Question What kind of stable overclocks are you running on r7 3700x/3800x?

[Sipan9000]

I have a 3800x going at 4.4ghz all core @ 1.325 volts and it's running flawlessly about 30c idle and 70c max load with Cryorig r1 ultimate cooler, I haven't tried going lower on volts or higher frequency & I'm curious how mine stacks up to other overclocks.

How are your overclocks looking on these chips? I feel like I can push it to 4.5ghz or go under 1.3 volts.

Also has anybody here tried a single core overclock?

 

[zx128k]

You have won the silicon lottery with that 3800x. Most people consider 1.325 volts to be safe. [1] If you can run prime95 8k fft and are stable that is great. To reach 4.5GHz all cores you will enter the danger zone for voltage which is above 1.325 volts. [1] FIT safe voltage reddit It could take months to find out if your cpu degrades or not and if this is indeed correct.

Matisse	SSE Frequency	AVX2 Frequency	Voltage	% Capable
3700X	Not Tested	4.05GHz	1.237V	100%
3700X	Not Tested	4.10GHz	1.250V	Top 74%
3700X	Not Tested	4.15GHz	1.262V	Top 21%
3800X	Not Tested	4.20GHz	1.275V	100%
3800X	Not Tested	4.25GHz	1.287V	Top 53%
3800X	Not Tested	4.30GHz	1.300V	Top 20%

Silicon lottery website

An all core 4.4GHz overclock is very good. 1.325 volts vcore seems safe. Getting to 4.5Ghz all core could take a lot more voltage.

Personally I would overclock the IF to 1900, RAM to 3800 and tighten the RAM timings to CL15 or CL14. Hopefully you have a good Samsung b-die kit. If you are lucky you will get 11600+ in time spy cpu.

You will get more performance from the RAM overclocking, than increased core frequency.

This is my 3dmark score with just a RAM overclock, this should be considered a value towards the maximum. Time Spy CPU score 11460 and details link
With PBO limits set to motherboard and Scalar set to x10. Time spy cpu score on very cold day. 11628

Note temperature has a big affect on how far the cpu will boost. Scores between 11260 and 11500 are normal at stock with my RAM overclock. You have a faster all cores clock speed than I have, by at least 50-100MHz. I will drop to 4.3MHz or lower in the last part of the time spy cpu test which is the part that generates the final score.

My settings.

 

[drea.drechsler]

I have a 3800x going at 4.4ghz all core @ 1.325 volts and it's running flawlessly about 30c idle and 70c max load with Cryorig r1 ultimate cooler, I haven't tried going lower on volts or higher frequency & I'm curious how mine stacks up to other overclocks.

How are your overclocks looking on these chips? I feel like I can push it to 4.5ghz or go under 1.3 volts.

Also has anybody here tried a single core overclock?

How long does you hold with Prime95 small FFT's?

 

[drea.drechsler]

...Most people consider 1.325 volts to be safe..... the danger zone for voltage which is above 1.325 volts. [1] FIT safe voltage reddit ...

I have read where Stilt pretty much regrets having ever said that since it's been so completely misunderstood, and still used incorrectly.

 

[zx128k]

I have read where Stilt pretty much regrets having ever said that since it's been so completely misunderstood, and still used incorrectly.

This post,

"
AMD_Robert

Technical Marketing33 points · 4 months ago · edited 4 months ago

No, 1.325V is NOT THE MAX. This is conjecture by some random person on the internet for 24/7 overclocks. If the chip is running in stock condition, not overclocked, then boost up to 1.5V is completely normal. Please read the detailed brief linked in the PDF.
Discussions of overclocked voltages and stock operation should not be mixed. Users who see random boost up to 1.5V should not go set a static core voltage in response. This is not helpful. It's actually dangerous because it raises the average voltage the chip experiences over a 24H period.
Let me repeat for anyone who might read this in the future: voltages of 0.200V to 1.500V is normal for Ryzen. You will see anything in this range at the desktop. Any application is capable of triggering boost. Mouse input can trigger boost (user input is high-priority!). Your browser can do it. Your game launcher can do it. This is normal behavior. This is what the CPU is designed to do.

The-Stilt

52 points · 4 months ago

1.325

Speaking of the "broken telephone" effect...
The person who originally quoted my post either didn't fully understand what was being said, or alternatively didn't read it completely. Since then the misunderstood / misquoted post has been gaining momentum here at Reddit.

What was actually said:
"According to FIT, the safe voltage levels for the silicon are around 1.325V in high-current loads
and up to 1.47V in low-current loads (i.e ST), depending on the silicon characteristics."

Few key things here to understand:
"Depending on the silicon characteristics" is the most important one, and unfortunately it has been completely omitted from the quotes. The figures I mention in the sentence above are based on the average of the CPU specimens I had access at the time. Each and every single piece of silicon has its own unique characteristics and no two pieces of silicon are fully identical. Some of the silicon characteristics will greatly affect how much voltage the silicon, or in this case the CPU will require at given frequency, and especially how much voltage it can endure, without causing a permanent damage. Because of that the figures I mention might or might not fully apply to your specific CPU specimen.

Similar to previous Ryzen CPU generations, Ryzen 3000-series CPUs are also equipped with FIT-feature. FIT is an integrated silicon fitness monitoring feature, which exists in some form on most modern CPU and GPU designs, regardless of the vendor. Its function is to ensure that the performance of the product can be maximised, without allowing the reliability to decrease below a desired threshold. Reliability for the most part is voltage related. FIT will limit the CPU voltage below a point, where the silicon reliability would fall below the desired threshold. So unless the user manually overrides the voltage, there is no way for the CPU to operate at unsafe voltage levels.

A high-current load refers to a situation, where the CPU is essentially running "full beans". Generally very few consumer workloads can be considered as such. Consumer workloads, which come close to such scenario are mostly video encoding (e.g. X264 / X265) and rendering (e.g. Blender).

In low-current load scenarios, where 1-2 CPU cores are being utilised, both of the factors which are essential to and basically solely define the silicon reliability are drastically different: the current flowing through the CPU and the temperature which the CPU operates at, both which are significantly lower compared to a situation where the CPU is running "full beans" or even close to it.

In idle conditions, where the CPU cores, or at least parts of them are power gated most of the time (i.e. sleeping), the voltage becomes even less relevant. When an application or even a service running in the background wakes up a CPU core and causes it to boost to provide optimal performance, obviously there will be an increase in the voltage. Ryzen 3000 CPUs can run at extremely low voltages, but like on every single CPU in existence the higher frequencies require higher voltage to be fed to the CPU. Whenever there are aspects that can be improved in the current sleep-wake behavior remains to be seen. Personally, I have never seen any obvious anomalies in this regard.

In short: Leave the CPU alone, it knows what its doing. There is no guesswork involved when it comes to something as essential, as determining the default silicon characteristic specific voltage-frequency curve of a CPU."

 

[zx128k]

"Note that Caseking has pushed further than Silicon Lottery, having stock of a Ryzen 7 3700X processor cranked up to 4.3GHz at the time of writing – although the retailer doesn’t specify the voltage used, other than noting that it doesn’t exceed the safe limit of 1.4V. " https://www.techradar.com/news/amd-ryzen-3000-processors-could-disappoint-overclockers

There is a possibility I could do 4.5GHz all cores @ 1.4 volts vcore with a high LLC. That's time spy cpu 11700-11800 score range for my cpu and RAM OC.

Caseking english
"On the packaging of each pre-tested CPU, we also specify the core voltage specified by us and necessary for stable OC operation, with a tolerance range of approximately +/- 30 mV depending on the voltage supply of the mainboard used. However, in order to protect the silicon, the maximum core voltage is at most at harmless 1.4 volts. If you have questions about overclocking, the King-Mod service will be gladly available by e-mail after your purchase. "

 

[MrN1ce9uy]

In short: Leave the CPU alone, it knows what its doing. There is no guesswork involved when it comes to something as essential, as determining the default silicon characteristic specific voltage-frequency curve of a CPU."

This is the conclusion I've come to. My 2700X and 3700X both seem to perform best at stock settings.

 

[zx128k]

Considering my Ryzen 7 3700X runs at 1.5V out of the box, I'd say someone is either very wrong or my motherboard is very messed up. Come to think of it, I will be contacting ASUS and AMD to inquire about this.

The cpu will boost to high frequencies for single thread and light multi thread loads. That's were you get higher voltages like 1.5 volts appear. The main issue, what is a safe 24/7 overclock voltage were the voltage remains at the same level. Different sources have different ideas. The lowest value is 1.3-1.325 volts.

 

[MrN1ce9uy]

The cpu will boost to high frequencies for single thread and light multi thread loads. That's were you get higher voltages like 1.5 volts appear. The main issue, is what is a safe 24/7 overclock voltage were the voltage remains at the same level. Different sources have different ideas.

Yes, I read your other post in this thread and now I understand why I see 1.5V in HWMonitor and 1.47V average voltage while not gaming.

I can't get 4.4GHz all-core manual OC with my 3700X, and couldn't get 4.3GHz with my 2700X without setting either's voltage past 1.40-1.45V, so I just left OC'ing alone. *Using NZXT Kraken X62.

 

[zx128k]

Yes, I read your other post in this thread and now I understand why I see 1.5V in HWMonitor and 1.47V average voltage while not gaming.

I can't get 4.4GHz all-core manual OC with my 3700X, and couldn't get 4.3GHz with my 2700X without setting either's voltage past 1.40-1.45V, so I just left OC'ing alone. *Using NZXT Kraken X62.

You have to get lucky to get an all core 4.4GHz overclock with a 3700x. It is possible but unlikely.

3800x @ 4.65GHz

3800x 4.5GHz

 

[drea.drechsler]

"Note that Caseking has pushed further than Silicon Lottery, having stock of a Ryzen 7 3700X processor cranked up to 4.3GHz at the time of writing – although the retailer doesn’t specify the voltage used, other than noting that it doesn’t exceed the safe limit of 1.4V. " ....

If I recall, CaseKing also uses less 'stressful' stress test parameters by using a Prime95 version prior to the heavy AVX instructions and using fairly large FFT data sets. That's a lot more 'real-world', in that it more accurately reflects 'heavy' useage than the AVX-rich small FFT's does.

CaseKing alos uses (award winning overclocker) der8auer to do their binning though! so you got that going for you.

I seem to remember Silicon Lottery saying it's pretty much pointless to further bin Ryzen 3000 for overclocking. It seems AMD has already binned so well there's terribly little left for them.

 

[zx128k]

If I recall, CaseKing also uses less 'stressful' stress test parameters by using a Prime95 version prior to the heavy AVX instructions and using fairly large FFT data sets. That's a lot more 'real-world', in that it more accurately reflects 'heavy' useage than the AVX-rich small FFT's does.

CaseKing alos uses (award winning overclocker) der8auer to do their binning though! so you got that going for you.

I seem to remember Silicon Lottery saying it's pretty much pointless to further bin Ryzen 3000 for overclocking. It seems AMD has already binned so well there's terribly little left for them.

From the link I provided above.

"In order to explore the maximum clock values of the binned CPUs and to ensure their stability, der8auer tests each individual copy for at least one hour under Prime95 26.6 with 1344K. " - Caseking english

 

[DUAL33s]

I have a 3800x going at 4.4ghz all core @ 1.325 volts and it's running flawlessly about 30c idle and 70c max load with Cryorig r1 ultimate cooler, I haven't tried going lower on volts or higher frequency & I'm curious how mine stacks up to other overclocks.

How are your overclocks looking on these chips? I feel like I can push it to 4.5ghz or go under 1.3 volts.

Also has anybody here tried a single core overclock?

I own a 3700X and I run it on an Asus ROG Strix X570-E Gaming motherboard. I am able to run my chip at 4.3ghz @ 1.35v with LLC of 3 or 4. My chip is fully stable and performs best at LLC 3, but I can also push it as low as 4 if I feel like it. Also, this is on an 'air-cooled' system.

I suspect my chip can probably do 4.35ghz+ if I push a little more voltage into it. But Im happy with 4.3ghz/1.35v for now.

 

[DUAL33s]

This post,

"
AMD_Robert

Technical Marketing33 points · 4 months ago · edited 4 months ago

No, 1.325V is NOT THE MAX. This is conjecture by some random person on the internet for 24/7 overclocks. If the chip is running in stock condition, not overclocked, then boost up to 1.5V is completely normal. Please read the detailed brief linked in the PDF.
Discussions of overclocked voltages and stock operation should not be mixed. Users who see random boost up to 1.5V should not go set a static core voltage in response. This is not helpful. It's actually dangerous because it raises the average voltage the chip experiences over a 24H period.
Let me repeat for anyone who might read this in the future: voltages of 0.200V to 1.500V is normal for Ryzen. You will see anything in this range at the desktop. Any application is capable of triggering boost. Mouse input can trigger boost (user input is high-priority!). Your browser can do it. Your game launcher can do it. This is normal behavior. This is what the CPU is designed to do.

The-Stilt

52 points · 4 months ago

Speaking of the "broken telephone" effect...
The person who originally quoted my post either didn't fully understand what was being said, or alternatively didn't read it completely. Since then the misunderstood / misquoted post has been gaining momentum here at Reddit.

What was actually said:
"According to FIT, the safe voltage levels for the silicon are around 1.325V in high-current loads
and up to 1.47V in low-current loads (i.e ST), depending on the silicon characteristics."

Few key things here to understand:
"Depending on the silicon characteristics" is the most important one, and unfortunately it has been completely omitted from the quotes. The figures I mention in the sentence above are based on the average of the CPU specimens I had access at the time. Each and every single piece of silicon has its own unique characteristics and no two pieces of silicon are fully identical. Some of the silicon characteristics will greatly affect how much voltage the silicon, or in this case the CPU will require at given frequency, and especially how much voltage it can endure, without causing a permanent damage. Because of that the figures I mention might or might not fully apply to your specific CPU specimen.

Similar to previous Ryzen CPU generations, Ryzen 3000-series CPUs are also equipped with FIT-feature. FIT is an integrated silicon fitness monitoring feature, which exists in some form on most modern CPU and GPU designs, regardless of the vendor. Its function is to ensure that the performance of the product can be maximised, without allowing the reliability to decrease below a desired threshold. Reliability for the most part is voltage related. FIT will limit the CPU voltage below a point, where the silicon reliability would fall below the desired threshold. So unless the user manually overrides the voltage, there is no way for the CPU to operate at unsafe voltage levels.

A high-current load refers to a situation, where the CPU is essentially running "full beans". Generally very few consumer workloads can be considered as such. Consumer workloads, which come close to such scenario are mostly video encoding (e.g. X264 / X265) and rendering (e.g. Blender).

In low-current load scenarios, where 1-2 CPU cores are being utilised, both of the factors which are essential to and basically solely define the silicon reliability are drastically different: the current flowing through the CPU and the temperature which the CPU operates at, both which are significantly lower compared to a situation where the CPU is running "full beans" or even close to it.

In idle conditions, where the CPU cores, or at least parts of them are power gated most of the time (i.e. sleeping), the voltage becomes even less relevant. When an application or even a service running in the background wakes up a CPU core and causes it to boost to provide optimal performance, obviously there will be an increase in the voltage. Ryzen 3000 CPUs can run at extremely low voltages, but like on every single CPU in existence the higher frequencies require higher voltage to be fed to the CPU. Whenever there are aspects that can be improved in the current sleep-wake behavior remains to be seen. Personally, I have never seen any obvious anomalies in this regard.

In short: Leave the CPU alone, it knows what its doing. There is no guesswork involved when it comes to something as essential, as determining the default silicon characteristic specific voltage-frequency curve of a CPU."

Also, when most system monitoring apps report "1.48v" or "1.5v" it is onlt taking into consideration the HIGHEST voltage currently being pushed into any given core. Therefore under single threaded workloads, "1.5v" is only going to a single core/thread and therefore is not as "dangerous" as some ppl suspect. It's actually completely normal, when running your chip on 'auto' from what I understand.

 

[drea.drechsler]

Also, when most system monitoring apps report "1.48v" or "1.5v" it is onlt taking into consideration the HIGHEST voltage currently being pushed into any given core. Therefore under single threaded workloads, "1.5v" is only going to a single core/thread and therefore is not as "dangerous" as some ppl suspect. It's actually completely normal, when running your chip on 'auto' from what I understand.

Read back a few, the linked comments by the Stilt kind of explains what's happening.

When volts peaks into the 1.45-1.5 volt range it's in very low current density periods, with high clock speeds but processing a light load. High voltage with low current density is not what's harmful, it's the high voltage concurrent with high current as with an extremely heavy processing load that's harmful. The algorithm, everything in AUTO, will dial the volts and clocks right back to what most manual overclocks are when that happens.

A manual overclock with low voltage means you can't get the high clocks the processor is capable of with light bursty loads. That's the kind of load it sees in games... the time you want a high clock to address those 1% low FPS moments that can lead to stutter.

 

[zx128k]

Also, when most system monitoring apps report "1.48v" or "1.5v" it is onlt taking into consideration the HIGHEST voltage currently being pushed into any given core. Therefore under single threaded workloads, "1.5v" is only going to a single core/thread and therefore is not as "dangerous" as some ppl suspect. It's actually completely normal, when running your chip on 'auto' from what I understand.

If the maximum safe 3800x constant vcore voltage is 1.3 volts vcore then there is no point in a manual overclock but if the maximum constant vcore is 1.4 volts then the 3800x could with a gold sample trade blows with most 9900ks cpu's. 4.4GHz+ all core overclock, IF 1900 and 3800 RAM CL15. You can get a time sky cpu score of 11700 which is the same as the average 9900ks cpu score. Average 9900ks time spy cpu score 11734 | 3800x 4.4GHz 1.325 volts LLC3 11711 bench run

 

[DUAL33s]

Read back a few, the linked comments by the Stilt kind of explains what's happening.

When volts peaks into the 1.45-1.5 volt range it's in very low current density periods, with high clock speeds but processing a light load. High voltage with low current density is not what's harmful, it's the high voltage concurrent with high current as with an extremely heavy processing load that's harmful. The algorithm, everything in AUTO, will dial the volts and clocks right back to what most manual overclocks are when that happens.

A manual overclock with low voltage means you can't get the high clocks the processor is capable of with light bursty loads. That's the kind of load it sees in games... the time you want a high clock to address those 1% low FPS moments that can lead to stutter.

Whoops...yes i actually forgot to mention the low current / light workload aspect of it as well. Thank you. And yeah i will have to go back and re-read the whole thread and article.

But yeah, many ppl are shocked when they see voltage readings near 1.5.... but in the end, it is nothing to worry about (especially with Ryzen 3000) and Ive seen multiple well respected members of the commuity comment on this subject and they all agree.

Thanks for the correction.

 

[drea.drechsler]

... You can get a time sky cpu score of 11700 which is the same as the average 9900ks cpu score. Average 9900ks time spy cpu score 11734 | 3800x 4.4GHz 1.325 volts LLC3 11711 bench run

I'm not really sure I should care. Even my 3700X scores so close to a 9900ks it's indescernible in games. And it already beats one in content creation or productivity (except a few Adobe apps).

And what's more, play it in a realistic setup...meaning NOT on a $1200 GPU at absurdly low 1080P resolution...it will score the same. It's the GPU that makes the difference; even with a 5700XT playing at 1440P my 3700X is just loafing along at 30-40% MAX utilization in Ghost Recon BP.

There's just no compelling reason to enter into the grief and bother of balancing the knife edge of an all-core, low voltage overclock.

 

[zx128k]

I'm not really sure I should care. Even my 3700X scores so close to a 9900ks it's indescernible in games. And it already beats one in content creation or productivity (except a few Adobe apps).

And what's more, play it in a realistic setup...meaning NOT on a $1200 GPU at absurdly low 1080P resolution...it will score the same. It's the GPU that makes the difference; even with a 5700XT playing at 1440P my 3700X is just loafing along at 30-40% MAX utilization in Ghost Recon BP.

There's just no compelling reason to enter into the grief and bother of balancing the knife edge of an all-core, low voltage overclock.

3700x can't match a 9900ks in games . You can only reach the average 9900ks on time spy cpu with a 3800x. There is a lot more performance to get out of the 9900ks than the 3800x. You can get 3600 RAM with tightened timings and hit over 12k time spy cpu score. The 3700x is just slower overall, it can't match the 3800x in any way on average because of the binning process. Many 3700x CPU can't do the higher IF frequencies for example. I don't know why reviews state the 3700x overclocked is a 3800x, its not.

That a side, you can overclock the crap out of a 3700x and a gold sample can do 4.4GHz all cores. Also some can do IF 1900. Every with 3733 RAM with very tight timings you can hit high scores. Here, https://www.3dmark.com/spy/9271166 this is a 3700x with a CPU Score 12 187. This score is number 7 in the world and should not be taken as a 24/7 score. Some people do bench runs with high voltage.

With both the 3800x and the 3700x the RAM timings matter a lot more than the clock speeds. The 9900ks can be overclocked the same way.

What I don't get is how the #1 guy got a 14k time spy cpu score with a 3700x with 3200 RAM.

 

[drea.drechsler]

3700x can't match a 9900ks in games . ..

I'm sorry, but I can't see just what a 9900ks could possibly do for me when my 3700x is loafing along at 30-40% utilization playing GR-BP at 1440p. At 4k it would just get worse, even with a 2080ti. So when you say "can't match....", it's purely academic.

And yes, everything's VH to ultra: i love the eye candy.

Same goes for overclocking a 3800x to try and bridge the small performance gap opened by an overpriced dinosaur processor. As I said, show me something that benefits me from the effort. I'm simply not interested in playing 'who can p**s further' like a 12 yo at scout camp.

 

[zx128k]

I'm sorry, but I can't see just what a 9900ks could possibly do for me when my 3700x is loafing along at 30-40% utilization playing GR-BP at 1440p. At 4k it would just get worse, even with a 2080ti. So when you say "can't match....", it's purely academic.

And yes, everything's VH to ultra: i love the eye candy.

Same goes for overclocking a 3800x to try and bridge the small performance gap opened by an overpriced dinosaur processor. As I said, show me something that benefits me from the effort. I'm simply not interested in playing 'who can p**s further' like a 12 yo at scout camp.

I would advise trying to pick a fight with someone else. This is an overclocking forum which implies we are seeking the most performance possible. GR Breakpoint is a really bad game, PC gamer gave it only 40% out of 100% and it has little relavence on this thread what performance you feel you have in that game. Your post is completely off topic and basically a personal attack on people that overclock.

 

[MrN1ce9uy]

I would advise trying to pick a fight with someone else. This is an overclocking forum which implies we are seeking the most performance possible. GR Breakpoint is a really bad game, PC gamer gave it only 40% out of 100% and it has little relavence on this thread what performance you feel you have in that game. Your post is completely off topic and basically a personal attack on people that overclock.

The main purpose of overclocking in today's world of CPUs is personal gratification. There are no widespread, large performance gains from overclocking. I didn't see it as a personal attack, just someone stating their opinion; one which a lot of people can agree. Intel is faster in gaming. If you are a professional gamer who relies on it for income, maybe those extra frames per second help you. If you are a regular gamer, they only count if you are counting because it doesn't affect the gaming experience for the average gamer.

 

[drea.drechsler]

I would advise trying to pick a fight with someone else. This is an overclocking forum which implies we are seeking the most performance possible. GR Breakpoint is a really bad game, PC gamer gave it only 40% out of 100% and it has little relavence on this thread what performance you feel you have in that game. Your post is completely off topic and basically a personal attack on people that overclock.

My post is perfectly on topic since IMO, overclocking done correctly should result in improved performance in real computing activities. And I'm not picking a fight, just saying it like it is: manual overclocking Ryzen 3000 really offers incredibly little benefit so people reading who think like I do (about the purpose behind overclocking) and want to go after those thousand point gains (in a highly synthetic benchmark) don't think it really means anything. Especially when you can't be sure they're not pumping maximum voltage to the CPU, dunking their radiator in ice water and spraying the VRM's with canned air to keep it cool in order to post it.

I apologize if you don't like being compared to a 12 yo in scout camp.

And BTW, BP is a 'bad game' because of game play, not because of the 3d world. It's a bonafide triple-A for it's graphics content and brings most systems to their knees in high settings with a huge, highly varied, open world and highly detailed graphics. Even if not 'playing the game' it offers a lot of interesting areas to explore, which is what attracts me as much as killing a few enemies.

 

[zx128k]

Overclocking the cpu is about getting the most from your silicon and a lot of people are competitive about getting the highest score possible. If you don't like that then I would recommend not posting in overclocking threads. My system is overall approx. up to 30% faster from overclocking in lots of different games.

If you want to seek to insult other forum posters that's your reputations problem, I am not going to engage in a flame war with you.

 

[COLGeek]

Civility is a must. This is the only warning that will be issued.

Attack ideas, with facts and sources, and not each other.

Thank you.