Hynix CJR vs Samsung B Die

[Krit]

I have question can a overclocked Hynix CJR die reach speed of Samsung B die for example 3200MHz CL14 or 3600MHz CL16 when XMP is enabled.

How fast is Samsung B die when XMP is enabled ?

For comparison there is Hynix CJR die 3800MHz CL16-21-21-36 @1.41v

 

[TheoneandonlyMrK]

They are quite similar apparently, those are quite close to mine, bit higher in fact i get 52000 read 29000write 50000 copy at a similar speed.

 

[Krit]

Hmm that's very interesting i was thinking that Samsung B die is faster because it costs much more.

Samsung B die ~ 140eur
Hynix CJR die ~ 90eur

 

[TheLostSwede]

Have a look here. He only went down to CAS 15 on the B-dies though and 16 on Hynix CJR at 3800MHz. Keep in mind those tests are with a 3900X.
It seems Samsung might have a very slight edge, but as you point out, it's generally not worth the price premium.

AMD Ryzen 3000 DDR4 Scaling – Part II – Samsung B-Die vs Micron E-Die vs Hynix CJR – English version

Samsung B-Die vs Micron E-Die vs Hynix CJR One of the main changes that came with the launch of the X570 chipset, along with the launch of the third generation Ryzen processors, was the way …

lab501.ro

I have my CJR's at 16-19-16-19-36, 1.36V and paid the equivalent of US$89.

 

[Krit]

I have my CJR's at 16-19-16-19-36, 1.36V and paid the equivalent of US$89.

That's very good result i think but is it 100% stable ? DRAM Voltage seems very low.

 

[TheLostSwede]

That's very good result i think but is it 100% stable ? DRAM Voltage seems very low.

Yeah, not having any problems at all.
I only got the 3600 kit, but your 3800 kit is rated to run at 3800 at 1.35V, so not sure why you're pushing your Voltage so high.
Also, don't use XMP.

 

[Krit]

I'm using Patriot Viper Steel 3600 CL17 2x8GB. Dram Calculator recomends 1.41v - 1.45v for this type of memory.

 

[TheLostSwede]

I'm using Patriot Viper Steel 3600 CL17 2x8GB. Dram Calculator recomends 1.41v - 1.45v for this type of memory.

Right, exactly the same kit then.

I just decided not to worry about the Voltage. Keep in mind that the settings are recommendations to get the broadest possible range of memory to run at the speeds that it spits out. It doesn't take individual IC variances into account, so it's possible that you need higher Voltage than me, but have you tried lowering it?

 

[Krit]

I think at lower voltage it will crash. My current timings are 16-21-18-21-36-56-504 @1.41v.

Previously when i tested memory at 16-21-17-21-36-56-504 it crashed at 1.41v but now with optimized timings it won't crash anymore. I mean heavy testing for memory.

 

[TheLostSwede]

I think at lower voltage it will crash. My current timings are 16-21-18-21-36-56-504 @1.41v.

Previously when i tested memory at 16-21-17-21-36-56-504 it crashed at 1.41v but now with optimized timings it won't crash anymore. I mean heavy testing for memory.

Think...
But have you actually tried? If I have the same RAM, although admittedly different hardware for the rest of the system and can run the RAM at 0.5V lower than you, at least it's worth a try, no?
Also, try changing the ProcODT, if you haven't, mine is at 40 something. Could be the UEFI implementation from MSI as well.

 

[Krit]

Oh i done a lot off ram testing just one step bellow and it chrashes becouse voltage is not enought.

ProcODT is 43.6

16-21-18-21-36-56-504 @1.41v. Stable
16-21-17-21-36-56-504 @1.41v. Unstable

 

[TheLostSwede]

Oh i done a lot off ram testing just one step bellow and it chrashes becouse voltage is not enought.

ProcODT is 43.6

16-21-18-21-36-56-504 @1.41v. Stable
16-21-17-21-36-56-504 @1.41v. Unstable

Fair enough, I guess your hope is on MSI delivering a better UEFI then.
Could potentially be the memory controller as well, but at least you're at 1900 1:1, which should be as good as it gets.

Oh i done a lot off ram testing just one step bellow and it chrashes becouse voltage is not enought.

ProcODT is 43.6

16-21-18-21-36-56-504 @1.41v. Stable
16-21-17-21-36-56-504 @1.41v. Unstable

Well, look at this, B450 comparison. Looks like you're actually hitting the limit on B450, or at least you're really close. If you look at page 4, he's on similar a Voltage as you too.

AMD Ryzen 3000 DDR4 Scaling – Part III – AMD Ryzen 5 3600X + B450 – English version

Ryzen 5 3600X + B450 Hello… it’s me again… yes, the RAM guy. You remember, right? Back in July, AMD released the third Ryzen generation, together with a new chipset, X570, …

lab501.ro

 

[GorbazTheDragon]

I would expect b-die to do better particularly on RCDs, and I think the subtimings are probably going to be a bit tighter too. I'm pretty sure any of the stuff you are running right now would beat B-die XMP, but once you work b-die down there's not really any competition. Rev. E is only really competitive at getting max frequencies, but like CJR the timings aren't going to go as low.

 

[oxrufiioxo]

I picked up some Team T Force legend 3200 14-14-14-31 (Bdie) for my Ryzen 3900X build slightly before it launched for $114 with promo code on newegg.

Stock Xmp sucks on the Aorus Master but they're pretty easy to tune and tighten up at 3600MTs. On an Asus Tuf x570 the same kit ran awesome with just base xmp.

To me they were more than worth the extra $20-$25 over hynix/micron kits

 

[GorbazTheDragon]

Usually the 3200c16 sticks are around $75... So that would be closer to $45, and that's quite a cheap price for b-die.

 

[Cowboystrekk]

One of the major differences is that the Samsung can run tRFC a lot lower. Appart from CL, tRFC is often the most important timing for gamingperformance. How much tRFC does for the 3000-series I don`t know, but for the 1 and 2-series it could mean a lot (up to several percents) in some games. At 3600 B-die often can do around 270-300 tRFC, CJR can do 400-500 and E-die can do 500-600.

 

[oxrufiioxo]

Usually the 3200c16 sticks are around $75... So that would be closer to $45, and that's quite a cheap price for b-die.

I have a lesser 3200C16 kit on my 9900K system that overclocks like crap..

CORSAIR Vengeance RGB Pro 32GB 288-Pin DDR4 3200 Desktop Memory - Newegg.com

Buy CORSAIR Vengeance RGB Pro 32GB (4 x 8GB) 288-Pin DDR4 DRAM DDR4 3200 (PC4 25600) Desktop Memory Model CMW32GX4M4C3200C16 with fast shipping and top-rated customer service. Once you know, you Newegg!

www.newegg.com

This kit specifically, I would definitely not buy them again over the Legend variant.


Comically I felt like I was being cheap not going with a 3600 CL16 kit they were going for around 190$ at the time. I'm pretty happy I went with what I did over those kits.

Total build cost was over $3000 So in my case 50$ is pretty pointless to save.

On a 3600 and maybe even a 3700X build Some Micron E die or Hynix CJR Make a ton of sense.

 

[moproblems99]

My 3900x on x570 is doing 16 19 19 19 39 1.35. not sure other timings at the moment as I have just been trying to get a stable system. Granted the timings are what it is rated for. Once I get cooling under control, tweaking comes next.

Edit:. This is CJR

 

[TheLostSwede]

So far no-one has done areal apples to apples comparison though.
We're seeing different size kits, different AGESA's, different "tweaking" and so on, so it's hard to draw any real conclusions.
Price wise I would say CJR is the way to go, but sure, if you want to squeeze every last drop of performance out of your system and have the money to burn, by all means, get a high-end B-die kit.

 

[oxrufiioxo]

So far no-one has done areal apples to apples comparison though.
We're seeing different size kits, different AGESA's, different "tweaking" and so on, so it's hard to draw any real conclusions.
Price wise I would say CJR is the way to go, but sure, if you want to squeeze every last drop of performance out of your system and have the money to burn, by all means, get a high-end B-die kit.

Those last 2-3% are always the most expensive... You got a pretty sweet kit I wish my Corsair kit did similarly.

 

[TheLostSwede]

Those last 2-3% are always the most expensive... You got a pretty sweet kit I wish my Corsair kit did similarly.

It seems I might've gotten lucky as well, at least compared to a lot of other people running the same kit. That said, some of it seems to be UEFI and board dependent as well.
To be fair, it wasn't running like this initially, although with each major UEFI update from Gigabyte, it has improved and I'm really stoked with this kit, especially considering how little I paid for the performance I get out of it.

 

[Vario]

Samsung B die will overclock better than Hynix CJR. Micron/Crucial E Die is a bit closer to B-Die than CJR.

The B-Die I have can do 3400 CL14 and 4300 CL19 but that is with Intel 8600K not Ryzen.
Here's the validation for the 4300:

Intel Core i5 8600K @ 4801.15 MHz - CPU-Z VALIDATOR

[qpyumu] Validated Dump by Vario (2018-01-19 22:05:35) - MB: ASRock Z370 Taichi - RAM: 16384 MB

valid.x86.fr

Overall I would buy whatever is cheapest because its not that noticeable for the extra cost.

I likely could get even higher speeds than that, such as 3600 CL14 or over 4300 but I didn't bother doing much overclocking testing with this build. I didn't do any substantial overvolting or fine tuning of the secondary and tertiary timings. I haven't really done any significant benchmarking / overclocking since the first week I built it back in 2018, due to my lack of interest.

I ran Aida just now here is the ram at my daily 3200CL14. Probably my numbers aren't too helpful for comparing for Ryzen.

 

[oxrufiioxo]

It seems I might've gotten lucky as well, at least compared to a lot of other people running the same kit. That said, some of it seems to be UEFI and board dependent as well.
To be fair, it wasn't running like this initially, although with each major UEFI update from Gigabyte, it has improved and I'm really stoked with this kit, especially considering how little I paid for the performance I get out of it.

Not gonna spend a lot of time messing with sub timings at 3800 but this was a quick and dirty 1.4v 3800. I tried to make the gaming benchmarks as cpu limited as possible but likely need to swap in my 2080 ti to get a better picture on how much it matters going from 3200 to 3800... Those read/write/copy though... I think the bigger picture though is how easy with Bdie it was to get 3200/3600/3800 working didn't even need to manually set the NB clock.


3200 XMP


3200 manual timings 1.35v


3600 manual timings 1.35v


3800 16-16-16-32 auto subtimings. 1.4v

 

[Vario]

Huh? You're talking Intel here, the OP has an AMD CPU and CJR is much better on Ryzen than Micron E-die.
Maybe try and grasp the context of the thread before posting something that isn't quite relevant?

I am aware it is a largely Ryzen thread, stated that above. However he does not specifically ask for Ryzen info in his original post. My intent was to show that B die can overclock quite high above its XMP.

 

[TheLostSwede]

I am aware it is a largely Ryzen thread, stated that above. However he does not specifically ask for Ryzen info in his original post. My intent was to show that B die can overclock quite high above its XMP.

Which is why I removed my reply...