DDR3 DRAM FAQs And Troubleshooting Guide

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.
> This is because it takes more testing for the manufacturers to match up modules for a four-module set.
Please, could you express that? At least with a link to an article maybe, or some words. I can't remember my previous builds, but the last time I bought 4 separate 4 GB modules instead of a 4x4 set because of cheaper price. I thought the overpricing was due to some marketing or something. Was I wrong?
 


Doing so is much cheaper but no guaranty they will work together,
proven / testify by me, I need to RMA 2x to get the all 4 works together,
luckily I'm long time consumer so they allowed it...

Even they are same spec, using same chip, "same" production number/lines..
The one who want to avoid all those headache should buy sets...

edit: individually no problems with each stick..
 

________________________

The manufacturers test the sticks that go into a package to ensure they all play together. For a single stick, if it works it's good to go. Finding two sticks that play nice is relatively easy, three sticks may take a bit more testing, 4, 6 or 8 sticks can take a fair amount of time. At times you can find the sets of more sticks at a better price, but usually say a 4 stick set will cost a bit more than a pair of two stick sets. We often see folks in the forums that pick up two sets of 2 sticks rather than buy a 4 stick set thinking they are getting over on the manufacturer only to find their two sets won't play nice, then might end up paying return postage, restocking fees, etc.
 

_______________

Beat me to the response, thanks 😉

 


Whoa, thanks for the info. My coworker has just picked documents referring to our latest office upgrade done by us: 5 PCs all having separately bought 4x4 sticks and no complaints from users. They are not dumb users at all, but what do you mean "headache"? Maybe there are problems that they think are not worth to report? Honestly I always thought problems with RAM are always sort of descreet: either the PC works smooth or it obviously glitches. Keeping in mind what you've told there gotta be something wrong with any of that machines.

UPD. Thank you for your answer, Tradesman1, just read it. And the article, too.
 
Headaches could be anything from a rig won't start at all, to occasional BSODs, running slow, occasional freezes, etc. Also mixing can work, a lot of things can play in, lower data rates i.e.1066, 1333, 1600 are more prone to play nice than higher data rate DRAM. Same with sets using looser timings at a given data rate (which you are likely to see in an office environment. Additionally getting all for a give rig at the same time has a better chance to play than those bought at separate times. However at the same time you can get identical packages at the same time, that simply won't play (I've gotten sets straight from manufacturers that won't play. A good Tech doing the installs may come across incompatibilities that are 'fixed' with minor voltage/timing adjustments (I help people with this about daily in the forums).
 
Usually if it don't match, either it don't boot, or runs at lower speed/latency than spec/advertised...
the headache is from trying figuring out which one/two/three that caused that and need to be replaced (RMA).. (they are fine individually)

In fact last time because of hurry and budget, I bought the cheapest brand (Hynix rebrand) I can get...
that also contribute why they don't work at first time setup..
(actually first RMA they work together but much higher CL, second RMA because the replacement don't work to together)
 
Problem for me is that I'm still running an old socket 1366 setup and triple-channel kits are barely sold now (at least at reasonable prices) so my only option really is picking a 4GB stick and setting quantity to 3 in the basket. Am I still likely to have problems if I'm not overclocking and just sticking to stock clocks?
 
Thank you again, guys, very much for the info. Now I feel like investigating this all, test timings, see logs, poll users. Never heard neither of BSs nor application crashes from them for years, the would've mention this. Hope it ends up lowering frequency or rising voltage/CL, not finding corrupted data or something breaking work process.
PS. Maaaybe buying far not the cheapest memory and some extreme luck played roles... I'm gonna get cheaper brands but sets if >2 in the future anyway.
 
Nice article!. It explains some of the concepts I have been confused on. It also explains why the memory timing numbers are getting "slower" when compared to what I remember from the DDR1-2 days.

regarding "virtual memory" -- take out mention of "virtual disk" as it will just confuse people.
(also referred to as the swap file, or virtual disk)
Click to expand...
instead try:
(also referred to as the swap file, or paging file)
Click to expand...

 

________________

That's one of the reasons we see such high CLs in DDR4, DDR3 is a forward progression from DDR2 and only planned to go up to 1600/11 then (in the original plans they would pick up DDR4 at 1866/12 (2133/13 etc), but people wanted faster DRAM and the manufacturers responded and tried to keep the CLs low which is why in DDR3 2133/9 is common, 2400 is normally CL10-11 etc. IN part the higher initial CLs are also coming in from the lower voltgae being employed, so I believe down the road DDR4 will see the timings tighten up as the market broadens.

And you're right on the virtual disk/page file/swap file/etc many use different names, same as with DRAM, some call it DIMMs, or Memory Modules, then there are the more common slang terms of sticks, blocs, bricks, etc
 
So today we 're all smart enough to understand what CPUs are, how they work, how they 're manufactured, basically we all have knowledge of a person with pHD in electrotechnics but none of us have enough knowledge to understand how DRAM works? XDD Misleading title incoming!
 


Sadly but it's true, Memory like an orphan child (no offense) often miss looked ...
 
Very good article but 1 stupid question. Shouldnt faster ram run on tighter timings on lower speeds? For example lets take two Kingston 16GB sets
HX324C11SRK2/16 DDR3-2400 CL11 has two profiles:
XMP Profile #1: DDR3-2400 CL11-13-14 @1.65V
XMP Profile #2: DDR3-2133 CL11-13-13 @1.6V
HX321C11SRK2/16 DDR3-2133 CL11 has two profiles:
XMP Profile #1: DDR3-2133 CL11-12-12 @1.6V
XMP Profile #2: DDR3-1600 CL9-9-9 @1.5V

I would expect that 2400 CL11 ram would be faster on 2133mhz then 2133 CL11 ram , but according to xmp it is actually slower/more lose timings. Is this only due to weird profile setup chosen by Kingston or my logic is at fault here telling me that faster ram should be running much better (ie more tight timings) on lower speeds?
Cause I was thinking to buy 2400 CL11 for AMD FX83xx/63xx and let it run on either 2133 or 1886 (depending what i can get from CPU) as I assumed it could run better then just "native" 1866 CL9

Maybe my logic is fine and you need to do manual timings to achieve it and not rely on XMP profiles?
 
On the Kingston set 2400/11 you show the 2400/11 is faster, you could however manually drop to 2133/10 IF you keep the voltage at 1.65 (they have dropped it to 1.6 per the XMP profile - either way the 2400 is still higher performance. With the 2400/11 sticks you can sort of count on running 2133/10, 1866/9 or 1600/8 if you set up manually.

Depending on the CPU you go with chances are you can run at 2400 with a 8370 and maybe 8350. I used my 8370 rig for the testing on some 32GB 2400 sets here:

http://www.tomshardware.com/reviews/dram-benchmark-fluctuations,4080.html
 
Thanks Tradesman. I've learned a lot just by following you in threads. I appreciate the education. I think I'll need to re-read this several times before I can start catching on enough to feel confident.
 
The links to each section on the first page of this article are broken. I reported this a long while back to the forum community managers and they were supposed to get with whoever on correcting this issue. These are still referenced articles so fixing them would be nice to see get done. Thanks.

@Jsimenhoff @SHaines
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom