What is your max infinity fabric on ryzen 5000? September 2021 edition

[FilipM]

Freshly tested yesterday, swapped the sticks around (A0-A2-A0-A2 -> A2-A0-A2-A0)

4x8

 

[Merritt1232005]

Probably not going to happen with 4x8, maybe take out a pair and try again

Not excited about that but it's worth a try to see.

 

[P4-630]

September 2021 edition

It's almost 2023....
Just sayin..

 

[Merritt1232005]

It's almost 2023....
Just sayin..

Oh wow, it is!

 

[The King]

Morning community,

I just updated my setup with a new memory kit and was working to adjust everything on my rig and happen to come across this post. I thought I might see if I can get some assistance/guidance on a good configuration for everyday use. I put everything back to default in bios and have only enabled XMP profiles.
Current PC Specs:
CPU: Ryzen 5600x
CPU Cooler: Artic Freezer II 240 AIO
HDD:
OS:Samsung 980 PRO 1TB Win 10 64
Spare HDD:970 EVO 500GB
Spare HDD: 3 SATA Mec drives
Memory: 4x8=32 GSKILL Trident Z RGB 4000mhz cl15 Kit=F4-4000C15Q-32GTZR
MB: MSI MAG x570 Tomahawk Wifi
GPU: Sapphire Nitro+ 5700xt
850w PSU
Bios Settings: All default setting except below.
XMP on set to Profile 1 DDR4 4000mhz 15-16-16-36
FCLK: 2000mhz

Everything boots and runs normally but getting WHEA errors after OCCT mem testing. Looking to adjust setting to see if I can get it stable. Kinda new in the OC arena. I read the setting in previous comments but nit sure I have them with my setup.
Thanks!

That is some really good bin Samsung B-die.

If its boots which it does you are more than halfway there to get it WHEA free.

Two voltages that need to be adjusted is VSOC and IOD.
FOR VSOC you maybe need between 1.2V upto 1.25V
IOD between 1.1V-1.15V

I would not run higher than those values on a daily setup. On one of my boards I need at least 1.21V VSOC and 1.15V IOD to be WHEA free at 3933 CL15.

 

[Cowboystrekk]

Morning community,

I just updated my setup with a new memory kit and was working to adjust everything on my rig and happen to come across this post. I thought I might see if I can get some assistance/guidance on a good configuration for everyday use. I put everything back to default in bios and have only enabled XMP profiles.
Current PC Specs:
CPU: Ryzen 5600x
CPU Cooler: Artic Freezer II 240 AIO
HDD:
OS:Samsung 980 PRO 1TB Win 10 64
Spare HDD:970 EVO 500GB
Spare HDD: 3 SATA Mec drives
Memory: 4x8=32 GSKILL Trident Z RGB 4000mhz cl15 Kit=F4-4000C15Q-32GTZR
MB: MSI MAG x570 Tomahawk Wifi
GPU: Sapphire Nitro+ 5700xt
850w PSU
Bios Settings: All default setting except below.
XMP on set to Profile 1 DDR4 4000mhz 15-16-16-36
FCLK: 2000mhz

Everything boots and runs normally but getting WHEA errors after OCCT mem testing. Looking to adjust setting to see if I can get it stable. Kinda new in the OC arena. I read the setting in previous comments but nit sure I have them with my setup.
Thanks!

Many WHEA19 or a few? If you get like 1 a minute minor voltage adjustment can fix it, if you get 10+ you should lower fclk. 3800/1900 is WHEA19 free for most.

 

[Merritt1232005]

Many WHEA19 or a few? If you get like 1 a minute minor voltage adjustment can fix it, if you get 10+ you should lower fclk. 3800/1900 is WHEA19 free for most.

Not sure because as soon as I seen the first error within 8 seconds of the testing I stopped it and turned it off. I can retest again to check it.

That is some really good bin Samsung B-die.

If its boots which it does you are more than halfway there to get it WHEA free.

Two voltages that need to be adjusted is VSOC and IOD.
FOR VSOC you maybe need between 1.2V upto 1.25V
IOD between 1.1V-1.15V

I would not run higher than those values on a daily setup. On one of my boards I need at least 1.21V VSOC and 1.15V IOD to be WHEA free at 3933 CL15.
View attachment 272144

I'll try the voltage change to see if that clears it up.

Not sure because as soon as I seen the first error within 8 seconds of the testing I stopped it and turned it off. I can retest again to check it.


I'll try the voltage change to see if that clears it up.

Ok I just changed the voltage on both and seems be be running without WHEA errors for 55 mins now using OCCT memory tester on AVX2.
Current voltage settings in bios: Everything else is default.
SOC Volts: 1.25
IOD Volts: 1.175
Promising but I'll keep the test running and if it stays stable I should be good with that except I'm going to try and lower the IOD volts closer to 1.150 Volts. I was still getting WHEA erros @ 1.150 Volts so I raised it a bit and seems to have done the trick.

So at this rate if everything checks out then that means I'm running everything 1-1 ratio with IF @ 2000mhz with all 4 ram slots loaded. Seems harder for most to hit this target with the 5600x. Would this because the memory kit or maybe Silicon lottery?

 

[The King]

Not sure because as soon as I seen the first error within 8 seconds of the testing I stopped it and turned it off. I can retest again to check it.


I'll try the voltage change to see if that clears it up.


Ok I just changed the voltage on both and seems be be running without WHEA errors for 55 mins now using OCCT memory tester on AVX2.
Current voltage settings in bios: Everything else is default.
SOC Volts: 1.25
IOD Volts: 1.175
Promising but I'll keep the test running and if it stays stable I should be good with that except I'm going to try and lower the IOD volts closer to 1.150 Volts. I was still getting WHEA erros @ 1.150 Volts so I raised it a bit and seems to have done the trick.

So at this rate if everything checks out then that means I'm running everything 1-1 ratio with IF @ 2000mhz with all 4 ram slots loaded. Seems harder for most to hit this target with the 5600x. Would this because the memory kit or maybe Silicon lottery?

Consider also running 3933 FCLK 1966. That should require less IOD than 1.175V which is on the high side.

 

[Merritt1232005]

Consider also running 3933 FCLK 1966. That should require less IOD than 1.175V which is on the high side.

I was able to modify the FCLK to 1967Mhz and 3933 for Ram and dialed back on the IOD voltage to 1.150 volts and everything is stable without WHEA errors.

Appreciate the guidance!

 

[Kapone33]

Updated to 3600 14-15-15-36

 

[izy]

Not sure because as soon as I seen the first error within 8 seconds of the testing I stopped it and turned it off. I can retest again to check it.


I'll try the voltage change to see if that clears it up.


Ok I just changed the voltage on both and seems be be running without WHEA errors for 55 mins now using OCCT memory tester on AVX2.
Current voltage settings in bios: Everything else is default.
SOC Volts: 1.25
IOD Volts: 1.175
Promising but I'll keep the test running and if it stays stable I should be good with that except I'm going to try and lower the IOD volts closer to 1.150 Volts. I was still getting WHEA erros @ 1.150 Volts so I raised it a bit and seems to have done the trick.

So at this rate if everything checks out then that means I'm running everything 1-1 ratio with IF @ 2000mhz with all 4 ram slots loaded. Seems harder for most to hit this target with the 5600x. Would this because the memory kit or maybe Silicon lottery?

My 1933MHZ OC was stable without WHEA in OCCT AVX2 and some other stress tests but when i tested with OCCT SSE i was getting many WHEAs so maybe you should check SSE instead.

 

[freeagent]

IOD @ 1.15v

 

[3x0]

SOC 1.25v is a bit concerning too

 

[Merritt1232005]

My 1933MHZ OC was stable without WHEA in OCCT AVX2 and some other stress tests but when i tested with OCCT SSE i was getting many WHEAs so maybe you should check SSE instead.

Just passed 10min mark on the SSE testing and no errors so I'll let that run for a bit. Fingers crossed.

 

[izy]

Just passed 10min mark on the SSE testing and no errors so I'll let that run for a bit. Fingers crossed.

Sounds good , mine had wheas in sse after like 5 min but no whaes in avx2, i cant get it stable whatever voltages im throwing at it on 1900+.
Edit: i mean its stable and works just fine but cant get rid of wheas.

 

[Merritt1232005]

Sounds good , mine had wheas in sse after like 5 min but no whaes in avx2, i cant get it stable whatever voltages im throwing at it on 1900+.
Edit: i mean its stable and works just fine but cant get rid of wheas.

What is your setup? Specifically what motherboard, cpu and memory specs are you running?

 

[izy]

This is what i have stable ( i tried with the RAM from my other computer and no luck)

What is your setup? Specifically what motherboard, cpu and memory specs are you running?

Edit: Im running 4 sticks 32GB in general but when testing some OC i only keep 2 sticks.

 

[Merritt1232005]

This is what i have stable ( i tried with the RAM from my other computer and no luck)
View attachment 272249


Edit: Im running 4 sticks 32GB in general but when testing some OC i only keep 2 sticks.

What's the specs on the memory?

 

[izy]

What's the specs on the memory?

Here you go: https://files.fm/thumb_show.php?i=xmrsq92at
This RAM has 1802 XMP ^_^

Ram is fine , found on Reddit .. someone did 4400MHZ

 

[The King]

What's the specs on the memory?

His RAM is Mircon Rev. E and yours is Samsung B-die.

Micron Ram can generally do higher MCLK with less voltages.
Samsung B-die can do lower CL with tighter sub-timings. Some good bins like yours may also do high MCLK.

Two completely different IC manufacturers that behave differently.
ProcODT and other settings are different between these two sticks so don't copy settings between the two.

 

[Merritt1232005]

Here you go: https://files.fm/thumb_show.php?i=xmrsq92at
This RAM has 1802 XMP ^_^

Ram is fine , found on Reddit .. someone did 4400MHZ

Do you know if the voltage you're MB is putting to the socket is accurate? I know it differs by manufacturer but if you set the SOC to 1.150 Volts in bios ect it could be reading it wrong with the sensors and actually be close to 1.500 or higher in some case's but again that just what I've been made aware of. You're milage may vary on that and the only way to know for sure is with a volt meter to the back of the MB socket/VRM area while it's running and under a load. If your not comfortable with doing that I would advise against it cause you could brick your system if you are not up to the task. Outside of tweaking the settings one way or the other it will likely come down to how good the parts (CPU,Ram) are with your build. I'm by no means an expert but I'm aware that all the above factors in to you achieving your goals of higher clocks and how your system functions with the commands you give it.

 

[Cowboystrekk]

This is what i have stable ( i tried with the RAM from my other computer and no luck)
View attachment 272249


Edit: Im running 4 sticks 32GB in general but when testing some OC i only keep 2 sticks.

Nice tuning! If I were to change something I would try lowering voltages, timings are very good for rev E. I would try SOC 0.95-1.05, IOD 0.9-1.0, CCD and VDDP 0.8-0.9. Try highest suggested first, then lower. It will make IO-die consume way less and make cores boost better. Do you have curve optimizer active?

 

[izy]

Nice tuning! If I were to change something I would try lowering voltages, timings are very good for rev E. I would try SOC 0.95-1.05, IOD 0.9-1.0, CCD and VDDP 0.8-0.9. Try highest suggested first, then lower. It will make IO-die consume way less and make cores boost better. Do you have curve optimizer active?

Everything is on auto other than the RAM voltage, my FCLK is stuck at 1867, works till 2000 or something but with WHEAs. Thanks
Edit:
IOD 0.9 no problem (can i go even lower?), maybe i should give it a try for 1900 with lower IOD , only tried with higher IODs.
Do i need to check RAM too after lowering IOD?
CPU is boosting 4650 - 4675 in OCCT AVX2.

UPDATE: I could get VDDP , CCD and IOD at 0,9v , seems stable in OCCT extreme setting. I tried to get them at 0.85 but im getting WHEAs and i didnt try anything between , SOC doesnt want to get any lower , i left it on AUTO.

 

[Cowboystrekk]

Everything is on auto other than the RAM voltage, my FCLK is stuck at 1867, works till 2000 or something but with WHEAs. Thanks
Edit:
IOD 0.9 no problem (can i go even lower?), maybe i should give it a try for 1900 with lower IOD , only tried with higher IODs.
Do i need to check RAM too after lowering IOD?
CPU is boosting 4650 - 4675 in OCCT AVX2.

UPDATE: I could get VDDP , CCD and IOD at 0,9v , seems stable in OCCT extreme setting. I tried to get them at 0.85 but im getting WHEAs and i didnt try anything between , SOC doesnt want to get any lower , i left it on AUTO.
View attachment 272491

Run aida to see if performance is the same. Too low IOD usually doesn't affect stability, but it will reduce performance. Example on my setup:
4000 IOD aida ns:
1.04v 53.7ns
1.03v 54ns
1.02v 55-56ns
1.01v 57-59ns

100% stable, just slower.

Kerp 1867 is you get WHEA19 above There is little benefit above 1900 since you must raise VDD18 which can't be done on all MBs. I used to run 4000 (I never get WHEA19), but it reqUIRES 1.88V VDD18, which increases CPU-temp by 5-10C vs 1.6v VDD18 I can run at 3800. Performance with 4000 is only about 2% better so not worth it.

Ypu could try lowering WR/RTP btw: keep WR twice of RTP.
14/7
12/6
10/5
May work and give slight boost. RFC and RC may also go lower. RC maybe 52-54, RFC 520, 528 or 536. My previos rev E kit could do 522 at 3733.

 

[izy]

Run aida to see if performance is the same. Too low IOD usually doesn't affect stability, but it will reduce performance. Example on my setup:
4000 IOD aida ns:
1.04v 53.7ns
1.03v 54ns
1.02v 55-56ns
1.01v 57-59ns

100% stable, just slower.

Kerp 1867 is you get WHEA19 above There is little benefit above 1900 since you must raise VDD18 which can't be done on all MBs. I used to run 4000 (I never get WHEA19), but it reqUIRES 1.88V VDD18, which increases CPU-temp by 5-10C vs 1.6v VDD18 I can run at 3800. Performance with 4000 is only about 2% better so not worth it.

Ypu could try lowering WR/RTP btw: keep WR twice of RTP.
14/7
12/6
10/5
May work and give slight boost. RFC and RC may also go lower. RC maybe 52-54, RFC 520, 528 or 536. My previos rev E kit could do 522 at 3733.

Ill check AIDA later , i tested CB single and im getting better results while the multi are the same.
Yes ill stick to 1867 , my RAM couldnt do stable CL14 at 1900 anyway (have to do CL15 with some timings increased) and i prefer a stable system for when im working on this system, im only trying to push its limits for fun but ofc if i find something stable i wont throw away the extra performance