News AMD Ryzen 7000 Burning Out: Root Cause Identified, EXPO and SoC Voltages to Blame

[tristanx]

I'm interested seeing a rereview after the fixes are out. I'm sure Zen 4 will be nerfed.

 

[TerryLaze]

I'm interested seeing a rereview after the fixes are out. I'm sure Zen 4 will be nerfed.

They can just find ram that will reach the same speed with less volts.
And with they I mean reviewers as well as normal users.

 

[Maebius]

So, if both AMD and Intel insist on this 'enabling EXPO/XMP "may" void the warranty' nonsense, maybe ALL hardware outlets should test their products ONLY with the stock settings from now on.

 

[digitalgriffin]

I checked the Asus website and sure enough a spanking new Bios version 1412 dated today was there - my previous version was 1409. Did my usual flash procedure, enabled the EXPO tweaked and AI overclock settings, and ran some new Cinebench with monitoring from HWiNFO64. Scores seem unaffected. The SOC volts were slightly reduced from around 1.35 volts to 1.335 volts locked in.

Not much of a change from my previous results, but I do feel more comfortable that I won't run into flaming (or bulged) CPU issues. I also appreciate Asus jumping to get the new Bios out quickly.

Apparently, Gigabyte is worried to. They posted a new BIOS for the Aorus Elite AX B650 yesterday.

 

[hotaru.hino]

I'm interested seeing a rereview after the fixes are out. I'm sure Zen 4 will be nerfed.

Zen processors by default have always been pushed to the edge of what's safe to run these things at and a lot of people suggest undervolting them anyway. More often than not, this results in increased sustained performance.

AMD's been playing this frequency pushing game too hard.

 

[digitalgriffin]

Zen processors by default have always been pushed to the edge of what's safe to run these things at and a lot of people suggest undervolting them anyway. More often than not, this results in increased sustained performance.

AMD's been playing this frequency pushing game too hard.

Having a higher voltage by default allows for a greater percentage of chips to be good. A borderline margin chip can often be saved with a little more juice. It's not like AMD or Intel can release chips where some run 1.1V and others 1.0 for the same chip model.

 

[danny009]

Hey'all, just wanted to let you folks know, my 5+ years old PC have absolutely zero problems and then my asus gaming laptop had PLENTY problems just recently purchased.

Or, y'know, stop buying a new GPU/iphone every year so those big tech dont get motivated to do such things. They arent your friend, and they will do EVERYTHING to get your dollars.

But what do I know? I'm just an old fart living in older times, but hey, at least my computer working.

 

[hotaru.hino]

Having a higher voltage by default allows for a greater percentage of chips to be good. A borderline margin chip can often be saved with a little more juice. It's not like AMD or Intel can release chips where some run 1.1V and others 1.0 for the same chip model.

Or they could tone down the clock speed a little. That voltage is only needed for light workload turbo, and if they need to drop 200 MHz from a ceiling of 5.2GHz or whatever, is anyone really going to notice that performance drop?

It's good for reviews and bragging, but that's about it.

 

[TJ Hooker]

It's not like AMD or Intel can release chips where some run 1.1V and others 1.0 for the same chip model.

Different individual chips, even of the same model, may in fact run at different voltages.

"Individual maximum VID values are calibrated during manufacturing such that two processors at the same frequency may have different settings within the VID range."

https://cdrdv2-public.intel.com/337344/337344_8th-gen-core-family-datasheet-vol-1_Rev009.pdf

 

[cryoburner]

So, if both AMD and Intel insist on this 'enabling EXPO/XMP "may" void the warranty' nonsense, maybe ALL hardware outlets should test their products ONLY with the stock settings from now on.

Overclocking just about any modern CPU tends to be rather pointless these days, both for Intel and AMD, since the chips are binned at the factory and are already pushed near their limits at stock settings. The performance benefits of overclocking are going to be imperceptible on pretty much any of these CPUs, and the significant increases to power draw and heat output are not at all worth the few percent or less performance gains that overclocking typically brings. The only real benefit to leaving them unlocked at this point is to show people that they are in fact operating optimally at stock settings, and are not being artificially limited much for product segmentation, at least in terms of per-core performance.

 

[InvalidError]

Overclocking just about any modern CPU tends to be rather pointless these days, both for Intel and AMD, since the chips are binned at the factory and are already pushed near their limits at stock settings.

For the cores, sure. For memory though, the official specs only go up to 5200MT/s at 1.15V for AMD and 4800MT/s for Intel, anything beyond that is also considered overclocking for the purpose of potentially voiding your warranty and there are some fairly substantial performance gains to be had from running DDR5 at 6000+MT/s with tighter timings, especially on things like 1% lows in games.

 

[TerryLaze]

Overclocking just about any modern CPU tends to be rather pointless these days, both for Intel and AMD, since the chips are binned at the factory and are already pushed near their limits at stock settings.

Even automated overclock can have a huge impact, cinebench can gain 20% even just by the automated overclock (MCE) and increasing the power limits (Having it "default" where most mobos will push it as high as they can) .
It's not worth it to most people because it only benefits a very limited number of software and because of the increased power.
https://www.pugetsystems.com/labs/a...ermals-and-Content-Creation-Performance-2375/

 

[Maebius]

Overclocking just about any modern CPU tends to be rather pointless these days, both for Intel and AMD, since the chips are binned at the factory and are already pushed near their limits at stock settings.

As @InvalidError explained, enabling XMP/EXPO yields substantial performance gains compared to "official supported" settings.
If we don't touch anything else, these highly advertised functions shouldn't be considered overclocking.

 

[jacko.mj4]

Intel: "Lets see what AMD do when they see we have faster CPUs". AMD: "Push the voltages to the max and beyond, we have to match intel, wait... oh no no no no!" Intel: "Ha ha ha ha ha ha, now no one will be buying AMD. Well done boys". AMD: "We'll sort it and people have very short memories, no problem".

 

[Sergei Tachenov]

To be fair, it's not like they pushed the voltage to make their CPUs faster. They're perfectly running with the same performance at lower voltages, so it's more like a stupid design oversight.

 

[TerryLaze]

To be fair, it's not like they pushed the voltage to make their CPUs faster. They're perfectly running with the same performance at lower voltages, so it's more like a stupid design oversight.

As @InvalidError explained, enabling XMP/EXPO yields substantial performance gains compared to "official supported" settings.
If we don't touch anything else, these highly advertised functions shouldn't be considered overclocking.

They lose a lot of performance without the RAM OC.
And it probably wasn't AMD that pushed the voltages but the mobo makers.
If there is any fault with AMD it would be them not telling the mobo makers, and the world, that their new CPUs immediately break at anything above 1.3V.
Doing that would have caused the sales to be horrible though so of course they wouldn't do that.

 

[Sergei Tachenov]

Enabling EXPO does yield substantial performance gains indeed. What they have missed completely is that you don't need to push VSOC that high just to make it stable! In fact, I actually managed to make it stable by bringing VSOC down from 1.3V to 1.25V. Currently running at 1.2V, works fine too. And that's with EXPO 6000 CL30, which is the famous sweet spot for the 7000 series.

While at 1.3V it worked too, I had serious issues with sleep mode. They're gone completely after lowering VSOC. No performance loss whatsoever, just a perfectly running PC.

So that default VSOC setting was nothing but a major screw-up on the vendor part, be it the mobo vendor or AMD, doesn't really mater in the end.

 

[TerryLaze]

Enabling EXPO does yield substantial performance gains indeed. What they have missed completely is that you don't need to push VSOC that high just to make it stable! In fact, I actually managed to make it stable by bringing VSOC down from 1.3V to 1.25V. Currently running at 1.2V, works fine too. And that's with EXPO 6000 CL30, which is the famous sweet spot for the 7000 series.

While at 1.3V it worked too, I had serious issues with sleep mode. They're gone completely after lowering VSOC. No performance loss whatsoever, just a perfectly running PC.

So that default VSOC setting was nothing but a major screw-up on the vendor part, be it the mobo vendor or AMD, doesn't really mater in the end.

Yeah but the point of XPO is for it to be easy, one click on one setting and done, that's why it has to use higher values because it has to make sure that it will work even on RAM that isn't quite as good and needs a little bit more V.

 

[CeltPC]

Enabling EXPO does yield substantial performance gains indeed. What they have missed completely is that you don't need to push VSOC that high just to make it stable! In fact, I actually managed to make it stable by bringing VSOC down from 1.3V to 1.25V. Currently running at 1.2V, works fine too. And that's with EXPO 6000 CL30, which is the famous sweet spot for the 7000 series.

While at 1.3V it worked too, I had serious issues with sleep mode. They're gone completely after lowering VSOC. No performance loss whatsoever, just a perfectly running PC.

So that default VSOC setting was nothing but a major screw-up on the vendor part, be it the mobo vendor or AMD, doesn't really mater in the end.

Yes I lowered my SOC to 1.25 volts as well, and does just fine with 6000 CL30.

 

[greenreaper]

For regular use 1.25V SoC should indeed be plenty, but if you are overclocking Infinity Fabric (FCLK) I suggest checking with y-cruncher - the last test selected by default, 17 IIRC - as it is great at picking up the kind of vectorised instability that will cause e.g. computation errors in BOINC. I needed this to go over 2100Mhz (likewise, 1.20V turned out to be insufficient for 2040Mhz, although that instability would only rarely appear in general use).

 

[CeltPC]

With the latest Bios version 1616 (AGESA version to Combo AM5 PI 1.0.0.7.a) for my board and doing nothing but enabling "EXPO Tweaked" and AI overclock, I get 1.235 volts for SOC. That is lower than what I was previously manually setting. No instability, so I am just letting everything run as is.

The performance is great, so I would agree with Steve at Gamers Nexus that it is a resolved issue with Asus AM5 MB's. Combined with Asus also making the fixes to it's policies and warranty, I am comfortable with my system and happy with the build choices.

 

[STbob]

I'm not sure what you are getting at. What you stated above is not a reason why, all it says is that you overclocked something.

In my original post I said "I don't understand why users overclock".

I have not overclocked since x58 Intel 920 chip Since then I use servers or Workstations with ECC ram I have zero interest. But...

On the consumer end of things its popular and manufacturers are pushing it. I recently bought two AMD systems to build for kicks and I just went with the recommendations from Micro Center on the RAM type etc. I bought a locked chips (x3d chips) are locked from overclocking but the DDR is not.

It says official pc3200 ram is supported. I buy
TridentZ Neo (I did not want colors but its all they had in 32GBx2)
DDR4-x2 XMP 2.0
CL16-18-18-38 1.35v

That is approved memory for Asus MB.

Notice it does not say 1.2v (DDR4 spec) its says 1.35v on the ram.

When you boot the profile comes up automatically with the settings on the sticker. EZ. But.... They just tricked you into over clocking your DDR4 sneaky eh.

Does it say 1.2v spec anywhere on the chip? No. No warnings on overclocking either.

So if my system blows up they could say "not covered" but in a court of law they would likely loose. I doubt very much they will not cover warranty on what they term as "factory overclocked ram" They don't call it that on the box.

 

[Sergei Tachenov]

So if my system blows up they could say "not covered" but in a court of law they would likely loose.

There is some good news on this. Asus apparently gave up and officially confirmed that warranty does cover EXPO/XMP/DOCP:

ASUS BIOS updates and warranty coverage for ASUS AM5 platform motherboards with Ryzen™ 7000 series processors

press.asus.com

 

[STbob]

There is some good news on this. Asus apparently gave up and officially confirmed that warranty does cover EXPO/XMP/DOCP:

ASUS BIOS updates and warranty coverage for ASUS AM5 platform motherboards with Ryzen™ 7000 series processors

press.asus.com

Well they don't want to pull a Bud Light and loose all their customers. They wanted to save money and deny people but the bean counters found over all they might loose more if they don't. They did not have a strong legal leg to stand on. They had a leg of greed to save from having to replace warranty items but it was too risky a business move.
They choose wisely.

 

[TerryLaze]

Well they don't want to pull a Bud Light and loose all their customers. They wanted to save money and deny people but the bean counters found over all they might loose more if they don't. They did not have a strong legal leg to stand on. They had a leg of greed to save from having to replace warranty items but it was too risky a business move.
They choose wisely.

Yeah but that still only covers any damage on the mobo directly caused by the XMP ,which is practically a zero percent chance to happen, if your ram burns out or your CPU destroys your socket you are still out.