Figuring out G.Skill's SNs

[Don_Dan]

I thought it was about time to make a complete list of all the different codes G.Skill uses to mark different ICs and their manufacturers. A lot of valuable info is scattered over many different threads, I tried to collect as much info as I could, please feel free to add more numbers or dispute what I wrote!

Let's start:

G.Skill serial numbers always start with the year and week, for example 913 is the 13th week of 2009 while 1240 is the 40th week of 2012. Pretty straightforward.

After that there are four digits which stand for different ICs/manufacturers (the meaning of a letter (C, G, L, M, P) instead of a 0 in the fourth position is unknown at this point):

0000 - Review/Test sample (no IC specified)
01xx - Micron
0140
0150
0160
02xx - Hynix
0230
0240
0260
03xx - Samsung
0340/034P
0360
05xx - ??
0511
06xx - Elpida
0640/064L/064M
0650
0660/066L
0690 (MxH-E Hyper on kits with typical Hyper specs (Perfect Storm 2000 7-8-7-20 or 2200 8-8-8-24), on other kits it might be BBSE/EBSE)
07xx - Elixir (/Nanya)
0740
08xx - Nanya (sometimes eTT/uTT)
0840
0850
0860
09xx - ProMOS eTT/uTT
0900
0940/094G
10xx - Powerchip/PSC
1030
1040/104G/104L/104P
11xx - SpecTek (/Micron)
1120
1140
12xx - Qimonda/Aeneon (sometimes eTT/uTT)
1240/124M
1250/125M
1260
13xx - ?? (possibly blank IC)
1340
14xx - ??
144C

Since April 2012 G.Skill is using new numbers:

AB40 - Samsung/Hynix (seems to be both, only used for a short time in April 2012 (week 17 2012), mainly seen on Trident X 2400 CL10 kits where it is used for Samsung 4Gb B-Die)
000x - Hynix? (only on samples?)
0000

x1xx - Nanya eTT/uTT
DDR1 only?: 6100
DDR3 only?: 2100/210L/2150
x2xx - Powerchip/PSC (sometimes eTT/uTT)
DDR3 only?: 1200
x3xx - Micron
DDR4 only?: 3300
x4xx - Hynix
DDR3 only?: 1400/14HL/2400
DDR4 only?: 3400/B400
x5xx - Samsung
DDR3 only?: 1500/2500/250B
DDR4 only?: A500
x6xx - SpecTek (/Micron)
DDR3 only?: 1600
DDR4 only?: B600
x7xx - ??
DDR3 only?: 1700
x8xx - Elpida
DDR3 only?: 0800/1800
DDR3 only?: 3800 MxH-E Hyper (possibly same situation like 0690 in old scheme)
x9xx - Elixir
DDR3 only?: 29B0

The first digit of the new serial numbers and the last two digits of the old serial numbers might indicate the speed bin originally assigned by the IC manufacturer.

These four digits are followed by six (on DDR4 from 2016/2017 seven) more digits which seem to be consecutive numbers assigned to each kit.

This is it for now!

Of course credit for this goes to many people like Tapakah, stunned_guy and many more!

If you need any more info about a specific memory kit, please visit the RAMlist at i4memory.


Since August 2017 G.Skill is using new codes that indicate the IC:

Samsung
DDR3:
03133X2810P = Samsung 2Gb P-die
03160R4810B = Samsung 4Gb B-die
03160M4810Q = Samsung 4Gb Q-die
03160R4810Q = Samsung 4Gb Q-die
DDR4:
04240X4810D = Samsung 4Gb D-die
04213X4810E = Samsung 4Gb E-die (DDR4-2133)
04240M4810E = Samsung 4Gb E-die (DDR4-2400)
04213X8810B = Samsung 8Gb B-die
04240M8410C = Samsung 8Gb C-die (DDR4-2400)
04266M8410C = Samsung 8Gb C-die (DDR4-2666)
04240M8810C = Samsung 8Gb C-die
04320X8810D = Samsung 8Gb D-die
04320MS410A = Samsung 16Gb A-die

Hynix
DDR3:
03160R4420M = Hynix 4Gb MFR
DDR4:
04213X4820A = Hynix 4Gb AFR
04240X4820B = Hynix 4Gb BJR
04240H48211 = Hynix 4Gb 1JR
04213X8821M = Hynix 8Gb MJR
04240R8823A = Hynix 8Gb AJR
04213X8821A = Hynix 8Gb AJR
04213X8821C = Hynix 8Gb CJR
04213X8821D = Hynix 8Gb DJR
04213X8825J = Hynix 8Gb JJR
04266X8820J = Hynix 8Gb JJR
04266XS820M = Hynix 16Gb MJR
04320XS826A = Hynix 16Gb AJR
04240BS821C = Hynix 16Gb CJR
L4240RS821C = Hynix 16Gb CJR
04320X8820C = possibly mislabeled 16Gb CJR

Micron/SpecTek
DDR4:
04240H4833A = Micron 4Gb A-die
04440X8833B = Micron 8Gb E-die (design ID Z11B)
L4320XS833B = Micron 16Gb B-die
04440XS833A = Micron 16Gb B-die (design ID Z22A)

Powerchip/PSC
DDR3:
03133X1841D = PSC 1Gb D-die
DDR4:
04213X4841A = PSC 4Gb A-die

Nanya
DDR4:
04320X8850C = Nanya 8Gb C-die

The first two digits stand for the memory type, DDR3 (03) or DDR4 (04). The meaning of 'L' instead of '0' is unknown. The next three digits probably indicate the JEDEC spec (133 = DDR3-1333, 160 = DDR3-1600, 213 = DDR4-2133, 240 = DDR4-2400, 266 = DDR4-2666, for Micron chips the factory bin has also been seen 440 = DDR4-4400). The meaning of the 6th digit is unknown at this point. The 7th digit indicates IC density (4 = 4Gbit, 8 = 8Gbit, S = 16Gbit) and the 8th digit might indicate organization (usually 8 = x8 or 4 = x4). The 9th digit indicates the IC manufacturer (1 = Samsung, 2 = Hynix, 3 = Micron/SpecTek, 4 = Powerchip Semiconductor Corp., 5 = Nanya) and the 11th digit indicates the die revision (B = B-die etc.) or the design ID for some Micron chips. The meaning of the 10 digit is still unknown.

Thank you to Mickulty for the first info about the new system.


Since the release of DDR5 in 2021 G.Skill is using an adapted version of the one above:

Samsung:
DDR5:
0M48AXS810B = Samsung 16Gb B-die
0M48KRS810B = Samsung 16Gb B-die
0R48AXS810B = Samsung 16Gb B-die

Hynix
DDR5:
0R48KXS820M = Hynix 16Gb M-die
0T56KXS820A = Hynix 16Gb A-die
LT64AXR820M = Hynix 24Gb M-die

Micron/SpecTek
DDR5:
0R48RXS830A = Micron 16Gb A-die

The meaning of the first two digits is unknown (0M, 0R). The next two digits probably indicate the JEDEC spec (48 = DDR5-4800, 56 = DDR5-5600), while the meaning of the 5th and 6th digit is unknown. The 7th digit indicates IC density (S = 16Gbit, R = 24Gbit) like above and the 8th digit might indicate organization (usually 8 = x8 or 4 = x4). The 9th digit indicates the IC manufacturer (1 = Samsung, 2 = Hynix, 3 = Micron/SpecTek?) and the 11th digit indicates the die revision (B = B-die, M = M-die etc.). The meaning of the 10 digit is still unknown.


Edit 02.11.12: I've added some info from OCX, I see Tapakah has updated his thread tonight as well. ^^
Edit 18.11.12: Update for x100 and x600
Edit 20.11.12: Update for 2100, added 0800
Edit 21.11.12: Decided to put 2100 and 2150 into different categories
Edit 08.12.12: Update for 1240 -> Qimonda
Edit 16.12.12: Small corrections
Edit 22.01.13: Added 1340 -> ProMOS, 09xx changed to "no idea"
Edit 09.03.13: Added 3800 -> Elpida MxH-E Hyper
Edit 25.03.13: Added 0940 -> PSC eTT/uTT
Edit 12.04.14: Added 0690 -> EBSE
Edit 31.05.14: Fourth spot changed to variable
Edit 16.09.14: Added 0150 -> Only D9GTR? (1600 7-7-7 @ 1.9V kits)
Edit 06.10.14: Added 3400 -> DDR4 only so far, marked in italic
Edit 17.11.14: Added 3300 -> DDR4 only so far, marked in italic, seen on F4-2400C15Q-16GRB kit
Edit 12.10.20: Clarified 0690 -> not always MxH-E Hyper, added new DDR4 scheme
Edit 20.01.21: Added example for new scheme for DDR3
Edit 17.05.21: Changed 09xx to ProMOS, thanks @redux, 13xx changed to "no idea"
Edit 01.06.21: Added 144C -> "no idea", seen on Ripjaws F3-12800CL9D-4GBRL, 1207144C180xxx
Edit 29.06.21: Added 14H0 and 03160R4420M, added B400 on DDR4, added 2150 to Nanya, changed x10x to x1xx, added DDR2 0511, 0900 and 0230
Edit 18.07.21: Added 6100, seen on a kit of F1-3200PHU2-2GBNS from Jan 2014
Edit 20.09.21: Added 03133X1841D, seen on a kit of F3-12800CL7D-4GBPI-B and F3-17600CL7D-4GBPIS from 2018, added 04213X8825J, 04213X8821A and 04266M8410C seen on Aegis kits, added 04240H4833A seen on F4-2400C15D-8GNT
Edit 28.11.21: Added 04213X8821M, seen on a kit of F4-3200C16D-32GVK, added 04240BS821C seen on a kit of F4-4400C19D-32GVK
Edit 21.01.22: Correction for 0650 - [strike]BBSE[/strike] Elpida
Edit 04.08.22: Added 03160M4810Q, seen on a F3-2133C9D-8GAB kit from November 2017
Edit 05.08.22: Added 0M48KRS810B, seen on F5-6000U3636E16GX2-TZ5S from week 41 2021 and 0M48AXS810B, seen on F5-6000J4040F16GX2-TZ5S from week 03 2022
Edit 06.10.22: Added 04320XS826A, seen on F4-4800C20D-32GTES from week 25 2021
Edit 11.10.22: Added Added 14HL seen on F3-1600C11S-4GIS from week 03 2015, also added 064M & 066L, 124M & 125M as variants for Elpida or Qimonda
Edit 12.10.22: Added 104G seen on F3-17000CL9D-4GBXL from week 43 2011, 1143104G139598
Edit 13.10.22: Added 0R48AXS810B seen on F5-6000J3636F16GX2-TZ5RS from week 03 2022, added 0R48KXS820M seen on F5-6400J3239G16GX2-TZ5RS from week 17 2022, formatting
Edit 17.10.22: Added L4240RS821C seen on F4-4400C19D-32GTZR from week 25 2022
Edit 19.10.22: Added 0T56KXS820A seen on F5-6600J3440G16GA2-TZ5RK from week 37 2022
Edit 14.12.22: Added 04213X4841A seen on F4-2133C15S-4GNT from week 31 2018, added 03160R4810Q seen on F3-2400C10D-16GTX from week 49 2021, added 034P seen on F3-2400C10D-8GZH from week 16 2012 (April 2012, old scheme), added 04320MS410A seen on F4-2666C19D-64GVK from week 39 2022, added L4320XS833B seen on F4-3600C19D-32GTRG from week 35 2022 and added 0R48RXS830A seen on F5-5200U4040A16GX2-RS5K from week 41 2021 (Micron Z9ZSB)
Edit 05.02.23: Added 04320X8820C seen on F4-4400C19D-32GVK from week 40 2022 (16GB SR modules), added 210L seen on F3-1866C9D-16GSR from week 26 2015, added 04320X8850C seen on F4-3200C16D-32GVK, added 04440XS833A seen on F4-3600C16D-32GVKC from week 02 2023 (C9BLJ, CPG marking, design ID Z22A, DDR4-4400 factory bin), added 04440X8833B seen on F4-3600C16D-16GVKC from week 05 2023 (C9BLM, CPG marking, design ID Z11B, DDR4-4400 factory bin)
Edit 07.02.23: Added 04320X8810D seen on F4-3600C19D-16GSXKB from week 42 2022
Edit 05.03.23: Added 03160R4810B seen on F3-2400C10Q-32GTX from week 18 2018, added 104P seen on F3-16000CL9T-6GBTD from week 13 2010 and week 29 2010, added 1700 seen on F3-10600CL9D-2GBNQ from week 39 2012, added 0850 seen on F3-12800CL8D-2GBTD from week 10 2010
Edit 26.04.23: Added 04240R8823A seen on F4-3000C16D-16GISB from week 50 2018, added LT64AXR820M seen on F5-8000J4048F24GX2-TZ5RS from week 14 2023, added 29B0 seen on F3-12800CL9D-4GBXL from week 43 2015

[websmile]

Good to see this information now also available on XS, nice work, Dan

[Don_Dan]

Good to see this information now also available on XS, nice work, Dan

Thanks mate!

[basco]

thanks very much

[Leeghoofd]

I hate to say it, but good job Daniel :p

[Reefa_Madness]

Very useful and now it is bookmarked. Thanks.

[Don_Dan]

thanks very much

Very useful and now it is bookmarked. Thanks.

You're welcome!

I hate to say it, but good job Daniel :p

Thanks Alby ( a.k.a. old man )!

[Shiranui Gen-An]

New kit Trident X 2666C10 Samsung based:

[Don_Dan]

New kit Trident X 2666C10 Samsung based:

Thanks for the post!

[SonDa5]

New kit Trident X 2666C10 Samsung based:

Newegg has them for $180. Seems kind of pricey for 8gb or RAM. What are they capable of as far as over clocking is concerned?


Slightly slower CL 11 2666mhz kit can be had for around $120.

The cheapy Sammy 8GB kits can be had for $40.


What makes these so special?

[Shiranui Gen-An]

Newegg has them for $180. Seems kind of pricey for 8gb or RAM. What are they capable of as far as over clocking is concerned?


Slightly slower CL 11 2666mhz kit can be had for around $120.

The cheapy Sammy 8GB kits can be had for $40.


What makes these so special?

If the Dominator Platinum 2666C10 sticks are any indication, A LOT.

[felix_w]

What is it, DS or SS ?

[Reefa_Madness]

New kit Trident X 2666C10 Samsung based:

But which Sammies? Weren't there some kits found that were sporting HYK0 chips, instead of HCH9?

[Don_Dan]

What is it, DS or SS ?

The kit must be double-sided, Samsung 4Gb ICs can't do those clocks.

But which Sammies? Weren't there some kits found that were sporting HYK0 chips, instead of HCH9?

I think there's no way to determine which ICs are used without removing the heatspreader. In my opinion it doesn't matter anyway, we know it must be some kind of Samsung 2Gb D-rev chip, after binning is done by G.Skill the difference between HCH9 and HCK0 should be negligible.

[Alex-Ro]

Whats the next rev after d?20nm chips from samsung should be extremely interesting,wonder when they show up....

[Reefa_Madness]

I think there's no way to determine which ICs are used without removing the heatspreader. In my opinion it doesn't matter anyway, we know it must be some kind of Samsung 2Gb D-rev chip, after binning is done by G.Skill the difference between HCH9 and HCK0 should be negligible.

I respectfully disagree with you, my young friend, as to it not mattering. It does to me.

I'm have a pretty good idea as to what HCH9 chips can do at various voltages. The experience with HCK0 is pretty restricted, mostly to Samsung generic, so we don't really know whether they have been limited by such things as the layout of the PCB used by Samsung on the LV / LP modules, or whether they don't clock as high, but if you sample results from known HCH9 based stuff, compared to known HCK0 the odds are that the former will best the latter. For the same dollars I'd rather get HCH9 in my modules.

[Zeus]

Please allow me to add some confusion Daniel:



Pretty sure it's some sort of Samsung chip:



Sorry for messing things up, i think Gskill is to blame.

[Reefa_Madness]

i think Gskill is to blame

Without a doubt.

[Alex-Ro]

Please allow me to add some confusion Daniel:



Pretty sure it's some sort of Samsung chip:



Sorry for messing things up, i think Gskill is to blame.

how do they clock,single or double-sided?

[Don_Dan]

Whats the next rev after d?20nm chips from samsung should be extremely interesting,wonder when they show up....

K4B2G0846E is already listed on Samsung's site as "MassProduction", but I haven't seen these chips yet.

I respectfully disagree with you, my young friend, as to it not mattering. It does to me.

I'm have a pretty good idea as to what HCH9 chips can do at various voltages. The experience with HCK0 is pretty restricted, mostly to Samsung generic, so we don't really know whether they have been limited by such things as the layout of the PCB used by Samsung on the LV / LP modules, or whether they don't clock as high, but if you sample results from known HCH9 based stuff, compared to known HCK0 the odds are that the former will best the latter. For the same dollars I'd rather get HCH9 in my modules.

Well, I agree with you that the data about HYK0 is pretty limited because most people test their Samsung LV/LP modules for energy saving settings at low volts, but I've seen some binned modules do very well at higher voltages. Please check this thread. You can see that the best modules are comparable to the Samsung value sticks with HCH9 we have seen. I don't think I'd prefer HCH9 over HYK0, but we might see more HCH9 just because they're cheaper. Another problem might be that most people don't bother to remove the heatspreader, for them it's good enough to know they have Samsung chips, so our data is limited on that side as well.

Please allow me to add some confusion Daniel:

Pretty sure it's some sort of Samsung chip:



Sorry for messing things up, i think Gskill is to blame.

Great, of course they look entirely different than the chips on "2100"!



I have the same question as Alex, double-sided or single-sided? I have never seen those chips to be honest... Can you tell me their size? And maybe some OC results?

[Reefa_Madness]

Well, I agree with you that the data about HYK0 is pretty limited because most people test their Samsung LV/LP modules for energy saving settings at low volts, but I've seen some binned modules do very well at higher voltages. Please check this thread. You can see that the best modules are comparable to the Samsung value sticks with HCH9 we have seen. I don't think I'd prefer HCH9 over HYK0, but we might see more HCH9 just because they're cheaper. Another problem might be that most people don't bother to remove the heatspreader, for them it's good enough to know they have Samsung chips, so our data is limited on that side as well.

HCH9 is cheaper than HYK0? Maybe for you, but not here. The generic HCH9 stuff sells for about the same price, or a few more $ (US) here than the HYK0 Samsung LV stuff. The HYK0 based Greens were recently on sale for $36 shipped per 2x4GB kit at the egg, or individually for $17 per module. The generic HCH9 sells for about $22 per module, plus shipping at the only two places that I know it can be had.

As for preferences, I still believe that as a general rule your chances of higher clocks are with the HCH9. Not saying that there isn't HYK0 that clocks good, just that in my mind I think the odds are better with HCH9. If GSkills and others start binning HYK0 and begin to mass produce high-end modules with them (not just a few here and there), then I won't be too proud to admit I was wrong. At my age I've had plenty of practice at admiting I had it wrong.

[XBrytanX]

I'm with you on this one and that's why i don't own single HYK0 ,HCH9 is what rides in high-end sets

[Reefa_Madness]

I guess that these types of discussions and opinions are probably best left for another thread, as there is no need to clutter this one up with what amounts to personal preferences, so I will do my part and refrain from making any more editorials...or at least I'll try real hard.

[stunned_guy]

Got TridentX 8Gx2 F3-2400C10D-16GTX with 12451500..... and RipJawsZ 8Gx2 F3-2133C11D-16GZL with 12341500.....
Will look after ic next day.

[Zeus]

Daniel, The sticks are double sided and do 2000 C9 but there's no scaling after that. In other words: it's cr@p. For dimensions of the chip i'm gonna have to pull the spreaders again. I will if you insist though.