[SOLVED] B-Die CL16 vs D-Die CL15 - Ryzen 3000

[Miller2000]

Hi all,

My first thread in the forum. I have searched some but not been able to find anything directly related to my question.

Just upgraded to a Ryzen 3800x on a Asrock Taichi x570 motherboard. A final thing to decide on is the memory.

I bought the following memory: D-Die G.skill 3600 Mhz CL15 (F4-3600C15D-16GTZ) running with XMP settings giving the following latency: 16-15-15-35. (CL is automatically being increased to 16 due to Ryzen not accepting uneven values for CL (according to G.skill support themselves).)

I have run a few tests and am getting bad latency results and generally mediocre to bad results for the memory tests. Now I am trying to figure out why.

Results from MaxxMem and PerformanceTest V9.0:
PerformanceTest Results
MaxxMem Results

I have ordered a pair of B-Die G.skill 3600mhz CL16 (F4-3600C16D-16GTZR ) as well, which are supposed to be B-die. However I am thinking that the CL15 rated memory theoretically should perform better even if they are not "supposed" to be run with an AMD. I want to make sure I have given the CL15 D-dies the correct settings for making a proper comparison.

Is there any reason for the B-dies to be expected to perform better than the D-dies at the original rated speed and latency?

 

[Roland Of Gilead]

Look for 'gear down mode' or equivalent in bios settings. This will enable you to use the CL15 setting, and you can test to see if it performs better.

 

[Miller2000]

Look for 'gear down mode' or equivalent in bios settings. This will enable you to use the CL15 setting, and you can test to see if it performs better.

Hi Keith,

Thanks for you reply. I actually did try this based on G.skill's on suggestion. I activated "gear down" but it did nothing to the CL. Still stuck at 16.

 

[Roland Of Gilead]

Hey there,

Have you updated the bios? This is essential. Once done, do a CMOS clear, then try again in bios.

It's possible that even if you have done the bios update, without the CMOS clear, it could still not work. Let us know how that goes.

 

[Roland Of Gilead]

Also, worth looking at Ryzen DRAM calc. To see if you can loosen the CL, and tighten other timings to make up for the difference between CL15-CL16.

 

[Miller2000]

Hi again. Great inputs.

I did update the BIOS... But did not clear CMOS! Will definitely do this right away.

Regarding the DRAM Calc, this was what started my whole investigation. I get an error message when trying to calculate things when choosing D-Die. It simply says "not supported". So that could be seen as a downside of the D-die's already.

 

[TJ Hooker]

What exactly makes you your memory performance is lower than it should be? I'd run AIDA64 memory bandwidth and compare to review results, e.g. https://www.techspot.com/review/1891-ryzen-memory-performance-scaling/

 

[Miller2000]

What exactly makes you your memory performance is lower than it should be? I'd run AIDA64 memory bandwidth and compare to review results, e.g. https://www.techspot.com/review/1891-ryzen-memory-performance-scaling/

Based that on a few numbers I managed to find online. Comparing to that Techspot performance scaling my results are not even close to those charts.

Read at 31455 mb/s
Memory latency: 101 ns

 

[DMAN999]

I have Hynix DJC (D Die) G.Skill Ripjaws V 16GB (F4-3600C16D-16GVKC) @ 3733 MHz 16-19-19-19-36-56 in my 3700x rig and this is the best I could get so far:

https://imgur.com/a/jLdkoRa

View: https://imgur.com/a/jLdkoRa

https://imgur.com/bcrCfPl

View: https://imgur.com/bcrCfPl


I OC'd it to 3733 and had to tigthten up a few timings including tRFC, tRFC2 and tRFC4 to get there.

PS
The C16 Samsung B die kits are Not as well binned as the C14 B die kits and do NOT OC or perfrom as well.
I have G.Skill TridentZ RGB16GB 3200 MHz (F4-3200C16D-16GTZR) @16-17-17-17-36-56 (Samsung B die) in my 2600 rig and they do Not OC or perform anywhere near as well as the C14 B Die kits do.

https://imgur.com/cwvzV6z

View: https://imgur.com/cwvzV6z

 

[Miller2000]

Thanks DMAN for your inputs. It seems Bios had been reset for some reason. Changed the clocks manually to 3600 and enabled XMP again. One thing I noticed in the first run above was that timings actually showed as 15-15-15-35 - but at the base clock of 2133 mhz. Is there any particular reason it could set the timings correctly for the lower clock?

I got better results this time:

 

[DMAN999]

If you Disable GearDown Mode it can register the C15 without bumping it up to C16.
With GearDown Mode Enabled it will force it to C16.

PS
This was the best I could get out of my Samsung B Die 3200 kit:

https://imgur.com/DGFKqIu

View: https://imgur.com/DGFKqIu

 

[TJ Hooker]

Based that on a few numbers I managed to find online. Comparing to that Techspot performance scaling my results are not even close to those charts.

Read at 31455 mb/s
Memory latency: 101 ns

Hmm, maybe those techspot numbers were after they manually tweak timings or something. Here's another review where they're using 3600 MHz CL16 RAM and get similar results to yours. https://www.tweaktown.com/reviews/9051/amd-ryzen-3900x-3700x-zen2-review/index3.html

 

[DMAN999]

Read this:
https://www.pcgamesn.com/amd/ryzen-3000-memory-overclocking-sweet-spot

https://imgur.com/XPgb3em

View: https://imgur.com/XPgb3em

 

[Miller2000]

I got he B-Die kit and have done some bench mark tests to try to figure out which to choose of this kits.

I made a presentation that includes some of the tests I've done. And honestly the difference is not that big, and unfortunately results are not very consistent either.

See results here: https://drive.google.com/open?id=1s0MncDK9fUPi6HAIPvglCE1iUmcNG0cB

The D-Dies are approx 50 USD (or 500 Swedish Kr) more expensive. Any reason not to choose the B-Die over the D-Die despite the CL15 rating?

 

[DMAN999]

You probably won't ever notice the difference between those 2 kits in real world use.
So I'd keep whichever kit You like better.

 

[blaQQ]

Hi all,

My first thread in the forum. I have searched some but not been able to find anything directly related to my question.

Just upgraded to a Ryzen 3800x on a Asrock Taichi x570 motherboard. A final thing to decide on is the memory.

I bought the following memory: D-Die G.skill 3600 Mhz CL15 (F4-3600C15D-16GTZ) running with XMP settings giving the following latency: 16-15-15-35. (CL is automatically being increased to 16 due to Ryzen not accepting uneven values for CL (according to G.skill support themselves).)

I have run a few tests and am getting bad latency results and generally mediocre to bad results for the memory tests. Now I am trying to figure out why.

Results from MaxxMem and PerformanceTest V9.0:
PerformanceTest Results
MaxxMem Results

I have ordered a pair of B-Die G.skill 3600mhz CL16 (F4-3600C16D-16GTZR ) as well, which are supposed to be B-die. However I am thinking that the CL15 rated memory theoretically should perform better even if they are not "supposed" to be run with an AMD. I want to make sure I have given the CL15 D-dies the correct settings for making a proper comparison.

Is there any reason for the B-dies to be expected to perform better than the D-dies at the original rated speed and latency?

They are NOT D-die.

G.skill 3600 Mhz CL15 (F4-3600C15D-16GTZ) are 100% B-Die


Latest versions of Typhoon Burner gets WRONG readouts on this memory kit. Older versions get correct memory reading on this kit.

WRONG read! - K4A8G085WD-BCPB - These do not exist neither on the internet or samsung DDR4 production page.

CORRECT read! K4A8G085WB-BCPB - Exists on the internet aswell as the samsung DDR4 production page.