• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."

Question Speaking of Ryzen 3000 CPUs longetivity and voltages

birdie

Member
Jan 12, 2019
98
71
51
Currently there are isolated reports that Ryzen 3000 CPUs are dying or crashing on people after going through some serious stress testing/intensive workload. I'd love to ask people with serious science/engineering background why the new 7nm AMD CPUs run at a such a high voltage (1.45V), also given the fact that e.g. the Sandy Bridge CPUs from eights ago, based a substantially less advanced node (32nm), run at ~1.2V. I'm an absolute no one in this field but from what I've heard, the higher the voltage the shorter the lifetime of a CPU is. So, what could this high voltage mean for the new CPUs lifetime?

Somewhere from the web:

I experienced a once or twice a day thermal shutdown. My cpu was running at 80°C. 24/7 at 4100-4200Mhz. I tried down clocking and it made no difference, even to base clock 3800Mhz. Finally I thought what I had I changed between the 2700X and the 3900X. Answer: I switched to the 256-bit AVX application because it was faster by 3 minutes than the SSE41 app which I had always used for 1800X or 2700X. The AVX was handicapped on the earlier Ryzens and the SSE41 app always beat it in performance. So I reverted to the SSE41 app and immediately dropped 15° C. I would like to run the AVX app but I cannot with my current cooling solution which is a 360mm rad and custom block. So SSE41 app it is. System has stayed crunching for over a day now with no issues.
 

EXCellR8

Diamond Member
Sep 1, 2010
3,887
788
136
Smaller die = less tolerance to higher voltage, is what I figure. There could be quite a number of factors however.

I cut off my forced voltage at around 1.42v because that's already much higher than what AMD recommends for a 'safe' constant voltage.
 

Hitman928

Diamond Member
Apr 15, 2012
3,576
3,939
136
1st, you quoted story doesn't make any sense. You can't down clock and not have temperatures go down unless the CPU was thermal throttling at both clocks. There's no reason it should thermal throttle with a custom water setup unless something is wrong with the setup. Since it's a custom block and we now have this chiplet design, it's very possible that the custom block isn't covering the chiplet properly and leaving hotspots somewhere across the processor.

2nd, Without getting into the physics of it all too much, yes, as a general rule the higher the voltage the shorter the lifespan. With that said, modern processes are pretty durable. The key though is the duration the voltage is applied. Higher voltage in short time spans is perfectly fine. Higher voltage applied constantly = early death. So AMD set up the boost to be able to apply high voltages to try to reach high frequencies for lightly threaded workloads. Modern schedulers will bounce around the load across various cores as time goes forward so you never have a sustained high voltage load on a single core for an extended period of time. If you are maxing out the cores, the voltage should be lower as you are constantly using all cores. If you manually overclock you should be more in line with the all cores used boost voltage rather than the single core used voltage.
 

Tup3x

Senior member
Dec 31, 2016
526
392
136
Hardware Unboxed guys did manage to kill their 3900 a little too easily. It might have been a dud though. I haven't seen enough reports about Ryzen 3000 series dying to be we worried. I'm pretty sure that AMD knows what they are doing. That being said there may be buggy BIOSes on some boards that may cause nasty stuff. There are still some bugs that needs to be squashed.
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
21,415
9,476
136
I run mine 24/7@100% and vcore runs average 1.27v. I have 2 3900x's, not a big sample, but no problems.
 
  • Like
Reactions: Drazick

EXCellR8

Diamond Member
Sep 1, 2010
3,887
788
136
I think we're talking upwards of 1.4v though... 1.27 is fine for all cores @ 100%
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
21,415
9,476
136
I think we're talking upwards of 1.4v though... 1.27 is fine for all cores @ 100%
Yes, without PBO enabled (I think thats it) I run all cores@100%@1.27 vcore running 4 ghz speed for 142 watts,95a TDC, 140a EDC.

Its faster than any of my 2700x's or 2xxx series Threadrippers. 240mm AIO@75c
 
  • Like
Reactions: Drazick

Thala

Golden Member
Nov 12, 2014
1,256
569
136
I'd love to ask people with serious science/engineering background why the new 7nm AMD CPUs run at a such a high voltage (1.45V)
Thats because 1.45V is required to reach the advertised turbo frequency whenever no other limitation applies. The voltage required might change from die to die and is fused at factory testing.

I'm an absolute no one in this field but from what I've heard, the higher the voltage the shorter the lifetime of a CPU is. So, what could this high voltage mean for the new CPUs lifetime?
1.45V in itself should not damage the chip as long as the temperatures are within reasonable ranges.
 

Topweasel

Diamond Member
Oct 19, 2000
5,384
1,582
136
Thats because 1.45V is required to reach the advertised turbo frequency whenever no other limitation applies. The voltage required might change from die to die and is fused at factory testing.



1.45V in itself should not damage the chip as long as the temperatures are within reasonable ranges.
Easy to say that. But let's remember that it was anything above 1.35 that saw rapid deterioration of Zen CPU's. Not to say these are the same CPU's, different design, at different process, from a different manufacturer. Still I wouldn't be confident that is a safe voltage for consistent usage even if the CPU's spike that high with PBO.

Honestly I don't see a huge benifit to OC'ing. Even PBO isn't going to make much of a difference. PBO isn't going to make something seem faster and by the time you need the extra performance the CPU will be slow for the task either way. Maybe I am just an old timer. When I overclocked it really helped out. Now not so much. I should be getting my board today, when I get it built, I'll be turning on PB2 (if not on by defualt) and calling it a day.
 

Thala

Golden Member
Nov 12, 2014
1,256
569
136
Honestly I don't see a huge benifit to OC'ing. Even PBO isn't going to make much of a difference. PBO isn't going to make something seem faster and by the time you need the extra performance the CPU will be slow for the task either way. Maybe I am just an old timer. When I overclocked it really helped out. Now not so much. I should be getting my board today, when I get it built, I'll be turning on PB2 (if not on by defualt) and calling it a day.
I agree to this - i do not advertise overclocking anyway as the returns (e.g. the effective gains) are typically very low. And of course chances of local current spikes that are higher than with lower voltages that could damage the chip are higher of course.
 

DrMrLordX

Lifer
Apr 27, 2000
17,467
6,474
136
Currently there are isolated reports that Ryzen 3000 CPUs are dying or crashing on people after going through some serious stress testing/intensive workload. I'd love to ask people with serious science/engineering background why the new 7nm AMD CPUs run at a such a high voltage (1.45V), also given the fact that e.g. the Sandy Bridge CPUs from eights ago, based a substantially less advanced node (32nm), run at ~1.2V. I'm an absolute no one in this field but from what I've heard, the higher the voltage the shorter the lifetime of a CPU is. So, what could this high voltage mean for the new CPUs lifetime?
First off, the voltages that run in default mode seem to be inflated on non-AVX/AVX2 workloads. When I run Prime95, my voltages drop like a rock. That's stock (no PBO) on a 3900x. So whatever that quoted person was doing . . . was strange, and probably involved static overclocking. Even with PBO, I can't get voltages to run very high in Prime95 small FFTs.

When I did run Prime95 small ffts @ 4.25 GHz and 1.25v it got hot, fast.

You can't down clock and not have temperatures go down unless the CPU was thermal throttling at both clocks.
Depends on cooling. On my setup the CPU doesn't go up in temp by all that much when I raise clocks in a static OC. Voltage is another story.

Approxmiate data when running CBR20 MT and 1.25v:

4 GHz: 55C
4.1 GHz: 56C
4.2 GHz: 56C
4.3 GHz: 57C

With some minor variation.

There's no reason it should thermal throttle with a custom water setup unless something is wrong with the setup.
CPU temp seems pegged to hotspot sensors (I think?) so it's easier than you might think. I have mine with a nice block (Heatkiller IV pure copper) on a massive rad (MO-RA3) and two d5 pumps, and the CPU runs hotter than my Radeon VII!

Hardware Unboxed guys did manage to kill their 3900 a little too easily.
What did they do?

I run mine 24/7@100% and vcore runs average 1.27v. I have 2 3900x's, not a big sample, but no problems.
Is that running stock, or is that a static OC? If you're doing 4 GHz you can probably run that with a lower voltage.
 

Hitman928

Diamond Member
Apr 15, 2012
3,576
3,939
136
Depends on cooling. On my setup the CPU doesn't go up in temp by all that much when I raise clocks in a static OC. Voltage is another story.

Approxmiate data when running CBR20 MT and 1.25v:

4 GHz: 55C
4.1 GHz: 56C
4.2 GHz: 56C
4.3 GHz: 57C

With some minor variation.
There's many factors in cooling a CPU, but as an example you're showing a smaller clock spread and have 2 C of temperature delta. The reported post had a higher clock spread with 0 C delta. Again, not possible unless the CPU is actually throttling down below the 3.8 GHz tested. I guess the only other option would be if you had the cooling system set to ramp right at 80C which would be a crazy high temp to have an inflection point at, especially running on water.
 

Tup3x

Senior member
Dec 31, 2016
526
392
136
So, you have no idea.. got it.
Mmm? Just saying their sample died during their testing. Can't check why because on my phone and it will be a while until I can. Feel free to read TechSpot's Ryzen article (it was probably mentioned in there too) or watch the Hardware Unboxed video review.

Edit: Was easy enough to Google.
On an even more disappointing note, we somehow managed to end the life of our 3900X sample at this stage of the review. We don’t recall exactly what settings were applied, but we know we hadn’t manually adjusted voltages yet. We believe after testing the 4.3 GHz overclock with auto voltage, we increased the LLC to see what impact that had on temperatures and during our first CB20 pass the system crashed and reset, and never booted up again.
TechSpot Ryzen 3900X review.

So they aren't exactly sure what happened. Obviously I can't be either.
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
21,415
9,476
136
Mmm? Just saying their sample died during their testing. Can't check why because on my phone and it will be a while until I can. Feel free to read TechSpot's Ryzen article (it was probably mentioned in there too) or watch the Hardware Unboxed video review.

Edit: Was easy enough to Google.

TechSpot Ryzen 3900X review.

So they aren't exactly sure what happened. Obviously I can't be either.
And it could even be the motherboard....NOT the CPU. maybe we find out later.
 
  • Like
Reactions: Drazick

Hans de Vries

Senior member
May 2, 2008
260
589
136
www.chip-architect.com
People will continue to kill CPU's because they don't take the usual proffesional precautions against electrostatics.

I remember an extreme case of a collegue killing a CPU in its socket just by pointing at it (in front of my eyes). A 3 cm spark between the top of his finger and the CPU did the work.
 
Last edited:

CHADBOGA

Platinum Member
Mar 31, 2009
2,043
700
136
People will continue to kill CPU's because they don't take the usual proffesional precautions against electrostatics.

I remember an extreme case of a collegue killing a CPU in its socket just by pointing at it (in front of my eyes). A 3 cm spark between the top of his finger and the CPU did the work.
Was your colleague Sheev Palpatine?
 

B-Riz

Golden Member
Feb 15, 2011
1,481
609
136
Yes, without PBO enabled (I think thats it) I run all cores@100%@1.27 vcore running 4 ghz speed for 142 watts,95a TDC, 140a EDC.

Its faster than any of my 2700x's or 2xxx series Threadrippers. 240mm AIO@75c
That's about what I get on the 3900X all core heavy load.

There is nothing wrong with these chips (I have 3600 and 3900X), I think a lot of people are new / used to how Intel chips work and OC.
 

DrMrLordX

Lifer
Apr 27, 2000
17,467
6,474
136
There's many factors in cooling a CPU, but as an example you're showing a smaller clock spread and have 2 C of temperature delta. The reported post had a higher clock spread with 0 C delta. Again, not possible unless the CPU is actually throttling down below the 3.8 GHz tested. I guess the only other option would be if you had the cooling system set to ramp right at 80C which would be a crazy high temp to have an inflection point at, especially running on water.
Okay, you have a point there. There are some people out there with temp-based fan profiles though. I certainly wouldn't put 80C as my target temp.
 

Tup3x

Senior member
Dec 31, 2016
526
392
136
And it could even be the motherboard....NOT the CPU. maybe we find out later.
Could be (combined with 7nm process, probably doesn't like extreme voltage spikes - LLC can do pretty hc stuff). It will take a while for things to work perfectly. There are quite a bit of little issues and weird behaviour.
 

Abwx

Diamond Member
Apr 2, 2011
9,150
996
126
Somewhere from the web:
That s all BS, either incompetence if not plain viral marketing, Total system power is 216W with Cinebench MT and 199W for Prime 95 including AVX2, so how could the guy have higher temp with AVX2..?.

https://www.computerbase.de/2019-07/amd-ryzen-3000-test/4/#diagramm-test-leistungsaufnahme-max-avx-last-windows-10-komplettes-system


Btw, we can also see that AMD s FPUs are vastly more efficient than Intel s, guess that it s what motivated this denigrating campaign, desperate times require as much desperate answers..
 

ASK THE COMMUNITY