• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Buildzoid's 3700x Static Overclock Degrades Processor

You can do this indirectly by capping current while having middle scalar values (x3-5) and low level Vdroop (high LLC setting).
PS: look at it this way, there is a vid table in which FIT upper limit is presumably higher than the base level of the cpb table.
Sometimes I wonder just who writes these software for AMD. Same with their gpus. Is it hard to run them at near threshold voltage when you have adaptive clock stretching because all sorts of issues are going to be resolved at the lower spectrum of stock boundaries. Just who runs 1.2v when the score would be the same at 1.05v with a clear conscience it wouldn't break anything?
 
Last edited:
Dont remember if TDP ever was anywhere close to what the chips actually use.
its more or less an indicator whats needed to run it properly (no oc) when it comes to power/cooling,
and easier to drop in an upgrade as long as tdp is similar.

and even with strong chip binning by amd, lottery etc will still have a big impact on what V is safe, or isnt.
the first 3600 i got was running 1.44 under all core full load (max power), the second one is doing same clocks but at 1.33v
all with pbo/xfr etc off, only PB on(as per pbo=off), everything on auto except ram.
no difference which board i used (msi gaming or aorus ultra).
doubt the second one would survive long going with the same V the first one needed.
 
Dont remember if TDP ever was anywhere close to what the chips actually use.
its more or less an indicator whats needed to run it properly (no oc) when it comes to power/cooling,
and easier to drop in an upgrade as long as tdp is similar.
For sure it’s not the total power draw of the entire package.
TDP for Intel CPUs is what the chip will dissipate towards IHS when runs at base clock.
TDP for AMD is what CPU will dissipate towards IHS on average boost clocking (not base clock, not max all core clock).

My 65W TDP rated 3600 is allowed (by stock limits) to draw 88W (PPT).
The higher 105W TDP SKUs are limited to 142W (PPT). And I think the 95W TDP ones are limited to 125W (PPT).
 
My 65W TDP rated 3600 is allowed (by stock limits) to draw 88W (PPT).
The higher 105W TDP SKUs are limited to 142W (PPT). And I think the 95W TDP ones are limited to 125W (PPT).

This is correct.
 
I'm not exactly surprised. Ryzens (2) are pushed to the top from the factory. Pushing them over leads to degradation. Who'd have thought?

This. And note... Intel is not going to be very different when they do move to a smaller node. We saw this coming already since 22nm Ivy Bridge that couldn't hold an OC candle to 32nm, even just because concentration of heat on smaller surface area. Add smaller transistors and you've got a nasty cocktail of heat and power.
 
I totally agree, my 3770K is by far the hottest cpu that I’ve owned, and puts out very little actual heat.
 
Healthy long-term allowed voltage is going down the smaller you go with the manufacturing process. The other side of voltage issue is that the minimum voltage to make transistors work tends to not decrease as fast or increase, making the viable voltage range smaller and smaller.

I seem to remember there was an actual document behind 1.325V maximum recommended for TSMCs 7nm but when trying to Google it now, I cannot find these references any more. In an empirical way, this seems to be about the right number.
 
He knows that, and he posted it in a screenshot up above. Upper 1.3V range.

Please read threads before replying to them...
That's not extreme.
This is a daily OC:


This is extreme:


Pretty sure 1.37v is for a daily OC. :rolleyes:

Also, Buildzoid has no extreme subs with 3700x that I can see.
 
Last edited:
Easy..... let's play nice.
 
Let the record show Buildzoid has no extreme subs with 3700x.

Doesn't look like he was abusing it to me.
 
Did I say it was or are you just misrepresenting me so you can flaunt your irrelevant CPU-Z-peen?

As far as CPU-z-peen goes, that's not even close to a semi-.
That's jump into a cold lake in winter cpu-z-peen.
 
Last edited:
That's not extreme.
This is a daily OC:


This is extreme:


Pretty sure 1.37v is for a daily OC. :rolleyes:

Also, Buildzoid has no extreme subs with 3700x that I can see.

They key difference here is your daily example is a 2700x which used an entirely different process and has no direct meaning to a 3700x.
 
That is not 1.37v, but presumably as high as 1.47v due to msi defaulting to FPU vid.

AMD says up to 1.5 is OK.
The 3**** chip I ran briefly registered up to 1.46 on all auto settings.

They key difference here is your daily example is a 2700x which used an entirely different process and has no direct meaning to a 3700x.
So they're not both AMD 8-core CPUs only 1 generation apart?
 
Forgive my ignorance, but I was under the impression that the TDP was merely "defined" by the type of stock cooler AMD ships with each processor.
AMD says up to 1.5 is OK.
The 3**** chip I ran briefly registered up to 1.46 on all auto settings.


So they're not both AMD 8-core CPUs only 1 generation apart?

Yes but one has 7nm lithography and the other is refined 14nm lithography. Using the same IHS means that there is a vast difference between the two chips. A better comparison would be the 1700 vs the 2700x those are much closer than the 3700x to either of them.
 
Yes but one has 7nm lithography and the other is refined 14nm lithography. Using the same IHS means that there is a vast difference between the two chips. A better comparison would be the 1700 vs the 2700x those are much closer than the 3700x to either of them.
I would use a 3000-series chip, but still trying to figure out how to shut off cores with this one like that one can do.
I see other people doing it.
 
So they're not both AMD 8-core CPUs only 1 generation apart?

As said, one is 12nm the other is 7nm. Early first gen 7nm at that. So, big difference.

I would use a 3000-series chip, but still trying to figure out how to shut off cores with this one like that one can do.
I see other people doing it.

The options are right in my UEFI for shutting down cores.
 
I would use a 3000-series chip, but still trying to figure out how to shut off cores with this one like that one can do.
I see other people doing it.

Use Ryzen Master you can disable cores directly in there. Even though I don't understand why you would want to do that.
 
Dont remember if TDP ever was anywhere close to what the chips actually use.
its more or less an indicator whats needed to run it properly (no oc) when it comes to power/cooling,
and easier to drop in an upgrade as long as tdp is similar.

and even with strong chip binning by amd, lottery etc will still have a big impact on what V is safe, or isnt.
the first 3600 i got was running 1.44 under all core full load (max power), the second one is doing same clocks but at 1.33v
all with pbo/xfr etc off, only PB on(as per pbo=off), everything on auto except ram.
no difference which board i used (msi gaming or aorus ultra).
doubt the second one would survive long going with the same V the first one needed.

What EMF maximum for any semiconductor needs to be taken with caution. I dont expect a 14/10 nm/7nm part to handle volts my 8350 can handle. Just as I dont expect my chip to handle volts a 8088 can...
 
AMD says up to 1.5 is OK.
The 3**** chip I ran briefly registered up to 1.46 on all auto settings.


So they're not both AMD 8-core CPUs only 1 generation apart?
No it does not apply to a 3000 what we know about Ryzen 1000/2000. As @kapone32 says 1000 and 2000 are similar achitectures and with similar dynamics. 3000 is a completely new stuff and apparently unkown to most.

While on auto these chips can hit 1.475~1.5V but this is only on idle or very low load and very low current. If you had gone through all posts here you would have read what I previously said.
What voltage anyone see when the CPU is on auto has nothing to do with a static voltage on a static speed. If the CPU on auto settings for a given load and speed is feeding cores with 1.35V it does not mean that you can set a static 1.35V for that speed. Even less (1.325V) voltage is not safe. The FIT controller of the CPU is flactuating speed and voltage hundreds of times within a sec to regulate silicon stress. Have it on static speed/voltage and this is like taking down all defenses. Degradation is at hand...

I keep saying... What users know about past CPUs and OC do not apply to Ryzen 3000. There is no free extra performance on these chips. Forget it...
The only things one can do is to further cool it down as low as possible (50C if possible) and play with PBO EDC values to restrict current and silicon stress. Only then it boost more by its own and still by a few %.
 
No it does not apply to a 3000 what we know about Ryzen 1000/2000. As @kapone32 says 1000 and 2000 are similar achitectures and with similar dynamics. 3000 is a completely new stuff and apparently unkown to most.

While on auto these chips can hit 1.475~1.5V but this is only on idle or very low load and very low current. If you had gone through all posts here you would have read what I previously said.
What voltage anyone see when the CPU is on auto has nothing to do with a static voltage on a static speed. If the CPU on auto settings for a given load and speed is feeding cores with 1.35V it does not mean that you can set a static 1.35V for that speed. Even less (1.325V) voltage is not safe. The FIT controller of the CPU is flactuating speed and voltage hundreds of times within a sec to regulate silicon stress. Have it on static speed/voltage and this is like taking down all defenses. Degradation is at hand...

I keep saying... What users know about past CPUs and OC do not apply to Ryzen 3000. There is no free extra performance on these chips. Forget it...
The only things one can do is to further cool it down as low as possible (50C if possible) and play with PBO EDC values to restrict current and silicon stress. Only then it boost more by its own and still by a few %.

Iirc ryzen 3000 is a mcm design.
 
Iirc ryzen 3000 is a mcm design.
Except that, even if it was a monolithic one, it would still be as discribed above. Its the 7nm proccess node primarily and the fact that AMD took advantage all headroom for speed and voltage. In a way that does not damage/degrades the silicon of course.
 
Back
Top