DDR3 IC thread

[Gunslinger]

Anyone?

[TaPaKaH]

Corsair ver8.1x is Nanya 2Gbit, never seen any of these do more than DDR3-2133.

[Reefa_Madness]

I've always found it odd for Corsair to have used Nanya chips in their Dominator 1866 and 2000 kits. Why use "run-of-the-mill" ICs in what is supposed to be your top line? Cost savings can't possibly justify loss of face/reputation could it?

I can understand them being used in some of the other (lower) product lines, but on their high-end it never has made sense to me.

[filmbot]

I've always found it odd for Corsair to have used Nanya chips in their Dominator 1866 and 2000 kits. Why use "run-of-the-mill" ICs in what is supposed to be your top line? Cost savings can't possibly justify loss of face/reputation could it?

I can understand them being used in some of the other (lower) product lines, but on their high-end it never has made sense to me.

Unfortunately they're always going to say if the IC will do the "high end" specs they will be sold at, they are fine to use.

[Reefa_Madness]

Unfortunately they're always going to say if the IC will do the "high end" specs they will be sold at, they are fine to use.

While I understand that is their position, I've always thought that it was counter productive. The upper end of their high-end stuff is good, but I've always felt that they taint the name/reputation of their high-end products by using these types of IC at the lower end of those product lines. Then again, they probably figure that those that buy the lower binned products of their high end line know very little about what ICs are better and just buy the "name".

Personally, when asked, I strongly suggest to people to stay away from the 1866 / 2133 Doms (except for the 2133s ver4.13 kit with 9-11-10-27 timings). Generally very little headroom on those kits.

[Dr. Vodka]

So, I have these lowly (boring?) kingston KVR1333D3N9/4G, double sided 4GB 1333 9-9-9-24, made in mid 2011 in my P67/2500k rig. I've been able to get them up to 1866 10-10-10-27 1.5v for 24/7 use. Through some bclk oc I've seen these get to 2000 at similar timings and 1.65v (too bad I don't have z77 and an ivy cpu to use that divider) but for the life of me I can't get these to even boot at 2133. Tried relaxing timings, tried more voltage up to 1.7v and not a sign of life at those speeds.



Any idea what these ICs are? Maybe if they can be identified I could get a hint on what to do. I've been researching them yet there isn't much out there. Am I expecting too much out of them, or out of my 2500k? Thanks.

[TaPaKaH]

It looks like you have Elpida 2Gbit BCBG, so your results are normal. I wouldn't expect these to even do 2000.

[Hyperhorn]

Any info out there about Micron D9RLT, e. g. regarding capacity? (Upper line says 3RAI8 and on the side I can read QLF3.) I've seen it as SSD cache, but not on RAM sticks yet.

[TaPaKaH]

I made a slight update to the first post featuring more obscure ICs.

[Don_Dan]

Thanks for updating the thread, Sam!

1Gbit revision D (K4B1G0846D)
OEM code: M378B5673DZ1 for 2GB modules

To be found on:
Some 1.8V+ rated kits with specs of 1800 9-9-9, 1866 9-9-9 or 2000 9-9-9 and 1.65V rated 2000 9-9-9 kits from late 2008 - early 2009.
Kingston also used these on 1800 8-8-8 HyperX sets manufactured in late 2008 - early 2009.
Corsair ver4.1

Benching relevance:
None.

How does it overclock:
Does not like voltage above 1.8V. At 1.8V decent kits can do 1000MHz 8-9-8 and 1100MHz 9-9-9.

Not that it matters much, but I don't agree about the overclocking part completely.
In my opinion HCF0 scale up to 2+V and they can do 900MHz 7-7-7 and 1000MHz 8-8-8, 1100MHz 9-9-9 already required 2.1V in my testing, but anything over 1000MHz isn't that reliable on X48 anyway. On modern systems the required Vdimm is probably lower. By the way, which system do you use as reference for the given binning parameters?

I know you had some interesting results with HCF0 on OCZ FlexII, like 1000MHz 9-9-9-24 @ 1.52V and 1000 MHz 8-8-8-24 @ 1.76V on X48, but that's so much better than anything I've seen on my two kits ( Corsair 2000C9D ). Anything I missed or just a superb kit?

[TaPaKaH]

I was filling information on old ICs mostly from personal experience and back in X48/X58 days I wasn't much of a RAM expert. I had many kits with 1Gbit-D but not a single one, even the FlexII above would do 900MHz CL7 at any voltage, though I had multiple kits up to and above 1000MHz CL8/CL9 on X58 under 1.8V, hence why I thought these do not scale too well.
If you say so, I'm going to up it to ~2V for now. May be I will try to buy a kit in the future to get a better impression on the ICs

[TaPaKaH]

Recently bought a bunch of Samsung 4Gbit revision D OEM modules just to see what the ICs are like. The best pair can do this:

[Don_Dan]

I was filling information on old ICs mostly from personal experience and back in X48/X58 days I wasn't much of a RAM expert. I had many kits with 1Gbit-D but not a single one, even the FlexII above would do 900MHz CL7 at any voltage, though I had multiple kits up to and above 1000MHz CL8/CL9 on X58 under 1.8V, hence why I thought these do not scale too well.
If you say so, I'm going to up it to ~2V for now. May be I will try to buy a kit in the future to get a better impression on the ICs

Well, my experience with HCF0 has been mixed. I had only two 2000C9DNV kits, I haven't tested the 2133C9D kit, I bought recently, yet.
Of the first kit, one stick was dead, the other could do 1100 MHz 9-9-9 @ 2.10V, 1040 MHz 8-8-8 @ 2.00V, 914 MHz 7-7-7 @ 2.00V. Then I got a second kit that could do 1000 MHz 9-9-9 @ 1.96V, 950 MHz 8-8-8 @ 2.10V, 890 MHz 7-7-7 @ 2.10V. ( 32M stable on X48 )
As you can see, my board doesn't like dual channel of HCF0 at all, the second kit could do 1066 MHz 9-9-9 @ 1.69V on Z68 ( by websmile ). I guess the higher voltage requirements are either due to me using S775 or rather my board in particular. I'd like to try my HCF0 on Z87 as well, I think it hasn't been done yet by anyone. It would be nice if I could compare to your results!

Recently bought a bunch of Samsung 4Gbit revision D OEM modules just to see what the ICs are like. The best pair can do this:

Looks like Samsung 4Gb ICs are getting worse.

[tiborrr]

I have three PSC X that can do 1200 7-11-7-28-1T @ 1.83V (A3G-A), the two others (one X A3G-M, one T A3G-U) are unable to do that (they stop at around ~ 1140). I just want to run 4 sticks.

Since this is not the material you are looking for I am asking whether anyone is willing to sell me a stick to match my other three?

[websmile]

I sell an OCZ PSC kit at classifieds, XEX A3G-A, but I do not have a single stick -maybe someone else can help you

[TaPaKaH]

It's not new that PSC ICs were licensed by Elpida but I haven't seen any IC that would carry both markings ... until today
http://item.ebay.de/291116558373
http://abload.de/img/_57l9utu.jpg

[Reefa_Madness]

Thanks for sharing that pic.

[darkage]

hi
i have some gskill psc x kits, wonder if there are great diferences between xd/xe and xf ics
2x2gb 12800cl8d-4gbrm - xd ics /9451040
2x2gb 16000cl6d-4gbpis - xe ics /10401040
3x2gb 12800cl7d-6gbpi - xf ics /11021040
can they be mixed in x79 ? ;-)
thanks for your help

[TaPaKaH]

In light of recent HOT contest, here is some info on Kingston IC detection.

Method 1:
Look at this bit of the sticker on the packaging, there should be a combination of numbers or letters.

02 indicates Elpida, 08 indicates Micron, 16 indicates Samsung, 32 indicates Hynix, KTC indicates "Kingston Technology Centre" so you can expect anything.
No further IC detection is possible using the sticker.

Method 2:
11-symbol vertical code on the module stickers.

Fourth symbol (letter) indicates ICs: "E" for Elpida, "H" for Hynix, "M" for Micron, "N" for Nanya, "Q" for Qimonda, "S" for Samsung.
Firth and sixth symbol (numbers) indicate number of ICs: "08" is for single-sided, 16 is for dual-sided modules.
Seventh symbol (number or letter) is month of production: "1" for January, .., "A" for October, "B" for November, "C" for December.
Eighth and nineth symbols (numbers) indicate year of production.
...other ones I haven't yet figured out.

Of course, knowing that modules are, say, Elpida is nice but it's not very precise since Elpida made many different DDR3 ICs. It is possible to go one step further in the detection.
Take a look at this (or a similar code). I don't know the internal name for it, so let's call it batch number for now.

If you order the kits according to these batch numbers, you will roughly get the order, in which they were produced/sold/shipped.
Because Kingston mostly used whatever ICs they had available at the time, knowing this batch number (or, alternatively, production date) and comparing it to close batch numbers with known ICs, can give you a hint about your own set.
Of course, for this you need a IC+batch database. HERE is an Excel file of my personal IC observations, which you can use for data mining.

For example, if you're looking for BDBG/BDSE you'll see that it's very common on Elpida-based kits with batch numbers in 4500000-5000000 range, whereas on batches prior to 4500000 you're likely to get BABG and on batches beyond 5000000 you're likely to get BFBG/BFSE.

[Reefa_Madness]

@Sam,
Thanks for that post...lots of good info in there.

[Shiranui Gen-An]

Samsung 4Gbit Q-die on Patriot Viper 3 2x8GB DDR3-2400C10.

[Don_Dan]

How do they clock?

[Shiranui Gen-An]

Not very well lol I got them to boot once at 1300 11-13-13 but haven't been able to replicate it. Basically they aren't any different from the 4Gbit B-die stuff we've seen before.

[Don_Dan]

Thanks for the info, just what I expected.

[websmile]

Evil Don_Dan - Samsung Q are nothing to be happy about