasel]

My brother's current Z97 setup currently has 8gb (2x4gb) using this RAM: G.Skill Ripjaws 2x4gb DDR3 1866

He's thinking about upgrading to 16GB and ideally would simply purchase an identical 8GB kit, providing a total of 4x4GB.

Are there issues running 4 sticks of RAM in Z97 boards? I know this could be a problem in the past due to the extra stress it puts on the memory controller. I don't think he overclocks, so the RAM would be running at the stock speed or lower.

I know Z97 doesn't support quad-channel, so would the memory be running in a two pair x dual channel configuration?

 

[LTC8K6]

It will run in Dual Channel mode and it should be no problem.

 

[Sheep221]

Unless you would heavily increase the voltage, that could potentially damage the CPU onboard mem controller, but running 4 sticks in stock is what the CPU is designed to work with so no problem there.

 

[BlueWeasel]

Thanks guys!

 

[BonzaiDuck]

My brother's current Z97 setup currently has 8gb (2x4gb) using this RAM: G.Skill Ripjaws 2x4gb DDR3 1866

He's thinking about upgrading to 16GB and ideally would simply purchase an identical 8GB kit, providing a total of 4x4GB.

Are there issues running 4 sticks of RAM in Z97 boards? I know this could be a problem in the past due to the extra stress it puts on the memory controller. I don't think he overclocks, so the RAM would be running at the stock speed or lower.

I know Z97 doesn't support quad-channel, so would the memory be running in a two pair x dual channel configuration?

Does anyone recall this as a "problem in the past?" I'd never had a problem running four sticks going back to DDR[1] days. How would it stress the memory controller without an increase in voltage to it, and without an increase in the vDIMM voltage? Depending on how you run them but at mostly stock latencies, increases to the VCCIO/VTT or IMC would be miniscule if necessary at all.

 

[dave_the_nerd]

Shouldn't be a problem running stock.

The only thing to remember is that some motherboards that support, say, DDR3-1866 might only support DDR3-1600 with all four slots populated. (Backing off one step with all four slots populated isn't atypical.) So don't freak out if that happens.

The motherboard manual will probably spell it out for you.

 

[BlueWeasel]

Does anyone recall this as a "problem in the past?" I'd never had a problem running four sticks going back to DDR[1] days. How would it stress the memory controller without an increase in voltage to it, and without an increase in the vDIMM voltage? Depending on how you run them but at mostly stock latencies, increases to the VCCIO/VTT or IMC would be miniscule if necessary at all.

I mean to say "potential issues if overclocked." I've never actually used all 4 slots before and base my question on recommendations from back in the early days of DDR.

 

[myocardia]

...

 

[mv2devnull]

The memory controller was not in the CPU "in the past".

Does anyone recall this as a "problem in the past?"

As hearsay. On some Forum a poster did complain that a P45-board from vendor X could not drive 4*4GiB sticks (the max for P45's memory controller) from vendor Y, while 3*4GiB was ok. The poster was mainly blaming the Y's quality, but also referring to "tales of lacking power supply in P35 generation boards". Hearsay of hearsay. INET legend?

On Xeons -- IMC and three channels of ECC RAM -- Intel documents that the controller drops memory speed if there are more than one module per channel. That compensates for the additional power needed. Those Xeon boards usually have three memory slots per channel. I have not seen similar document for consumer products, which usually have two slots per channel.

 

[LTC8K6]

I don't recall a "4 sticks" problem, but I do recall long ago speed limits if you used 4 sticks.

Such as the board might be rated for 1600, but if all 4 memory slots were occupied, it might drop to 1333.

Dual channel mode used to be picky, too.

Nowadays Intel's controllers will even run in dual channel mode with mismatched ram and even with 3 sticks of ram.

Not sure about AMD.

 

[BonzaiDuck]

I don't recall a "4 sticks" problem, but I do recall long ago speed limits if you used 4 sticks.

Such as the board might be rated for 1600, but if all 4 memory slots were occupied, it might drop to 1333.

Dual channel mode used to be picky, too.

Nowadays Intel's controllers will even run in dual channel mode with mismatched ram and even with 3 sticks of ram.

Not sure about AMD.

[And also responding to myocardia, Blueweasel, dave_the_nerd and mv2devnull]

I stand corrected more or less. I had only recently (re-)discovered that a 2x8 + 2x2 arrangement won't allow for CR=1. Even with CR=2, I had to bump up (not the vDIMM or vCCIO but) the vCORE by 12 mV.

But when I saw "stress to the memory controller," I was thinking of either shorter CPU lifespan or some degradation to hardware that wouldn't occur with 2 sticks.

I'd never seen something like that happen. And of course, it was the break with LGA-775 when the MC became integrated with the processor, so we're also talking about two different hardware regimes.

 

[thilanliyan]

I am using 4x4GB sticks in my Z87 board. No issues.

 

[VirtualLarry]

I don't recall a "4 sticks" problem, but I do recall long ago speed limits if you used 4 sticks.

Such as the board might be rated for 1600, but if all 4 memory slots were occupied, it might drop to 1333.

Dual channel mode used to be picky, too.

Nowadays Intel's controllers will even run in dual channel mode with mismatched ram and even with 3 sticks of ram.

Not sure about AMD.

My AMD FM1 board(s) are rated for DDR3-1866, "one DIMM per channel". With two DIMMs per channel (requires a four-slot mobo), the maximum DIMM frequency drops to 1600.

 

[BonzaiDuck]

I am using 4x4GB sticks in my Z87 board. No issues.

Hannibal Lector might say . . . "Have ya done things to 'em?"

Seriously, though. You configured to the stock settings, didn't you?

If you didn't, did you try to "customize" your settings with a single 2x kit?

As far as I can see, given the market for parts at Egg and many elsewhere, You'll see an equal showing of 4x kits for either quad-channel or dual-channel boards. So two 2x kits should run at their stock or spec settings.

Not a great deal of hardware passes through my hands compared to some other members. But I'd mentioned before a model of Ripjaws 1600's @ 9-9-9-24 as a 2x4 kit. You could, of course, run them with CR=1. But you could also overclock a 4x4 of those sticks without changing VDIMM --to 1866 and 10-10-10-30. And I'm pretty sure they would run at CR=1.

But you'll not likely choose RAM with that much information or with that in mind. You might not know what you'd get for overclocking possibilities.

I'm more inclined to buy a large 2x kit -- 2x8GB my current choice. Can't say about the Kingstons; I've had Corsair XMS that will likely do it, but I pretty much expected that a 2x kit of G.SKILL would work fine at CR=1.

 

[bigboxes]

In building the rig in my sig I had problems with four sticks of ram. First, I had defective ram slots and RMA'd the mobo back to Amazon. The replacement board worked fine. However, one set of ram sticks failed memtest and I RMA'd the pair back to G.Skill as it was past the 30 days to send it back to the seller. The return pair of ram worked fine. 4790k oc's to 4.7 without issue. I haven't pushed it further as I just don't care about high oc's any longer. I keep it set to a straight 4.4 on all four cores.

 

[myocardia]

...

 

[BonzaiDuck]

In building the rig in my sig I had problems with four sticks of ram. First, I had defective ram slots and RMA'd the mobo back to Amazon. The replacement board worked fine. However, one set of ram sticks failed memtest and I RMA'd the pair back to G.Skill as it was past the 30 days to send it back to the seller. The return pair of ram worked fine. 4790k oc's to 4.7 without issue. I haven't pushed it further as I just don't care about high oc's any longer. I keep it set to a straight 4.4 on all four cores.

It says something about complexity. I knew a Swiss chef who always said "Simple is best." The more factors and instances you introduce, the more likely some problem with any one of them will cost time, trouble and even uncertainty about the cause of a problem.

And it also makes some sense to find the upper bound of possible or workable clock speeds, and run the CPU slower than that because it's a performer anyway. All four cores IS an overclock for a chip that will turbo to 4.4 by default. A lot of folks chose to do it that way with Devils Canyon.