SOLVED My G.Skill TridentX came in today! Questions already!

[MongGrel]

Yeah, but mine should be 7-7-7-21 in that case, I'm not doing it right I guess.

Still need to mess with a few things on these more they pretty new, and have still pondered trying others.

Might try underclocking em and tighting up the timings again for grins and giggles, bought em as they seemed pretty good for the L5639 when I stuck it in here.

Working pretty nicely in general actually.

But have bought enough things lately, he he.

 

[Woomack]

Don't look at CPU-Z readings. From some time they're often wrong ( especially SPD/XMP table ). This memory is working fine at XMP profiles on 2 boards that I was testing and is overclocking pretty good.
I was testing 16GB TridentX 1866 couple of days ago -> http://www.overclockers.com/forums/showthread.php?p=7630930 , for every setting there are also timings in table and I added info about voltage so it should be easy to set something more.

 

[EarthDog]

Is there any merit to "adding up" the three 'front end timings' to reach the last value?

I know on my sticks (11-13-13-31) they sure as hell do not. Is that specific to an IC or just some interwebs nonsense that happened to be a mere coincidence that then propogated out to the public?

 

[Ajay57]

Sure got you scratching your head about adding things up ED!!! There is some merit in a basic way to begin with, but like you and others we all want to run tighter ram timings.

So in that context its never going work, just take it as a basic guide for beginners!

Not for us WAR HORSES who overclock everything in our set ups.

AJ.

 

[keny]

Hypers do, I know that, but as for the rest I'm not sure, I think someone has just noticed it on a few sets and said "hey they add up" and so the rumour begins!!!

That's my guess anyway

 

[Woomack]

For new kits it's generally like tRAS = ( tCL+tRCD+tRP ) -2 but for higher clocked kits, manufacturers are leaving it lower so it's often like tRAS = (tCL+tRCD+tRP ) -4 or more.
Samsung can work at lower tRAS while Hynix usually needs higher value but manufacturers make the same base settings for all IC so they could switch used chips without changing all product line. When you look at available memory then you find out that most 1866 kits have tRAS = 27, 2133 have tRAS = 28 and 2400 have tRAS = 31 while all these kits are on mixed Samsung, Hynix or for lower clocks also Micron or Nanya IC.

There is some lower value at which IC is still stable and as you can see in overclocking results, many samsung IC are going to work at tRAS = 21-28 for 2133-2800 clocks while Hynix needs 27-35 for the same clocks. There is no special limit for upper value so better is to set higher for all kits while it won't really affect performance but will be for sure stable.

I'm not sure on what base were made all these "timing rules" that are around the web but adding first 3 timings to get 4th is working almost only for older IC and mainly Elpida.

 

[EarthDog]

I'm not sure on what base were made all these "timing rules" that are around the web but adding first 3 timings to get 4th is working almost only for older IC and mainly Elpida.

I see where its coming from here, LOL!

So in that context its never going work, just take it as a basic guide for beginners!

Its not good advice at all it seems. I heard that, years ago, and knew, from the several kits I had, that doesn't work. Did it even work on your kit?




Thanks keny (EDIT: and woomack)... just trying to make sure facts reign supreme...and not blanket statements that do not apply to all. The sticks I have are the Kingston Hyper X Predators 2666Mhz 11-13-13-31. So those look like -8, LOL!

 

[Woomack]

I think it's changing for some kits as I was sure that 2666+ had tRAS 35 so I checked specs and there is no official info how high it is, only 11-13-13:
http://www.kingston.com/datasheets/KHX26C11T2K2_8X.pdf
2800 12-14-14 kit has tRAS 32 but Avexir 2666 based on the same IC has 35 ( the same as Geil and Team Group 2666 C11 ).

 

[Desync]

Thanks everyone for all the help. Learned a lot about memory this go around.

Also, after I set the speed in bios to 1866 and set the timings to 8-9-9-24 and the volts to 1.60 my CPU was able to be OCd again to 4.0ghz and is stable in P95.

 

[Ajay57]

Actually it did until as you already know i have changed every setting since, but if you do not like, what i have posted. For the good of the site i am only willing to DELETE IT!

AJ.

 

[EarthDog]

I took exception to it because I thought it wasn't true for all ICs. Me liking something or not has nothing to do with what is posted at the site. I am a proponent of facts... and what was posted appears to have been true with certain IC's at one time in life and perhaps a few now, but across all sticks. So its a terrible 'general rule' it seems as Woomack has essentially confirmed. It wouldn't work for me for example... on any of the past kits I have had.

(Actually) "Knowing (and understanding what one THINKS they know) is half the battle".... "GO JOE!"


Quote (sans parenthesis) is from GI Joe cartoon I watched as a kid in the 80s..

 

[Ajay57]

Fine ED then someone should go onto YouTube and tell them to stop posting video guides telling you that it works with Z77 platforms and others in general.

But i do understand that the factual side has to be correct, but this has spread over the Web and in some instance is common knowledge!

Regards,

AJ.

 

[EarthDog]

I have no control over youtube, LOL! I post at these forums. When I see things that are out of line, as any forum goer does/should do, they are questioned, hashed out, and corrected if needed.

 

[Desync]

 

[EarthDog]

I guess I am out of thanks for the day...

PRICELESS! "Bon jur"

 

[Ajay57]

Congrats on your up grade and promotion ED!!

AJ.