X370 Taichi VCore nasty spikes

lowdog wrote: Read the reddit post again where AMD_Robert explains about the "spikes" being normal and part of the technology/specifications of the processor. There is NO problem with the Asrock board or firmware in regard to the voltage spikes with auto settings in bios.

I agree with you the things you said. I won't chant my typical mantra here, though. I anyone wants to see it, all they have to do is to scroll up a few pages.

When I get to put a new UEFI version I'll be sure to update you guys, so you'd know how it went. That's in case anyone cares.

Hey to you too, Onox!

First things first. Please do NOT give up on custom PC building! One hiccup doesn't necessary mean a catastrophic failure. Well, sometimes it does, but that's life. Assembling PCs yourself is a rewarding and fun thing to do; do NOT stress about it.

I've been assembling PCs way before I could earn my own money, and sometimes it wasn't as smooth as I wanted it to be.

The fact that you've successfully assembled your first very own PC and it boots mean you've done NOTHING wrong with it! If your temps are as they should and PC works, then whatever the fault is, it's not with you.

As it turns out, the fault appears to be with ASRock in this case. I didn't even receive some sort of reply from these guys. Guess they care till you give them your hard earned cash, no longer than that. TBH, that's my first ASRock build since they splintered off from ASUS. I'm also planning to be my last. I just jumped wagons for the lulz but got more issues than I anticipated.

The X370 Taichi seems very well engineered but the state of their software/firmware is abysmal. I've seen better work from trainees in some IT companies. Sorry, but I have.

Now, on the case - my issues somewhat differ from yours, although we share the same skepticism and worries.

As you see, I'm still on P3.10, so is another user, datonyb (who seems to be perfectly happy with his board, though). My issues are: I get bad voltage spikes on auto settings; I get high temps on idle with auto settings(!); my Sapphire Boost is not working. Well, these are the ones that bug me the most but the list is nowhere near short.

None of the offset or manual settings produced desirable results. So did the LLC settings. On all auto I had spikes around 1.58, 1.59V!!! The hell these are normal! Only thing that put the voltage where I set it was the P-states. They were fixing the voltage exactly where I wanted it to.

Problem with P-states is that they still produce sub-optimal voltages. And this bugs me beyond words.

Now I've settled for my broken Sapphire Boost. CPU caps at 3400MHz but at least I never get above 1.212V. Vcore also drops to 0.8V when idling. This is far from the XFR but at least it behaves like a normal PC, albeit slower. In this Ryzen build of mine stability is more important than sheer speed.

Oddly enough I haven't seen such issues with ASUS or MSI. Gigabyte already fried some Ryzens with high voltage, so I don't even consider them. I know because I took an extensive search on the webz but nowhere you can see such behavior en masse as here.

Also, you should go for Tdie when reading temps. HWiNFO reports that OK for me. There is 20C offset from Tdie & Tctl. That's by AMD design. ASRock replaced the reported temp on other boards, my guess is that they did it silently with the Taichi too.

All my fans are blasting @ 100%, so I don't really care about smooth RPM control. Your hand is the best indicator as you noted. IMO 55C is just fine with your cooler. I think you installed it correctly.

To conclude, leave all on auto and wait for a better times (and UEFI). I think your case is really close to the intended operation of your system, you should be just fine. It's also noteworthy that no public software can monitor the real-time voltages and currents. CPU's internal circuity is switching these way too fast. Even if you see 1.55V for 1-2s it means that the CPU went there for a fraction of a second, but monitoring SW can't get to that time resolution.


Edited by zlobster - 28 Dec 2017 at 2:11am

Thanks everyone for the help!

When I set the VCore manually I got rid of the spikes, PC now runs cooler and is again stable as rock. \m/ Do you see what I did there?

With manual voltage I see CPB & XFR are still kicking in, which I think it's strange because I've set VCore to 1.22V. Under heavy load it droops to 1.19V. Funny thing is PC runs well even with 1.19V VCore; under load it just droops even lower. Odd, no?

Custom P-states also work well. I'll tune them better after I find my 1700X limit.

Finally, I can't say the problem is gone. For now I've just worked it around with your help. @Moderators - you can close the thread if you wish, but please forward this to ASRock engineering teams. To me it's still a problem. When one buys a car, one doesn't tune the gas pedal manually, just because it gives too much fuel. That's factory's job!

I was just researching what datonyb said. These custom P-states are very powerful thing! It's as close to metal as you can get. With these you can program the CPU on a low HW level, pretty much as you wish. 2000MHz for idle? You gots it! 4000MHz for the lulz? You gots it! Thanks, AMD.

I don't think I've noticed the temp spikes you're mentioning. Then again, I don't use AIDA64 for reading or stress-tesitng. HWiNFO64 reports good and stable temps for me. My advice to you is to check if your CPU cooler is seated properly and the TIM is spread well. Improper cooler contact with the CPU IHS is a frequent reason for the thing you're mentioning.

I will definitely try tweaking the UEFI these days and I'll report back the results.

On a side note, I'd really like to hear ASRock's engineering justification behind these crazy auto VCore voltages, and see a fix in the future.

ive had set 3.9 ghz and 1.3 volts fixed voltage and llc 2

heres the taichi guide

https://www.youtube.com/watch?v=52Tw-wcT7o4

my p states settings are

9c
8
27

that will make sense when you watch the video