Discussion Ryzen 3000 CPUs overclocking: exercise caution

VirtualLarry

No Lifer
Aug 25, 2001
56,327
10,035
126
Hmm, I'm unconvinced. If his was an "early batch", maybe just bad silicon, bad luck.

My 3600 is from "early batch" (bought it the night of, or the night after release), I've run it @ 127C (according to Ryzen Master) for several hours, using a fixed-clock OC, around or above 4.0Ghz, @ 1.38V or so, it needed that much for my PrimeGrid workload.

Then I basically set defaults for CPU clock, had RAM (TridentZ RGB "for AMD" 3600) @ XMP 3600, FCLK 1800 on my Asus B450-F ROG STRIX.

Had weekly crashes/lockups.

Recently, set all-core OC again, using AGESA 1.0.0.4 patch B / UEFI BIOS 3003, to 4.0GHz @ 1.3685V. Seems stable for mining software on CPU and GPU.

Although, for the first time since setting all-core OC using this AGESA version, I couldn't wake monitor from sleep this morning, and had to manually use physical RESET button on front-panel to reboot. Could have been my RX 5700 too, I recently OCed the VRAM again from 1750 to 1800. It seems over time (12+ hours), VRAM temps creep up, or something, and things start to crash. Now running at 1775 VRAM clock in Wattman/Tuning / ADR2020.

Anyways, I'll let the forum know if my R5 3600 gets knackered due to all-core OC. Temps are OK, 82C max, under 240mm AIO WC (CM MLL 240).

It's unclear to me whether buildzoid's 3700X was under air or water from that linked thread.

There was mention that he set Vddg (is that the same as Vsoc?) to 1.15V, which, Vsoc is only supposed to go to 1.10V max, I read.

I also wonder about degradation due to LLC settings, sometimes overly-aggressive LLC, combined with fixed-voltage / clock OC, will cause voltage overshoot.
 

Arkaign

Lifer
Oct 27, 2006
20,736
1,377
126
The initial 3k launch, myself and friends experienced multiple issues. A DOA CPU and another with a single working memory channel. Since then, I've installed many of them without issue. Speaking purely on the CPU side.

Now, OC on 3000 series appeared largely pointless before long, so I have been running them stock + tight Ram timing, so I think problems should be relatively rare hopefully.
 

DrMrLordX

Lifer
Apr 27, 2000
21,620
10,829
136
VDDG is not the same as vSoC. There are five common settings you set for Matisse: vcore, VDDG, VDDP, vSoC, and (of course) vDIMM. Though vDIMM is for the RAM, it's still intertwined with VDDP and VDDG (you mess with those two in RAM OC).

The_Stilt warned us early on that vcore in excess of 1.325v or so was risky. That was based on one individual CPU in his possession, and it was based on current levels of the workload. 1.325v with low current is safer than 1.325v with high current (Prime95, Blender, stuff like that). The highest I will go is maybe 1.34v, and that's only for workloads of Blender intensity and below.

It's quite possible that Buildzoid was also using a lackadaisical method for monitoring vcore. CPU-z consistently provides the highest voltage readings of any software I use to monitor vcore, not counting Ryzen Master which consistently gets confused by vdroop and voltage offsets.
 

coercitiv

Diamond Member
Jan 24, 2014
6,187
11,858
136
Hardcore overclocker pushes 1.375V through a CPU that is known to safely do 1.325V on average. CPU degrades. Anandtech forum member creates cautionary thread specifically tailored for Zen 2 overclocking as result.

Under no circumstance should any reader search the web for any other CPU degradation story, be it from AMD or Intel. Exercise caution when buying Zen 2 CPUs, they are highly sensitive to FUD.
 

amrnuke

Golden Member
Apr 24, 2019
1,181
1,772
136
I once towed a 1500 lb Uhaul trailer with my 1999 Ford Contour and broke 2nd gear. Had I been towing with a 1999 Ford Taurus, may not have been an issue.

So this guy took a 3700X and tried to get it to do things it wasn't designed for. Sometimes that'll work and sometimes you'll break you transmission.

In other news, water is wet.
 

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,328
4,913
136
The_Stilt warned us early on that vcore in excess of 1.325v or so was risky. That was based on one individual CPU in his possession, and it was based on current levels of the workload. 1.325v with low current is safer than 1.325v with high current (Prime95, Blender, stuff like that). The highest I will go is maybe 1.34v, and that's only for workloads of Blender intensity and below.

That was the FIT voltage for his one specific chip. Depending on the characteristics of the silicon used in your chip, 1.325V may not be safe at all for your sample.

Manual OC is pretty much for only extreme OCers at this point anyways. You can get acceptable results tweaking the parameters for the stock boost without the risk.
 

yeshua

Member
Aug 7, 2019
166
134
86
Hardcore overclocker pushes 1.375V through a CPU that is known to safely do 1.325V on average. CPU degrades. Anandtech forum member creates cautionary thread specifically tailored for Zen 2 overclocking as result.

Speaking of "that is known to safely do 1.325V on average". I would love to get a quote on that from the official AMD website, e.g. www.amd.com

Under no circumstance should any reader search the web for any other CPU degradation story, be it from AMD or Intel. Exercise caution when buying Zen 2 CPUs, they are highly sensitive to FUD.

I said nothing about buying or not buying AMD CPU. I also did not try to spread FUD as I just quoted BuildZoid - I apologize for doing that and creating this topic. I still don't understand the need to ridicule me.
 

coercitiv

Diamond Member
Jan 24, 2014
6,187
11,858
136
Speaking of "that is known to safely do 1.325V on average". I would love to get a quote on that from the official AMD website, e.g.
You want either AMD or Intel to put a safe value for overclocking on their websites? The same overclocking that voids warranties for both brands?

I also did not try to spread FUD as I just quoted BuildZoid - I apologize for doing that and creating this topic. I still don't understand the need to ridicule me.
Because you advised people to exercise caution with a specific CPU family based on the experience of a known hardcore overclocker:

1582843046159.png
 

Ottonomous

Senior member
May 15, 2014
559
292
136
I said nothing about buying or not buying AMD CPU. I also did not try to spread FUD as I just quoted BuildZoid - I apologize for doing that and creating this topic. I still don't understand the need to ridicule me.
Yeshua your entire post history is a defense of the hardware vulnerabilities in Intel's chips, how AMD would be totally irrelevant with an 8-core TGL proc, how your 3700 is not achieving optimal frequencies, etc, you're essentially here with an agenda

Its obvious why people would be a bit suspicious.
 

Topweasel

Diamond Member
Oct 19, 2000
5,436
1,654
136
We have seen AMD CPU's be sensitive to overvolting in the past. Not surprised if that is the case here. Its important to remember that AMD's single core load at nearly 1.5v and all core loads are different. But its the same story as always, I can see pushing for more sustained all core performance when OC'ing. But realistically Zen has spent its whole life now at the edge of its silicon window and AMD has been working hard on solutions to get that last 1%. So without a chiller I am not sure I see a real point in OCing these chips.
 

Kenmitch

Diamond Member
Oct 10, 1999
8,505
2,249
136
So without a chiller I am not sure I see a real point in OCing these chips.

I guess it depends on one's work loads in the end. If a person is just doing normal everyday things there really isn't any advantage to overclocking.

Apps that can utilize all the cores can show some benefit from a all core overclock. The end user would need to do some experimenting with all core clocks vs stock vs pbo to see what works best for them in the end.
 

eek2121

Platinum Member
Aug 2, 2005
2,930
4,026
136
We have seen AMD CPU's be sensitive to overvolting in the past. Not surprised if that is the case here. Its important to remember that AMD's single core load at nearly 1.5v and all core loads are different. But its the same story as always, I can see pushing for more sustained all core performance when OC'ing. But realistically Zen has spent its whole life now at the edge of its silicon window and AMD has been working hard on solutions to get that last 1%. So without a chiller I am not sure I see a real point in OCing these chips.

it’s not the voltage that kills, but rather, the current. I have had my chip as high as 1.7V.
 

chrisjames61

Senior member
Dec 31, 2013
721
446
136
Speaking of "that is known to safely do 1.325V on average". I would love to get a quote on that from the official AMD website, e.g. www.amd.com



I said nothing about buying or not buying AMD CPU. I also did not try to spread FUD as I just quoted BuildZoid - I apologize for doing that and creating this topic. I still don't understand the need to ridicule me.


It smells of FUD. One guy, Buildzoid who is a reckless but in a good way ruins his cpu. Seems pretty much an isolated incident.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,542
14,496
136
It smells of FUD. One guy, Buildzoid who is a reckless but in a good way ruins his cpu. Seems pretty much an isolated incident.
A known overclocker admits to overclocking without any caution, and somebody has to post it here saying "exercise caution" ? I think somebody has an agenda.
 
  • Love
Reactions: spursindonesia

Arkaign

Lifer
Oct 27, 2006
20,736
1,377
126
People grasp at any straw to knock Ryzen cpu's. Unfortunately for them it isn't working very well lol!

Good products speak for themselves :) It's been epic seeing the releases go from laughable (later FX series) to really promising (Zen) to quite excellent (Zen+) to fundamentally superior in most respects (Zen2).

However, it seems like a few people may need to adjust how they think about enthusiast CPU experiences with regards to the Zen2 in particular. I find it basically counterproductive to bother with any manual OC or overvolting. Gains are absolutely minimal short of subambient class cooling setups, and often you will see REDUCED performance by screwing with it. They are exceptionally optimized out of the box from the word go in that respect, and the chief way to enhance your performance a bit is to combine it with friendly DDR4 and carefully follow guides and the Ryzen memory calculator to achieve best results for bandwidth, latency, and real world performance.

It's just a different way of doing things. Perhaps not as exciting as it once was where we'd get excited as enthusiasts about particular steppings and how much we could tune and OC systems 10, 20, even as much as 50 percent beyond stock performance. That just isn't the way Zen2 behaves. And that's fine, arguably it's superior due to not needing silicon lottery and heavy effort and combining with OC style mobos to achieve the results you want.

7nm TSMC is extremely dense, and limits seem based around areal density causing hotspotting that cannot be effectively radiated away from such miniscule surface areas. AMD has tuned and binned these SKUs to not leave basically anything on the table from stock. Treating a Zen2 like it's Sandy Bridge or even Pile-driver, and yoinking the volts and current up in a futile pursuit of a meaningful OC is just risking degradation. You can't outrun basic physics, silicon electromigration is very real. And at a certain point any CPU die will degrade, as with SNDS and other memorable examples.

It's a new paradigm. Zen2 SKUs are class leading without the need or even practical benefit from any manual OC, especially in regards to any more than stock power. The tinkerer and tweaker community among us will have to be content with amazing CPUs that are already optimized for peak performance out of the box, and focus on areas where effort actually pays, eg; DDR timing, GPU tuning, etc.
 

chrisjames61

Senior member
Dec 31, 2013
721
446
136
Good products speak for themselves :) It's been epic seeing the releases go from laughable (later FX series) to really promising (Zen) to quite excellent (Zen+) to fundamentally superior in most respects (Zen2).

However, it seems like a few people may need to adjust how they think about enthusiast CPU experiences with regards to the Zen2 in particular. I find it basically counterproductive to bother with any manual OC or overvolting. Gains are absolutely minimal short of subambient class cooling setups, and often you will see REDUCED performance by screwing with it. They are exceptionally optimized out of the box from the word go in that respect, and the chief way to enhance your performance a bit is to combine it with friendly DDR4 and carefully follow guides and the Ryzen memory calculator to achieve best results for bandwidth, latency, and real world performance.

It's just a different way of doing things. Perhaps not as exciting as it once was where we'd get excited as enthusiasts about particular steppings and how much we could tune and OC systems 10, 20, even as much as 50 percent beyond stock performance. That just isn't the way Zen2 behaves. And that's fine, arguably it's superior due to not needing silicon lottery and heavy effort and combining with OC style mobos to achieve the results you want.

7nm TSMC is extremely dense, and limits seem based around areal density causing hotspotting that cannot be effectively radiated away from such miniscule surface areas. AMD has tuned and binned these SKUs to not leave basically anything on the table from stock. Treating a Zen2 like it's Sandy Bridge or even Pile-driver, and yoinking the volts and current up in a futile pursuit of a meaningful OC is just risking degradation. You can't outrun basic physics, silicon electromigration is very real. And at a certain point any CPU die will degrade, as with SNDS and other memorable examples.

It's a new paradigm. Zen2 SKUs are class leading without the need or even practical benefit from any manual OC, especially in regards to any more than stock power. The tinkerer and tweaker community among us will have to be content with amazing CPUs that are already optimized for peak performance out of the box, and focus on areas where effort actually pays, eg; DDR timing, GPU tuning, etc.



I look at overclocking as a hobby vs of any practical value today.
 

eek2121

Platinum Member
Aug 2, 2005
2,930
4,026
136
Good products speak for themselves :) It's been epic seeing the releases go from laughable (later FX series) to really promising (Zen) to quite excellent (Zen+) to fundamentally superior in most respects (Zen2).

However, it seems like a few people may need to adjust how they think about enthusiast CPU experiences with regards to the Zen2 in particular. I find it basically counterproductive to bother with any manual OC or overvolting. Gains are absolutely minimal short of subambient class cooling setups, and often you will see REDUCED performance by screwing with it. They are exceptionally optimized out of the box from the word go in that respect, and the chief way to enhance your performance a bit is to combine it with friendly DDR4 and carefully follow guides and the Ryzen memory calculator to achieve best results for bandwidth, latency, and real world performance.

It's just a different way of doing things. Perhaps not as exciting as it once was where we'd get excited as enthusiasts about particular steppings and how much we could tune and OC systems 10, 20, even as much as 50 percent beyond stock performance. That just isn't the way Zen2 behaves. And that's fine, arguably it's superior due to not needing silicon lottery and heavy effort and combining with OC style mobos to achieve the results you want.

7nm TSMC is extremely dense, and limits seem based around areal density causing hotspotting that cannot be effectively radiated away from such miniscule surface areas. AMD has tuned and binned these SKUs to not leave basically anything on the table from stock. Treating a Zen2 like it's Sandy Bridge or even Pile-driver, and yoinking the volts and current up in a futile pursuit of a meaningful OC is just risking degradation. You can't outrun basic physics, silicon electromigration is very real. And at a certain point any CPU die will degrade, as with SNDS and other memorable examples.

It's a new paradigm. Zen2 SKUs are class leading without the need or even practical benefit from any manual OC, especially in regards to any more than stock power. The tinkerer and tweaker community among us will have to be content with amazing CPUs that are already optimized for peak performance out of the box, and focus on areas where effort actually pays, eg; DDR timing, GPU tuning, etc.

With Zen 1 you could definitely see gains (example: my 1950X can be overclocked to 4.25 GHz if I wanted to). However, with Zen+ and especially Zen 2, it's less productive. Even with Zen 1 I can't be bothered to overclock anymore. It doesn't surprise me that Buildzoid has degraded his CPU. All guidance I've seen indicates that the voltage needs to be kept under 1.325V, and ideally under 1.3V, for safe operation of the CPU.