some ddr3 binning

Hiya folks!

Finally getting around to some ddr3 binning (and learning more about ram in the process). The goal is to figure out the best 2-4 sticks for each platform I own and get rid of the rest. Eventually plan to take them reasonably cold (-45C or -100C) but, at minimum, will know the best air sticks for benching.

Starting with socket 775. Here's what I'm looking at:

Micron D9GTR/S (bin at 900C7):
- CellShock 1600C7-7-7
- Crucial Ballistix 2x1 1600C8-8-8
- Corsair Dominator 2x1 1800C7-7-7

Hypers (bin at 1000C7):
- Corsair Dominator 3x2GB 2000C7-8-7
- Kingston HyperX 3x2GB 2000C8
- Kingston HyperX 2x2GB 2133C8
- SuperTalent 2x2GB 2133C8

Micron D9GTR/S.
Lowest possible voltage to run 900MHz at 7-6-6-18 1N (with tight subs) for each single stick. Testing is 1 pass of super pi 32. Aiming for 2.0v or less.

Cucial Ballistix 1600 8-8-8-24, 1.65v (default)
Stick #1: 1.988v
#2: 2.1v
#3: 2.2v
#4: 1.965v
#5: 2.0v
#6: dead

Cell Shock 1600 7-7-7-21, 1.65v (default)
#1: 2.0v
#2: 2.13v

Corsair Dominator 1800 7-7-7-20, 2.0v (default)
#1: 2.15v
#2: 2.13v

What do you think about the voltages? 7-6-6 at 900 is pretty good I think (is it?). Figure keep the lower voltage ballistix and the dominators for the next round of tbd testing. Think the corsair's might be worth keeping since the delta from stock (2.0v) isn't too shabby?

Next up are the Elpida Hypers. Pretty sure they should do much better for 775. Haven't found good, tight binning settings yet so could use thoughts there.
Update: Going to use 7-7-7-20 with tight subs. Aiming for 1.80v or less.

As always...appreciate any thoughts, recommended binning settings (for pass 2 d9 and hypers), comments, etc.

Thanks!

Something to think about.
Currently running PSC (G,Skill Eco's) @ 6-7-6-20 1T @ 900-930 ish @ 1.7v on 775.

Thanks for the info Mr.Scott!

I'm definitely no pro but that seems like some outstanding ('golden'?) psc. Correct or would that be considered just 'above average'?

Hmmm....now I'm wondering....should I:
- stick with my current plan of figuring out which sticks need the lowest voltage for 1000@7-7-7-20 then focus on those sticks and tighten or,
- go for 900@6-7-6-20 then work my way up speedwise on the ones that needed the lowest voltage for pass 1?

Based on a recommendation to quickly id the strongest hyper sticks...tried a few at 1000 6-7-5-20 1T with 1.8v...none of them passed

funsoul said:

Thanks for the info Mr.Scott!

I'm definitely no pro but that seems like some outstanding ('golden'?) psc. Correct or would that be considered just 'above average'?

Hmmm....now I'm wondering....should I:
- stick with my current plan of figuring out which sticks need the lowest voltage for 1000@7-7-7-20 then focus on those sticks and tighten or,
- go for 900@6-7-6-20 then work my way up speedwise on the ones that needed the lowest voltage for pass 1?

Based on a recommendation to quickly id the strongest psc sticks...tried a few at 1000 6-7-5-20 1T with 1.8v...none of them passed

Click to expand...

I don't consider it outstanding. I have a couple sets of Flares that do slightly better, but all in all I think they're just average. None of my PSC sets will do much more than 1000 at reasonably tight settings.
Elpida Hypers running fast seem to produce better 2D results.

Most above average PSC will make ~900MHz 6-7-6. Most in good memory kits were rated at 800MHz 6-8-6 1.60-1.65V.
It's like:
1600 6-7-6 , 6-7-5
1800/1866 6-7-6, 6-8-6
2000 6-9-6, 7-9-7
On air they are usually scalling up to 1.85V. Many PSC don't like 1.9V+ without better cooling ( some don't like it at all ).

You can probably push these microns up to 6-6-6 1800 but at high voltage like 2.2V. Similar results were on Qimonda chips - my OCZ 3x1GB 1600 7-7-7 kit could make 1800-1900 7-6-6 ~2.0V. Elpida looks best for older platforms like 775. BBSE and Hypers can make 6-6-6/7-7-7 1800-2000 and require lower voltage than microns.

Thanks Woomack!

So....for this first pass to figure out the best hyper sticks to focus additional effort...I should be fine trying to find the ones that can do 1000 7-7-7 at 1.80v or less or should I be trying something different?

I couldn't get any of the microns to pass with c6

OK...now...based on a recent post...remembered that the 400 strap on x48 is broken. That's the strap I used for the micron testing...does this mean I should re-test all those sticks using the 200 or 266 strap instead OR, since I was at least consistent, am fine keeping the 2 ballistix and 2 dominator sticks as the best of the micron bunch?

I would sell all those Microns and keep all Elpida ( in case if you kill any kit ). Should be good for older platforms like 775, 1155 ( mainly SB ), 1156 and 1366.

So...how much you think the micron sticks are worth these days? Also...on the hypers...to quickly identify the stronger from the weaker...can I just find which sticks need the least voltage for 1000 7-7-7 (or maybe 6-7-6?)?

Struggled for a few hours yesterday :/ Was unable to get 1000 7-7-7-20 working with the Kingston HyperX 2000C8 (tried only 1 stick) nor the SuperTalent 2133C8 (tried both sticks) at any voltages from 1.65 to 1.85v.

Right before shutting down for the night tried 1 stick of the Kingston HyperX 2133C8 at 1.75v and got spi32 to pass. Hoping that I'll be able to tighten it/them some more (and hopefully pass with lower voltage).

If anyone has time or inclination...could you review all my timings and sub-timings and make suggestions on what settings to use before further attempts? Again...trying to quickly separate the 'worth further testing/tweaking sticks' vs. the 'go ahead and sell those sticks'.

Tia for your thoughts!

Well....either my hypers kinda stink or I've got something wrong in my subtimings. So far...have only gotten 2 sticks to pass at 1000 7-7-7-20. One kingston 2133c8 made it with 1.68v, the other stick passes with 1.77v. All the other sticks so far are unable to pass with voltages tested from 1.65v-1.85v (most of the sticks throw a bsod at or above 1.8v). Am I doing something wrong? Any suggestions (incl. subtimings)?

Found this elsewhere....

Sam OCX said:

My rule of thumb is test 1000 7-7-6 at 1.65V first. If it passes, try to minimise voltage (best kits will need 1.55-1.56) and then to proceed to other stuff, such as 900 6-6-5 or 1100 7-x-x. If it doesn't, the kit is not likely worth the attention.

Click to expand...

Man...he clearly gets/has MUCH better sticks I think. I don't seem to have anything that'll do those timings at anywhere near those volts (again, unless I've got something set wrong).


Ideas? Thoughts?

Remember that this memory can overclock better on newer platforms. Usually on X58 or P67 tighter timings at 2000 were easier to set at lower voltage.

You can set tRFC lower ( 88 or something near ) but I don't really remember settings for Hypers. My only Hyper based kit is 2000 8-8-8 HyperX and it's not working at 7-7-7 on any newer platform but runs up to 2300 at 8-8-8 and higher voltage.

Thanks Woomack!

OK...wrapping up the hypers/775 and going to set up for 1150 and psc testing. Any thoughts on what I should using for timings/subtimings for binning? I've been told 2666 8-12-8-28 under 1.85 but whatcha think? Seem right? Subtimings?

Tia for all the patience and thoughts everyone!

OK...finished up the initial binning of the 775/hypers (1000 7-7-7-20). Tested from 1.65v - 1.84v.


kingston 2000c8
stick 1: fail...would not pass with up to 1.77v, throws bsods at 1.78v or higher
2: 1.73v
3: 1.72v

super talent 2133c8
1: fail...1.8v or higher throws immediate bsod
2: fail

kingston 2133c8
1: 1.77v
2: 1.68v....this one looks decent?

Guess I should see how much tighter I can get the 4 passing kingstons and sell the rest?

I don't think you will need so many hypers, especially that PSC can be not much worse and work much better on every newer platform than 775. Keep best 4 sticks and sell all others. Probably won't be worth much nowadays but I'm not sure what are the prices in US.

Re PSC, on 1150 keep sub timings at auto and try main timings + voltage up to 1.9V. Start to play with sub timings when you stuck at weirdly low clock and something will clearly hold you down.
Depends from modules, my PSC liked voltage between 1.82-1.95V on ambient cooling. Better kits could make 2600-2666 8-12-8 or 9-12-9 1.85-1.95V. Worse couldn't pass tests @2600+ CL9.
Z77 is better for PSC overclocking while on Z87/Z97 boards more depends from IMC if you can make tighter timings or not. For instance 2400 7-11-7 is already good result while on haswell CL7 probably won't work and you have to set 8-11-7.

On PSC you try to get as low tRCD as you can as it helps more than CL and is much harder to lower than CL. So 2600-2666 8-12-8 is good starting point and 8-11-8 would be great but nearly impossible on most, even good PSC. You can also try to push them up to 2800 and check if 9-12-9 will work. As I already said low CL will depend from IMC so if you can't set CL7 above 2400 then it doesn't mean that memory is bad.

At the beginning I would try to make something like 2400 7-10-7/7-11-7 and 2600/2666 8-11-8/8-12-8. On SB I would try 2133 6-10-6/7-10-7. Most above average PSC will make 2000 6-9-6/7-9-7 so won't be much worse than hypers for older platforms.

Wow...thanks for the massive info Woomack!

Finished up the hyper testing (for now). Have 4 kingston sticks that'll do 7-7-6 1T (other timings in the pic below)...2 2000c8 and 1 2133c8 take 1.72v, the other 2133c8 needs 1.69v.

Hmmm...should I go ahead and use 1155 for psc testing? Also still have samsung and bbse sticks to test, too.

Woomack said:

I don't think you will need so many hypers, especially that PSC can be not much worse and work much better on every newer platform than 775. Keep best 4 sticks and sell all others. Probably won't be worth much nowadays but I'm not sure what are the prices in US.

Re PSC, on 1150 keep sub timings at auto and try main timings + voltage up to 1.9V. Start to play with sub timings when you stuck at weirdly low clock and something will clearly hold you down.
Depends from modules, my PSC liked voltage between 1.82-1.95V on ambient cooling. Better kits could make 2600-2666 8-12-8 or 9-12-9 1.85-1.95V. Worse couldn't pass tests @2600+ CL9.
Z77 is better for PSC overclocking while on Z87/Z97 boards more depends from IMC if you can make tighter timings or not. For instance 2400 7-11-7 is already good result while on haswell CL7 probably won't work and you have to set 8-11-7.

On PSC you try to get as low tRCD as you can as it helps more than CL and is much harder to lower than CL. So 2600-2666 8-12-8 is good starting point and 8-11-8 would be great but nearly impossible on most, even good PSC. You can also try to push them up to 2800 and check if 9-12-9 will work. As I already said low CL will depend from IMC so if you can't set CL7 above 2400 then it doesn't mean that memory is bad.

At the beginning I would try to make something like 2400 7-10-7/7-11-7 and 2600/2666 8-11-8/8-12-8. On SB I would try 2133 6-10-6/7-10-7. Most above average PSC will make 2000 6-9-6/7-9-7 so won't be much worse than hypers for older platforms.

Click to expand...

Based upon this info I don't know if I have good or bad PSC sticks.

I couldn't make anywhere near 2000+ (1000MHz+) the last time I tried.

Started at 800-850MHz 6-8-6-24 1T 1.55v (just in case it needed the additional voltage). Ended up at 900MHz 8-9-8-22 1T (forgot to write down the voltage, somewhere between 1.55v and 1.65v I think), and 950MHz 10-10-10-24 1T (voltage somewhere between 1.65v and 1.75v). I run 800MHz 6-8-6-20 1T at 1.5v most of the time.

I never went past 1.77v though. I was testing on LGA1155 at the time.

I see PSC 2x2GB kits selling for $20-35 on ebay, when I see them. They don't seem to be worth much these days, good for buyers, not so good for sellers though.

For me Z77 was working better than Z87/97 but had limited max clock. Anyway I could easier set CL6/7 than on haswell. I don't know if it's worth to change platform for tests if you already set everything on haswell.

Re PSC, most PSC were in kits like 1600 6-8-6/7-8-7 1.55-1.65V and 2000 6-9-6, 7-9-7, 8-9-8, 9-9-9 1.55-1.65V. If I remember good, PSC on X58 could make 2000 6-9-6 or 7-9-7 and that was what all were looking for at reasonable voltages ( count up to 1.75V ).
Better PSC from late batches were in kits like 1600 6-8-6, 2133 7-10-7/7-10-10 and 2400 8-11-8/9-11-9. Considering that fact, if you are looking for better PSC then it should make at least something like that.

I have couple of PSC kits:
- 2 kits of 2x2GB Dominators GT 2000 C8 - can make 2000 6-9-6, 2133 6-10-6, 2400 7-11-7, 2600 8-12-8 but they don't like higher voltages, max on air ~1.83V
- G.Skill RipjawsX 2x2GB 2133 9-10-9 1.50V - can make about the same at 2000-2400 but I could bench it up to 2666 CL8-12-8 or 2800 9-12-10 and max 3100+ for valid on AMD or Z77
- ADATA 3x 2000 9-9-9 1.55-1.65V - can make 2000 7-9-7 at about 1.65V, 2133 7-11-7, 2400 8-12-8, 2600 9-12-10, 2666+ was crashing when I was testing it last time <-- this is worse PSC

Trying to set up for psc (and other) ram testing on a msi z97 xpower ac with a g3258. Been a while since I played with haswell so could use a little guidance on getting things set up properly.

Have tried quick tests with a single geil evo2, 1600 pi and 2300 pi all with the same (or very similar) lousy result. Boot the system and go into bios. Have set vccsa to 1.25v and left everything else on auto. For the evo2 and 2300pi, I selected the psc 2x2 2666 1.90v mode2 profile. For the 1600pi I used the 2x2 psc 2600 1.90v profile. All sticks booted to windows but spi32 fails within moments of starting.

Below are my voltage and ram timings. Any thoughts, guidance, etc would be much appreciated. Hope to start testing tonight but think I must be missing something :/

Tia!!!

Finally got 1 of the 2300 pi sticks to pass at 2600 using the 2x2 psc 2600 1.9v preset. Definitely struggling...know I'm missing something, have an incorrect/insufficient setting(s) or something

Could also use suggestions on timings/subtimings for quick binning purposes.

Ideas?

That's some good progress. Now start minimizing voltage and tightening timings. For tight air PSC you'll need tWCL6, tRFC max 100, preferably lower.
Try to pass a run while changing to 2 values above. If passes, try lowering voltage. GL!