[HowTo] SO-DIMMs Extreme: Boost 1600 to 2133 and beyond / JEDEC & XMP

When you say 8GB do you mean 2x4GB? If yes then its fine. You can find ready profiles for Corsair 1600 to work like 1866 and give us profiles from your stock RAM.

 

Well since it has 4 RAM slots, I was going to have both, 2x4gb Corsair and 2x4gb of the stock for a total of 4x4gb.
I assume I'd have to use a different profile for each pair?

 

Well, actually people used to flash Corsair profiles on Kingston HyperX RAM. You can try this out too. But who knows...

 

Thank you for your help! I'll just try them out when the laptop gets here.

 

has anyone achieved 2400mhz with samsung memory?

 

I tried reading most of this thread, but couldn't find out: has anyone attempted overclocking 8gb sodimms on new ivy bridge based platforms?

 

same here.

 

john maybe you will be the first?

to answer the above i don't think anyone has tried 2400mhz. might be possible in ivy mobo since it should have the 2400mhz multiplier for ram.

Otherwise me, John and another baseclocked to 2260mhz-2300+ the later being John's overclock.

Mine easily got to 2275, didn't try harder than that because I mostly reached a hard to use baseclock. it starts to get hard to keep 106.8+ baseclock to stay and not reset as you keep pushing it up higher. John got 108 but its same theory. you climb in increments hoping each time it sticks and doesn't freeze as you apply the next raise... in the end it can take a few tries to get it to stick. Great thing is you can set it and forget it once it sticks.
an annoyance is setting baseclock can cause you overclock mode to disable requiring your multis to be set againand also your ram settings can get reset. so sometimes i have to set baseclock then reset your cpu/ram.
I wish we could add the 2400mhz ram multi to our M18x R1 bios.

 

my laptop should be coming today ill see if i can try 2400mhz on my ivy 3920xm

 

RAM will Die on 2400. Or will be close to it.

 

sounds about right maybe ready a fire extinguisher

 

it will only die if you dont buy higher rated ram and the board supports it. so buying ram at 2133 should in theory be able to do 2400...

 

lol but when are they going to release that been waiting on Kinston CL9 -11-9-27 for a year or more. Well i guess with Ivy out there is finally a demand for it.

 

you mean...after we beta tested it and they seen it working. now they start really testing..

 

Ram will only die to temperature or voltage.

 

John here is a bunch of pics and comparisons you asked for.

Spoiler

2266 13 13 13 36


2266 13 13 13 36 winsat


2266 13 13 13 36 winrar


2266 13 13 13 36 on IGP


2266 13 13 13 36 maxmem better


2133 9 10 10 25 even better


2133 9 10 10 25 better winsat


2133 9 10 10 25 5ghz winsat


2133 9 10 10 25 5ghz better


2266 memtest


2266 bios

update here is 2266 12-12-12-36 @1.65v
I am going to memtest and report back and keep working down to see what kind of timings I can get.

Spoiler

 

thanks brother marc!

edit:
is that [email protected] in bios or in xmp?

 

I've managed 10-10-10-27-1t at 975mhz 1.5v

 

bios ie spd ext... my XMP1 2133 CL9 works via bios or XMP but I tried to load cl12 on XMP2 and at 1.65v and it wasn't working. this was before i got to change the THP. So I loaded XMP2 with 1.65v set on both volts in that profile to finsd it still wouldnt load... so for now I am about to test on plain SPD ext with mem in bios manually set to 1.65v so far not a prob at all... but strange I xan't get XMP going I think one of those volts in there cant be set to 1.65v I don't feel like tinkering more yet. gonna test in memtest and once I found my best I can do i'll try again to get the XMP to work.

Meaker awesome job congrats you did some awesome work. I think Nando should try those speeds on his samsung. he cant set volts or baseclock but he could try the 10-10-10-27 i suggested. Nice work Meaker.

 

i found that my xmp wasnt working right because i had it set to single channel when it should have been dual channel. now it's working...need to get around to dropping cl states. right now it's cl11 1.55V

 

Final tweak to sub timings, what do you think?

 

can someone tell me how the samsung 30nm sodimms compare to the corsair vengeance memory when overclocking?

 

Samsung 1600 CL11 are much more overclockable. Corsair is up to 1866. Samsung is up to 2200

 

sweet! thats the stock stuff im getting in my laptop and i would like to get it to 2133 at least

 

just the man i needed to ask....what is your take on the 8 gig corsair modules? frys and newegg has them. anyone test these yet...meaning...you in particular??

 

Clevo does force a downclock on 1600 8gb Corsair to 1333, once more than 2 modules are used...that maybe an indication of what to expect.
They will mostly do 1866, but maybe not go much higher.

 

are you able to overclock passed that 1333 limit when using more the 2 modules?

 

Is it possible to have 4x4GB @ 1866MHz? These are the sticks rated at 1866Mhz from Corsair.

And the BIOS dumps are here. I put them as a zip file because NBR doesn't accept the .thp file from Thaiphoon. I also included the dump from each stick, although to me they appear identical. They might be, but eh it's 1KB so who really cares anyways

And a question, when I put these in to test them, I found something interesting in CPU-Z:
These sticks came from Ebay. (listed new, the guy didn't know how to make his computer use the XMP profile to run at 1866MHz so he was getting slower sticks)

And these came from Newegg.

Again identical, but just this time to show that from each place the sticks appear the same. Now, in CPU-Z they are listed as JEDEC for 1866MHz. Is this an error, or indeed correct? The reason they're at 800MHz (lol) is because my computer is a piece of **** from HP that can only use memory up to 800MHz (despite the fact that they advertised and sold me a laptop with DDR3 1066 memory).

 

CPU-Z is pants at reading jedecs.

Try using (even the trail version) of thiaphoon burner. It's read will tell you all the programmed settings.

 

Well I don't want to shut my computer off again (busy atm) and swap the RAM, so I'll just believe Corsair's website that they're XMP.

 

does anyone know if the read times would be faster off 2x8gb modules over 4x4gb modules even know that ivy bridge is dual channel?

 

congrats Meaker your getting those very tight on timings. Congrats thats total win where you are at. In comparison your around my 2133 9-10-10-25 in how tight you got that. you latency number seems better even though bandwith is lower 50 latency is great.

Great keep it up John awesome. Thats at 2133 stock baseclock or are you over 2133? Nice though i've found 2133 CL12,11,10 take about 1.55v but if you dip into CL9 I needed 1.6v according to Alienware Preboot assesment not even memtest... metest had passed at 1.55v...

i'll have an update on 2266mhz DDR3 soon been going good so far hit walls already so i'll be on my tightest timings soon.

 

Is it possible to run 4x4GB sticks at 1866MHz? I don't want to overclock RAM past it's specified speed/timings, I only want to make it work as originally intended (but with all 16GB and not just 8, if possible).

 

1866mhz at 4sticks of 4gb a piece for 16gb total is easily possible and works great.

 

Is it possible to change the XMP profiles of Corsair Vengeance to JEDEC. I was told it is and directed to this thread. See my post on the previous page for info (lots of pictures)

 

Tyranids i have the same corsair that you do exact same pair. They are rated at 1866 10 10 10 27 at 1.5v. Cpuz isnt listing the other Jedec profiles. From the pic your just seeing 3. This is fine and normal. There is in fact a jedec in it for 1866 10 10 10 27. Im not sure of your hp laptop but when 1866 is set in our alienwares manually they boot to 1866 cl10 properly and run on 1.5v as sold. They are plug and play and by default they do not in fact have xmp profiles as is. You can add XMP profiles with a program like Thaiphoon Burner. There is a Corsair Vengeance thread and has a thp with two xmp profiles. One of us could make one for you. You want xmp to jedec (plug and play spd ext.) well what do you mean. Do you have a thp that has xmp and want one of the xmp speeds set as the main plug and play speed ext (jedec)?

 

I was under the impression that Corsair Vengeance RAM used XMP profiles. These sticks are for a Sager NP150 (Clevo P150EM) which doesn't support XMP. If these RAM modules really are JEDEC then I won't have issues, I just wanted to make sure before the Sager arrives that they're all set and ready to go at 1866MHz. My HP I tested them in can't do over 800MHz, which I knew. I just had to put them into something to get their dumps and info.

 

yup they should be great in your sager in that case. and yes they are how they should be. correct XMP is not part of those Corsair Vengeance unless some one mods them. They will plug and play to 1866 CL10 when the mobo is capable and or if you set the system to 1866 manually.

 

Oh perfect haha, newegg that I bought them from lists them as XMP ram so I was skeptical. That's awesome though, thanks a ton =D

 

Anytime I hope you enjoy your new incoming system with the 1866 ram. Should be nice and snappy.

 

So with this taiphoon burner program I can achieve 2133Mhz CL10 with my 4x4GB 1866Mhz CL9 Corsair Vengeance ram in my alienware m18x?

Is it totally stable?

Can I go even higher than 2133Mhz?

 

has anyone tried the Mushkin Enhanced Essentials 8gb modules? they run at 1.35V it makes me thing that they may be running 30nm tech but i cant find anything on them at all

 

It all depends. You need to test your ram individually. None of my corsair vengeance 1866 cl10 run at stable at 2133. Try to check in each post which ram sticks they are using... Model company etc etc.

The above results I have posted are on Samsung 1600mhz 11-11-11-28 chips at 4gb density and I am using them on Dual Channel at 8gb total. 16gb is possible for 1866 and 2133mhz... It depends on the handling of each ram stick as just one could be not a good overclocker and would then cause errors mixed with the rest atincreased speed. There are sticks operating at 1866 CL9,10 and up. There are ram sticks some of use programmed over their normal speed and some are handling 2133 CL13,12,11,10 and CL9. What kind of volts they need depends on testing individual sticks also. RiRi-FiFi got 10-10-10-24 on 1866 which is a very fast timing. My Samsung chips are handling more than a pair of Hynix, Kingston and Corsair that I have tested.

For me 2133 CL13-CL10 has only required 1.55v. My 2133 9-10-10-25 requires 1.6v. When memtest was used I passed at that 9 10 10 25 on 1.55v for 35 hours. By using my "pre boot assesment" under diagnosis heading of my boot menu on my Alienware I ran it and allowed the extended memtest to run too which revealed to make my chips fully stable at that I needed 1.6v. For that timing unlike 2133 CL13-10 I need my baseclock at 99.8 or 99.72 via bios whichever makes Windows show 99.6 instead of 99.7.

As for over 2133mhz me and a few others have ran 2133 mhz and raised our baseclocks. John got 108mhz+ * 2133 he got into 2300mhz+ on his ram. we/he hasn't said that's stable but it was interesting none the less. He stated that 108 baseclock may not be stable but perhaps 107 or 106 could be for him as he wan't sure without testing. So far I started testing again and have stabalized 106.33 * 2133 on my M18x R1 at so far, 13-13-13-36, 12-12-12-36, 11-11-11-36, 11-11-11-33, and currently running tests for 10-11-11-33. I'm running 1.65v though to ensure long term stability inside the ram. So far so good still working my timings down.

IGP is stable on 2133mhz even my tight timings of 9-10-10-25. I found as long as you use 1.6v or less you can boot and use IGP on even these speeds and timings. For a second I tried 2266 on IGP at 1.65v and causes a shutdown as soon as desktop attempts to load. But I did successfully play for a bit on 2266mhz when I had 1.6v set with no shutdowns. As mentioned long term may not have been fine at 2266mhz on the 1.6v but it allowed it to work with the IGP no probs.

 

*double post sorry twas ma phone that did it.

 

was not stable...would crash during gaming.

but if i ran 2133 or even 2200 it was stable. the 2300 was strictly for speed and to see how high the bclk and ram could go.

 

That's what I thought thanks for elaborating John. Good news if you end up testing more as above I am gaming stable at 2266 1.65v 10-11-11-33. Still working down it takes so long to be sure your stable, but you know how it goes John. When I get as far as I get I'll post my THP for you to see how yours does on my Cl9 2133 and wherever I get on the 2266 cheers friend here's to the coolest project ever lol.

I've been testing 7970m during day when I have time and leaving ram tests going overnight when I sleep.
Cheers this is a lot of fun.

 

You could always do it the other way round (imo a more useful method). What you do is you push it to a point where you know you'll get errors and crashes after a few minutes and then you work down. Once you've found a frequency that works for an hour or so you knock the frequency down a tiny bit more (probably use bclk in this case), or increase the voltage a little bit and then you know it's stable. That way you save hours of validation and you can be pretty damn sure you've got something that won't fail.

 

Well lol I am doing that my 2133 won't go tighter. 2267mhz is my limit at CL13. As follows 2133 at 106.54 failed. 2133 at 106.42 via bios is what I am actually on. Giving me what I am testing now 2267 to be exact. I can't use higher mhz. And I have no more than 1.65v to work with. I got here very quickly the way I'm testing. I use memtest and preboot assesment from alienware. it's been faster than it seems but you have to test for hours or you don't know it's stable. I've had them fail out into 10-12 test passes at nearly 12-15 hours in. My 2133 I did the 35 hours on my final testing when I was as far as I go to be absolutely sure only. I test on preboot assesment then memtest. Preboot will throw errors for me and or complete within 15min. Next I test in memtest etc and the next day I bench, stress and game to see how they handle.

Here is where I am doing what you suggested already. This is how I went up till now. First I set 13-13-13-36 the slowest CL timings my system boots (no 14t ? And up) and tested fastest mhz ie 2267 is where I ended up. Next I tested and passed error free 12-12-12-36. Then 11-11-11-36... Then 10-10-10-36 and hit first wall even losest 10's wouldn't pass without throwing errors. So I moved back up to 11-11-11-33 and that passed. Now I'm on 10-11-11-33 and the 10 is the tightest it can be ie one before CL9. It only took a few days to get to here and this way I can guarantee stability I'm not guessing on anything... I tried that on the beggining and the fastest accurate way has been what I described. You can see how I can go after that. I'll get the first 3 numbers as low as possible but only changing one at a time and confirming stability. Next the 4th and 5th number are tied together... You add the third CL number in nanosecs to the 4th number and that gives you the only value you can use for the 5th number. Now The 6th nimber is the only other one I adjust and I do that one last. So after 1,2,3,4 and dth timong is tightest as possible and mhz and volts are as far as you can I lower the 6th as far as possible in increments till it errors again and set it where it's stable. The tertiary/secondary timings like the whole right side column in Thaiphoon burner timing table editor/XMP editor leave those as is as the wil need to be what they are already set to. 2133,1866 and all other speeds will have a whole different right column. Any change I've done there always causes errors. So for my 2133 I started at CL13 and worked down for a stable 9-10-10-25-(30 something)-(30something)-88T (at 2133 that value defaults to 160T)

I think if anyone follows the methodolagy above they could find max of their ram without wondering if it's stable.

Just my opinion and I respect any other methods just the same. I just got into ram overclocking I have no previous experience. I'd say everyone in these ram threads has done outsatanding. Maybe some of you can benefit from this method... Maybe not Cheers though it's plenty fun.

 

It is interesting but I'm happy with 1866 9-10-10-27. I think the amount of hassle for the amount of performance gained is just not worth it. I dont think I'd even be able to tell the difference between 1866 and 2133 in any everyday apps

 

That's a great timing for 1866. I agree the hassle is high when trying to determine ranges. In my benches I've tried when the latency is low enough 2133+ is way faster than 1866. I'm sure your system likes 1866 CL9 and is working fine.
Interesting your timing is similar to mine. 1866 9-10-10-27 vs 2133 9-10-10-25... I think it's pretty cool you found those timings.

Update for my project: now working on 9-11-11-33 @ 2267 mhz just keeps getting lower

I'd say most of us shooting for 2133 are enthusiasts. I didn't post my findings to get everyone to do the same... Just figured my experience with my own fun could help others atleast I hope.

It's simple for any one else. Pick a already safe Jedec speed and set it and forget it.

I personally am working to get tightest timings I can because 2267 @ cl13 it is not faster than some 2133 and 1866

When I tested 2133 at CL13-CL9 the speed goes up by a very large amount. If a ram set can run your 1866 Cl9-10-10-27 but 2133 only runs on CL13 the obvious choice is 1866. For some of use we could run 2133 9-10-10-27 in which case 2133 is well worth it over 1866 etc.

I hope more keep finding speeds and timings that work good. I am seeing a lot of 1866 Cl9 variations and that's great... My Corsair Vengeance last I tested wouldn't even tighten one latency by one single adjustment without error... So it's nice to see this progress from everyone. Ri-Ri Fi-Fi did a nice job and so did many others.
Your testing has finished and others testing has just be begun. I'd have to say most people push boundaries because other people inspire and push each other... That's truly what makes the tech community so great.

Latency and ram speed gohand in hand. I noticed 2267 at CL11-11-11-36 had more bandwith than 2133 CL9-10-10-25 and matched in latency... Maxmem tested around 23gb on 2133 above and 52 latency. 2267 at the timing mentioned received 24gb at 52 latency. As I get it lower that latency will go even lower and bandwith will go up maybe a hair more.

If you get a chance download if you care to the winrar app and go to tools benchmark and run it for a minute and post pic... That will show what your cpu+ram can average in compression per second. 1866 should get atleast 6mb/s and when I checked my 2266 at the cl11 it was already nearing 10mb/s. This sounds like a little amount but it's a lot compared to other setups. A synthetic bench may not show everything in real world use but it does show a great example of how everyday tasks speed up the higher the speed and lower the latency. So you can imagine the differences in file transfers, bus communications and much more. Latencies for anything your system does would improve. Some people do a lot of work on their comps and what takes a pc a few mins to complete... time is shaved off that. If you have a long term project a faster system could save you maybe a few minutes in a day. Not much individually but out of our lives some sit and wait for everything to complete and when that equates to a few hours in a month it's definately a benefit. I'd rather a rendering project finish early if it can.

I respect everyone here and that includes what anyone chooses to do with their own system. I do hope though that if others want to try that they feel comfortable doing so and knowing what the benefits are if there are any.