Two Corsair kits, same shelf, same price. One is CMK16GX4M2B3200C16. The other is CMK16GX4M2E3200C16. One letter apart. Every forum answer to this question is a guess, so we checked it against our own catalogue of 813 RAM kits.
The short version: that letter changes nothing you can measure, and the thing that does change the memory chips is printed on the module, not in the part number.
Reading a Corsair part number
Take CMK16GX4M2B3200C16 apart:
| Segment | Means |
|---|---|
| CM | Corsair Memory |
| K | Product line. K is Vengeance LPX |
| 16G | 16GB total in the kit |
| X4 | DDR4. X5 is DDR5 |
| M2 | 2 modules in the kit |
| B | Version letter |
| 3200 | Rated speed in MT/s |
| C16 | CAS latency 16. The C is a profile marker, see below |
Product line letters across the Corsair kits we list:
| Prefix | Line |
|---|---|
| CMK | Vengeance and Vengeance LPX |
| CMH | Vengeance RGB |
| CMW | Vengeance RGB Pro |
| CMG | Vengeance RGB RS |
| CMT | Dominator Platinum RGB |
| CMP | Dominator Titanium |
| CMS | Vengeance SODIMM |
The version letter: we found 19 test cases, and it changed nothing
Our catalogue holds 19 groups of Corsair kits whose part numbers are byte-for-byte identical apart from that one letter. Some groups have three variants.
| Part numbers | Letters | Speed | CL | True latency | Our speed score |
|---|---|---|---|---|---|
| CMK16GX4M2B3200C16 / CMK16GX4M2E3200C16 | B, E | 3200 | 16 | 10.0ns | 8.0 |
| CMK32GX4M2D3200C16 / CMK32GX4M2E3200C16 | D, E | 3200 | 16 | 10.0ns | 8.0 |
| CMK32GX5M2B6000C36 / D / E | B, D, E | 6000 | 36 | 12.0ns | 7.3 |
| CMH32GX5M2B6000C36 / D / E | B, D, E | 6000 | 36 | 12.0ns | 7.3 |
| CMH64GX5M2B6400C32 / N / Y | B, N, Y | 6400 | 32 | 10.0ns | 8.8 |
| CMT32GX5M2D6000C36 / E / X | D, E, X | 6000 | 36 | 12.0ns | 7.3 |
| CMP64GX5M4B6400C32 / X | B, X | 6400 | 32 | 10.0ns | 8.8 |
In all 19 groups the rated speed, CAS latency, true latency and our speed score are the same. Where a colour or a voltage differs between two entries, that is our data being incomplete on one of them, not the kits differing.
So the version letter is a SKU marker. Corsair issues a new one when the product gets re-released, repackaged, or re-qualified. It is a bookkeeping code, not a performance grade. A newer letter is not a better kit.
The letter that does tell you something: C or Z
The letter immediately before the CAS number is different. That one is worth reading.
We checked all 227 Corsair kits in our catalogue that end in a C or Z followed by a CAS number:
| Letter | Kits | List an XMP profile | List an EXPO profile |
|---|---|---|---|
| C | 181 | 181 | 11 |
| Z | 46 | 13 | 46 |
Every single Z kit carries AMD EXPO. Every single C kit carries Intel XMP. A handful of C kits carry both, but if you are on AM5 and you want a profile that just works, Z is the letter to look for.
So in CMK32GX5M2B6000Z30, the Z is telling you it is EXPO tuned. In CMH32GX5M2B6000C36, the C is telling you it is an XMP kit.
The number that actually changes the chips is not in the part number
Here is the part nobody puts in an article.
Look at the white label on the module itself. There is a line that reads ver 4.32 or ver 5.32 or similar. That number is not in the part number, it is not on the box listing, and it is not in most retailer descriptions. It is what tells you which memory chips are inside.
Our own catalogue entry for CMK16GX4M2B3200C16 records it plainly: ver 3.xx is Micron, ver 4.xx is Samsung, ver 5.xx is SK Hynix. Same part number, three different manufacturers of memory chip depending on when the kit was built.
This is confirmed on Corsair's own support forum, where a user tracked the same CMK16GX4M2B3200C16 shipping as Samsung B-die in ver 4.32 and SK Hynix A-die in ver 5.32. Corsair's own staff response in that discussion is unambiguous:
We cannot mix and match 2 different version of RAM kits on a system since this will cause a different kind of issues with your system.
Read that again in context. Two kits with the same part number but different ver numbers are not guaranteed to run together. People in that thread hit blue screens doing exactly that, and the only fix offered was to turn XMP off and tune by hand.
What this means when you are actually buying
- B versus E is the wrong question. Whichever one is cheaper and in stock is the right one. The specs are identical and the letter is a SKU code.
- C versus Z is the right question if you are on AMD. Z means EXPO. On an AM5 board that is one fewer thing to fight in the BIOS.
- Buy the total capacity you need in one kit. Not because of the letter, because of the ver number you cannot see before it arrives. Two 16GB kits bought separately may carry different chips and may refuse to run at rated speed together. One 32GB kit is guaranteed matched at the factory.
- If you must add to an existing kit, photograph the label on your current modules, read the ver number, and try to source the same one. Corsair does not let you filter by it, so this usually means a seller who will check the sticker for you.
- If you plan to tune timings by hand, the ver number decides your ceiling, not the letter. SK Hynix A-die and Samsung B-die have very different headroom. Our guide to memory ICs and overclocking headroom covers which chips do what.
This is not only a Corsair habit
Silently changing components under a fixed part number is standard practice across the industry. We wrote the storage version of this problem in SSDs that change parts under the same name, where drives got slower after launch reviews with no change to the model number.
The memory version is gentler, since a different die is usually not slower at the rated profile, but it bites in exactly one place: mixing. If you are pairing modules, the ver number is the only thing that matters and it is the one thing nobody advertises.
The kits in this article
- Corsair Vengeance LPX CMK16GX4M2B3200C16, 16GB DDR4-3200 CL16, 16-18-18-36 at 1.35V
- Corsair Vengeance LPX CMK16GX4M2E3200C16, the same kit with a different SKU letter
- Corsair Vengeance LPX CMK32GX4M2E3200C16, the 32GB version
- Corsair Vengeance CMK32GX5M2B6000Z30, a DDR5 Z kit at the AM5 6000MT/s sweet spot
Compare any two of them side by side in our RAM comparison tool, or read true latency versus speed rating if you want to know why CL on its own tells you nothing.