Samsung B-die RAM finder(for choosing RAM for Ryzen systems)

[kalniel]

I'm going to test something about this - in my new 3900X build. I have obtained:

Corsair CMK16GX4M2B3200C16 Vengeance LPX 16 GB (2 x 8 GB) DDR4 3200 MHz RAM.

I have also secured:

Corsair CMK16GX4M2Z3200C16 Vengeance LPX 16 GB (2 x 8 GB) DDR4 3200 MHz RAM, AMD Ryzen Optimised.

Neither of these would appear to be B-die, and on the CAL and other specification data they appear identical. However, one is specifically described as AMD Optimised.

So, when the rig is up and running, I'll test both sets of memory to determine if there is actually much that is different between them or if there is anything that needs to be concerned about.

The real rated SPDs are quite different:

CMK16GX4M2B3200C16 is 2133MHz 15-15-15-36 at 1.2V

CMK16GX4M2Z3200C16 (ryzen one) is 2666MHz 16-18-18-35 at 1.2V

Corsair then 'overclock' both to 3200Mhz 16-16-16-36 at 1.35V

[Helios451]

The real rated SPDs are quite different:

CMK16GX4M2B3200C16 is 2133MHz 15-15-15-36 at 1.2V

CMK16GX4M2Z3200C16 (ryzen one) is 2666MHz 16-18-18-35 at 1.2V

Corsair then 'overclock' both to 3200Mhz 16-16-16-36 at 1.35V

By the way, where do you find out the real rated SDPs?

Despite that original SPD data, the current settings would seem to make them identical, so does that original setting imply the silicon is different or there is something baked in that expresses itself to like Ryzens?

[kalniel]

By the way, where do you find out the real rated SDPs?

It was at the corsair site for each product, eg:
https://www.corsair.com/uk/en/Catego...tab-tech-specs

and

https://www.corsair.com/uk/en/Catego...tab-tech-specs

You can also look up the values for your RAM using CPUZ.

Despite that original SPD data, the current settings would seem to make them identical, so does that original setting imply the silicon is different or there is something baked in that expresses itself to like Ryzens?

It means the silicon is likely different. But if you use XMP or whatever then they'll run at the same speeds even if the underlying silicon has different capability. Corsair will have verified that they can run at the advertised (same) settings - hence the 'tested' latency, voltage and speed on the above pages.

[Domestic_Ginger]

According to thaiphoon burner, the uploaded data (this is all there is in the database) for the SKU's are as follows:-

CMK16GX4M2B3200C16

Hynix TFC
1067 MHz 16 15 15 36 50 4 6 6 23
1067 MHz 15 15 15 36 50 4 6 6 23
933 MHz 14 13 14 31 44 4 5 5 20
933 MHz 13 13 14 31 44 4 5 5 20
800 MHz 12 11 12 27 38 3 5 5 17
800 MHz 11 11 12 27 38 3 5 5 17
667 MHz 10 9 10 22 31 3 4 4 14
667 MHz 9 9 10 22 31 3 4 4 14

and

SAMSUNG E-DIE
1067 MHz 16 15 15 36 50 4 6 6 25
1067 MHz 15 15 15 36 50 4 6 6 25
933 MHz 14 13 13 31 44 4 5 5 22
933 MHz 13 13 13 31 44 4 5 5 22
800 MHz 12 11 11 27 38 3 5 5 19
800 MHz 11 11 11 27 38 3 5 5 19
667 MHz 10 9 9 22 31 3 4 4 16
667 MHz 9 9 9 22 31 3 4 4 16

CMK16GX4M2Z3200C16

SAMSUNG B-DIE
1333 MHz 16 18 18 35 53 5 8 8 28
1200 MHz 15 17 17 32 48 5 8 7 26
1067 MHz 14 15 15 28 43 4 7 6 23
1067 MHz 13 15 15 28 43 4 7 6 23
933 MHz 12 13 13 25 38 4 6 5 20
800 MHz 11 11 11 21 32 3 5 5 17
800 MHz 10 11 11 21 32 3 5 5 17
667 MHz 9 9 9 18 27 3 4 4 14

[CAT-THE-FIFTH]

Corsair are a lottery though - only some of their highest and models might be more consistent.

If you want E-die,it's best to get Crucial Ballistix kits,which tend to have those.

[Helios451]

It was at the corsair site for each product

You can also look up the values for your RAM using CPUZ.

Yes, sorry, I should have though of doing doing that! As I haven't built the rig yet I can't use CPUZ, but I'll certainly exploit it to the full when testing the modules.

Corsair will have verified that they can run at the advertised (same) settings - hence the 'tested' latency, voltage and speed on the above pages.

So it's a bit of marketing talk to say that one is AMD Ryzen optimised and the other not...

[kalniel]

Yes, sorry, I should have though of doing doing that! As I haven't built the rig yet I can't use CPUZ, but I'll certainly exploit it to the full when testing the modules.



So it's a bit of marketing talk to say that one is AMD Ryzen optimised and the other not...

*shrug* I don't know if the sub timings are different. Or it could just be that Ryzen works better when you stick closer to JEDEC values - it's quite possible to get JEDEC rated 3200 speed RAM which doesn't need overclocking by Corsair or whoever, and maybe they think that works better on Ryzen. Or maybe you can clock that RAM higher, which makes it a better fit up to 3600 or so where Ryzen really likes it.

[kalniel]

Hynix TFC

Hmm at https://www.skhynix.com/products.do?...=&rk=&rc=#tg02 they don't have TF speed for any 'C' dies (CFR/CMR), only A and B. It's likely to be H5AN8G8NAFR, TF speed.

I agree about Samsung B die for the other sticks though, all Hynix chips are 19-19-19 for 1333/2666

But CPUZ will report the manufacturer these days so we'll find out

[DanceswithUnix]

I just ordered 2 sticks of https://www.amazon.co.uk/gp/product/B07BGDG5ZP/

I gather from some Googling that Micron E die is OK, so these are basically underclocked performance ram with ECC to tell you you have gone too far? I usually don't overclock and these are dual rank, but I'm tempted to have a little play

[Helios451]

[QUOTE=DanceswithUnix;4208212]I just ordered 2 sticks of https://www.amazon.co.uk/gp/product/B07BGDG5ZP//QUOTE]

I like it that you're not into bling!!

[Helios451]

*shrug* I don't know if the sub timings are different. Or it could just be that Ryzen works better when you stick closer to JEDEC values - it's quite possible to get JEDEC rated 3200 speed RAM which doesn't need overclocking by Corsair or whoever, and maybe they think that works better on Ryzen. Or maybe you can clock that RAM higher, which makes it a better fit up to 3600 or so where Ryzen really likes it.

Give me a couple of weeks and after doing my build log I'll do a parallel test of the two pairs of 8GB sticks - and see which ones are the most stable and effective, and which OCs the best on the Ryzen rig.

[cptwhite_uk]

I just ordered 2 sticks of https://www.amazon.co.uk/gp/product/B07BGDG5ZP/

I gather from some Googling that Micron E die is OK, so these are basically underclocked performance ram with ECC to tell you you have gone too far? I usually don't overclock and these are dual rank, but I'm tempted to have a little play

Buildzoid on Youtube reckons Micron E will clock above 4000, it's in one of his videos somewhere.

[CAT-THE-FIFTH]

Crucial UK sell 3600C16 Micron E-die kits for £90,and you can get 2.4% cashback on it too.

[cptwhite_uk]

The Crucial Ballistix using Micron E have "AES" towards the end of their part code.

[CAT-THE-FIFTH]

The Crucial Ballistix using Micron E have "AES" towards the end of their part code.

They have replaced those apparently with newer kits:
https://www.reddit.com/r/Amd/comment...is_edie_works/

Pretty much, while it's not 100% technically true to call it E-die, it's just as good if not better than the die you would find in the AES parts. We only use Micron tuned die in our Crucial Ballistix line, that's something we're committed to.

Everything Ballistix branded that has been manufactured since ~1.5 years ago has used only Micron die. Our latest generation of Crucial Ballistix parts continue this trend.

While it's not 100% correct to call what's in our latest memory E-die, it's for all intents and purposes the same E-die you would find in parts like our previous AES sport line. Our E-die however isn't like what you would find in E-die in other vendors parts, since we're Micron we specifically tune our die for performance with Crucial Ballistix parts, that's why our stuff consistently OCs and runs so well with Ryzen

[DanceswithUnix]

Buildzoid on Youtube reckons Micron E will clock above 4000, it's in one of his videos somewhere.

Nice, though I don't expect to push anywhere near that far. You don't buy ECC ram to live on the edge