Jump to content
HWBOT Community Forums
bolc

Set of timings for DDR3 Elpida Hyper

Recommended Posts

Hi Folks,

I am looking for a "good" tight set of subtimings for DDR3 Elpida Hypers, run on air. For some Dom GT and Kingston HyperX 2000C8.

The 666 18 1t and 777 21 1t, 888 24 1t and alike. I looked, I looked, but haven 't found a  screenshot with Asrock timing configurator :D

Thanks. Cheers

 

  • Like 1

Share this post


Link to post
Share on other sites

Different platforms sometimes seem to have different subtimings in bios. Or they're renamed or linked to other timings.

775 timings look a lot different to 1150 is what I was trying to say.

Edited by Noxinite

Share this post


Link to post
Share on other sites

If my target is 2000 mhz i start with 1600 on auto and lock all of those before going to 2000.

Like this i am sure i have a good start . Rtl’s always stay on Auto.Once you reched you’re desired freq and want to make finetuning ,you can take the subtimings one by one and put them on auto,see if after trainig the value is going lower than what it was set before.

I hope it makes sense to you.

  • Like 1

Share this post


Link to post
Share on other sites
10 hours ago, suzuki said:

These are from that run with 990x.

Dimm voltage was 1.90v .

E43B8BB4-F9B4-4DCF-8153-5F24B69B743C.jpeg

thanks I will try to translate that into ivy/haswell lanuages :D

but it goes fine I think
tddWrTWr = tWRWR dd  in haswell terms
tsrWrTWr = TWRWR I guess
tdrRdTWr = TWRRD dr
etc

overall i have much lower values, except for first three 8 8 14 which I have at 8 8 8 , and derived form the auto. the 3 values i haven t lowered yet.  
would you share your secondaries, I am interested in tWR in fact. currently at 8. it does1800 mhz with at the moment :
6 6 5 17 1t 3 56 8 3 20 3 5 5 for 2ndaries, as listed in asrock configurator order

spi32m did 18 loops out of 24, not far from the stability  :D bump at 1.78 V

Edited by bolc

Share this post


Link to post
Share on other sites
10 hours ago, Noxinite said:

Different platforms sometimes seem to have different subtimings in bios. Or they're renamed or linked to other timings.

775 timings look a lot different to 1150 is what I was trying to say.

yep I got you, this is clearly different form the g3258 / m7 series language i am used to :D

I tried some 1800  6 6 5 17 1t trfc=56 with tight subtimings.
I know some tWRRD dr dd and tWRWR dr dd need to be = 5 or it won t post.

I will post some screenshots and if you have advices guys, let me know. the next question will be: is it more efficient to have these low values or should I increase some back ...

max safe voltage on hypers is 1.85 ...? (if I want to keep those sticks :D) ; on psc i can bench up to 1.95 but i rarely use more than 1.85-1.88 in life given i rarely use 2600+ cas8...

 

Edited by bolc

Share this post


Link to post
Share on other sites

is TRDWR supposed to be the sum of TRDRD and TWRWR ?
and same for their respective dr and dd ?

and TWRRD = TWRWR + TRDWR ?

Edited by bolc

Share this post


Link to post
Share on other sites

You can probably run: 6-6-5-18-1T 4-48-10-4-26-4-4-6 4-5-4-15-4-4-4-4-4-8-8-8.

I tried to find my 2000MHz 6-6-4 32M run (with hypers on phase), but couldn't find it. :(

Edited by Noxinite

Share this post


Link to post
Share on other sites

looks like i stumbled on a fairly good one...
IO-L rtl was set to 4, but now it stucks to 5 and if I force 4,. it will not post... any ideas what I am missing? but it does even bette at spi32m with 5...

Edited by bolc

Share this post


Link to post
Share on other sites
23 minutes ago, Noxinite said:

You can probably run: 6-6-5-18-1T 4-48-5-4-26-4-4-6 4-5-4-15-4-4-4-4-4-8-8-7.

I tried to find my 2000MHz 6-6-4 32M run (with hypers on phase), but couldn't find it. :(

thks i will give it a try but it does not like me putting 4s instead of 5s in the tertiaries

nope it won t do the listing a decent voltage, or at all. :)

trfc52 goes fine though. trrc48 boots but will need more voltage, at 1.82 the spi32m time goes longer.


i need to check the impact of faw and tWRRD on performance, then trfc and tras, and i will be done for now :D

and next, 776 20 ishish so i can redo my 2500k spi32m score :D
 

Edited by bolc

Share this post


Link to post
Share on other sites
1 hour ago, Noxinite said:

Maybe I mis-remembered. :/

thanks, this is helpful anyways ;)

so far so good :D

 

 

tRDWR_dd at 7 is a no go.

i have increased FAW and TWRRD to check what id does ; except the initial value calc, it makes it slower overall

1800C6_#1-trfc52-iol4.PNG

Edited by bolc

Share this post


Link to post
Share on other sites

Talking about hypers - what voltages do you usually run/what is the safe limit at ambient? I usually tend to chicken out around 1.84V with my kit (which appears to barely manage 2040 7-7-7 at that voltage), but I heard a couple people are running their kits at >2V for benching? And one last question - is it worth running them on Dice or even LN2? 

Edited by ground1556

Share this post


Link to post
Share on other sites

If your Hypers really need 1.84V for 1020MHz 7-7-7 then they are a truly tragic kit that doesn't deserve to live :D Most of these are in the 1.625-1.675V range for 1000 CL7 with linear scaling up to 1100-1150MHz.

I've been using the same pair of GTX2 for all my benching and binning for the last 7 years. Doing countless powerdowns, running at 1.90V+ loaded for a multiple hours every month,testing up to 2.10V if necessary. They're still alive.

Edited by TaPaKaH
  • Confused 1
  • Sad 1

Share this post


Link to post
Share on other sites
39 minutes ago, TaPaKaH said:

If your Hypers really need 1.84V for 1020MHz 7-7-7 then they are a truly tragic kit that doesn't deserve to live :D Most of these are in the 1.625-1.675V range for 1000 CL7 with linear scaling up to 1100-1150MHz.

I've heard others say the exact same thing about my kit :( But at least on my x58a-oc and Rampage II Extreme I could get none of the sticks to finish 32m above 1020 7-7-7 at 1.84V. Might as well push them to 2 V when required then... Only positive thing about this kit is that its not weak in tRCD, no difference in required voltage between 7-8-7 and 7-7-7...

Share this post


Link to post
Share on other sites

You might be limited by the platform, as in, mainboard setting some subtiming way to tight, hence necessity for 1.84V. Usually even Kingston Hypers are not so bad.

On tRCD - 1020 hardly tells you anything. tRCD 6 would be extremely impressive (only Hicookie's GTX2 were capable of such a feat), tRCD 7 is standard on 90-95% of sticks. If tRCD 8 is necessary at 1020 then kit is garbage.

Share this post


Link to post
Share on other sites
20 minutes ago, TaPaKaH said:

You might be limited by the platform, as in, mainboard setting some subtiming way to tight, hence necessity for 1.84V. Usually even Kingston Hypers are not so bad.

Hm, I just sorta killed my last non-1366 DDR3 board last weekend(lets say it refuses working on cold completely and on ambient it can only boot a single stick of ram at really loose timings), so that testing will have to wait for a while now. Gonna try finding a second kit to compare against, but others appear to get far better results with the 2000 C8 Dom GT kits on x58 from what I've seen. Will probably just see how far they go at 2.05V and if they don't last I'll find another kit I guess...

Whats the typical tRCD limits I can test for?

 

Sorry for derailing the OP a bit with my own kit ;)

Share this post


Link to post
Share on other sites

@Ground1556 - there is a big difference between binned kits of hyper and common ones like dom 2000 c8.

For example at 2200 7-7-6-60-1T The Dom C8 made spi32m at 1.88v while my best stick of DomC7 needed only 1.70.

If you are planning to stay a while on 1366/775/1156 than you should find a binned kit and forget about the rest.

Share this post


Link to post
Share on other sites

ZeltL81.png

okay, at slightly less comfortable voltage I can at least manage decent timings at good clocks. Probably gonna use those if possible... 

2000 6-7-6/6-6-6/6-7-5/6-6-5 refuses to work even at 2.0V, so it remains pretty terrible...

(yes efficiency is absolute garbage, I know, thats just my test OS which is filled with all kinds of crap nor an optimised 32m run)

Edited by ground1556

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×