• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Overclocking B-die on Ryzen 5 2600

Joined
Jan 8, 2020
Messages
834 (0.43/day)
Location
Maryland, USA
Processor Ryzen 5 5600X
Motherboard MSI MPG X570S Carbon Max Wifi
Cooling CPU: bequiet! Dark Rock 4. Case fans: 2x bequiet Silent Wings 3 140s, 2x Silent Wings 3 120s
Memory 2 x 8 GB Patriot Viper Steel DDR4-4400 C19
Video Card(s) Sapphire NITRO+ RX 5700 XT
Storage 2TB Mushkin Pilot-E M.2, 1 TB SK Hynix P31 M.2, 1 TB Inland Professional, 500 GB Samsung 860 Evo
Display(s) MSI Optix MAG271CQR 1440p 144Hz, MSI Optix MAG241C 1080p 144Hz
Case Lian Li Lancool III
Audio Device(s) Philips SHP9500, V-Moda BoomPro, Sybasonic Better Connectivity USB DAC/Amp
Power Supply EVGA SuperNOVA G3 80+ Gold 750W
Mouse Glorious Model D Wireless
Keyboard Custom Qwertykeys Navy QK80: Sarokeys Strawberry Wine switches, GMK CYL DMG3 keycaps
System specs are up date but for convenience, the relevant components are:
  • Ryzen 5 2600
  • MSI B450 Gaming Pro Carbon AC (Latest BIOS - updated today)
  • 2x8 GB Patriot Viper Steel DDR4-4400 C19
Assuming my IMC isn't absolute garbage, what can I reasonably expect to hit?

This GitHub guide seems to imply I should be able to do 3400, but wondering if I should go for something lower like 3200 with tighter timings. Haven't really messed around with it much.

Any tips right off the bat would be appreciated.
 
use aida 64 memory test benchmark stock and then do a mild oc and go from there.
 
Start with finding your infinity max - heres mine
1661636668443.png

Going above 1.5v on mem is no problem but you should consider active cooling (a fan on the mem)

Btw you should not expect to be able to run 4 dimms on that chipset, but have set several setups at CL14 with MSI Mortar and Tomahawk boards
 
Hi,
This kit is rated for 4400c19 at 1.45v and 4266 at 1.35v so this is pretty wild

 
System specs are up date but for convenience, the relevant components are:
  • Ryzen 5 2600
  • MSI B450 Gaming Pro Carbon AC (Latest BIOS - updated today)
  • 2x8 GB Patriot Viper Steel DDR4-4400 C19
Assuming my IMC isn't absolute garbage, what can I reasonably expect to hit?

This GitHub guide seems to imply I should be able to do 3400, but wondering if I should go for something lower like 3200 with tighter timings. Haven't really messed around with it much.

Any tips right off the bat would be appreciated.

3200CL14 should be a walk in the park at 1.35V and 3600CL14 should be doable if your 2600 is feeling it? Though for 3600CL14 I reckon closer to 1.5V than 1.45V, 4400CL19 is not exactly a top B-die bin and they behave as such.

Binning is a bit all over the board for the CL19 Vipers but 4400 was their top dog for a while and it's a little more consistent. The two 4400CL19 kits I've had are about a year apart and perform about the same.

Your board shouldn't be an obstacle. It's an old 4-layer 4DIMM but you have single rank DIMMs. Performance for Ryzen 1000/2000 hasn't been changed for like 3 years now, BIOS shouldn't make any difference.

The Patriots seem to be a little picky when it comes to running 4/4/16/4/8/12 tRRDS/RRDL/FAW/WTRS/WTRL/WR. Needed more VDIMM than I was willing to give it. tRFC I've had them down to about 131-133ns before they started crapping out, but you might get luckier.

iirc Ryzen 2000 doesn't have much in the way of CLDOs. But I'm sure @ShrimpBrime -retired knows how best to push them beyond 3200

4333cl16 timings.png
 
Start with finding your infinity max - heres mine
View attachment 259635
Going above 1.5v on mem is no problem but you should consider active cooling (a fan on the mem)

Btw you should not expect to be able to run 4 dimms on that chipset, but have set several setups at CL14 with MSI Mortar and Tomahawk boards
I'm on Zen+: FCLK is locked to MCLK.

3200CL14 should be a walk in the park at 1.35V and 3600CL14 should be doable if your 2600 is feeling it? Though for 3600CL14 I reckon closer to 1.5V than 1.45V, 4400CL19 is not exactly a top B-die bin and they behave as such.

Binning is a bit all over the board for the CL19 Vipers but 4400 was their top dog for a while and it's a little more consistent. The two 4400CL19 kits I've had are about a year apart and perform about the same.

Your board shouldn't be an obstacle. It's an old 4-layer 4DIMM but you have single rank DIMMs. Performance for Ryzen 1000/2000 hasn't been changed for like 3 years now, BIOS shouldn't make any difference.

The Patriots seem to be a little picky when it comes to running 4/4/16/4/8/12 tRRDS/RRDL/FAW/WTRS/WTRL/WR. Needed more VDIMM than I was willing to give it. tRFC I've had them down to about 131-133ns before they started crapping out, but you might get luckier.

View attachment 259640
Thanks for the insight. Should I just leave the secondary and tertiary timings on Auto to begin? Guide recommends "Set primary timings to 16-20-20-40 (tCL-tRCD-tRP-tRAS) and tCWL to 16" and increasing frequency til I can't boot into Windows.
1661638915522.png


And the point about disabling CPU overclock: obviously you don't want another potential source of instability while trying to fine tune the memory. Would it be better to OC CPU first and find where that's stable, disable it to find memory, then re-enable? Sorry if I'm overthinking or asking dumb questions.
 
I'm on Zen+: FCLK is locked to MCLK.

Thanks for the insight. Should I just leave the secondary and tertiary timings on Auto to begin? Guide recommends "Set primary timings to 16-20-20-40 (tCL-tRCD-tRP-tRAS) and tCWL to 16" and increasing frequency til I can't boot into Windows.

And the point about disabling CPU overclock: obviously you don't want another potential source of instability while trying to fine tune the memory. Would it be better to OC CPU first and find where that's stable, disable it to find memory, then re-enable? Sorry if I'm overthinking or asking dumb questions.

No, that advice is for generic troubleshooting. You already know it's B-die. I don't see why 3200 14-14-14 shouldn't be easily doable at 1.35-1.4V for a kit at this bin. At 3600 you might have to stop down to 14-15-15 or 14-16-16 to lessen the VDIMM needs. 1.1V VSOC should be a good place to start for 3200.

Yeah just leave the secondaries for now and just work on booting with tCL/tRCD/tRP/tRAS. Leave CPU at stock, come back to it later.

If you're feeling really cautious, just start at 3200 16-16-16 1.35V.
 
Hi,
Those timings are more for micron and maybe hynix
B-die is a different beast.
 
No, that advice is for generic troubleshooting. You already know it's B-die. I don't see why 3200 14-14-14 shouldn't be easily doable at 1.35-1.4V for a kit at this bin. At 3600 you might have to stop down to 14-15-15 or 14-16-16 to lessen the VDIMM needs. 1.1V VSOC should be a good place to start for 3200.

Yeah just leave the secondaries for now and just work on booting with tCL/tRCD/tRP/tRAS. Leave CPU at stock, come back to it later.

If you're feeling really cautious, just start at 3200 16-16-16 1.35V.
3200 14-14-14-28 at 1.35v yielded errors in TM5. Bump up to 1.4v? I'd be comfortable daily-ing 1.45v if needed. Looks like maybe I need to set LLC for SOC voltage too
1661642655976.png
 

Attachments

  • 1661642649307.png
    1661642649307.png
    50.6 KB · Views: 127
Well, if you plan to run them at 1.4V, why bother testing stability at 1.35V ?
Just pick a voltage that won't cause memory to overheat in your environment and it should be fine.
Also, for when secondaries come into play :
PS. That tRC is just... WOW.
 
Well, if you plan to run them at 1.4V, why bother testing stability at 1.35V ?
Just pick a voltage that won't cause memory to overheat in your environment and it should be fine.
Also, for when secondaries come into play :
PS. That tRC is just... WOW.
Figured if I could run 3200C14 at 1.35v then I'd just do that but appears I need more voltage to be stable. Yeah, was watching that yesterday, interesting stuff for sure. My tRC? It's set to auto lol
 
3200 14-14-14-28 at 1.35v yielded errors in TM5. Bump up to 1.4v? I'd be comfortable daily-ing 1.45v if needed. Looks like maybe I need to set LLC for SOC voltage too
View attachment 259649

Just turn on Geardown and stop trying to run pure 1T. Or run 2T. Or bump VDIMM quite a bit to where it's stable if you're bent on pure 1T. Trying to run pure 1T on VDIMM alone really isn't sustainable above 3200 anyway.

For the secondaries you can just copy some of mine in the screenshot on for now. Start with 42 for tRC. 4/6/16/4/12/12.

For tRFC try to get down around 140ns for 3200 (though last I recall it's really hard to run tighter tRFC at really low speeds)

Reous tRFC list v21.png



You really don't need LLC for SOC...you don't have an APU. Just set VSOC higher accordingly. If Fabric isn't crashing then leave it alone
 
Last edited:
Run TaiphoonBurner. If the PCB revision is A2, then you'll be running 2T command rate.
 
Run TaiphoonBurner. If the PCB revision is A2, then you'll be running 2T command rate.
Isn't better at 1T w/ GDM enabled?

I'm curious of how is the DRAM performance between those 2 configurations.

I'm not even considering a 2T w/o GDM enabled against 1T w/ GDM and never really tested
 
Run TaiphoonBurner. If the PCB revision is A2, then you'll be running 2T command rate.

you have a link to this info? Being A2/B1 hasn't forced me into 2T only before

Seems pretty hard to find A0 nowadays. I have only one kit (4133CL19 Viper) and it clocks like trash at the upper end (seems to be characteristic of A0). Maybe they're right that A0 runs tighter at lower speeds

Isn't better at 1T w/ GDM enabled?

I'm curious of how is the DRAM performance between those 2 configurations.

I'm not even considering a 2T w/o GDM enabled against 1T w/ GDM and never really tested

GDM and 2T are sometimes similar, sometimes GDM is a bit faster. Sometimes 2T is a bit more stable. Not much point nitpicking between those two, pure 1T is the only real outlier
 
Almost all Samsung B-die should be able to do 3200 CL14 with 1.35V with GDM enabled 1T.

Ocing RAM on ZEN 2 and ZEN 3 can vary abit.

I think 3600/3733 should be possible maybe 3800. I recommend getting GMD enabled 1 T stable first if you are new to OCing RAM then
you can try for GDM disable 1T without or with AddrCmdSetup 56 which will make your life easier. This kit can do 4000/4066 CL16 on my B450 with ZEN 3.
PV4400 3600 CL14 15 15 15 TM5 Pass 1.45V crop.jpg
PV4400 3600 CL16 TT TM5 Pass Low Voltages crop 2.jpg
 
Almost all Samsung B-die should be able to do 3200 CL14 with 1.35V with GDM enabled 1T.

Ocing RAM on ZEN 2 and ZEN 3 can vary abit.

I think 3600/3733 should be possible maybe 3800. I recommend getting GMD enabled 1 T stable first if you are new to OCing RAM then
you can try for GDM disable 1T without or with AddrCmdSetup 56 which will make your life easier. This kit can do 4000/4066 CL16 on my B450 with ZEN 3.
View attachment 259700View attachment 259702
Thanks for stopping by. I'm on Zen+ so the higher frequencies aren't as obtainable. Currently trying to tighten stuff at 3200 CL14 and will see how much higher I can go
 
Thanks for stopping by. I'm on Zen+ so the higher frequencies aren't as obtainable. Currently trying to tighten stuff at 3200 CL14 and will see how much higher I can go
My bad made the classic Ryzen 2XXX is ZEN 2 error. :slap:

My 1700X which is ZEN1 had no problem doing + 3600 on B450. It could even do 3933 single channel.
3600 CL14 RCD 17 crop 1700X.jpg
t
 
Samsung B-die, or Micron?
 
you have a link to this info? Being A2/B1 hasn't forced me into 2T only before

Seems pretty hard to find A0 nowadays. I have only one kit (4133CL19 Viper) and it clocks like trash at the upper end (seems to be characteristic of A0). Maybe they're right that A0 runs tighter at lower speeds



GDM and 2T are sometimes similar, sometimes GDM is a bit faster. Sometimes 2T is a bit more stable. Not much point nitpicking between those two, pure 1T is the only real outlier
Yeah, OCN I do quite a bit of reading there. I supposed I could dig up a bunch of links. Buildzoid may have mentioned it in a video as well.

Sometimes 2T is a bit more stable. "" yes, that's true for ddr4, ddr3 ect ect. But the penalty was much higher on earlier ddr revisions.

And last, speaking from experience. My 4267mhz kit B-Die is also A2. Seems to clock real easy on 2T.

And what are we talking about for the OP???

Oh, Zen+ memory controller on top of that?

Yeah it ain't gonna get anywhere close to 4000mhz stable even with your kits.
 
Thanks for stopping by. I'm on Zen+ so the higher frequencies aren't as obtainable. Currently trying to tighten stuff at 3200 CL14 and will see how much higher I can go
I could do 3466 CL14 (also B-die) when I still had a 1600X on my B350 motherboard.

My improvement suggestion is the following at 3200CL14:
tRCDRD => 15
tRAS => 30
tRC => 44
tRRDS => 4
tRRDL => 6
tRDRDSCL => 4
tWRWRSCL => 4
tFAW => 16
tWR => 12
tCKE => 1
tRFC + tRFC2 + tRFC4 => 320

If this works fine, then increase the memory speed to 3466MHz.
 
Last edited:
Isn't better at 1T w/ GDM enabled?

I'm curious of how is the DRAM performance between those 2 configurations.

I'm not even considering a 2T w/o GDM enabled against 1T w/ GDM and never really tested
Gear Down Mode. If you want to run an odd number Cas latency. Otherwise the board will default to an even number 18 opposed to 17, otherwise it does nothing.

Don't tell me you guys SERIOUSLY use shit like AIDA64 for "ACCURATE" comparisons.

If that's the case, I'll back away slowly and cautiously.
 
3200 14-14-14-28 at 1.35v yielded errors in TM5. Bump up to 1.4v? I'd be comfortable daily-ing 1.45v if needed. Looks like maybe I need to set LLC for SOC voltage too
View attachment 259649
Try to loosen tRCDRD i found that going under 16 gave problems at my settings
 
Don't tell me you guys SERIOUSLY use shit like AIDA64 for "ACCURATE" comparisons.
I don’t have a dog In this fight I just OC my 3600 cL16 to 3800 CL18 and literally just changing the CAS. I refuse to go down the timings rabbit hole.
Anyways AIDA64 is the “go to” test. What is your alternative?
 
Back
Top