[SOLVED] 3700X 4GHZ@1.218v for all time use?

[Jayant Arora]

Hey guys!!
I'm running 3700X on Air with stock cooler[Wraith Prism].
I just wanted to know if it's safe to run i run 3700X set to 4GHZ @1.218v all times considering temps are way better than default. I'm not sure what actually are safe voltages for this CPU ?
Yea i know i loose some single core performance, i can live with that as temps are 10'-20C lesser.

I have good airflow with my case.

default temps PBO Off 1smus plan:
Idle 45-55'C
Gaming 45-70'C
Heavy Load[Rendring and Coding [50-90% load]: 55-85'C

4GHZ OC 1.218v PBO Off 1smus plan:
Idle 35-45'C
Gaming 45-55'C
Heavy Load[Rendring and Coding [50-90% load]: 55-70'C

 

[CosmicDance]

Yes that voltage is fine. Anything up to about 1.45 volts is good for that chip.

Andy

 

[King_V]

I just wanted to know if it's safe to run i run 3700X set to 4GHZ @1.218v all times

Why do you want to run that speed at all times? What's the point? You should let it run slower if the system is not doing much.

 

[Phaaze88]

Here's something you should try out.
Take your current bios settings and save them - there should be an option to save those settings as a profile somewhere.

Now, do a pass of:
Cinebench R20 - record the score
Passmark performance test(free trial) - record the score
Cpu-Z Bench - record the single and multi scores

Then go back and restore the optimal defaults - don't forget to re-enable A-XMP/DOCP so the ram is running at the advertised speed.
Run those same 3 tests again, record and compare the scores to your steady 4.0ghz setting.

 

[ohio_buckeye]

I agree before you settle on the overclock. Not that your overclock is a bad number, but I just upgraded a 3600. A few places I read said to leave the cpu at stock that you wouldn't gain a lot by overclocking. In fact, some things I read said you might even introduce stuttering into your games and reduce your performance a bit as far as how smooth everything was. So I would try it again and see.

The other thing I found, keep in mind I had a 1700x before my 3600, was that now you should try using Ryzen Master to monitor temps. Other programs like core temp and hwmonitor were reporting higher temps at times. When I got my airflow right in my case and left the 3600 stock, it runs pretty well. I don't even think I really need to overclock it to be honest.

 

[Jayant Arora]

Why do you want to run that speed at all times? What's the point? You should let it run slower if the system is not doing much.

Idle temps are bit higher with default settings and at times it did boost to 4+ ghz at idle anyways on its own which causes higher temps.
I want a sweet spot between temps and performance.

 

[Jayant Arora]

Here's something you should try out.
Take your current bios settings and save them - there should be an option to save those settings as a profile somewhere.

Now, do a pass of:
Cinebench R20 - record the score
Passmark performance test(free trial) - record the score
Cpu-Z Bench - record the single and multi scores

Then go back and restore the optimal defaults - don't forget to re-enable A-XMP/DOCP so the ram is running at the advertised speed.
Run those same 3 tests again, record and compare the scores to your steady 4.0ghz setting.

I did that already except passmark, here are results of 3 times avg:
With Default settings no PBO:
Cpu Z : single core: 530.1 multi core : 5598
Cimebench R20: CPU multi thread: 4882
Max Temp 70'C

With 4Ghz @1.218v
Cpu Z : single core: 505 multi core : 5490
Cimebench R20: CPU multi thread: 4731
Max Temps: 63.5'C

My Ram is samsung B die set to 3600@16-16-16-32-54

 

[drivinfast247]

Nothing wrong with that at all. Just a little loss of performance and lower temps and power usage.

 

[King_V]

Idle temps are bit higher with default settings and at times it did boost to 4+ ghz at idle anyways on its own which causes higher temps.
I want a sweet spot between temps and performance.

First, if it's only boosting above that at certain times, then the overall heat production is probably less than leaving at 4.0 continuously.

Second, if the higher temps from the occasional boost past 4.0 are actually causing problems, then it seems likely that you have inadequate cooling. Your readings in your first post seem very improbable.


Maybe what you want is to simply limit the boost to 4.0 maximum. Why would you want it to NOT drop below that when the extra performance isn't needed? You're not loosing any performance when speeds drop lower.

 

[Jayant Arora]

First, if it's only boosting above that at certain times, then the overall heat production is probably less than leaving at 4.0 continuously.

Second, if the higher temps from the occasional boost past 4.0 are actually causing problems, then it seems likely that you have inadequate cooling. Your readings in your first post seem very improbable.


Maybe what you want is to simply limit the boost to 4.0 maximum. Why would you want it to NOT drop below that when the extra performance isn't needed? You're not loosing any performance when speeds drop lower.

And How am i suppose to do that ? Set Max clock multiplier to 40 ? and Not 40 as constant?

 

[Jayant Arora]

And How am i suppose to do that ? Set Max clock multiplier to 40 ? and Not 40 as constant?

If you see this, is what i am trying to do to tame temperatures:

View: https://www.youtube.com/watch?v=2wM3obN2pAE

 

[King_V]

And How am i suppose to do that ? Set Max clock multiplier to 40 ? and Not 40 as constant?

If that's a setting option available, then yes.

 

[Phaaze88]

I did that already except passmark, here are results of 3 times avg:
With Default settings no PBO:
Cpu Z : single core: 530.1 multi core : 5598
Cimebench R20: CPU multi thread: 4882
Max Temp 70'C

With 4Ghz @1.218v
Cpu Z : single core: 505 multi core : 5490
Cimebench R20: CPU multi thread: 4731
Max Temps: 63.5'C

My Ram is samsung B die set to 3600@16-16-16-32-54

Yep. Not surprised by that at all. Even though there's a ~7C temp drop - which is significant - performance dropped as well... just a measly 3% though.

This doesn't apply to you it seems, but some people are all too eager to OC their Ryzen 3000 cpus even when they've been advised against doing so.
The usual Intel method of just increasing and locking the core frequency does not work; the secret lies in FCLK OC and tightening the memory timings.

I have good airflow with my case.

default temps PBO Off 1smus plan:
Idle 45-55'C
Gaming 45-70'C
Heavy Load[Rendring and Coding [50-90% load]: 55-85'C

Just get a tower cooler and that one scenario of 85C won't be an issue anymore? Sure, it's an added expense, but at least you're not having to limit the cpu?

Ryzen 3000 on it's own whims is very 'bursty', and no cooler can react to that behavior fast enough, which brings the following comparison:
Ryzen 3000 has higher idle temps than the Intel counterpart, but lower max load temps.

Now that you've put a limiter on it: you've sacrificed some performance to get the idle and load temps down, but power use has remained the same - or may have actually gone up slightly.

 

[drea.drechsler]

...
Now that you've put a limiter on it: you've sacrificed some performance
...
Ryzen 3000 on it's own whims is very 'bursty', and no cooler can react to that behavior fast enough, which brings the following comparison:
Ryzen 3000 has higher idle temps than the Intel counterpart, but lower max load temps.
...

I'd be willing to say it's sacrificed a LOT of performance since it can now never boost single threads to 4.4Ghz in 'light bursty' loads which predominate in most desktop workloads, and with gaming.

And FYI... my 3700X gets 100 pts higher in Cinebench R20 multi-thread test (5088). All I've done is let it boost eagerly by enabling PBO.

Getting a bigger cooler is always a good idea, even consider how good Ryzen's stock cooler is. But the way to 'control' temps is to ignore them until they get up to about 70C. Just set the fans to run at a fixed speed that's just barely 'not loud' up to about 70C, then ramp it up from there. Don't let it get airplane mode loud until about 85 or 90C.

 

[Jayant Arora]

Yep. Not surprised by that at all. Even though there's a ~7C temp drop - which is significant - performance dropped as well... just a measly 3% though.

This doesn't apply to you it seems, but some people are all too eager to OC their Ryzen 3000 cpus even when they've been advised against doing so.
The usual Intel method of just increasing and locking the core frequency does not work; the secret lies in FCLK OC and tightening the memory timings.


Just get a tower cooler and that one scenario of 85C won't be an issue anymore? Sure, it's an added expense, but at least you're not having to limit the cpu?

Ryzen 3000 on it's own whims is very 'bursty', and no cooler can react to that behavior fast enough, which brings the following comparison:
Ryzen 3000 has higher idle temps than the Intel counterpart, but lower max load temps.

Now that you've put a limiter on it: you've sacrificed some performance to get the idle and load temps down, but power use has remained the same - or may have actually gone up slightly.

You know you won't believe my stock wraith prism on full blast reaches 3700 RPM and why ? cause i have l two top intake fans direct firing over it @2000RPM at right angle.
Personally i don't like aesthetics of tower cooler. What do you suggest ? i go for custom loop or AIO like arctic freezer or something?

 

[Jayant Arora]

I'd be willing to say it's sacrificed a LOT of performance since it can now never boost single threads to 4.4Ghz in 'light bursty' loads which predominate in most desktop workloads, and with gaming.

And FYI... my 3700X gets 100 pts higher in Cinebench R20 multi-thread test (5088). All I've done is let it boost eagerly by enabling PBO.

Getting a bigger cooler is always a good idea, even consider how good Ryzen's stock cooler is. But the way to 'control' temps is to ignore them until they get up to about 70C. Just set the fans to run at a fixed speed that's just barely 'not loud' up to about 70C, then ramp it up from there. Don't let it get airplane mode loud until about 85 or 90C.

I like ryzen cause it's cheap and value for money for performance it gives but...
see I'm no CPU expert here but i'd say, that 4.4 GHZ bursts for light threaded workload is a gimmick, to just fulfill paper specifications, you might don't believe me but hey i'd tested that on my own:
Consider CPU Z and cine bench R20 scores i have:

#1 Default settings and PBO ON i can get best:
Single 531 and Multi 5601
CinebenchR20 4901

#2 I tried OC'ing my CPU and MAX stable i can push is 4.25GHZ all core @1.35V it yielded me
Single 533 and Multi 5800
CinebenchR20 5053

That is only 2 point difference in single core, but interesting thing is, i also tried to

#3 OC it to 4.4GHZ all core @1.45v which was not stable but i was able to complete some benchmarks:
Single 541
Multi ~6000 [*5998.6] [That DBZ scouter blew Off..bam!! It's Over 6000..............]
Cinebench 5223


So if you see #3 single core score is equivalent to a 9900K 100% match which is 543 on CPU-Z reference comparison and 113% on Multi-core score, it might blew your mind off,
Also if you see when you didn't have those 1ms 4.4GHZ bursts before 1.0.0.4 BIOS updates and after benchmark scores were all same and no difference,
Trust me when you see it running actually at 4.4Ghz for a longer period than a burst, what i had seen when i overclocked it #3, it changed the picture.

I strongly suggest what i did, you should not do that...but hey? I think you all guys should know..what is what and i may be not correct..

Here's my CPU Z Validation when i did 4.4 OC: https://valid.x86.fr/5qnxuy
Also note Temp it mentions it's only 66.9'C for a 4.4GHZ OCand it was on air..

 

[Phaaze88]

You know you won't believe my stock wraith prism on full blast reaches 3700 RPM and why ? cause i have l two top intake fans direct firing over it @2000RPM at right angle.

Oh, I believe it. The problem is, few people can tolerate fans running that high.
Does this top intake config yield better LOAD temps than the traditional front intake, rear exhaust, and single top exhaust? [Idle temps are irrelevant.]

Personally i don't like aesthetics of tower cooler. What do you suggest ? i go for custom loop or AIO like arctic freezer or something?

Hmm, how do I put this...
Aesthetics is the ONLY reason to go the custom loop or AIO route with Ryzen 3000; they are just THAT power efficient over their Intel counterparts.
Users can cool a 3900X/3950X reasonably well with an NH-D15 or Dark Rock Pro 4, but those same coolers will struggle with a 9900K, especially when overclocked.

You will pay for those aesthetics as well... custom loop is obvious, but AIOs are deceptive; they are more expensive than they appear.
If you are a looks > performance person, well... to each their own, I guess.

 

[drea.drechsler]

I like ryzen cause it's cheap and value for money for performance it gives but...
see I'm no CPU expert here but i'd say, that 4.4 GHZ bursts for light threaded workload is a gimmick
...

If your workload is always something like CineBench models... image rendering using Cinema 4D... I totally agree going a big all-core overclock can be great. But only IF you can keep it stable which is paramount when you use your computer to make money. Take my word for it, it's very annoying when a crash 3 hours into a 4 hour rendering makes you miss a deadline and a bonus.

But real-world workloads for us mundane users isn't like that, and if it is only rarely. We do things that are 'bursty', like a game where a thread has a sudden need to render out a game scene without lagging FPS (don't pretend to be expert at this, but it seems even multi-threaded games use a single rendering thread). Being able to boost to 4.4G when it needs is what helps reduce those '1% FPS lows' that we see as stuttering in busy action scenes when it's least helpful. If you lock it in at an overclock less than 4.4G you can't get the benefit of that boosting potential, but worse it doesn't show up in most benchmarks. Just because you can't see it (until in a situation where stuttering commences) doesn't mean it's not there.

All that said and done, i do agree the heavy lifting the processor does is at lower clocks, usually at 4.375-4.35Ghz just a few seconds after onset but all the way down to 4.1-4.05Ghz all-core when I'm rendering a handbrake. Funny thing though, it only completes a 4k UHD scene about 5 sec's shorter when I set up in a (stable and thermally manageable) 4.3Ghz OC. That's not much to get excited about in a 278 second encoding.

@Phaaze88: what do you mean when you say AIO's are more expensive than they appear? I've always felt they can be CHEAPER than they appear. That's because they come with (usually) good fans that you'd have to purchase separately to provide adequate cool air to the cooler so it doesn't end up trying to cool the processor with hot air.

 

[Jayant Arora]

Oh, I believe it. The problem is, few people can tolerate fans running that high.
Does this top intake config yield better LOAD temps than the traditional front intake, rear exhaust, and single top exhaust? [Idle temps are irrelevant.]


Hmm, how do I put this...
Aesthetics is the ONLY reason to go the custom loop or AIO route with Ryzen 3000; they are just THAT power efficient over their Intel counterparts.
Users can cool a 3900X/3950X reasonably well with an NH-D15 or Dark Rock Pro 4, but those same coolers will struggle with a 9900K, especially when overclocked.

You will pay for those aesthetics as well... custom loop is obvious, but AIOs are deceptive; they are more expensive than they appear.
If you are a looks > performance person, well... to each their own, I guess.

Yes!! It's better than when i have top as exhaust, also i modded my case a bit to add another rear exhaust to balance too much +ve air pressure. I have :
120mm x 2 @ 2000RPM Top intake
120mm x 2 @ 2000RPM Front intake
140mm x 1 @1700RPM Front Intake[I removed the DVD drive cages to add 140mm fan at front]
120mm x 1 @ 2000RPM Rear Exhaust
120mm x 1 @1300 RPM Rear Exhaust

Didn't i mention i have good airflow.. :/

 

[Jayant Arora]

If your workload is always something like CineBench models... image rendering using Cinema 4D... I totally agree going a big all-core overclock can be great. But only IF you can keep it stable which is paramount when you use your computer to make money. Take my word for it, it's very annoying when a crash 3 hours into a 4 hour rendering makes you miss a deadline and a bonus.

But real-world workloads for us mundane users isn't like that, and if it is only rarely. We do things that are 'bursty', like a game where a thread has a sudden need to render out a game scene without lagging FPS (don't pretend to be expert at this, but it seems even multi-threaded games use a single rendering thread). Being able to boost to 4.4G when it needs is what helps reduce those '1% FPS lows' that we see as stuttering in busy action scenes when it's least helpful. If you lock it in at an overclock less than 4.4G you can't get the benefit of that boosting potential, but worse it doesn't show up in most benchmarks. Just because you can't see it (until in a situation where stuttering commences) doesn't mean it's not there.

All that said and done, i do agree the heavy lifting the processor does is at lower clocks, usually at 4.375-4.35Ghz just a few seconds after onset but all the way down to 4.1-4.05Ghz all-core when I'm rendering a handbrake. Funny thing though, it only completes a 4k UHD scene about 5 sec's shorter when I set up in a (stable and thermally manageable) 4.3Ghz OC. That's not much to get excited about in a 278 second encoding.

@Phaaze88: what do you mean when you say AIO's are more expensive than they appear? I've always felt they can be CHEAPER than they appear. That's because they come with (usually) good fans that you'd have to purchase separately to provide adequate cool air to the cooler so it doesn't end up trying to cool the processor with hot air.

Every thing now manufactures in china and If you are comparing US and USD prices then they are deceptive at times, if you pick custom loop parts or say AIOs from china directly they are cheap as hell 1/3th of price. For like $150-$180 you can build a custom loop even with a D5 pump.

 

[drea.drechsler]

Every thing now manufactures in china and If you are comparing US and USD prices then they are deceptive at times, if you pick custom loop parts or say AIOs from china directly they are cheap as hell 1/3th of price. For like $150-$180 you can build a custom loop even with a D5 pump.

LOL... sounds like a truism for on-line shopping in general!

 

[Phaaze88]

@drea.drechsler
Hmm, what's a good comparison... perhaps the Dark Rock 4 VS the H100i Pro - over a duration of 10 years?
Let's say for simplicity's sake, that the parts last the duration of their warranties. [This isn't really fair, for some AIOs, namely the cheaper models, like the MasterLiquid ML240R's fans of only 1 year...]

H100i Pro:
125USD - retail
$125 - pump is good for 5 years, after which one must replace the entire unit. This will have been done at least once in a 10 year period.
$25 x 2 - 2x ML 120 Pro, which are good for 5 years
??? - Yes, while we know the chance is low(single digit), the possibility of leaking is still there. The cost depends on what is damaged.
= $300, or more(if your luck is bad) after 10 years

Dark Rock 4:
$75 - retail
$22x 3 - Silent Wings 3 pwm are good for 3 years
= $141, after 10 years

That may not have been a great comparison, but I hope I got my point across.

120mm x 2 @ 2000RPM Top intake
120mm x 2 @ 2000RPM Front intake
140mm x 1 @1700RPM Front Intake[I removed the DVD drive cages to add 140mm fan at front]
120mm x 1 @ 2000RPM Rear Exhaust
120mm x 1 @1300 RPM Rear Exhaust

That seems like way too much intake, but the case design does affect what setups work best.
But while your current setup may have helped cpu temps, I'm willing to bet gpu temps aren't as low as they could be unless you're running a blower style gpu.

 

[drea.drechsler]

....
That may not have been a great comparison, but I hope I got my point across.
....

I see your point... AIO's have a shorter service life. While air coolers, being a bit of an anvil in how they do their job, have an extremely long service life potential.

True enough and a valid consideration for those who don't include the cooling solution in the upgrade cycle of their rigs. AIO liquid coolers are much improved since inception but I'll hazard a guess most all current models will lose an appreciable measure of effectiveness at 10 years of service life simply due to deposit build up on the micro-fins of the water block. In my case, by then I'm pretty sure I'll have moved on to something else.

 

[Jayant Arora]

@drea.drechsler
Hmm, what's a good comparison... perhaps the Dark Rock 4 VS the H100i Pro - over a duration of 10 years?
Let's say for simplicity's sake, that the parts last the duration of their warranties. [This isn't really fair, for some AIOs, namely the cheaper models, like the MasterLiquid ML240R's fans of only 1 year...]

H100i Pro:
125USD - retail
$125 - pump is good for 5 years, after which one must replace the entire unit. This will have been done at least once in a 10 year period.
$25 x 2 - 2x ML 120 Pro, which are good for 5 years
??? - Yes, while we know the chance is low(single digit), the possibility of leaking is still there. The cost depends on what is damaged.
= $300, or more(if your luck is bad) after 10 years

Dark Rock 4:
$75 - retail
$22x 3 - Silent Wings 3 pwm are good for 3 years
= $141, after 10 years

That may not have been a great comparison, but I hope I got my point across.


That seems like way too much intake, but the case design does affect what setups work best.
But while your current setup may have helped cpu temps, I'm willing to bet gpu temps aren't as low as they could be unless you're running a blower style gpu.

I did my homework on air flow when i set those fans.
The second exhaust i mentioned sits rear end of pcie mount cover which i removed to make space for fan. So no issue for air trapping around gpu or mobo.
I have gtx 970 asus strix which i will soon replace with upcoming 3000
Nvidia series. I have never bothered because temps are so good just so if you want to know with silent fan mode 65C max and custom fan profile 55 C max. I run it on max OC to it limits allowed in asus gpu tweak, it never dwindled.