XMP\voltage question

[Niku-Sama]

OK i am stumped by this one. and i am not sure where it goes so i'm sticking it in general for the time being

I got my b650e Taichi and 7800x3D going and i am testing a few things out and i noticed something.
When i change my XMP profile after a bios update (from initial release to 1.18) it changes alot of voltage options.

However

Before the bios update it would only change about 3 or so, mainly VDDP and IO something (name escapes me at the moment)
Both versions supported running the XMP profile out of the box of 6400 32-39-39-102 @ 1.4v with no issue

my issue comes from when running super pi 2.1, 32M run before the bios update with 6400 XMP on i would get up to about 51C in a nekkid case
After update about 30 minutes later with 6400 XMP i get to 62C, same case

i feel like its setting voltages for things it doesn't need, what are your thoughts?
10C+ between the two seems a bit excessive

 

[EarthDog]

Are you sure it's only with XMP profile/XMP profile is doing that?? Did you try with it disabled and see if the behavior repeats?

after a bios update (from initial release to 1.18)

There can be a lot of differences from launch day bios to the latest.....I don't believe the launch day bios for that board fully supported that X3D chip. If you look on the website, it likely shows 7800X3D "bootable" with the board's release bios, but, actual verified support of 7000 X3D series doesn't come until March 2023 (when the chip was released)/what you updated to, 1.18. So, I'd venture to guess it's normal and how your CPU should be behaving in the first place using a/the latest BIOS that fully supports the processor.

EDIT: Moved to AMD mobos.

 

[Woomack]

Typical G.Skill DDR5-6400 CL32 1.40V temperature under longer high load is around 60-65°C. If you had 51°C, then the RAM load wasn't high, or there is really good airflow.
There is a chance that the motherboard on the new BIOS is bumping other voltages related to IMC and PMIC, but most settings affect the CPU, not RAM.

 

[EarthDog]

I thought this was about CPU temperature... never monitored RAM temps, lol.

Still, moving from a BIOS that doesn't fully support the CPU to one that does, could be a lot going on behind the scenes.

 

[Niku-Sama]

yes this is about CPU. ram is cool to the touch, its not hitting it that hard in super pi

And also yes, when i reset the bios to defaults the voltages all to auto and run it its maxing out at about 51C
when i enable XMP and thats the only thing i am physically changing, the change list when i go to save and exit is much longer than just timings and it lists the voltages it changes too.

before the bios update though, i would enable XMP and the temp wouldnt change but the ram speeds would. Ram voltage and i think VDDP IO.
i'm going to track down the voltages that change on the new bios the best i can tonight and if i feel up to it, flash it back to the initial and see what they are there.

so i'm not the only one that thinks its strange to get 10C on the cpu hotter after only changing the XMP profile after a bios update?

 

[EarthDog]

so i'm not the only one that thinks its strange to get 10C on the cpu hotter after only changing the XMP profile after a bios update?

I don't think it's strange considering the move from lack of support to supported. In theory, it's behaving like it should now. You just don't like the outcome of higher temps (nor do I blame you, note, lol).

The name of the game is to use the least amount of voltage for stability... time to play the game!

 

[Niku-Sama]

well considering what i was getting before the bios update i'll grab the voltage settings changes from the initial bios when xmp is enabled, and apply them to the updated bios. enable xmp reset voltages and go from there. It seems i'm already pretty lucky to run 6400 out of the box it seems, maybe i finally wont the lottery.
i'm not ready to overclock just yet, i still have some things to get for this build but nothing major. still sourcing a larger, all copper, AIO and hopefully after i move i can go back to my big case. the silent base is ok just a little... cramped

i would run the initial bios but its got the long post bug others have reported on AM5

edit:

ok it took a few hours but i pinned down the voltages between bios versions. it doesn't help that between the two they also moved and relabeled them

in short:
initial launch bios (1.07) i can enable XMP and set all vdd p\g\io\misc voltages back to auto and boot and run super pi
latest bios (1.18) i can enable XMP and set all vdd p\g\io to default, misc to 1.15 (from 1.12) and boot and run super pi

ram is at its xmp value of 1.4v, as are some others (vddio, vddq), vpp is 1.8

for what ever reason in ver 1.18 when i run vdd misc at what it wants to default at , 1.12v, it gives me garbage and hard locks on swirly dots for windows boot. 1.15v is ok.
doing it this way did drop my temps back to about where they were yesterday, i did see a few times up to 53 but its warmer in the house today too. it maintained about 4.8ghz the whole time

mind you this isnt super taxing on the CPU, its only running at about 10% load but i thought it was a good start to test for major manufacturer defects, figured they'd show up when i would run this.
right now its downloading 3dmark, a free stand alone non steam version so its a bit out of date but it'll work... i should really buy it next time it goes on major sale

 

[Woomack]

I thought you meant RAM temp as these CPUs run in idle at 45-50°C on 240AIO/large air coolers, not to mention in tests. Maybe you had a readings from unused cores as Spi is a single-threaded software. If you load even one core in Spi 32M then this core should have more than 51°C. Any higher load CPU test on Ryzen 7600 gives me 80°C+ on 240 AIO cooler, so I wonder what cooling do you have that it keeps the CPU at ~5x°C under load
Btw. SOC voltage usually bumps temps the most and at 6400 it's probably much higher than at stock/auto. You can set all voltages manually or the same value for SOC/IMC/VDD/VDDQ. This is where it's better to set it manually and lower some of these voltages. On the other hand, 6400 doesn't work in most cases. It seems stable but randomly crashes, so maybe ASRock checked what's required and in the new BIOS the motherboard sets higher voltages to keep stability.

 

[EarthDog]

i would run the initial bios

That bios does not fully support your chip. You need to run the latest so it does and you get the most out of that cpu. Why it's a consideration, I'm not sure.

 

[Niku-Sama]

I thought you meant RAM temp as these CPUs run in idle at 45-50°C on 240AIO/large air coolers, not to mention in tests. Maybe you had a readings from unused cores as Spi is a single-threaded software. If you load even one core in Spi 32M then this core should have more than 51°C. Any higher load CPU test on Ryzen 7600 gives me 80°C+ on 240 AIO cooler, so I wonder what cooling do you have that it keeps the CPU at ~5x°C under load
Btw. SOC voltage usually bumps temps the most and at 6400 it's probably much higher than at stock/auto. You can set all voltages manually or the same value for SOC/IMC/VDD/VDDQ. This is where it's better to set it manually and lower some of these voltages. On the other hand, 6400 doesn't work in most cases. It seems stable but randomly crashes, so maybe ASRock checked what's required and in the new BIOS the motherboard sets higher voltages to keep stability.

Yea frankly i'm surprised its working at 6400 my self. i only got this kit to kind of future proof my self a bit for when other, maybe faster cpus come down the line. i figured i'd be running at 6000 or 6200. I'll check later to see what auto is telling my other voltages to run at.

also i'm using a Fractal Design Celsius S36 with 2 extra ML140's on the front pushing, also these tests were run with a totally open case with no GPU and i would say it was probably about 66F in the room at the time of testing. you can see it here, there isnt much in it. I wanted to throughly test the system before buttoning it up because frankly the PURE BASE 500 is a decent case it is a bit small for some of these components. i'm sure once its not nekkid any more it'll be quite toasty

I ran 3dmark last night on IGP and a cooler "modified" 2070 and it got plenty warm but it still didnt crest 70c. again still an open case

That bios does not fully support your chip. You need to run the latest so it does and you get the most out of that cpu. Why it's a consideration, I'm not sure.

While that may be true it was working and boosting properly and gave me similar results between the two, about a 5'40" for 32m.

 

[Niku-Sama]

hate to double post but...

AMD Ryzen 7 7800X3D CPU Burns Up, Kills Itself & The ASUS X670 Motherboard

There are reports of AMD's Ryzen 7 7800X3D CPUs burning up and killing the ASUS X670 motherboard with it from the AMD subreddit.

wccftech.com

i think what i found and this are related, maybe i'm justified in researching it and manually adjusting voltages down.
i havent stressed the system yet, its been downloading games for days so when i get done moving i'll have something to do while i wait for the ISP to do their install thing.

just to make sure, what are the default voltages for this thing?

 

[Woomack]

I wasn't testing the X3D CPUs, but regular Ryzen 7k isn't much different regarding overclocking, and it doesn't require higher voltages to be overclocked. So if anyone burns them, then they probably ask for trouble pushing voltages way too high to see that 1% performance gain. When you manually OC these new CPUs, then it's actually recommended to undervolt them to keep lower temps, below the throttling point, and for the higher boost clock.

Either way, when I see redditor, a youtube channel or another Instagram model in hardware news/reports then, I don't take it seriously.

 

[mackerel]

The burning Ryzen 7000 CPUs have been investigated and a fix via AGESA has apparently been passed to mobo makers. Root cause seems to be high SOC voltages set by mobos when running high speed ram, and they will be capping it to 1.3v with the update. Does this affect high speed ram at all? I'm still not in DDR5 era yet.

AMD Issues Follow-Up On Ryzen Burnout, New BIOSes Limit SoC Voltages

Update your BIOS now.

www.tomshardware.com

 

[Zerileous]

I saw a video discussing this . I think the concern was high temps in the IO die caused by enabling EXPO. Hopefully this BIOS fix helps and it's strange that the IO die doesn't appear to have the same thermal protections as the cores.

 

[Niku-Sama]

well from what i've been reading when EXPO is enabled it causes other voltages to be adjusted too. everyone is still talking about just EXPO though, XMP does the same.
in fact it looks like thats what my concern was initially when starting this thread.
the IO die i think is mainly powered by the SOC voltage so yes maybe it cant handle the temps when its getting juiced up. mine wanted to run at 1.3 but i turned it down to 1.25 and i need to try and see if i can go lower. the other voltages i adjusted i can set to auto and they will run the stock V and not revert but it wont let me do that on this board with SOC and something else.

I've been busy moving right now so the only thing i've been doing with this system is downloading to play games offline when i can finally spend the night at my new place i'll have something to do other than unpacking.

i might look into this issue a little more on my own and see what i can come up with. if i find something extraordinary then i'll hotspot my phone to post but i doubt that'll happen

but i guess i did find this issue before they did...

 

[Woomack]

When you enable EXPO (or XMP), then 3-4 other voltages are adjusted, but it's mainly because motherboards set it this way. Depending on the motherboard, voltages are slightly different, but when you pass 6000, then motherboards usually force 1:1:1 voltages on AMD mobos, so VDD/VDDQ/SOC and also higher misc and one more voltage (I don't remember exactly now as I'm at work). In short, EXPO/XMP for 6000 CL30 or any higher memory is 1.40-1.45V, so all other voltages are 1.40-1.45V too.
If someone is overclocking manually, then don't set too high voltages and will leave SOC or other voltages at 1.20-1.30V. However, if it's really the only reason why X3D CPUs are having these issues, then I wonder why it wasn't even checked on the test stage (AMD internal tests or motherboard tests). I feel like we are once again beta testers for manufacturers. These things shouldn't even happen as 1.3V+ SOC is nothing really high, and we could only assume that if 1.5V+ is possible and not marked as unsafe, then someone actually tested it.

I didn't want to wait for the X3D CPU some months ago, so I have Ryzen [email protected] in my gaming PC and 7950X in my test rig. Somehow I'm glad I didn't spend extra money on my gaming PC, and I'm not really missing the 3D cache for tests. I'm not saying that X3D CPUs are bad, as the efficiency in games is great. I just don't feel "the pressure" of having one, even though I too often spend money on things that are sometimes a total waste of money or even stupid.

 

[Zerileous]

Is this new behavior for the 7000 series or DDR5? Did XMP always do this on Intel? My experience with DOCP/XMP on Ryzen DDR4 based systems was that it only changes the memory voltage and SOC voltage is independent. I can see how a bunch of people just enabled it not even considering it will be increasing voltages to the CPU itself.

 

[EarthDog]

I recall this happening in generations past, yes. Typically when those profiles are enabled, the board raises SOC or System Agent voltage depending on the kit and speed. The problem here is the AEGESA and AMD(?) seems to be adding to much hence the AEGESA updates for board partners.

 

[Niku-Sama]

i dont remember it being such a big jump on my 3600x, infact i dont think it changed any voltages when i enabled XMP on my 3600. but its a bit older, could be a new change between then and now too

 

[Niku-Sama]

i'm back bebeeee

also nothing strange afoot, nothing anomalous. i did notice a "beta" bios on asrocks website one day but its already gone. I think it had the 1.0.7.0 agesa but i cant remember, i see the 1.0.6.0 on the latest one from late march or april but i am going to hold off. my voltages haven't been spiking and i also dont let my computer idle so it should be doing any temp spiking like some of the others have reported.

to be safe the only other thing i've done since i last was able to post was set my pump and associated fans on my AIO to go full bore when ever its on. i was playing some offline games and in game i didnt see anything higher than 58C but they were also not very intense games.

i havent done any digging but i will soon, a new strange issue has come up, my input from keyboard mouse or controller quits some times out of the blue. i thought it was game related but it did it the other day when i was watching a DVD and i couldnt select anything on the menu. it wasnt lasting but it'll suck mid game online.