Six Of One, Half Dozen The Other?

Mar 25, 2006
I'm building an ASRock Deskmini to use strictly as my trading computer. I plan to run Trader Workstation from Interactive Brokers and Thinkorswim from T.D Ameritrade at the same time and maybe a few browser windows open and possibly Excel. Nothing that will require a discreet graphics card.

I originally bought a single 8GB stick of G.Skill Ripjaws Series memory

Now I'm thinking of buying another stick of the exact same RAM above, just to make sure I have enough.
But I see the G.SKILL Ripjaws Series is available in a 16GB (2 x 8G) kit.

I did some research but after reading the quote below I'm concerned that buying two separate sticks of the exact same model number may not be the same as buying a dual-channel memory kit shown above.

Should I return the single stick and buy the kit, or is there absolutely no difference between the kit and buying two sticks of the exact same ram separately?

Always always go for at least two matched sticks to ensure you're getting the dual channel mode of operation or you're looking at taking a pretty severe hit to overall system performance. Sure most people won't even notice it in day to day use but why waste the system's potential for performance that it's designed for in the first place. If you can't get 16GB as 2x8GB then trade the single 8GB stick for 2x4GB and go from there but always get at least 2 matched sticks for the system you're using.
There is no guarantee that two single DIMMs if the same SKU will use identical chips, even if the timings/voltages/speeds are identical.

That being said, if I were a betting man, I’d say they are the same more often than not and you’ll be ok.
Should I return the single stick and buy the kit, or is there absolutely no difference between the kit and buying two sticks of the exact same ram separately?
yes return/exchange it. for what your doing don't risk the chance of compatibility issues between sticks.
Well I just checked and I see it's too late to return the first stick even though I haven't opened it yet. I bought it from NewEgg's eBay store and used a coupon and some eBay Bucks that was about to expire but it's beyond the return window.

If I were to buy another identical 8GB stick, seems there's a good chance it'll be compatible. I mean don't a lot of folks add ram to their machines that way, by buying another stick of the same memory that's already installed?

If I did buy another, what do you recommend to test the two sticks for compatibility? Then again, for what I'm planning on using this for, maybe the best thing now is to just build it with the single 8GB stick I have and see if it's enough.
You'll want the dual channel for speed and 16gb for your usage. I would order from the same eBay listing and make sure everything thing matches, numbers, chips everything. Running stock should be ok. Use memtest to make sure they play nice, they should.
So I received the second stick and everything matches EXCEPT one number.
It's kind of hard to see but if you look in the yellow circled area, the number on the chip of one stick ends in 726 and the other ends in 723.
However, if you'll notice in the kit in the last pic, they both end 726. Not sure if that matters or not.



723 vs. 726 ... These sound like date codes, as in 23rd week for 2017 vs. 26th week of 2017.
Perhaps they are date codes of the chips. But if you look close, both sticks have "2017 Sep" printed below "2153" on the label.

I decided to send the one back and just go with a single 8GB stick. If I need more memory then I'll order a 16GB (2 x 8G) kit and just sell the single.
I just shoved 8x 16gb Corsair 3200mhx lpx into an x299 build and it gave me zero problems tonight.
Mixed pair of 64gb kits, sticks are all random serial #'s spread across 5x 64gb kits my buddy was using for a machine learning cluster.
v4.31 so it's Samsung.
Don't think it matters with an Intel build, and I couldn't be bothered reassembling sequential kits.

Don't worry much about it.

It's not like you are force feeding 5 petabytes of Apache header data thru MapReduce, then push the dataset into 20x 1070's.