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

Buildzoid's 3700x Static Overclock Degrades Processor

In my world I call this common sense. On TPU, we debate that common sense whenever someone gets fed a reality that doesnt sit well with them :)
Haha, when I talk about doing any sort of overclocking I jab myself a little bit. Make a monkey out of myself in my head, and when I talk about it. Preparing myself, I guess. In my mind whatever I'm doing is going to fry the chip. I abused my poor 2600 heavily... knowing I was fine with doing an impromptu upgrade. It ended up doing better than most I've seen... but I spent a lot of time up in the 90's pushing over 1.5v. You can't go poking around for voltage and thermal limits (which... if you're really overclocking, you're absolutely gonna) without also realizing that you're risking your hardware.

I think sometimes I wanted to fry it just so I'd have a reason to buy a new chip. But you always gotta know that. And maybe it's best to make that clear somehow when we talk about overclocking.

My 3900x is a different story. I don't want to buy another one of those. Plus... it's kinda already fairly badass at stock. I enjoy overclocking... sometimes more than casually, but with this one I just wasn't feeling it. I appreciate the engineering that went into the arch they built up and all of the interlocking systems for preserving it and exacting better performance than I might get from my most thorough and tedious overclock efforts. In stead of OCing it I just let it ride, see what it does, and learn more about it.

Overclocking is a way to engage with the hobby, I suppose. But there are so many other ways. You ain't gotta do it! And if you want to, you better know that's not what these are made for!
 
All heavy Overclocks and voltage increases eventually degrade a CPU regardless of who makes the CPU. But won't happen overnight. I've had a FX8350 heavily OCed with a voltage bump. Worked like a charm for more than 5 years sold it and it's still working like a charm or a champ. Lol
 
@Super XP what's that "(RDNA2 7nm+ Prototype) " in your system specs?....
 
As I said posts ago , there are a few here I could name have a few Ryzens set up with normal air cooling ,no overclock and they run these systems hard all day everyday, we will more than adequately hear about that IF and WHEN it happens.

Originally when I heard about 7nm it was just round the corner , it wasn't.

The 7nm we have now is not the same as that envisioned years ago ,plans adapt , the results of research further alter design and process.

No one is white knighting for him , he has a presence here , he could speak for himself.

But I'd wager most Here have tried your way , I didn't say it didn't work, I argued it's too faffy and is not a safe way long term, and not necessary. IMHo

And I don't think this drama regarding degradation while out of spec is necessary, you say you're an old school clocker ,all older chip's also degraded if you were far enough out of spec ie trying hard enough.
You are making an assumption there, and that assumption is that I am so much of a n00b and a (L)user that when I said that I had been experimenting with the Ryzen 3000 series of CPUs that I had NOT tested with an air cooler.

The air-cooler I used in my experiments was the Noctua NH-U12A and in my experimentation I discovered a little trick to increase the cooling performance of the cooler which you can see in the attached picture and that I will describe here and also share my experience with the cooler.

If you look at the picture, I attached the fans so that they were flush with the top of the cooler and underhanging the bottom of the fin-stack. This is because it prevents a pocket of dead air forming between the cold-plate and the fin-stack which would result in the cold-plate getting hotter than it would need to be. When I have seen others set up this cooler they always have the fans poking over the top of the fin-stack which is absolutely useless for cooling.

From my testing, I would say that if you want to have the best, and best value, cooling for a single chiplet Ryzen 3000 system then you should not be considering an AIO.

The reason for this is pretty simple, with an AIO less than 25% of the effective cooling potential (where the fins are that are transferring heat to the liquid) covers the hotspot of the single chiplet CPU. By contrast the Noctua NH-U12A has half of its seven heatpipes carrying heat away from the hotspot.

A 280 rad AIO will not be able to keep up with the NH-U12A

I say this because on the single chiplet CPU I had to test with - the 3600X - it was only when I used an AlphaCool Eisbaer 360 that I was getting slightly better cooling performance than with the air cooler. I have not seen any comparisons between a good air-cooler and an AIO with regard to cooling a single chiplet Ryzen 3000 series CPU.

I replaced the fans that come with the AlphaCool Eisbaer 360 with Noctua NF-a12x25 fans - probably the best fans you can get for quietness and performance.

The test I ran was quite simple, I cranked the 3600X to 4.125 GHz at 1.29375 Volts on all threads and ran Prime95 running with small fft for three hours on each and then monitored the temps. The TIM I used was TG Kryonaut. I also used the spread method to apply the TIM because the traditional pea sized or grain of rice sized blob on the centre of the CPU doesn't completely cover the hotspot caused by the fact that the chiplet of the Ryzen CPU is off centre.

The picture is of the test with the Noctua cooler, I didn't make one for the AIO because that was the one I tested second. There was around 4 °C in it with the AIO peaking at 80 °C CPU temp.

In each of the tests, the NF-A12x25 fans which I was using on both coolers were at max RPM
 

Attachments

  • 20191207_223914.jpg
    20191207_223914.jpg
    4.9 MB · Views: 231
  • prime95_4125_3hrs.PNG
    prime95_4125_3hrs.PNG
    812.5 KB · Views: 248
Last edited:
Can we all just stop arguing

You are making an assumption there, and that assumption is that I am so much of a n00b and a (L)user that when I said that I had been experimented with the Ryzen 3000 series of CPUs that I had NOT tested with an air cooler.

The air-cooler I used in my experiments was the Noctua NH-U12A and in my experimentation I discovered a little trick to increase the cooling performance of the cooler which you can see in the attached picture and that I will describe here and also share my experience with the cooler.

If you look at the picture, I attached the fans so that they were flush with the top of the cooler and underhanging the bottom of the fin-stack. This is because it prevents a pocket of dead air forming between the cold-plate and the fin-stack which would result in the cold-plate getting hotter than it would need to be. When I have seen others set up this cooler they always have the fans poking over the top of the fin-stack which is absolutely useless for cooling.

From my testing, I would say that if you want to have the best, and best value, cooling for a single chiplet Ryzen 3000 system then you should not be considering an AIO.

The reason for this is pretty simple, with an AIO less than 25% of the effective cooling potential (where the fins are that are transferring heat to the liquid) covers the hotspot of the single chiplet CPU. By contrast the Noctua NH-U12A has half of its seven heatpipes carrying heat away from the hotspot.

A 280 rad AIO will not be able to keep up with the NH-U12A

I say this because on the single chiplet CPU I had to test with - the 3600X - it was only when I used an AlphaCool Eisbaer 360 that I was getting slightly better cooling performance than with the air cooler. I have not seen any comparisons between a good air-cooler and an AIO with regard to cooling a single chiplet Ryzen 3000 series CPU.

I replaced the fans that come with the AlphaCool Eisbaer 360 with Noctua NF-a12x25 fans - probably the best fans you can get for quietness and performance.

The test I ran was quite simple, I cranked the 3600X to 4.125 GHz at 1.29375 Volts on all threads and ran Prime95 running with small fft for three hours on each and then monitored the temps. The TIM I used was TG Kryonaut. I also used the spread method to apply the TIM because the traditional pea sized or grain of rice sized blob on the centre of the CPU doesn't cover the hotspot caused by the fact that the chiplet of the Ryzen CPU is off centre.

The picture is of the test with the Noctua cooler, I didn't make one for the AIO because that was the one I tested second. There was around 4 °C in it with the AIO peaking at 80 °C CPU temp.

In each of the tests, the NF-A12x25 fans which I was using on both coolers were at max RPM

I use a Scythe Ashura myself, handles this "old hot chip"
 
Can we all just stop arguing



I use a Scythe Ashura myself, handles this "old hot chip"
I am not arguing, I am answering accusations.

There's a difference.

I use a Scythe Ashura myself, handles this "old hot chip"

The closest I come to being a FanBoi is with Noctua.

I bought one of the first fans they ever brought out about 15 years ago, and that fan (a 140mm that can can fit on 120 mm mount) is still running today in my friend's system and is as quiet today as it was back then.

I have had a number of fans from a number of manufacturers and I have had them fail or become rattly or in some other way noisy.

Of all the Noctua fans I have bought over the past 15 years I have never had even one get noisy, never mind fail.

I moved from the Noctua NF-F12 fans to the NF-A12x25 fans I now have for my systems and I have given away those fans to people, who use them and can't believe that I had been using those fans for five or more years.

I don't care about the colour, I care deeply about reliability, performance and above all else quietness. Noctua gives me the best of all those worlds.

Below is my current main system (i7-4790K transplanted into a Phanteks Evolv X case), all of the fans in there are NF-A12x25 and even under load, the system is silent.
 

Attachments

  • 20190119_034143.jpg
    20190119_034143.jpg
    4.5 MB · Views: 171
Last edited:
I am not arguing, I am answering accusations.

There's a difference.



The closest I come to being a FanBoi is with Noctua.

I bought one of the first fans they ever brought out about 15 years ago, and that fan (a 140mm that can can fit on 120 mm mount) is still running today in my friend's system and is as quiet today as it was back then.

I have had a number of fans from a number of manufacturers and I have had them fail or become rattly or in some other way noisy.

Of all the Noctua fans I have bought over the past 15 years I have never had even one get noisy, never mind fail.

I moved from the Noctua NF-F12 fans to the NF-A12x25 fans I now have for my systems and I have given away those fans to people, who use them and can't believe that I had been using those fans for five or more years.

I don't care about the colour, I care deeply about reliability, performance and above all else quietness. Noctua gives me the best of all those worlds.

Cool story, I like what I have, system I have will be from 6 years ago now.
 
Cool story, I like what I have, system I have will be from 6 years ago now.

You choose what you want and I choose what I want, or, as my grandfather used to say, "Everybody should be allowed to go to hell in their own way" :)

The thing is that I am still very content with my i7-4790K and there is a lot of useful life left in it, I don't NEED to build the 3950X system I have, I want to. The 4790K has been up and running now for just over 1,750 days 24/7 which is about 4.8 years. Even eight year old systems can still do a reasonably good job, never mind six year old systems.

I have been pretty lucky with regard to the timing of my purchases; for instance the RAM I have in the system I got for £105 and it is now £170, or the 1TB NVME M.2 drives which I got for £120 which are now £170. I also got a good deal on the motherboard which I managed to get for £580 as opposed to the normal price which is around £700.

I also managed to pick up an almost brand new EVGA 1080 Ti ICX FTW3 for £480.

Because I don't need the Ryzen 3000 system, I can have fun experimenting with it and can set it up to run optimally 24/7 when I finally switch over to it as my main machine.
 
You choose what you want and I choose what I want, or, as my grandfather used to say, "Everybody should be allowed to go to hell in their own way" :)

The thing is that I am still very content with my i7-4790K and there is a lot of useful life left in it, I don't NEED to build the 3950X system I have, I want to. The 4790K has been up and running now for just over 1,750 days 24/7 which is about 4.8 years. Even eight year old systems can still do a reasonably good job, never mind six year old systems.

I have been pretty lucky with regard to the timing of my purchases; for instance the RAM I have in the system I got for £105 and it is now £170, or the 1TB NVME M.2 drives which I got for £120 which are now £170. I also got a good deal on the motherboard which I managed to get for £580 as opposed to the normal price which is around £700.

I also managed to pick up an almost brand new EVGA 1080 Ti ICX FTW3 for £480.

Because I don't need the Ryzen 3000 system, I can have fun experimenting with it and can set it up to run optimally 24/7 when I finally switch over to it as my main machine.
One persons 24/7 doesn't always equal another's.

for example.
Setup like a Don.png
 
I use a Scythe Ashura myself, handles this "old hot chip"

The closest I come to being a FanBoi is with Noctua.

I bought one of the first fans they ever brought out about 15 years ago, and that fan (a 140mm that can can fit on 120 mm mount) is still running today in my friend's system and is as quiet today as it was back then.

I have had a number of fans from a number of manufacturers and I have had them fail or become rattly or in some other way noisy.

Of all the Noctua fans I have bought over the past 15 years I have never had even one get noisy, never mind fail.

I moved from the Noctua NF-F12 fans to the NF-A12x25 fans I now have for my systems and I have given away those fans to people, who use them and can't believe that I had been using those fans for five or more years.

I don't care about the colour, I care deeply about reliability, performance and above all else quietness. Noctua gives me the best of all those worlds.
 
the NF-A12x25 fans I now have for my systems and I have given away those fans to people, who use them and can't believe that I had been using those fans for five or more years.

The Noctua NF-A12x25 fans are available since May 2018......
 
The Noctua NF-A12x25 fans are available since May 2018......

Five years, two years, what's the difference? :laugh:

I do think they were referring to the NF-f12s and not the a12s.
 
Let me assure you, Nvidia wakes up with a cold sweat every morning all because of RDNA2. :D

Actually, I think that liquid is from half open bottles of the finest wine that were tossed aside after spending the night with the finest super models, could be the clap too I suppose, since they can practically print money. The only thing that Nvidia worries about is how dismal the actual AMD gpu product is compared to the hype.

At least AMD took a step forward with Navi. It might be a little too ambitious, or naive, to think they can claw it all back in this gen.
 
Actually, I think that liquid is from half open bottles of the finest wine that were tossed aside after spending the night with the finest super models, could be the clap too I suppose, since they can practically print money. The only thing that Nvidia worries about is how dismal the actual AMD gpu product is compared to the hype.

At least AMD took a step forward with Navi. It might be a little too ambitious, or naive, to think they can claw it all back in this gen.
So err ,nah.
 
So err ,nah.

If AMD can get remotely close, I'll likely buy their top GPU come later this year. But they are already like 40% or more behind the 2080 ti and they have to face everything that the jump to 7nm will bring. The only reason AMD is even competitive right now is because of the process advantage. AMD has a huge hole to climb out of. I would be pretty surprised, and excited, if they could pull it off this gen. I don't really expect parity (raw performance and efficiency) until the next gen.
 
Actually, I think that liquid is from half open bottles of the finest wine that were tossed aside after spending the night with the finest super models, could be the clap too I suppose, since they can practically print money. The only thing that Nvidia worries about is how dismal the actual AMD gpu product is compared to the hype.

At least AMD took a step forward with Navi. It might be a little too ambitious, or naive, to think they can claw it all back in this gen.
We will have to wait and see. Just don't underestimate RDNA2 and the importance of next generation gaming consoles that's going to use this GPU design.
 
One persons 24/7 doesn't always equal another's.

for example.
View attachment 147218

Thank you for giving me an idea.

Seeing your results it occurred to me to see what the performance would be if I confined the 3950X to just using one CCD (8 Cores/16 Threads) so after setting it up in the BIOS and in Ryzen Master, I then decided to see how high I could clock that virtual 3800X to run stably and then run CineBench on a loop.

As you can see from the picture, the CPU is using around 85 Watts and also that the cores are maxed out as you can see from the Task Manager. At the time I took the snapshot, CineBench had been running on a loop for over three hours (one run takes about 57 seconds) and it is cranked up to 4.425 GHz (although for some reason the Task Manager reads 4.41 GHz) at 1.29375 Volts.

My ambient temperature was at 25 °C as measured by a thermal probe just in front of the intake fans of the radiator.

That was an interesting experiment and I am not being snarky when I say thank you for planting that idea in my head.

If you had asked me beforehand I would have said that I would not have been able to go above 4.3 GHz because that is the limit when I am running it as a 16 Core/32 Thread CPU.
 

Attachments

  • Cinebench over 3 hours.PNG
    Cinebench over 3 hours.PNG
    1.1 MB · Views: 229
Last edited:
Thank you for giving me an idea.

Seeing your results it occurred to me to see what the performance would be if I confined the 3950X to just using one CCD (8 Cores/16 Threads) so after setting it up in the BIOS and in Ryzen Master, I then decided to see how high I could clock that virtual 3800X to run stably and then run CineBench on a loop.

As you can see from the picture, the CPU is using around 85 Watts and also that the cores are maxed out as you can see from the Task Manager. At the time I took the snapshot, CineBench had been running on a loop for over three hours (one run takes about 57 seconds) and it is cranked up to 4.425 GHz (although for some reason the Task Manager reads 4.41 GHz) at 1.29375 Volts.

My ambient temperature was at 25 °C as measured by a thermal probe just in front of the intake fans of the radiator.

That was an interesting experiment and I am not being snarky when I say thank you for planting that idea in my head.

If you had asked me beforehand I would have said that I would not have been able to go above 4.3 GHz because that is the limit when I am running it as a 16 Core/32 Thread CPU.
I'm glad you're happy.

Keep in mind that the screeny I showed is typical after two days of being at max load.

The coolant loop is both saturated as much as it gets and regulated for that 80 CPU max temp , for short term loads from zero at low ambient temperature it would boost higher.
 
Back
Top