Overclock.net banner
4,461 - 4,480 of 10,850 Posts
Finally stable using dram calculator but what I like to call 3800mhz timings on 3533mhz boot.

I've been following the guide correctly from TPU.

https://www.techpowerup.com/reviews/AMD/Ryzen_Memory_Tweaking_Overclocking_Guide/5.html

actually the ram wont boot 3266-3000-3333-3400 from dram calculator with rec, alt 1-3 including min to max volts. It does boot 3600 safe and debug but both fail prime95 after 4h. I decided to apply 3800mhz calculator onto 3600mhz but this failed prime95 overnight. Just woke up and found 3533 (w 3800 timings) passes 8h prime! The bandwidth on aida64 still much higher with loose 3533mhz ram than stock 3200mhz

Edit: also used tRFC calculator to get tRFC 2 & 4 values.
Frequency / 2000 x tRFC timing = tRFC ns

Image
 

Attachments

It's Read/Write/Copy performance is half of that of Ripjaws V 3200 CL16. Could be the the system used. Can't really compare latency cos they used an intel system.
That is because only a single module was tested, instead of a 2x8 GB kit.
 
Anyone?

Hello friends, my computer not booted with this new DRAM Ryzen Calculator 1.5.0.5

I've used the old calculator 1.4.1 and it worked perfectly.

What changed in calculator which not works for me now?

Here is the screenshot from 1.4.1 which worked for me, the selected BLACK is what i've changed on the bios.
Image


Here is the new screenshot from the 1.5.0.5 which my computer not booted.
Image
 
Finally stable using dram calculator but what I like to call 3800mhz timings on 3533mhz boot.

I've been following the guide correctly from TPU.

https://www.techpowerup.com/reviews/AMD/Ryzen_Memory_Tweaking_Overclocking_Guide/5.html

actually the ram wont boot 3266-3000-3333-3400 from dram calculator with rec, alt 1-3 including min to max volts. It does boot 3600 safe and debug but both fail prime95 after 4h. I decided to apply 3800mhz calculator onto 3600mhz but this failed prime95 overnight. Just woke up and found 3533 (w 3800 timings) passes 8h prime! The bandwidth on aida64 still much higher with loose 3533mhz ram than stock 3200mhz

Edit: also used tRFC calculator to get tRFC 2 & 4 values.
Frequency / 2000 x tRFC timing = tRFC ns

Image
Good job!

That is because only a single module was tested, instead of a 2x8 GB kit.
Oh i see. Did not even look. So, its about as fast as the Ripjaw V 3200 CL16. Did they test the rest Single, too?
 
Hi can someone can confirm that H5AN8G8N?FR-TFC is AFR ? Thx
Just run Thaiphoon - that will tell you everything.

Normally, there isn't a '?' - it would normally say MFR or AFR, but Thaiphoon will tell you.
 
It says ?FR if you have a new kit that Thaiphoon doesn't know yet. Later versions probably will recognize it as CJR, MFR or AFR.
Thanks Reous - was wondering why it was there.... :)
 
Honestly Samsung BDie can handle much higher voltages than other dies without adversely affecting them. I have had my set for 2.5 years and I start at 1.45v DRAM voltage. Although, I usually go for max daily performance which generally needs 1.5v. Performance of the DRAM has never degraded and ive ran them at 1.5v+ daily for well over a year at this point. With that being said, 1.45v is plenty safe and above that you should have either a small fan blowing directly on the ram or an aftermarket ram cooler. Staying between 20-30c is my sweet spot but I maintain stability above those temps. I made my own Ram cooler from an old gtx660 set of gpu fans and fabbed up a crude but effective way to mount them using some sheetmetal left over from removing a hard drive cage from a case then coated it with plastidip and when gently pressed onto the "fins" of the tridentZ heatsink the rubberized mounts pinch and hold onto the fins.



Note: One reason you get lower freqs is the combination of using a 4x8gb kit and a TR but try some higher dram voltages while monitoring the DIMM temps during a heavy memory test and if you dont exceed 35-40c you should be fine to run at 1.45v on BDie easily. Then you can just start testing soc voltages starting at 1v and moving up one increment everytime until you find the value that produces the least errors or you become stable. Then you can start optimizing timings and resistances.


Some timings to try: 14, 15, 14(15), 14(15), 30, 44, 4, 6, 24, 4, 8(10), 12, 0, 2(3), 2(3), 280(320), 14, 8, 7, 3, 1, 7, 7, 1, 5, 5, 1

Other settings: Proc_Odt - 53ohm(48)
Rtt_nom, Rtt_wr, Rtt_park - RZQ/7, off, RZQ/3
CAD_BUS DRV - 20(24)ohm, 20(24)ohm, 20(24)ohm, 20(24)ohm CAD_BUS AddrCmd, CsOdt, Cke - 0(1), 0(1), 0(1)
Gear Down Mode - Enabled
Power Down - Disabled
BGS - Disabled
BGS ALT - Enabled


Leave everything in Tweakers Paradise on AUTO except maybe CLDO_VDDP which you could try at 700

DRAM voltage - 1.45v depending on DIMM temps
DRAM Boot Voltage - same as DRAM voltage


Interleaving - Channel
Interleaving Size - 512kb
Interleaving Hash - Enabled
Memory Clear - Disabled
HW Prefetcher - Enabled
Opcache - Enabled
Spread Spectrum - Enabled (This will probably be found in the DIGI VRM menu when Switching frequency on the cpu is set to either AUTO or Manual IIRC)


Start at 2933mhz and if stable then bump up until you find max freq then you can tweak timings etc
You blue astralian, mate. I started with the lose timings and 2933 last night at around 3 am. Woke up by 7 and it was 0 errors I will continue testing :thumb:
 
Wrong thread. Apologies. :blushsmil

~Ceadder :drink:
 
You blue astralian, mate. I started with the lose timings and 2933 last night at around 3 am. Woke up by 7 and it was 0 errors I will continue testing :thumb:
Have you tried 3133MHz?
 
You blue astralian, mate. I started with the lose timings and 2933 last night at around 3 am. Woke up by 7 and it was 0 errors I will continue testing :thumb:
Good deal! Thats what thousands of hours of B Die overclocking netted me. I can usually get a set and have it stable in 10 min, providing the hardware is good. That set will do the tighter timings im betting. At least you have somewhere stable to start now.
 
Good job!







Oh i see. Did not even look. So, its about as fast as the Ripjaw V 3200 CL16. Did they test the rest Single, too?
Thank you. Any recommendations for hitting 3600?

Obviously the calculator wont get there but I wonder what timings can be applied to hit 3.6ghz

Sent from my SM-G960W using Tapatalk
 
@1usmus

Thank you so much for releasing and continuing to develop this tool. I have kinda crappy Hynix M-Die that was stuck at 2933MHz with loose timings for over a year on my system because the 3200MHz XMP profile didn't work with my CPU, I didn't know about the program and changing all the timings seemed very daunting so I loaded the XMP profile, lowered the frequency to 2933MHz, and called it a day.

Unfortunately that left me with 16-18-18-38-56 timings which definitely wasn't ideal. Thanks to your program coupled with Thaipoon Burner I was able to increase my memory speed to 3066MHz while lowering my main timings to 14-16-16-36-54 and decreasing my sub-timings as well. The memory stability testing tool that's built in also helped me figure out that 3200MHz wasn't gonna be worth it in my case because I had to set the timings way too loose.
Again thank you, you're a legend. :)








 
Did I miss the conversation I assume happened here about Samsung stopping B-die production with EOL 191Q? I wondered why I have had such a hard time finding what used to be common G. Skill 8GB B-die modules anywhere for sale.

I just learned about it now from here. Samsung Rings B-Die Memory Death Knell

So what is everybody going to use for equivalent speed and latency now that B-die is no more? Bad timing for the loss of B-dies and the advent of Ryzen 3000.
 
4,461 - 4,480 of 10,850 Posts