Ryzen DRAM Calculator by 1usmus

[bimmerdriver]

First thing to try if you're using xmp profile is manually set the voltage to what is on the ram sticker, sometimes xmp tries to use much less than you need. If your mobo boosts the voltage set it lower until you get it as close as you can to the sticker value. Mine is 1.45 in the picture below you can see how my mobo boosts the value I set. Yours is 2 values listed on corsairs own website as 1.2v and 1.35v so idfk I hear corsair isn't the greatest ram.

Nanya ram is micron I think so setting to micron should work. Also yes set to what thaiphoon burner lists. Best option is to return and get samsung b-die, I use Patriot Viper ram which is samsung b-die and not overpriced like most. I would recommend getting a faster kit than you need, like 3800mhz or more just in case you can clock it that high or somewhere in between.

oobymach is right, you need to manually set the Ram voltage to 1.40-1.45v in the bios.

I just read about your ram, if this doesn't work try a different set, something with E-die or older B-Die like previously recommended. Good luck!

Thanks for the very helpful posts.

When I originally enabled the XMP profile, I did not increase the DRAM voltage. I thought it was done as part of enabling the profile. So it's not surprising that it didn't work. Live and learn.

I increased the voltage to 1.35v (I set it to 1.35v and the MB actually set it to 1.38v. I didn't have to adjust the value myself.) and rebooted, then I reenabled the XMP profile, but at reduced timing. It rebooted so I kept increasing the speed and eventually got to 3200. The computer ran for several hours with no obvious problems, just a few degrees warmer. Later, I came back to it and the screen was black and it would not post. I left it powered off for a while, then tried again. It rebooted. I tried this a few times and it's pretty consistent. Unless the system is cool, this memory will not post, so I disabled the XMP profile.

I contacted Corsair. They told me the memory is not compatible with a Ryzen, which is strange since it's on the QVL for the MB with the picasso processor. Maybe I could get it to work with a higher DRAM voltage, but since Corsair said it's not supported, I will return it for a kit that is compatible.

My processor officially supports 2900 and the MB has up to 3466 on the QVL. Is 3200 a reasonable objective?

 

[purecain]

at least now you can get the modules with the best IC's for your cpu. I look forward to seeing your results. I hear E-Die is quite good.
Good luck with it !

 

[bimmerdriver]

at least now you can get the modules with the best IC's for your cpu. I look forward to seeing your results. I hear E-Die is quite good.
Good luck with it !

I'm back. The Gigabyte B450 AORUS M was on sale, so I exchanged the B540M DS3H for it. The B450 AORUS M isn't a high-end MB, but it's better than the B540M DS3H, especially for the same price. I found G.Skill Flare X 2X16GB 3200 CL14 locally (no shipping) at a decent price, so I picked it up. I know it's overkill for this CPU and MB, but at least it won't be the limiting factor...

I enabled the XMP profile and set the DRAM voltage to 1.35 v. The system runs fine, but it occasionally crashes when it's not being used. The only indication of a crash is a black screen and the Windows reliability history log entry has a message like, "The previous system shutdown at 11:25:56 PM on ‎2020-‎12-‎14 was unexpected."

Aside from enabling XMP and changing the voltage, the rest of the BIOS settings are default. I noticed that Power Down Enable is set to Auto. Should I try disabling it? Should I try increasing the DRAM voltage?

Looking at the other BIOS settings, according to the BIOS, tRFC is 312, but according to ZenTimings, it's 374. The other BIOS settings are consistent with ZenTimings. Is there an explanation for the discrepancy?

I tested the memory using MEMbench in DRAM Calculator for Ryzen. Should I be using something else to ensure the memory is working properly?

Thanks in advance for your suggestions.

 

[Tomorrow]

Black screen crash may be many things.

Power Down should always be disabled. It just adds latency for a miniscule power saving. Not worth it.
Yes you could try increasing the DRAM voltage. 1.5v is perfectly fine for daily but i suggest you try with 1.40-1.45v first.

tRFC is odd indeed. Infact all tRFC, tRFC2 and tRFC4 do not match between BIOS and Zentimings. My opinion is that Zentimings is correct. Not sure why BIOS is unable to display proper numbers for it tho. I have the same problem with a different kit and board.

The best memory test i have found is TestMem5: https://testmem.tz.ru/tm5.rar along with cfg file: https://pastebin.com/puPcPCAk (select download, remove .txt extension and place in the same folder as TM5.exe

This finds errors that other tests miss. Also suggested are OCCT and Prime95 tho those are primarily CPU tests.

 

[Taraquin]

I have the Gigabyte B450m DS3H and it runs 3733cl15 with further improves subs on Micron rev E without problems at 1.43V. 3600cl14 also works fine. 3733cl14 works at 1.5V, but that is beyond my comfort zone.

I'm back. The Gigabyte B450 AORUS M was on sale, so I exchanged the B540M DS3H for it. The B450 AORUS M isn't a high-end MB, but it's better than the B540M DS3H, especially for the same price. I found G.Skill Flare X 2X16GB 3200 CL14 locally (no shipping) at a decent price, so I picked it up. I know it's overkill for this CPU and MB, but at least it won't be the limiting factor...

I enabled the XMP profile and set the DRAM voltage to 1.35 v. The system runs fine, but it occasionally crashes when it's not being used. The only indication of a crash is a black screen and the Windows reliability history log entry has a message like, "The previous system shutdown at 11:25:56 PM on ‎2020-‎12-‎14 was unexpected."

Aside from enabling XMP and changing the voltage, the rest of the BIOS settings are default. I noticed that Power Down Enable is set to Auto. Should I try disabling it? Should I try increasing the DRAM voltage?

Looking at the other BIOS settings, according to the BIOS, tRFC is 312, but according to ZenTimings, it's 374. The other BIOS settings are consistent with ZenTimings. Is there an explanation for the discrepancy?

I tested the memory using MEMbench in DRAM Calculator for Ryzen. Should I be using something else to ensure the memory is working properly?

Thanks in advance for your suggestions.

Using A2-B2-slots? Dual rank us sometimes a pain, use latest bios and try dram calc safe at 3600 and try higher if it works.

 

[bimmerdriver]

Black screen crash may be many things.

Power Down should always be disabled. It just adds latency for a miniscule power saving. Not worth it.
Yes you could try increasing the DRAM voltage. 1.5v is perfectly fine for daily but i suggest you try with 1.40-1.45v first.

tRFC is odd indeed. Infact all tRFC, tRFC2 and tRFC4 do not match between BIOS and Zentimings. My opinion is that Zentimings is correct. Not sure why BIOS is unable to display proper numbers for it tho. I have the same problem with a different kit and board.

The best memory test i have found is TestMem5: https://testmem.tz.ru/tm5.rar along with cfg file: https://pastebin.com/puPcPCAk (select download, remove .txt extension and place in the same folder as TM5.exe

This finds errors that other tests miss. Also suggested are OCCT and Prime95 tho those are primarily CPU tests.

WRT the black screen crash, it only happened with XMP @1.35v DRAM voltage, not with defaults, but agreed, could be caused by something else.

I will send a question to gigabyte about the difference between bios settings and ZenTimings / Ryzen DRAM Calculator.

Thanks for the suggestions for memory tests. I will give them a try as soon as I get a chance.

 

[Zach_01]

I'm back. The Gigabyte B450 AORUS M was on sale, so I exchanged the B540M DS3H for it. The B450 AORUS M isn't a high-end MB, but it's better than the B540M DS3H, especially for the same price. I found G.Skill Flare X 2X16GB 3200 CL14 locally (no shipping) at a decent price, so I picked it up. I know it's overkill for this CPU and MB, but at least it won't be the limiting factor...

I enabled the XMP profile and set the DRAM voltage to 1.35 v. The system runs fine, but it occasionally crashes when it's not being used. The only indication of a crash is a black screen and the Windows reliability history log entry has a message like, "The previous system shutdown at 11:25:56 PM on ‎2020-‎12-‎14 was unexpected."

Aside from enabling XMP and changing the voltage, the rest of the BIOS settings are default. I noticed that Power Down Enable is set to Auto. Should I try disabling it? Should I try increasing the DRAM voltage?

Looking at the other BIOS settings, according to the BIOS, tRFC is 312, but according to ZenTimings, it's 374. The other BIOS settings are consistent with ZenTimings. Is there an explanation for the discrepancy?

I tested the memory using MEMbench in DRAM Calculator for Ryzen. Should I be using something else to ensure the memory is working properly?

Thanks in advance for your suggestions.

You should disable Power Down for DRAM.
Also if you havent touch anything else but enabling XMP profile and DRAM voltage then tRFCs would be on AUTO. What you see in BIOS (as 312) is the default (non XMP) setting and not the actual setting.

Here:



Its one of the JEDEC profiles and not the XMP

 

[bimmerdriver]

I have the Gigabyte B450m DS3H and it runs 3733cl15 with further improves subs on Micron rev E without problems at 1.43V. 3600cl14 also works fine. 3733cl14 works at 1.5V, but that is beyond my comfort zone.

That's impressive. What processor and memory are you running?

Using A2-B2-slots? Dual rank us sometimes a pain, use latest bios and try dram calc safe at 3600 and try higher if it works.

I installed the memory in slots 1 and 2, according to the manual.

 

[Taraquin]

That's impressive. What processor and memory are you running?

I installed the memory in slots 1 and 2, according to the manual.

View attachment 180808
View attachment 180809
View attachment 180807

Ryzen 3600 and Crucial Ballistix 3000cl15 2x8.

 

[bimmerdriver]

I'm using version 1.73 of DRAM Calculator, which has the parameter DRAM PCB revision. I'm not clear what to use for this parameter. According to Thaiphoon Burner, the Revision / Raw Card is 0000h / B1 (8 layers). This doesn't correspond to any of the settings for DRAM PCB revision, so I guess it's not the right information.

I attached a photo of the module. This seems to be consistent with other photos I've seen of the A2 module type. Is that the correct type to use?

 

[nvoid1]

The Micron 16GB B-die are a bit of a let down. I got 4x16GB and am thinking of returning them and getting 4x8GB E-die. Anyone know if the next DRAM calculator update will have settings for >3466Mhz for the 16GB B-die, or am I stuck at this freq and timings?

 

[Tomorrow]

I'm using version 1.73 of DRAM Calculator, which has the parameter DRAM PCB revision. I'm not clear what to use for this parameter. According to Thaiphoon Burner, the Revision / Raw Card is 0000h / B1 (8 layers). This doesn't correspond to any of the settings for DRAM PCB revision, so I guess it's not the right information.

I attached a photo of the module. This seems to be consistent with other photos I've seen of the A2 module type. Is that the correct type to use?

Appears to be A2 indeed:

 

[bimmerdriver]

Merry Christmas everyone. I hope Santa was good to you.

Yesterday, I enabled the XMP profile in the BIOS (3200 CL14 @1.35V). I didn't make any other changes, except disabling Powerdown Mode. I attached screen captures of ZenTimings and the BIOS. I noticed the BIOS shows different timings for tRDWR and tWRRD. Is there a reason for that? ZenTimings doesn't show different values.

After I made the changes, I ran membench multiple times and I also ran prime95 / Large FFTs multiple times, long enough to get the temps up. I did not observe any stability problems or errors. Unfortunately, the problem I reported above of black screens when the computer is idle is back. It ran all night but crashed around 5:30 AM (over 6 hours of idle) and again this morning (around an hour of idle).

Since the system is stable when it's being used, I'm thinking this problem is being caused by a dynamic voltage being reduced too far when the system is idle. The BIOS and SIV refer to CPU Vcore, CPU VDDP and VCORE SOC. I recorded these voltages using SIV. CPU Vcore frequently drops down to 0.42V and I noticed it even went as low as 0.39V. I also noticed that CPU VDDP occasionally drops to 0V. I'm wondering if that's possible or if it's a bug in SIV. I enabled OC using EasyTune and it increased VRCORE SOC by 0.114V. It didn't change Vcore. I will leave the system running with this setting to see if it makes any difference. Does anyone have a suggestion about this?

Thanks in advance.

 

[tabascosauz]

Merry Christmas everyone. I hope Santa was good to you.

Yesterday, I enabled the XMP profile in the BIOS (3200 CL14 @1.35V). I didn't make any other changes, except disabling Powerdown Mode. I attached screen captures of ZenTimings and the BIOS. I noticed the BIOS shows different timings for tRDWR and tWRRD. Is there a reason for that? ZenTimings doesn't show different values.

After I made the changes, I ran membench multiple times and I also ran prime95 / Large FFTs multiple times, long enough to get the temps up. I did not observe any stability problems or errors. Unfortunately, the problem I reported above of black screens when the computer is idle is back. It ran all night but crashed around 5:30 AM (over 6 hours of idle) and again this morning (around an hour of idle).

Since the system is stable when it's being used, I'm thinking this problem is being caused by a dynamic voltage being reduced too far when the system is idle. The BIOS and SIV refer to CPU Vcore, CPU VDDP and VCORE SOC. I recorded these voltages using SIV. CPU Vcore frequently drops down to 0.42V and I noticed it even went as low as 0.39V. I also noticed that CPU VDDP occasionally drops to 0V. I'm wondering if that's possible or if it's a bug in SIV. I enabled OC using EasyTune and it increased VRCORE SOC by 0.114V. It didn't change Vcore. I will leave the system running with this setting to see if it makes any difference. Does anyone have a suggestion about this?

Thanks in advance.

Membench is not a stability tester. The included HCI.exe runs only like 1 instance and verifies only enough stability to get through the test. No point running Prime95 LargeFFT before you've even done stability testing.

Use TM5 anta777 extreme, Karhu, or HCI (with multiple instances to fill up all RAM on the free version).

XMP profile only sets primary timings. All the rest are determined by the board on the fly, and can be affected by the choice of board and firmware.

Stop trying to change BIOS settings through Gigabyte utilities. Do it in the BIOS.

For idle crashing, try disabling CPU C-states, or set Typical Current Idle as opposed to Low Current Idle. Or find a newer/older BIOS if that fails. No reason to up VSOC at 3200 speeds.

 

[Zach_01]

Stop trying to change BIOS settings through Gigabyte utilities. Do it in the BIOS.

I agree...

 

[bimmerdriver]

Membench is not a stability tester. The included HCI.exe runs only like 1 instance and verifies only enough stability to get through the test. No point running Prime95 LargeFFT before you've even done stability testing.

Use TM5 anta777 extreme, Karhu, or HCI (with multiple instances to fill up all RAM on the free version).

XMP profile only sets primary timings. All the rest are determined by the board on the fly, and can be affected by the choice of board and firmware.

Stop trying to change BIOS settings through Gigabyte utilities. Do it in the BIOS.

For idle crashing, try disabling CPU C-states, or set Typical Current Idle as opposed to Low Current Idle. Or find a newer/older BIOS if that fails. No reason to up VSOC at 3200 speeds.

With XMP enabled (no other changes), I ran TM5 using regular settings without any errors. I ran it again using anta777 settings and there were a couple of errors.

Yesterday, I was reading about the power plan due to it being mentioned in discussions about idle crashing. I had been using Windows High Performance, because I don't care about power consumption. I found this thread on Reddit, Is the Ryzen Balanced power plan still relevant for 2nd generation Ryzen? "tl;dr: use balanced or ryzen balanced for Ryzen, 2nd Gen Ryzen, Threadripper, etc. It's fine." The chipset driver includes the Ryzen Balanced power plan, so I enabled it and set the slider to best performance with Sleep: Never. (Even with Sleep: Never, the power plan has a setting Allow hybrid sleep: on. Any idea what that is? Should I disable it?) Interestingly, the documentation for Ryzen Master says to use the Windows High Performance power plan. It seems like AMD is all over the place with this.

To be sure there wasn't some other problem, I restored the BIOS to defaults (XMP disabled) and ran TM5 using anta777 settings and there were no errors. I reenabled the XMP profile and set Power Supply Idle Control to Typical Current Idle and left it overnight. I was hoping it would make a difference, but it crashed again. Like the previous night, it lasted around 6-7 hours before crashing.

The idle crashing doesn't happen with XMP disabled, so it appears to be related to the memory settings, but there are various threads about idle crashing being related to things including power plans, C states and voltages. When TM5 is running, Vcore is around the DRAM voltage (1.35v), Vcore SOC is around 1.2 and VDDP is around 0.92V. When the system is idle, Vcore is all over the place, including dropping way down, so I can't help but wonder if it's related to the idle crashing. Wondering if pursue this from the memory settings first, since there are a few errors or from the voltage / power settings? If memory settings should I back off to 2933 safe (since the 3400G is rated for up to 2933) or try 3200 safe?

Thanks for your help.

 

[tabascosauz]

With XMP enabled (no other changes), I ran TM5 using regular settings without any errors. I ran it again using anta777 settings and there were a couple of errors.

Yesterday, I was reading about the power plan due to it being mentioned in discussions about idle crashing. I had been using Windows High Performance, because I don't care about power consumption. I found this thread on Reddit, Is the Ryzen Balanced power plan still relevant for 2nd generation Ryzen? "tl;dr: use balanced or ryzen balanced for Ryzen, 2nd Gen Ryzen, Threadripper, etc. It's fine." The chipset driver includes the Ryzen Balanced power plan, so I enabled it and set the slider to best performance with Sleep: Never. (Even with Sleep: Never, the power plan has a setting Allow hybrid sleep: on. Any idea what that is? Should I disable it?) Interestingly, the documentation for Ryzen Master says to use the Windows High Performance power plan. It seems like AMD is all over the place with this.

To be sure there wasn't some other problem, I restored the BIOS to defaults (XMP disabled) and ran TM5 using anta777 settings and there were no errors. I reenabled the XMP profile and set Power Supply Idle Control to Typical Current Idle and left it overnight. I was hoping it would make a difference, but it crashed again. Like the previous night, it lasted around 6-7 hours before crashing.

The idle crashing doesn't happen with XMP disabled, so it appears to be related to the memory settings, but there are various threads about idle crashing being related to things including power plans, C states and voltages. When TM5 is running, Vcore is around the DRAM voltage (1.35v), Vcore is around 1.2 and VDDP is around 0.92V. When the system is idle, Vcore is all over the place, including dropping way down, so I can't help but wonder if it's related to the idle crashing. Wondering if pursue this from the memory settings first, since there are a few errors or from the voltage / power settings? If memory settings should I back off to 2933 safe (since the 3400G is rated for up to 2933) or try 3200 safe?

Thanks for your help.

Lower your VSOC. 1.27V is obscenely high regardless of Ryzen generation. Try testing again with VSOC at 1.1V. Shouldn't even need that much for 3200, but I'm not familiar with Pinnacle CPUs.

If that doesn't solve the rebooting, step it down to 2933 and try the same timings.

As for errors in TM5, not enough DRAM voltage for the settings you chose. Up it a notch or two and test again.

 

[bimmerdriver]

Lower your VSOC. 1.27V is obscenely high regardless of Ryzen generation. Try testing again with VSOC at 1.1V. Shouldn't even need that much for 3200, but I'm not familiar with Pinnacle CPUs.

If that doesn't solve the rebooting, step it down to 2933 and try the same timings.

As for errors in TM5, not enough DRAM voltage for the settings you chose. Up it a notch or two and test again.

Thank you for your reply. I really appreciate your help.

I corrected my previous post, changed the reference from Vcore to Vcore SOC, which was around 1.24V while running TM5.

After the previous post, I used DRAM calculator with the A2 PCB revision to make safe settings for 2933. Loaded them into the bios and rebooted. It wasn't happy with the settings and downgraded them. I will try again, starting from the XMP timings to see if that makes any difference.

Right now, Vcore SOC is 1.16V. The only voltage setting that I've touched is the dram.

UPDATE...

Finally, some progress. I tried entering settings for 2933 based on XMP and A2 safe. The BIOS wouldn't accept either of them. Possibly because I was using 1.35V? Not sure.

I reset to defaults, then went back to the XMP profile at 3200 with the DRAM voltage bumped up to 1.36. Then ran a full pass of TM5 and there were no errors. That's a first.

The only other changes are that power down mode set to disabled and Power Supply Idle Control set to Typical Current Idle.

I'm using the AMD Ryzen Balanced power plan with sleep and hybrid sleep disabled. I will leave it like this to see if it lasts overnight.

I attached the current settings shown from ryzen master and a graph of the voltages while TM5 was running, showing when it stopped running. V DRAM is 1.38 (set to 1.36), VCORE SOC is 1.116, VDDP is 0.924 and Vcore drops down to 0.708.

I also attached a screen capture of SIV when TM5 started. VCORE SOC was in the range of 1.224 to 1.248, then dropped to 1.116 a couple of seconds before Vcore went up.

I attached screen captures of the safe settings based on imported XMP and A2 for 2933 and 3200. I was not able to boot with any of these settings. The BIOS always reset them to SPD / 2133.

If the computer runs overnight, I will try and tighten up the timings tomorrow.

 

[bimmerdriver]

System ran all night, was still running this morning, a first to run that long with XMP enabled. Left it idle for a few more hours, then it crashed, so not out of the woods yet.

Increasing the DRAM voltage definitely seems to have improved the stability under load. However, I'm not clear if overclocking the DRAM is the cause of the idle crashes or if something else is causing the crashes, which are only visible because of overclocking the DRAM. Should the next step be to increase DRAM voltage another notch or to try something else, such as disabling C-states? Or should I make more changes to the memory timings, since a few are slightly tighter than the SAFE settings.

 

[Paman]

Hi Guys,

I got myself some Corsair LPX 32Gb Modules, they use Micron B-Dies.

I tried DRAM Calculator, but it cant even tell me some timings with 3600Mhz.
Right now these modules are running stable at 4133Mhz with 1,4V

As you can see the timings are giving me mixed feelings.
Unfortunately the DRAM Calculator cant help me to figure out how to set the timings correctly.
Right now I just loaded the XMP profile, pushed the voltage to 1.4V and increased frequency to 4133Mhz.
The board is a Gigabyte B550 Aorus Pro AX.

Any idea how to proceed from here?

Thanks

 

[Zach_01]

Get the latest version of ZenTimings and show us what you got now

ZenTimings

ZenTimings is a simple and lightweight app for monitoring memory timings on Ryzen platform.

zentimings.protonrom.com

Also it would be nice to run AIDA64 memory & cache benchmark and see the memory scores.
I believe your memory subsystem's clocks is out of sync right now and produce a latency penalty.

 

[Paman]

Hello Zach_01, thanks for your reply.

 

[tabascosauz]

1. VSOC way too frickin high. All Renoir needs is 1.2V or lower, certainly for a pretty mundane speed like 4133. I only need 1.15v to maintain IF stability at 4000, up to 1.2v at 4400.

2. It looks like your FCLK is still in sync as it should be below DDR4-4400 so no worries there.

3. Go into BIOS and turn TSME off, then run AIDA again. I have a B550I AX, iirc its under AMD CBS/DDR4 controller/Security. TSME incurs an automatic ~5ns latency penalty.

4. You're not running 1.4v the way you think you are. Just open up HWInfo and look at your DRAM volts. The B550I AX overvolts VDIMM by 0.05-0.06V over what you set in BIOS.

5. Rev.B is too new for DRAM Calc, but it should be an excellent overclocker. Something tells me you don't actually need 1.46V for 4133/18.

 

[Zach_01]

Yeah he didnt state his system so I assume he had a desktop CPU thats why I also assume that FCLK, UCLK would be out of sync at MEMCLK 4133.
Which brings us to: @Paman you should fill your profile with system specs like most we have.

 

[Tomorrow]

4750G is a desktop CPU. Although yes it's restricted to OEM's it can still be bought in DIY market. And yes it needs to have the security features disabled to achieve optimal performance. Tho it is slower in (CPU limited) games than Zen 2 (direct comparisons are 3700X and 3800X). But it can run higher IF and memory clocks and lower latency once properly configured.
I know becvause i had one for testing. I managed to achieve 60,6ns latency with Hynix J-Die. With either Micron or Samsung B-die i would have gotten under 60ns (thats generally Zen 3 territory).

The calculator itselt needs a major update to support new chips. But im guessing 1usmus is busy wiith CTR program instead.
I suggest watching Buildzoid's videos instead. He's been messing around with Micron B-Die a lot recently.