Processors

[PiersJH]

After seeing the positivity and innovation surrounding Zen 3, I decided to purchase a 5900X. I've found the performance astonishingly good (especially with AVX2 workloads), however, there's a question I can't get a clear answer to, and when I do receive an answer, it's contradictory (i.e. my AMD hardware contradicts AMD's statements).

I've added space between the bullet points because it looked like a wall of text. Please note, a VID of 1.55000V often sees a Vcore of 1.55000V reported (VDDCR). I am generally not concerned by VID values so long as the delivered voltage stays within specification.

• AMD's Robert Hallock states 1.55000V is an acceptable and expected, albeit uncommon, VID/Vcore for the 59xx CPUs.

• Overclocker and Asus contributor The Stilt states that is not correct, and that the actual Vmax (factory-set limit) of my specific CPU is 1.50000V. This has been verified using a method I've been asked not to share publicly (I'm happy to PM this to an AMD employee).

• The motherboard is using "AGESA V2 PI 1.2.0.3 Patch C", the latest available at present for my ('Asus ROG Strix B550-E Gaming WIFI')

• This is being tested at stock configuration with PBO options disabled, manufacturer 'enhancements' disabled, and both with JEDEC-compliant memory and XMP/AMP/A-XMP/DOCP (take your pick of names for XMP).

• The sensors used on most B550 and X570 boards (that provide the SVI2 TFN reading) only support up to 1.55000V (source: The Stilt, and data sheets for the sensors), which means I have no idea how much voltage is truly going into my CPU. It appears to operate perfectly fine with an all-core AVX2 load reaching ~69°C @ ~1.18V and 4.275 GHz, and a single-core SSE4 (chosen for higher boosting behaviour) load reaching ~70°C @ ~1.55V and 4.850 GHz.

• Ryzen Master might be a better way of measuring this, but it doesn't provide a logging function (i.e. log X data for X minutes to a CSV). It's essentially useless.

• Using HWiNFO, I have logged data and found instances where at least 1.55V is reached (Vcore).

• I have spoken with Asus and they aren't sure (it was escalated to Taiwan).

• I have spoken with the author of HWiNFO, and he's looked into it and noticed similar reports but has no answer, or is bound by NDAs (which is understandable).

• No part of my entire system is older than ~3-4 months.

This leaves me regretting going with AMD again, as while 1.55V+ might be fine for 'bursty' workloads every day in the view of AMD, the three year warranty isn't very reassuring.

A final note, I returned to AMD after leaving due to poor software and technical support in 2009 (and moved over 1,800 SME devices to Intel systems because of my personal experience). It seems support has greatly improved, but AMD's hardware shouldn't be contradicting AMD's statements.

[ryzen_type_r]

Try a different make/model mobo if you think the CPU is getting too much voltage.  Mobo manufacturers often push certain parameters beyond what is recommended even with all 'default' settings.

[PiersJH]

That's entirely the opposite of the point.

[PiersJH]

I believe I've found a driver fix to the problem (which is possibly caused by a fault with AMD's Chipset drivers). Since applying it 4 weeks ago, effective VID has stayed at 1.50000V, as is the case with what appears to be the majority of chips. I've left it this long to reply in order to test whether the fix has worked.

A reply from AMD staff would be appreciated, but seems unlikely.