News AMD Issues Follow-up Statement On Ryzen Burnout Issues, Limits SoC Voltages

[Admin]

AMD has issued a second statement on the Ryzen 7000 burnout issues, citing that it will cap certain voltages to prevent the issue.

AMD Issues Follow-up Statement On Ryzen Burnout Issues, Limits SoC Voltages : Read more

 

[Vanderlindemedia]

I just dont understand how one in their right minds would have allowed 1.4V SOC.

 

[HKTacticblade]

I just dont understand how one in their right minds would have allowed 1.4V SOC.

A lot of motherboard makers actually dial it up automatically, which is just insane.

 

[TerryLaze]

AMD has issued a second statement on the Ryzen 7000 burnout issues, citing that it will cap certain voltages to prevent the issue.

AMD Issues Follow-up Statement On Ryzen Burnout Issues, Limits SoC Voltages : Read more

"These higher SoC voltages are needed to support higher memory overclocks, but AMD says that the 1.3V limit will not materially impact overclocking headroom, an important distinction."

That is not at all what the statement says, even if that could be what they meant.
The ability to overclock memory will still be there is what AMD says in the statement, there is no mention there about any impact on the range/headroom.
"None of these changes affect the ability of our Ryzen 7000 Series processors to overclock memory using EXPO or XMP kits or boost performance using PBO technology."

 

[atomicWAR]

My first bios dialled my SOC voltage to 1.35v for using EXPO. First thing I did was dial it down to 1.25v which was no brainer even before the exploding cpus. Then spent a few days getting my speeds at ddr5 6000, timings working at cl30 and ram voltage low(ish) but stable (also 1.35v down frtom 1.4v stock EXPO). Never trust xmp/expo profiles or your motherboard maker. Providing liberally high voltages/auto OC are frequently an issue for hardware makers users pay the price for.

 

[-Fran-]

Good. At least this is, hopefully, ending on a positive note and not with AMD and the ODMs burying their heads in the sand.

Also, this tidbit is important to me: "The EXPO profile itself does not increase the SoC voltage, the board vendors assign a pre-determined value of their own to support the increased speed of the EXPO profile".

AMD is at fault for not providing a more rigorous guideline or even checks within AGESA, but that comes with the territory of allowing ODMs to "get creative". Sadly, this is one where being too permissive have backfired horribly. Live and learn.

As for the Warranty... Gosh... If they'll still honour those in reality, why even put them outside of the warranty on paper? So darn stupid. AMD, Intel, please don't be so stupid on this one. If you want to support EXPO/XMP, then help memory makers to validate and make them "within spec" and not "overclocking", so they can be covered within your warranties. Or even better, push JEDEC to have a more comprehensive set of ranges for memory and timings! Come on!

Regards.

 

[setx]

AMD is at fault for not providing a more rigorous guideline or even checks within AGESA, but that comes with the territory of allowing ODMs to "get creative". Sadly, this is one where being too permissive have backfired horribly.

AMD is not at fault: Asus, for example, does the same on Intel. Even if CPUs there don't die as fast from this it doesn't mean that unreasonable high voltages for "auto" are ok.

 

[klavs]

My first bios dialled my soc voltage to 1.35v. First thing I did was dial it down to 1.25v which was no brainer even before the exploding cpus. Then spent a few days getting my speeds at ddr5 6000 l, timings working at cl30 and ram voltage low(ish) but stable (also 1.35v). Never trust xmp/expo profiles or your motherboard maker. Providing liberally high voltages/auto OC are frequently an issue for hardware makers users pay the price for.

So you set your SOC to 1.35v to get to 6000CL30. And AMD says you can do it at 1.3v?

I think AMD, RAM manufacturers and Motherboard makers MUST write in large letters that DDR5 better than 5600CL30'ish can cause burnout on AM5 platforms, because of the SOC it requires.

 

[TerryLaze]

AMD is not at fault: Asus, for example, does the same on Intel. Even if CPUs there don't die as fast from this it doesn't mean that unreasonable high voltages for "auto" are ok.

How should ASUS know what the reasonable level of voltage is if AMD doesn't tell them?
If AMD told them from the beginning that 1.35V might be enough to kill CPUs then do you really think that any mobo maker would go above that?!

As for the Warranty... Gosh... If they'll still honour those in reality, why even put them outside of the warranty on paper?

Yeah because they honor the warranty because it was their fault.
I still think they got a bad batch of components and are scrambling to try and fix it without having to do a mass recall.

 

[atomicWAR]

So you set your SOC to 1.35v to get to 6000CL30. And AMD says you can do it at 1.3v?

I think AMD, RAM manufacturers and Motherboard makers MUST write in large letters that DDR5 better than 5600CL30'ish can cause burnout on AM5 platforms, because of the SOC it requires.

No you misread my post. I set the SOC voltage to 1.25V from 1.35 that EXPO wanted to use as default in my release bios and in some later revisions but not my current bios which uses 1.25v as default SOC voltage. The Ram voltage wanted 1.4 volts default and that was tuned down to 1.35V. Ram voltage and SOC are not the same though both should be lower than they were by default with EXPO. The SOC voltage needs to be 1.3 or less according to AMD. I find 1.3V for SOC voltage high honestly though its on the edge of safe for overclocking. Regardless manufacurers trying to pump to much voltage through parts is nothing new IME. Point being never trust your auto settings in your motherboard or EXPO/XMP settings as it is common for them to pump to much voltage though parts. You should try to find the lowest stable voltages and only use your motherboard auto settings/expo/xmp as a starting point to get running.

 

[klavs]

No you misread my post. I set the SOC voltage to 1.25V from 1.35 [...] The Ram voltage wanted 1.4 volts default and that was tuned down to 1.35V.

Sorry, if I am asking a stupid question. But what kind of RAM voltage are you talking about? Can you show a pic of your bios? Are we talking about VDD, VDDQ or VPP? VPP I assume? I would be interested in seeing your settings. I thought, and may be wrong, that the SOC was significant for XMP/EXPO and not the VPP.

 

[thisisaname]

For the price of the boards and the CPUs they should have not been released with such a destructive flaw!

 

[btmedic04]

So you set your SOC to 1.35v to get to 6000CL30. And AMD says you can do it at 1.3v?

I think AMD, RAM manufacturers and Motherboard makers MUST write in large letters that DDR5 better than 5600CL30'ish can cause burnout on AM5 platforms, because of the SOC it requires.

I run ddr5 6000 30-38-38-28 on my 7950x3d with 1.35v vdimm and 1.25v vsoc. In the end, manual tuning is still the way to go for enthusiasts

 

[atomicWAR]

Sorry, if I am asking a stupid question. But what kind of RAM voltage are you talking about? Can you show a pic of your bios? Are we talking about VDD, VDDQ or VPP? VPP I assume? I would be interested in seeing your settings. I thought, and may be wrong, that the SOC was significant for XMP/EXPO and not the VPP.

No worries, its not stupid to ask specifics. Listed straight from my AsRock x670e Taichi.

VDDIO 1.35v
Dram VDD 1.35v
Dram VDDQ 1.35v
VDD_SOC 1.25V
VDD misc 1.1V
VDDP 1.15V

I also have some other small tweaks but this covers all the voltages I can think of and gives you the gist of getting DDR5 6000 CL 30 running on my system.

 

[Deleted member 2838871]

I also have some other small tweaks but this covers all the voltages I can think of and gives you the gist of getting DDR5 6000 CL 30 running on my system.

Thanks for that. I'm also running 6000 CL 30.

What's funny is mine blew up Saturday morning 3 days after I built PC... and I posted here about having installed the MSI beta BIOS for my Mag x670e board and I took some heat for it for installing the BIOS because it was a beta.

What can I say? I installed it because it said "optimized for 7000 series processors." It wasn't long after that the news broke that it was happening all over the place.

Then I went to the MSI boards and asked them and was told the BIOS is the one that I should have installed... even though it's dated 4/14.

https://www.msi.com/Motherboard/MAG-X670E-TOMAHAWK-WIFI/support

It does appear that all of the MSI AM5 boards have a BIOS for that date. Needless to say I find it a bit confusing.

I got new parts the next day and rebuilt the PC with a new CPU and mobo and it has been fine the last 4 days... with EXPO disabled. For now I'm leaving it disabled. I RMA'ed the dead parts.

I did go into the BIOS and it looks like the SOC is capped at 1.3 so that's good. I just don't know if I should enable EXPO or not because it was this BIOS and EXPO running when mine died. I heard a faint pop from the mobo... display went out... PC went dead and further attempts to power on were a quick fan spin and shut down.

I did not see any obvious damage to the cpu or the socket.

 

[atomicWAR]

Thanks for that. I'm also running 6000 CL 30.

Anytime. Yeah I knew as soon as I loaded the EXPO settings things needed tweaking from past expereince. If I were you I'd load expo to see the settings it tinkers with as some may not be as straight forward as voltages it my tinker with your ohms, etc. If you see anything sketchy like I did with high SOC or Dram voltages, turn them down before saving and rebooting (assuming your trying to hit ddr5 6000).

 

[10tacle]

Never trust xmp/expo profiles or your motherboard maker. Providing liberally high voltages/auto OC are frequently an issue for hardware makers users pay the price for.

Exactly. This is not a new issue either. I first experienced it with a 2011 Intel Sandy Bridge i5 2500K build and enabling XMP (and XMPII) profiles. It was an enthusiast ASUS board specifically for overclocking, but the factory overclocking profiles were just absurd as was even the default "base" XMP profile for running memory at the rated speed. If I remember correctly, it overvolted the CPU to 1.45 when low 1.3s were perfectly fine for a 4.5GHz all core O/C with a lot of manual fine tune setting. DRAM voltages were over the top too but can't remember those DDR2 settings.

I don't think it can be emphasized enough to warn new builders out there to stay far away from any motherboard's XMP profiler and especially overclocking software utilities by the motherboard. That's the first thing I warn about when giving new builders advice and instruction - besides warning that you are early killing the chipset as well as wasting electricity, especially in the summer when the PC is running hotter than it needs to be. Many unfortunately just do not have the patience or discipline or time (or all three) to take the time of learning the overclocking art.

 

[Deleted member 2838871]

Anytime. Yeah I knew as soon as I loaded the EXPO settings things needed tweaking from past expereince. If I were you I'd load expo to see the settings it tinkers with as some may not be as straight forward as voltages it my tinker with your ohms, etc. If you see anything sketchy like I did with high SOC or Dram voltages, turn them down before saving and rebooting.

First AMD system since 2001... so I have no past experiences. Will have a look when I get home. 👍

Many unfortunately just do not have the patience or discipline or time (or all three) to take the time of learning the overclocking art.

Makes sense. I actually have plenty of patience. Not all that interested in OCing my system because I'm definitely not all that experienced with it and honestly... why would it? This CPU and GPU are monsters.

In the past I've just enabled the ram speed via the mobo and left it at that.

 

[atomicWAR]

First AMD system since 2001... so I have no past experiences. Will have a look when I get home. 👍

My first one since Phenom was still a thing and my first primary daily driver since AMD FX64 chips ruled the roost shaming Pentium 4s/Ds across the board.

In the past I've just enabled the ram speed via the mobo and left it at that.

While tempting, it's generally a bad idea. I helped a lot of friends with Ry zen 1/2/3 with ram issues as default frequently didn't work for a lot of users, even with tuning (voltages/speeds/CL) it didn't always work which is part of why AMD got a bit of a reputation for ram compatibility issues and rightly so. This is a big reason EXPO exsists today. Many friends/family were use to using Intel and or loading defaults or XMP. I always encourage folks to see what default/xmp settings are, and then tweak them lower if anything looks to high (1.3-1.4V most frequently seen over volts depending on setting) or you want to decrease power usage.

 

[Deleted member 2838871]

My first one since Phenom was still a thing and my first primary daily driver since AMD FX64 chips ruled the roost shaming Pentium 4s/Ds across the board.

Athlon XP 1800+ with Windows XP.

While tempting, it's generally a bad idea. I always encourage folks to see what default/xmp settings are, and then tweak them lower if anything looks to high (1.3-1.4V most frequently seen over volts depending on setting) or you want to decrease power usage.

Yeah I'm definitely not doing the click it and forget it thing again. Will have a look at the settings later and see what EXPO adjusts and go from there.

Not losing any sleep over running with EXPO disabled anyway but it would be nice to get the full speed.

 

[Deleted member 2731765]

Could this be really true ? The creator of the Hydra tool for AMD Ryzen CPU tuning and optimization claims that AMD and board partners knew about this issue before launch, but AMD had asked them to remain quiet.

Check the following tweets by Yuri.

https://twitter.com/x/status/1650791399699218433

View: https://twitter.com/1usmus/status/1650791399699218433

https://twitter.com/x/status/1650793064653287424

View: https://twitter.com/1usmus/status/1650793064653287424

https://twitter.com/x/status/1650810800745922564

View: https://twitter.com/1usmus/status/1650810800745922564

 

[Deleted member 2838871]

Could this be really true ?

Ryzengate. (tm)

 

[Alvar "Miles" Udell]

Honestly I'm surprised this didn't happen with the 5800x3d and wouldn't be surprised if this keeps happening to the point of a lawsuit and recall.

AMDs PBO is garbage, and I can say this as the owner of 2 high end X Series AM4 motherboards and 3 AM4 CPUs. It is insanely aggressive with voltage, doesn't allow CPUs to reach advertised boost speeds, and hampers performance far below what they're actually able to do because it relies on "smart" systems.

 

[Deleted member 2731765]

Ryzengate. (tm)

Most likely, yes ! And also a bit controversial.

Oh btw, I just saw your reply/update under the PC gaming sub-forum thread. I will reply after some time. It seems you got 2 of the latest AAA games free with your new hardware purchase. 👍

And it sucks to know you were also one of victims of these Ryzen CPU burnout issues. Glad both parts were RMA'd though.

 

[setx]

How should ASUS know what the reasonable level of voltage is if AMD doesn't tell them?

The specs provided definitely listed the range of "default" voltages. As for how high you can go above that – of course official documentation won't say that as it's completely unsupported case.

If AMD told them from the beginning that 1.35V might be enough to kill CPUs then do you really think that any mobo maker would go above that?!

Definitely would. The whole "extreme overclock" nonsense is based on providing means to set voltages and the rest well outside the spec. How else would the motherboard makers justify the price of their top/higher-end boards? And that overclocking is further tuned for complete idiots so pressing one button/changing one value would provide significant overclocks because board would also silently change other values, including voltages to insane levels.