DRAM calculator for Zen 3

[Makaveli]

View attachment 186445maybe not.

I trust Thaiphoon burner which clearly says its B Die.

And this kit is 2x8GB SR Thaiphoon burner is pulling data for one stick that is why its listed as 8 and not 16.

 

[dgianstefani]

If you can read it says right there - updated January 2021.

I trust Thaiphoon burner which clearly says its B Die.

View attachment 186449

And this kit is 2x8GB SR Thaiphoon burner is pulling data for the one stick that is why its listed as 8 and not 16.

View attachment 186448

Looks like Thaiphoon and Bdie finder in agreement then.

 

[Makaveli]

If you can read it says right there - updated January 2021.


Looks like Thaiphoon and Bdie finder in agreement then.

Yes i'm not aiming for 3800/4000 speeds right now just trying to get 3600 cl14 with the fast profile and Geardown off. However as I said there is some instability so I will have to load 3600 safe instead of fast to see if I can figure out what is causing it. The B die you are using should be higher quality but I should still be able to get good clocks with this.

 

[dgianstefani]

Yes i'm not aiming for 3800/4000 speeds right now just trying to get 3600 cl14 with the fast profile and Geardown off. However as I said there is some instability so I will have to load 3600 safe instead of fast to see if I can figure out what is causing it. The B die you are using should be higher quality but I should still be able to get good clocks with this.

At least some of that instability can be fixed by voltage, if you already have the optimal settings. Every kit out there is safe at 1.5v, and the better quality ones sit at 1.6/1.7 for years with no issues.

Only rule is to keep it under 40c, 45 being where you start getting a lot of errors. 35c and under preferably (which is what I have under normal use non stress), and 20c is "ideal" temperature for B die.

I'm probably going to upgrade the thermal pads under my neo's heatsink to 13w/mk ones and see if I can improve those temps by a few more degrees. I already have 2 40mm ram fans, but I think the pressure/pads aren't ideal stock. Fine for stock voltages, but I like removing every bottleneck I can.

I'm at 1.65V 24/7 so that's probably a bit more than the stock heatsinks can do under full stress. I don't have any instability, but I also don't run software that stresses my ram 24/7.

 

[thesmokingman]

Yes i'm not aiming for 3800/4000 speeds right now just trying to get 3600 cl14 with the fast profile and Geardown off. However as I said there is some instability so I will have to load 3600 safe instead of fast to see if I can figure out what is causing it. The B die you are using should be higher quality but I should still be able to get good clocks with this.

Why are you specific about disabling GDM? In reality it only needs to be diaabled when running odd numbered CAS, otherwise it's really hard to disable on its own on Zen cpu.

 

[Makaveli]

Why are you specific about disabling GDM? In reality it only needs to be diaabled when running odd numbered CAS, otherwise it's really hard to disable on its own on Zen cpu.

Based on what i've been reading so far doesn't it reduce performance?

 

[tabascosauz]

Based on what i've been reading so far doesn't it reduce performance?

If you care about 0.3ns in AIDA benchmark, yeah. Are you going to feel that difference in literally everything else you do with your computer?

The stability boost is pretty significant with it on, and if you want to run with GDM off at higher speeds or tighter timings you need to compensate with quite a bit of extra voltage or get extremely lucky with CADBUS on your board. 1.5V, let alone above 1.5, is not so easily done below 50C when one does not have a sandwich case that protects everything on the board from the heat of the GPU.

 

[thesmokingman]

Based on what i've been reading so far doesn't it reduce performance?

You'd be the only human in the universe that could sense the difference?

 

[Makaveli]

If you care about 0.3ns in AIDA benchmark, yeah. Are you going to feel that difference in literally everything else you do with your computer?

The stability boost is pretty significant with it on, and if you want to run with GDM off at higher speeds or tighter timings you need to compensate with quite a bit of extra voltage or get extremely lucky with CADBUS on your board. 1.5V, let alone above 1.5, is not so easily done below 50C when one does not have a sandwich case that protects everything on the board from the heat of the GPU.

For now max voltage will be 1.45v I will need to take another look at the secondary voltage settings again for that 3600 Fast profile and test. If that failures then 3600 Safe test.

I will also test with GDM on.

 

[thesmokingman]

Fyi, I ignore the "misc settings" in Dram Calc. There's no need to follow any of its recomendations in that box.

 

[Chomiq]

Yes i'm not aiming for 3800/4000 speeds right now just trying to get 3600 cl14 with the fast profile and Geardown off. However as I said there is some instability so I will have to load 3600 safe instead of fast to see if I can figure out what is causing it. The B die you are using should be higher quality but I should still be able to get good clocks with this.

View attachment 186452

That "overclocking potential" is total BS and doesn't mean anything as it's a fixed value for set Die type.

 

[BossferatuEs]

If you Google around you'll find lots of people not being able to get above 3200mhz on MSI boards anymore with new BIOS. I haven't seen MSI acknowledge it as an issue yet anywhere

im using my msi x570 a pro with IF at 2000 having relation 1:1:1, they dont know to config NB

 

[Radi_SVK]

Hi guys, can I also please kindly ask someone to look at my numbers? I am a bit desperate here. My story is a bit of a bummer.. Just today I have received my 4x8Gb Corsair Vengeance RGB Pro paired kit,(Micron B-die) cl16 , 3600MHz , 16-19-19-36 default. I have decided to replace the exact RAMs, except I had 2 different 2x8GB paired kits cl18 , 3600MHz ,cl18 18-22-22-42 default. One kit was same Micron B-die and the other was some inferior Nanya, def not B-die. So those 2 kits mixed I have tried hard to fine tune, wasnt really after OC, but just to improve timing to the max to beat that not so great 72ns latency. Due to them kits not being paired, they didnt even run at rated 1.35V together in first place. Had to use DRAM 1.4V and SoC 1.1V The best I managed was cl16 16-16-16-38 and that gave me 68ns at best. Not being happy I have decided to sell these and get same 32GB,but as paired kit this time, as I still like the RAM and they fit in budget. So I got my above mentioned 4x8GB paired kit cl16 16-19-19-36 and to my disbelief and shock the best I can do atm is cl16 16-17-17-36 ?!! how is that possible? and the latency is disappointing too as it did not improve, got worse actually. How comes the mixed non paired set was performing way better? Can you kindly suggest some settings of my new kit based on the screenshots? what am I missing? By the way, I am running my 5800X all cores OC @ 4.5GHz and voltages are tight..

 

[evilhf]

Acho que ganhei na loteria de silício (IF 2020mhz 4 sticks 4x8gb gskill royal/ allcores 4,87ghz)

 

[Taraquin]

Acho que ganhei na loteria de silício (IF 2020mhz 4 sticks 4x8gb gskill royal/ allcores 4,87ghz)View attachment 191631

Impressive!

 

[Flexarius]

@evilhf

cool!

could you take a picture with ZenTimings? It's very helpful to showing all your settings.

Thanks

https://zentimings.protonrom.com/

 

[JawnyS]

hello, i have 4 stick of hyper x 4000mhz cl19 kit, there is also a 2nd xmp profile at 3600 17-18-18-18
i can do 3600 mhz with 1.365V 100 stable with hcl mem test 600 %, but anything above 3600mhz will give me error no matter what ive tried many different settings, voltage etc. there is no temp sensor on these sticks so i dont know if im making error because of temps or not, i recently change my board for an asus x570 hero wifi, before i was using an x570 asrock steel legend, with the steel leging i manage to get 3933 cl 16-19-19-19 at 1.45, but i was getting usb disconnect all the time, but the ram was stable 600% as well, but i cant make it work on this board, anybody has any advice, my ram is b-die but its not the pcbp one as you can see on taiphoon burner

 

[JGalland]

Hi guys,

I'm running a F4-3600C16D-32GTZR kit from G.Skill on an Asrock X570 Taichi w/ Ryzen 5 5600X.

For overall performance, mainly in gaming, do you think I should go for higher frequency (to bump the infinity clock to 1900 Mhz) or tighter timings ?

 

[Alyjen]

Hi guys,

I'm running a F4-3600C16D-32GTZR kit from G.Skill on an Asrock X570 Taichi w/ Ryzen 5 5600X.

For overall performance, mainly in gaming, do you think I should go for higher frequency (to bump the infinity clock to 1900 Mhz) or tighter timings ?

and your current timings are? overall I think it's easier (and still fast) to aim for something like 3600CL14, it's doable, doesn't require high voltage, is 24/7 stable etc. if you want to have it faster then next step is 3800CL15 BUT this requires some serious tweaking due to gear down mode, and also higher voltage, 3800CL14 is usually above 1.5V and 4000CL16 seems easy for memory but it's very hard due to IF limitations

I run my set at 3733CL14 at 1.48V but if I had 4 dims and not only two I'd have to tweak it a bit more and most likely I'd drop the frequency a bit, not that you'll see it anywhere in real life applications

 

[JGalland]

and your current timings are? overall I think it's easier (and still fast) to aim for something like 3600CL14, it's doable, doesn't require high voltage, is 24/7 stable etc. if you want to have it faster then next step is 3800CL15 BUT this requires some serious tweaking due to gear down mode, and also higher voltage, 3800CL14 is usually above 1.5V and 4000CL16 seems easy for memory but it's very hard due to IF limitations

I run my set at 3733CL14 at 1.48V but if I had 4 dims and not only two I'd have to tweak it a bit more and most likely I'd drop the frequency a bit, not that you'll see it anywhere in real life applications

Right now I'm running very conservative timings :

16 - 16 - 16 - 16 - 34 - 50
with tfaw @ 28 and tRFC @ 258

The Taichi isn't the best for memory overclocking afaik. I believe with 1.45V I could achieve CL14 ; I just wonder what DIMM temperatures are acceptable ?

 

[Alyjen]

Right now I'm running very conservative timings :

16 - 16 - 16 - 16 - 34 - 50
with tfaw @ 28 and tRFC @ 258

The Taichi isn't the best for memory overclocking afaik. I believe with 1.45V I could achieve CL14 ; I just wonder what DIMM temperatures are acceptable ?

Temps depends on the memory die used, e.g. b-dies can take more abuse than others, but the rule of thumb is that if it's not crashing then temps are ok, and errors caused by temperatures are usually easy to spot as they only appear later in the tests, or after you heat up your set by other tests or gaming (like memtest is working if you run it, but won't if you run it after 30 minutes of prime/occt)

 

[JawnyS]

ive managed to get 4 stick of samsung b-die but the bcrc one at 3800, these are the timings i have achieve below from what i was at before 3600, its 100 % stable, im getting about 57 ns on aida with these timings. im using a 5900x and a x570 hero wifi, for some reason i cant seems to go higher than 3800, my system is really really slow when i go higher than 3800, on my x570 asrock steel legeng i was able to get 3933 stable with 4 stick but with much looser sub timings, im using 1.42 V on dram, anyone here has experience overclock these b-die -bcrc

 

[oobymach]

Is your VDDG on auto? I had trouble with mine on auto, setting it to manual 0.95v seems to have stabilized it for me. Hope this helps.

 

[Zach_01]

ive managed to get 4 stick of samsung b-die but the bcrc one at 3800, these are the timings i have achieve below from what i was at before 3600, its 100 % stable, im getting about 57 ns on aida with these timings. im using a 5900x and a x570 hero wifi, for some reason i cant seems to go higher than 3800, my system is really really slow when i go higher than 3800, on my x570 asrock steel legeng i was able to get 3933 stable with 4 stick but with much looser sub timings, im using 1.42 V on dram, anyone here has experience overclock these b-die -bcrc

Very few systems can run stably over 1900MHz. Even 1900MHz is not guaranteed.
At this point (you are) its better to try to tight timings as far as they can go. How far it will depend on the DRAM voltage you are willing to run your b-dies and also on sticks quality (ICs and PCB)
From what I see there a lot of them that can be improved and should let latency come closer to 50ns. For sure under 55ns is achievable.

But personally I wouldn't start giving advise before asking...
How did you determine the system's stability?

The MEMbench on "DRAM Calculator for Ryzen" will give you a quick (2~3mins max or less) idea about DRAM errors. If it passes without errors is a very good sign. But thats not enough. For 100% stability you need longer RAM bench and I'm sure other users here are more experienced to suggest.
Also, gaming on different demanding games is nice RAM test for at least 1~2hours.

Do you have temp sensors on those sticks? You can see it with ThaiphoonBurner software(click Read button) and if yes HWiNFO64 sensors mode will give you the ability to monitor (dimm) temp values (Current, Min, Max, Avg).

 

[JawnyS]

Very few systems can run stably over 1900MHz. Even 1900MHz is not guaranteed.
At this point (you are) its better to try to tight timings as far as they can go. How far it will depend on the DRAM voltage you are willing to run your b-dies and also on sticks quality (ICs and PCB)
From what I see there a lot of them that can be improved and should let latency come closer to 50ns. For sure under 55ns is achievable.

But personally I wouldn't start giving advise before asking...
How did you determine the system's stability?

The MEMbench on "DRAM Calculator for Ryzen" will give you a quick (2~3mins max or less) idea about DRAM errors. If it passes without errors is a very good sign. But thats not enough. For 100% stability you need longer RAM bench and I'm sure other users here are more experienced to suggest.
Also, gaming on different demanding games is nice RAM test for at least 1~2hours.

Do you have temp sensors on those sticks? You can see it with ThaiphoonBurner software(click Read button) and if yes HWiNFO64 sensors mode will give you the ability to monitor (dimm) temp values (Current, Min, Max, Avg).

i did hcl mem test all 24 threads 29 gb total for 600 % on each thread , ive been gaming warzone without crashing which is quite an accomplishment if you ask me, and there is no temps sensor on these sticks unfortunatly. i dont think the temperature is a problem, it could since 4 sticks doesnt give much space between them, but my case is very well ventilated, with 2 140 intake in front, 1 in the back and two on top quite close to the ram stick, so there is no warm stagnant air in the case thats for sure, all of them are corsair ml pro, and the two fans on top are phantek fan