Download: DRAM Calculator for Ryzen 1.7.2 – what is new?

Been using this for a while. A few notes for this new version... In DRAM Calculator, tRCDWR is the second entry, and tRCDRD is the third. In my UEFI, those are reversed. Is my board just odd, or is this a labeling error in DRAM Calculator?

Second issue - I imported my RAM's XMP, generated by Thaiphoon. DRAM Calculator recommended a tRFC of 234, with an alternate of 368. However. the tRFC calculator in the "Additional Calculators" section of the program calculated a value of 513, based on a tRFC ns of 350 at 2933 MT/s. What's up with that?

And finally, clicking on the "Compare Timings" button often generated blank values, and sometimes caused a "Data Parsing" error message, which required me to kill DRAM Calculator via the Task Manager...

 

Are you high? It was always like that. There is no error if you pay attention to which exactly parameter you edit.

 

I've only used the program with one motherboard, so I wasn't sure. And no, I'm not high. If you can't be civil, ignore my questions and move on.

 

Motherboard has nothing to do with order labels in Calculator.

If my comment offended you, then I'm sorry.

 

Another reason I mentioned it is that, out of the dozens of labels in the program, those are the only two that aren't in the exact same order as the labels in my UEFI. Otherwise, I wouldn't have given it a second thought.

 

For me, this program never gave satisfactionary results, none of given timing sets work for my 2x16 GB G-Skill Trident Z B-die 3200 CL14 RAMs on 2700X with Asus Prime X470-Pro, computer do not start at all, even to BIOS.

 

Samsung?

The lower-end b die memories apparently differ significantly in quality compared to 3600's or higher though the timings for that kit still seems to be pretty impressive.
Motherboard can also have a bit of a effect but I suppose the higher-end Asus and otherwise boards would rely heavily on compatibility with high-speed memory kits including pushing these even more above spec compared to just toggling the XMP.
(Although results and build quality can differ even among the high-end X570's out now so exactly how big of a deal this is I suppose will vary and after a certain price range the returns and extra features are probably very diminishing anyway.)

Although I would probably use this more for reference for what could be possible and then dial in the values through the bios bit by bit, memory correction and overall testing for stability can be pretty complicated although going by user reports the 5700 GPU's and other Navi10 models are somehow exceptionally well suited for finding memory errors more so than testing software or running various demanding or RAM intensive games or applications ha ha.


Well it's like general overclocking I guess, there's just a bit of a variance for every hardware piece although I didn't know there was a difference between Samsung B on lower speed kits compared to high-speed ones and the price tag for Samsung dies and these lower timing modules are certainly a premium anyway even before hitting the upper 3000 speeds or higher so that was a surprise when doing some reading up on DDR4 and AMD Ryzen when building together this new computer system.

Micron E I think was the other runner up but both are mainly for a premium with additional overclock results over the existing XMP and since that seems to be more common the modules with these dies are price adjusted accordingly even against other 3200 - 3800 kits and 14 or 16 rated primary timing values.
(Samsung B in particular, makes early market DDR3 prices almost seem reasonable again. Almost.)


EDIT: Though I guess it shouldn't be a surprise that the lower rated kits gets the lower binned but still capable memory modules installed on them same as CPU or GPU or other computer hardware in general.
(Plus motherboard revisions stripping down or changing components around, thought that was no longer as big of a thing but it seems it still happens.)

Still a nice step up above official 2400 specs and timings at least but higher still isn't bad if it's doable.
(Especially with how AMD's hardware seems to really like high speed and low timings in particular for some really nice overall performance gains.)

 

Be aware that it has bugged RTT settings if you use all 4 slots !

(Select Memory Rank: 2 as a workaround)

---
It also seems to under-evaluate tCL on certain frequencies, for example it suggests on my memory CL 14 on 3733, and I know for sure my RAM is unable to do that while keeping voltage at safe levels (below 1.45)

I don't see any differences when using THAIPHOON imported XMP Report (aka, Manual mode), the results are the same as before, but now the "V" have been replaced with "PCB Revision" which seem to suggest different values than what comes from XMP import.

Overall, almost nothing changed, for me at least, same settings as before. (only CAD_BUS ClkDrv from 24 to 60... no idea what that does)

Here's my output, just in case anyone cares (Crucial Ballistix Extreme 16GB 3600 CL16, two kits - motherboard filled)



Observe the tDRWR = 9, the system doesn't boot if I set that to 8 as the calculator recommends. Not sure why ...
Second, my memory is at 1.43v*, works fine with that at 1900... 1.44 not needed (been like this for months, since I bought it)

With current settings, seems to work fine... it's been several hours since I set these values, played some games, did some work... all stable, no signs of problems.

---
*Hmm, which gets me thinking... maybe I can lower voltage even more ?
How low can I go and still maintain stability ?
Right now on 1.41v, and it booted fine, and I'm typing this... will see in a few hours if I got any crashes or other issues.

 

Hey guys. I am new to this forum and don’ t now much about overclocking RAM. I’ d really appreciate help. Thanks.

I just bought new PC with RAM overclocking potencial in my mind. I did OC GPU and CPU in the past but never RAM.

My Setup:

Ryzen 9 3900x
Asus Rog Strix RTX 2080 ti o11g
Asus x570 -e gaming
G. Skill Trident Neo 3600 CL16 (16 16 16 36) 2x16gb (it is B-die)
Seasonic Prime 850W

Here is a thing, I used DRAM calculator for Ryzen (v.1.7.1) imported XMP from Thaiphoon (in nanseconds) to get a correct info on the left side.

I went into bios and filled out everything that was available in my bios exactly as said in calculator. Although I made a mistake I didn’ t buy Crosshair motherboard because my bios is lacking DRAM Power supply options (current capability, power phase control ect.).

At first I started out with DOCP enabled and it didn’ t work. Then I read somewhere to load optimized defaults. Afterwards I tested safe present and fast present with recomended values from calculator. I also tried it out with max voltages in case low voltages were the problem.

Here is what happened:

In safe and fast present with recomended values my PC went in a multiple reboot (turning on and off for 4 times), than it went into safe mode and I had to press F1 to go back to bios.

When I went with higher voltages (max values in calculator), I got blue screen error and had to clear CMOS.

I really can not find a solution after searching for 15 hour straight. All forums all videos - nothing. People are just saying it works for them..

One more thing I also flashed latest bios in case it was corrupted

Am I missing something?
Is DRAM calculator for Ryzen wrong?
Is it my x570 -e gaming motherboard (it is missing some values), I think Crosshair has everything.
Is something wrong with my RAM sticks or I just lost silicon lottery? I think if it was a silicon lottery higher voltages would solve the problem, especially with safe present.

Here are some photos of a program calculations values. I also marked red all the values missing in my bios that I just was’ t able to fill out.

Thank you.

 

If possible monitor RAM temps, even more so with AIO.

 

I haven't gotten it to work either. I think its giving wrong RFC cause that's the one giving the biggest change it seems from stock 558 to 307.

 

No JJR support again, oh well

 

I've never had luck with this program I've tried every combo of ram timings and power options and I could never pass 3266 on my gskill 3200 bdie kit and r5 2600 I think it very cpu dependent on if your ram will overlook well or not my 2400g with same kit of ram will 3333mhz cl14 on both.

 

This version is the first time I've got stable results out of 4 dimms of 3600mhz b die. Kudos!

 

I really don't understand these craze of OC'ing stuff . All you get is bragging rights and thats;s about it.
OC videocard - 10 fps more
OC proc - I don't think you gain anything .
Oc ram - 3 - 5 fps more.
that's about it. ...
I done stuff like this in the era of Pentium 1 and Pentium 2 .... going DDR1 from 266 to 330 if I remember right. In that period of time it was more then worth it ... now I don;t really think it justified . Just my two cents

 

I did MEMbench and monitor temps with HWMonitor. I think temps are fine.

What do you mean “even more so with AIO”?

thank you

 

biggest change seems to be in the termination block, im gonna have to try those out to see if i can get 3800mhz instead of 3733 with 4 stick, otherwise my old settings are stable on the new agesa bios still

 

Why there's no A0 for Micron E-Die?

 

AMD's Ryzen CPU's and the infinity fabric speed makes it pretty useful, latency also matters more so hitting lower timings can still be useful even against the usual recommendations where the main thing is the clock speed and how that matters more than the timings and then there's the whole process with secondary timings and much more which I wouldn't be able to say much about, I can see the results though from a few of the comparisons or benchmarks available.

(This image is from Techspot.)





So from the reference DDR4 speeds of 2400 and CL15 primary timings at 126 FPS average which then when that's up to 3600 CL16 results in 135 FPS average that's nearly a 10% improvement in average framerate there for this game.

Push the infinity fabric speed higher (Even above the 1800 clocks.) with faster RAM and also retaining the lower timings and that's above a 20% increase for those top two results here.


Most of this besides the top performers here is gained just enabling the XMP and doing some light testing and comparisons checking stability and performance while ensuring voltage and such isn't under or overshooting if set to automatic instead of manual values but maximizing gains is pretty tricky although the results seem well worth it if it can be made stable.