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

[Mussels]

SoC shouldnt be 1.8v - either that's reading a different voltage or you've set your SoC *far* too high

 

[Taraquin]

Hi guys need help, I can't find VDD18 on my board, MSI Tomahawk X570, only VDD IP8 I can change value to 1.900, does any owners can tell, thx

Not all boards have access to VDD18. It us only important if you plan running above 1900fclk.

 

[Tatili]

Not all boards have access to VDD18. It us only important if you plan running above 1900fclk.

Yep, run above 1900fclk Whea19 free at my friend board MSI is my intention, he manage to run 3866 IF1933, but a lot Whea19 shown up instant due to Aida64 memory test!
Fixing CPU VDD IP8 voltage to 1.900v not help, found few dozen Whea19 in EventViewer!
We are looking to the primary VDD18 same as my Aorus board, let you bump it up over 2.000v as I was able to get rid of WHEA19 free!
Seems to me MSI boards doesn't have this option controller at all, have not found any info anywhere, thx!

Not all boards have access to VDD18. It us only important if you plan running above 1900fclk.

Thats VDD18 I'am talking about help me to get rid of thousands of WHEA19 seems like MSI X570 Tomahawk doesn't, but may help AORUS board owners !

 

[jesdals]

Sigh just found this link https://www.gigabyte.com/FileUpload/Global/multimedia/2/file/548/988.pdf after haveing the board for 2 years

 

[Mussels]

Sigh just found this link https://www.gigabyte.com/FileUpload/Global/multimedia/2/file/548/988.pdf after haveing the board for 2 years

That's actually got some really good general ryzen overclocking info in there, they state a lot of the max voltages for various settings that normally aren't well known

For a little extra boost in stability while overclocking we suggest you to try to change VCORE SOC up to 1.20-1.25 volts when using standard air or liquid cooling. If you use a PCIe 4.0 device try to avoid this adjustment on the VCORE SOC voltage. For CPU VDD18 you can adjust it up to 2.0 volts and for CPU VDDP up to +0.2 volts if this is required from the CPU you are testing. It wasn’t necessary for us. The PM_CLDO12, PM_1VSOC and the PM_1V8 you can adjust them up to 1.25V, 1.2V and 1.84V if your system is not stable without them at Auto

 

[Tatili]

Sigh just found this link https://www.gigabyte.com/FileUpload/Global/multimedia/2/file/548/988.pdf

That's actually got some really good general ryzen overclocking info in there, they state a lot of the max voltages for various settings that normally aren't well known

I fix Whea19 free guided from this line:https://www.gigabyte.com/FileUpload/Global/multimedia/2/file/518/939.pdf
"For a little extra boost in stability while overclocking we suggest you to change VCORE SOC up to 1.20-1.25 volts when using standard air or liquid cooling. For CPU VDD18, you can adjust it up to 2.1 volts and up to +0.2 volts for CPU VDDP if this is required from the CPU you are testing"

Sigh just found this link https://www.gigabyte.com/FileUpload/Global/multimedia/2/file/548/988.pdf after haveing the board for 2 years

Hi thx this is guide for X570 chipset, but I was able to fix Whea19 error with X470 guideline here:

https://www.gigabyte.com/FileUpload/Global/multimedia/2/file/518/939.pdf

"For CPU VDD18, you can adjust it up to 2.1v"
Also haveing this board 2 years ago

 

[Taraquin]

Yep, run above 1900fclk Whea19 free at my friend board MSI is my intention, he manage to run 3866 IF1933, but a lot Whea19 shown up instant due to Aida64 memory test!
Fixing CPU VDD IP8 voltage to 1.900v not help, found few dozen Whea19 in EventViewer!
We are looking to the primary VDD18 same as my Aorus board, let you bump it up over 2.000v as I was able to get rid of WHEA19 free!
Seems to me MSI boards doesn't have this option controller at all, have not found any info anywhere, thx!


Thats VDD18 I'am talking about help me to get rid of thousands of WHEA19 seems like MSI X570 Tomahawk doesn't, but may help AORUS board owners !

VDD18 2.0v or above is not something I would recommend, a bit risky, some report degredation. I ran 1.88v when using 2000fclk, but it drove temps up quite high. Usually raising SOC, IOD or VDDP can reduce WHEA19, and if the are few eliminate them.

 

[grammar_phreak]

Almost 2 years ago I got a 5950x for my Crosshair 8 motherboard. With single rank B-Die RAM I was able to get the Fclk to 2000 and the memory at 4000MT/s, but with dual-ranked RAM the most I got stable was Fclk to 1933mhz and the RAM at 3866MT/s. I tried the best I could with tweaking timings and latency.
The benchmark scores were actually higher with the dual-ranked memory running at 1900mhz and RAM at 3800MT/s.

 

[3x0]

The benchmark scores were actually higher with the dual-ranked memory running at 1900mhz and RAM at 3800MT/s.

Magic of Dual Rank

 

[Mussels]

Magic of Dual Rank

I'll never stop repeating to everyone how important 4 ranks is on Zen2/3 - it's such an uplift

Anyone seen their IF/RAM clocks changing in real time?

This is new, and i cant tell if its from the new BIOS, new chipset drivers, etc

 

[Kodehawa]

Anyone seen their IF/RAM clocks changing in real time?

This is new, and i cant tell if its from the new BIOS, new chipset drivers, etc

View attachment 286049View attachment 286050View attachment 286051

Yes, seeing that too.

I've seen it change on the CPU-Z window too, but it's pretty much random more than "in real time". Very hard to capture.

 

[izy]

I'll never stop repeating to everyone how important 4 ranks is on Zen2/3 - it's such an uplift

Anyone seen their IF/RAM clocks changing in real time?

This is new, and i cant tell if its from the new BIOS, new chipset drivers, etc

View attachment 286049View attachment 286050View attachment 286051

It does that because the BUS speed fluctuates and thats normal as far as i know if it doesnt go too low , you can try to disable spread spectrum in bios and it should fix that but on some Motherboards it didnt work for me disabling it and on some it did , not sure whats the catch

Edit: I just noticed (after seeing your post) with C-States enabled seems that it fluctuates more down to 98.50 , i am going to check BIOS for spread spectrum myself and see whats going on.

 

[Kodehawa]

Edit: I just noticed (after seeing your post) with C-States enabled seems that it fluctuates more down to 98.50 , i am going to check BIOS for spread spectrum myself and see whats going on.

Yeah, I saw it go down to 98 myself, but I couldn't get to picture it, changed inmediatly when I pressed the screen capture key.

 

[izy]

Yeah, I saw it go down to 98 myself, but I couldn't get to picture it, changed inmediatly when I pressed the screen capture key.

I tried with C-States on/off , SS on/off , setting the BUS manually to 100 (and all combinations) and it does the same thing in CPU-Z , fluctuates from 98-to 99 something but hwinfo shows 100 non fluctuating, maybe its something wrong with CPU-Z?

Edit: If i remember right , someone was saying that there are multiple spread spectrums on a motherboard and on some you cant turn off the spread spectrum that affects the BUS speed (this implicitly affects the RAM speed fluctuation) or it can be something related on how the motherboard is reacting to memory OC (i might be wrong on this).

Edit2: Here is a video, the clock seems to freeze during CPU-Z bench too now. (C-states on Auto , Spread Spectrum OFF , down to even 93mhz .. in hwinfo its rock solid 100mhz)

Edit3: Ok i got something ( if i remember right) , i think its related to Virtualization and Windows , to get a stable BUS clock you need to turn off Hyper-V in Windows or / and Virtualization in BIOS too. (i remember turning off Hyper-V in win10 was fixing this but i havent played with WIN11 , ive upgraded only a month ago and not sure what features of WIN11 wont work without Hyper-V , i am using vmware anyway)

Edit4: Seems that turning off Hyper-V didnt change anything (it used to work in WIN10) only disabling SVM (virtualization in bios) fixed it , if anyone knows what we can disable in WIN11 to keep the BUS clock stable with SVM enabled please let me know.
SVM off , SS disabled , C-States auto: (performance seems better, this makes me want to try and push my IF higher / CO lower and see if with SVM disabled it will work, sadly i need SVM )

 

[Kodehawa]

For me it doesn't go down under load. I'm seeing absolutely haywire values at idle tho, saw Uncore Frequency dip to 180 once, which doesn't happen on HWInfo lol

Maybe an issue on CPU-Z on there?

Also stuff like

 

[izy]

For me it doesn't go down under load. I'm seeing absolutely haywire values at idle tho, saw Uncore Frequency dip to 180 once, which doesn't happen on HWInfo lol

Maybe an issue on CPU-Z on there?

Also stuff like
View attachment 286081

In WIN11 it seems that you have to disable Hyper-V and Core Isolation to fix the BUS/DRAM frequency fluctuation when you have SVM on, for me after disabling them the BUS sits at 99.98Mhz (spread spectrum off in BIOS).

 

[Kodehawa]

In WIN11 it seems that you have to disable Hyper-V and Core Isolation to fix the BUS/DRAM frequency fluctuation when you have SVM on, for me after disabling them the BUS sits at 99.98Mhz (spread spectrum off in BIOS).

View attachment 286083

For me it only moves on CPU-Z, on HWInfo64 it doesn't (and I left it open for a while)

 

[izy]

View attachment 286084
View attachment 286085
For me it only moves on CPU-Z, on HWInfo64 it doesn't (and I left it open for a while)

It was the same for me , hwinfo 100 Mhz CPU-Z fluctuating, disable hyper-v and Core Isolation (if you are in WIN11) and that should fix the fluctuation in CPU-Z.
After doing that my benchmarks scores increased a bit.

 

[Taraquin]

2 things: always turn off power down mode (it will reduce perf and fluctuate fclk a lot) and spead spectrum if tuning as the latter can cause instability.

 

[sdip]

View attachment 272057

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

4x8

What's your 1.8v PLL set to in order to go that high? Can't stabilize 2000MHz FCLK

 

[freeagent]

What's your 1.8v PLL set to in order to go that high? Can't stabilize 2000MHz FCLK

Shouldn’t have to touch pll, it’s soc and vddgs

 

[Taraquin]

For 2000fclk you may need slightly higher VDD18 voltage to not get lower performance.

 

[sdip]

That's what I keep reading, but I get a flurry of error 0s and Bsod in Tm5 in a few seconds, even with the ram unsynced at a lower frequency. Tried bumping all voltages to 1.2v soc, 1.0 VDDP and ccd, 1.1 iod, 2.0v for the vdd1.8pll, but still fails and whea city

 

[Taraquin]

That's what I keep reading, but I get a flurry of error 0s and Bsod in Tm5 in a few seconds, even with the ram unsynced at a lower frequency. Tried bumping all voltages to 1.2v soc, 1.0 VDDP and ccd, 1.1 iod, 2.0v for the vdd1.8pll, but still fails and whea city

Error 0 is usually too low ram voltage. It is not due to SOC, VDDP etc usually

 

[Krit]

I had many RYZEN 3000/5000 CPUs usually they work stable at FCLK 1833/1867 that's average in my book.

Right now i'm using FCLK 1900 on RYZEN 5600 Zen 3 demans low voltages 0.980v soc, vddp 0.820v, vddg ccd 0.860v, vddg iod 0.900v (it's stable even at lower voltages i assume) Can post at FCLK 1933 but too lazy to make it stable.