[SOLVED] Ryzen 5 1600 @3.9Ghz

[winncootMKD]

So lately I wanted to OC my Ryzen 5 1600 and decided to go for 3.9Ghz for 24/7 use.

I am using HWMonitor to monitor the CPU temp, as AIDA's temp isnt accurate I think.
So firstly I started on Prime95 and when I selected Smallest FFT the CPU temp would go up to 77c (Stock Cooler) and crash, but on Blend all mode it was around 62/63c (Stock Cooler) and did not crash. So I dont know what are those two modes different. An explanation would be awesome!

But then as I've read on forums, AIDA64 was the best for CPU stress. So i downloaded it.

Core Clock: 3.9Ghz
VCORE: 1.356/1.368
Max temp: 68/69/70c (varies) (Stock Cooler)

It crashed once, but bumbped the VCORE and it was all good, used to crash around 5,6 minute mark and here I am 30mins in the middle of it, still running smooth, I opened Chrome to write down this thread and boom, it froze (crashed)
I'm sure the crash was caused by Chrome with 14 tabs waiting to load.
My question is, is that Core Voltage Safe for 24/7 use, and should I get Aftermarket cooler for Cooler CPU temp and if so, which one. I've read that better Cooling more CPU life?
Whats the best for the CPU to last longer while overclocked?

Any overclockint tips, I'm open to them

(Also an explanation for the prime95 modes would be appreciated) Thanks!

 

[TechyInAZ]

Yes that is a safe voltage to be at.

As for small FFTs, I have no clue. Are you running Prime95 without AVX instructions on?

I suggest you run ROG Realbench and OCCT (large data set) aswell to insure stability, run each one at a time once overnight

 

[winncootMKD]

Yes that is a safe voltage to be at.

As for small FFTs, I have no clue. Are you running Prime95 without AVX instructions on?

I suggest you run ROG Realbench and OCCT (large data set) aswell to insure stability, run each one at a time once overnight

Will defenetly do a longer stress test.
How are the temps? Are they fine with the stock cooler?
I am planning on getting an aftermarket one in the future, but in few months, so I'm asking if that is totally fine to run it for 24/7 use.
I mean I will hardly hit that temp, the CPU will not be 100% on full load, but I do stream, edit and render videos.

Also do you want me to run both programs at the same time?

 

[drea.drechsler]

Will defenetly do a longer stress test.
How are the temps? Are they fine with the stock cooler?
I am planning on getting an aftermarket one in the future, but in few months, so I'm asking if that is totally fine to run it for 24/7 use.
I mean I will hardly hit that temp, the CPU will not be 100% on full load, but I do stream, edit and render videos.

Also do you want me to run both programs at the same time?

You'll never get 'great' temps with a stock cooler when overclocked. But temperatures in mid-70's are not dangerous especially during a stress test (Prime95) that very little resembles a real-life work load.

Prime95 small FFT is the ultimate AVX-heavy CPU stress test for stability and helps to de-rate your overclock. De-rating helps to feel confident that it will remain stable in real-life useage a few years down the road as components age and dust accumulates. It's mix of AVX instructions in no way resembles any normal work load put on a system outside synthetic stress tests. Not even Prime95 run doing the job it was designed to do...finding Mersenne prime numbers...is going to stress it as much as the small FFT stress test will.

For a more realistic work load test try Asus RealBench. It puts a heavy real life multitasking load on the system with a much more realistic mix of the ultra-stressing AVX instructions that makes Prime95 (and a few others) so hard on the CPU. If it holds up a couple hours of RealBench stress test you can feel confident it's ready for 24/7 use. Then, if you want to, should it also hold up to 30 min's or so of Prime95 small FFT you can feel confident it will still be good in a couple years.

Also: AMD informs us voltage up to 1.425 is safe for long-term use, but I wouldn't go much if any over the voltage you're at until you put better cooling on.

 

[winncootMKD]

You'll never get 'great' temps with a stock cooler when overclocked. But temperatures in mid-70's are not dangerous especially during a stress test (Prime95) that very little resembles a real-life work load.

Prime95 small FFT is the ultimate AVX-heavy CPU stress test for stability and helps to de-rate your overclock. De-rating helps to feel confident that it will remain stable in real-life useage a few years down the road as components age and dust accumulates. It's mix of AVX instructions in no way resembles any normal work load put on a system outside synthetic stress tests. Not even Prime95 run doing the job it was designed to do...finding Mersenne prime numbers...is going to stress it as much as the small FFT stress test will.

For a more realistic work load test try Asus RealBench. It puts a heavy real life multitasking load on the system with a much more realistic mix of the ultra-stressing AVX instructions that makes Prime95 (and a few others) so hard on the CPU. If it holds up a couple hours of RealBench stress test you can feel confident it's ready for 24/7 use. Then, if you want to, should it also hold up to 30 min's or so of Prime95 small FFT you can feel confident it will still be good in a couple years.

Also: AMD informs us voltage up to 1.425 is safe for long-term use, but I wouldn't go much if any over the voltage you're at until you put better cooling on.

Awesome, so cooling is a solution too. What cooling do you suggest to me? Liquid or Air?

And also, can I change the BLCK on my Gigabyte AB350M Gaming 3. Its at 99,76 and not 100 so 3.9Ghz its actually 3890.72Mhz.

 

[winncootMKD]

And also, when doing stress test, does it just need to be that pogram open and me doing nothing on the pc, or can i browse chrome?
On RealBench do I select 4GB RAM or 8GB RAM.
In my pc I got 8

 

[TechyInAZ]

rog realbench 4GB of RAM, basically always choose half your memory.

I'd recommend you leave it as is when running, ROG RealBench will make all other programs sluggish.

 

[winncootMKD]

rog realbench 4GB of RAM, basically always choose half your memory.

I'd recommend you leave it as is when running, ROG RealBench will make all other programs sluggish.

Right, so I crashed on the settings I mentioned above on ROG realbench, but i was at 8gb of ram, that might be the problem. I’ll run with half the memory.
Im runing atm 3.85Ghz at 1.332v and its doing okay,
VRM temp is 63/64 which I think its safe.

One thing is, when doing the test sometimes I have cursor lag (mouse lag) when moving it. That normal right?

Will do more tests and info on what happened.

 

[drea.drechsler]

Awesome, so cooling is a solution too. What cooling do you suggest to me? Liquid or Air?

And also, can I change the BLCK on my Gigabyte AB350M Gaming 3. Its at 99,76 and not 100 so 3.9Ghz its actually 3890.72Mhz.

In general, DO NOT use bus-clock overclocking, ever, on AM4 platform. That is, unless you're ready to lose not only data but potentially even drives...NVME drives in particular. That's because altering the BLCK also alters SATA and PCIe bus speeds which can leave drives inaccessible...basically cute little gum stick sized bricks now.

I like liquid cooling with a front mounted 240mm radiator. The reason is air cooling, while very good, DOES require a much better thought out air-flow in the case. Since the heat of the GPU (which gets much, much hotter) is being dumped into the case along with that of the CPU you'll quickly be circulating hot air around the huge tower cooler limiting it's ability to cool. So you have to put some really good, quiet case fans on front and top/rear to move that air out and keep the processor from building heat. A front mounted radiator always operates with cool air blowing across it to cool the processor, and a GPU is much more tolerant of the heat from the CPU.

Also, when you buy a 240mm AIO it comes with two quiet fans already mounted so you only have to buy some equally good fans for rear/top exhaust. But an air cooler needs not only the fans for the rear/top but also those very important front fans that
push cool air into the case. So be careful how you compare cost of air coolers and liquid coolers.

 

[winncootMKD]

In general, DO NOT use bus-clock overclocking, ever, on AM4 platform. That is, unless you're ready to lose not only data but potentially even drives...NVME drives in particular. That's because altering the BLCK also alters SATA and PCIe bus speeds which can leave drives inaccessible...basically cute little gum stick sized bricks now.

I like liquid cooling with a front mounted 240mm radiator. The reason is air cooling, while very good, DOES require a much better thought out air-flow in the case. Since the heat of the GPU (which gets much, much hotter) is being dumped into the case along with that of the CPU you'll quickly be circulating hot air around the huge tower cooler limiting it's ability to cool. So you have to put some really good, quiet case fans on front and top/rear to move that air out and keep the processor from building heat. A front mounted radiator always operates with cool air blowing across it to cool the processor, and a GPU is much more tolerant of the heat from the CPU.

Also, when you buy a 240mm AIO it comes with two quiet fans already mounted so you only have to buy some equally good fans for rear/top exhaust. But an air cooler needs not only the fans for the rear/top but also those very important front fans that
push cool air into the case. So be careful how you compare cost of air coolers and liquid coolers.

I like the idea of liquid cooling, and by tour suggestion I should put the radiator on front.
Well, I do have 5 fans. 2x intake front, 2x Exhaust top and 1x Exhaust. The thing is, If i get liquid cooling and put it on front, I need to remove the drive slots, which i have no idea if they are removable or not.

Case is Deepcool Kendomen.

 

[winncootMKD]

News. At 3.9Ghz and 1.356/1.368v Test on AIDA64 is okay but crashing on ROG Realbench. Anyone knows why?

 

[drea.drechsler]

News. At 3.9Ghz and 1.356/1.368v Test on AIDA64 is okay but crashing on ROG Realbench. Anyone knows why?

Realbench is a system-level stress test. It hits a lot of stuff, simultaneously. It may be memory is the culprit...or maybe even the GPU. So if you've overclocked memory you might try taking that overclock down some out, or the same with GPU.

Andit might even be PSU isn't up to the task. Posting complete system spec would be helpful.

 

[winncootMKD]

Realbench is a system-level stress test. It hits a lot of stuff, simultaneously. It may be memory is the culprit...or maybe even the GPU. So if you've overclocked memory you might try taking that overclock down some out, or the same with GPU.

Andit might even be PSU isn't up to the task. Posting complete system spec would be helpful.

Only CPU is overclocked
CPU: Ryzen 5 1600
RAM: 8GB G.Skill 2133Mhz
GPU: XFX RX580 8GB GTS Black Edition
MB: Gigabyte AB350M - Gaming 3
PSU: Hantol 900W, 720W Real power 80+ Bronze

 

[drea.drechsler]

Only CPU is overclocked
CPU: Ryzen 5 1600
RAM: 8GB G.Skill 2133Mhz
GPU: XFX RX580 8GB GTS Black Edition
MB: Gigabyte AB350M - Gaming 3
PSU: Hantol 900W, 720W Real power 80+ Bronze

I'm not sure about that Hantol PSU. The rating seems good enough but that says nothing for how 'clean' the 12V power coming in is and whether the board VRM's can deal with excess ripple when it get's heavily load with both CPU and GPU simulataneously drawing heavy loads. Not saying it's the culprit, though.

Maybe someone else can comment on the PSU. In the mean time, you might want to drop the CPU OC to 3.8GHz leaving voltage the same and try RealBench again. If it passes, consider some upgraded cooling and increase volts to get it stable at 3.9. Remember, 1.425 is where AMD draws the line for processor longevity but that assumes it's well cooled.

BTW: I have that same motherboard and it has a pretty sorry VRM. It had to put out pretty high voltage to keep a 1700 stable at 3.8Ghz...that's 8 cores instead of 6, but still.

AND oh yeah...the tiny little block of aluminum the put on the VCore FET's, lacking any meaningful fins, in no way makes for an adequate heatsink. That means it heats up quickly and voltage to the CPU get's unstable. Making it worse is I was running an AIO so there's no airflow across that heatsink. The Wraith cooler does provide some airflow. I had to put a fan blowing right on it to keep it cool. It may not be an issue until you put an aftermarket cooler on, but keep that in mind when you do.

 

[winncootMKD]

I'm not sure about that Hantol PSU. The rating seems good enough but that says nothing for how 'clean' the 12V power coming in is and whether the board VRM's can deal with excess ripple when it get's heavily load with both CPU and GPU simulataneously drawing heavy loads. Not saying it's the culprit, though.

Maybe someone else can comment on the PSU. In the mean time, you might want to drop the CPU OC to 3.8GHz leaving voltage the same and try RealBench again. If it passes, consider some upgraded cooling and increase volts to get it stable at 3.9. Remember, 1.425 is where AMD draws the line for processor longevity but that assumes it's well cooled.

BTW: I have that same motherboard and it has a pretty sorry VRM. It had to put out pretty high voltage to keep a 1700 stable at 3.8Ghz...that's 8 cores instead of 6, but still.

So long story short, Voltage increase lowers the CPU life, but adding good cooling increases it again?

Also, im interested in your cpu, what voltage (bios setting) and ghz is it? Which cooling?

The motherboard is preety good considering the price

BTW: Tested at 3.9Ghz with 1.368 and 1.38v and it just crashes at 2min mark. Might be the cooling or PSU. I have dropped to 3.85Ghz at 1.332/1.344v and is going pretty well. Will test it for 3,4 hours tho.

 

[winncootMKD]

AND oh yeah...the tiny little block of aluminum the put on the VCore FET's, lacking any meaningful fins, in no way makes for an adequate heatsink. That means it heats up quickly and voltage to the CPU get's unstable. Making it worse is I was running an AIO so there's no airflow across that heatsink. The Wraith cooler does provide some airflow. I had to put a fan blowing right on it to keep it cool. It may not be an issue until you put an aftermarket cooler on, but keep that in mind when you do.

That is why I want air cooling and not liquid, but if i get liquid i do have a spare fan to blow on that. But im kinda sure the 5 fans I already have now will do some work, we will see tho

 

[drea.drechsler]

So long story short, Voltage increase lowers the CPU life, but adding good cooling increases it again?

Also, im interested in your cpu, what voltage (bios setting) and ghz is it? Which cooling?

The motherboard is preety good considering the price

BTW: Tested at 3.9Ghz with 1.368 and 1.38v and it just crashes at 2min mark. Might be the cooling or PSU. I have dropped to 3.85Ghz at 1.332/1.344v and is going pretty well. Will test it for 3,4 hours tho.

It's not quite that 'voltage increase shortens...add cooling increases'..the better way to look at it is keeping the CPU cool means it can tolerate the higher voltages.

I have to say that is a nice low voltage at 3.85. I'd go with the 1.368 voltage at that clock speed so long as temps under load are still in the 70's or lower. That's very safe volts and temp so if it holds for a couple hours like that only then start chasing even lower voltage.

 

[drea.drechsler]

That is why I want air cooling and not liquid, but if i get liquid i do have a spare fan to blow on that. But im kinda sure the 5 fans I already have now will do some work, we will see tho

What happens is VRM heatsinks are tucked right underneath the rear case fan. Even towere coolers don't always do it since the heatsink is also tucked underneath the lowest fins of the tower stack. That tends to create a 'dead air' zone right there...so lots of air movement between the CPU and the case fans, but none right in that little dead zone. My AIO just made it worse.

 

[winncootMKD]

It's not quite that 'voltage increase shortens...add cooling increases'..the better way to look at it is keeping the CPU cool means it can tolerate the higher voltages.

I have to say that is a nice low voltage at 3.85. I'd go with the 1.368 voltage at that clock speed so long as temps under load are still in the 70's or lower. That's very safe volts and temp so if it holds for a couple hours like that only then start chasing even lower voltage.

Earlier I tested 32mins with 3.85Ghz and 1.334/1.344v and it was 69c 70c MAX temp on CPU and 63-65 on VRM. Thats pretty good I think
Right now as of typing I am in the middle of the same test, if it holds up 3,4 hours I will leave it as be. Nice voltage, nice OC.

All testing on ROG Realbench

Also, thanks for help my dood, you’re awesome

 

[winncootMKD]

What happens is VRM heatsinks are tucked right underneath the rear case fan. Even towere coolers don't always do it since the heatsink is also tucked underneath the lowest fins of the tower stack. That tends to create a 'dead air' zone right there...so lots of air movement between the CPU and the case fans, but none right in that little dead zone. My AIO just made it worse.

I see. Whats your VRM temps on idle and load?
Im planning on 2fan 6pipe air cooler.

BTW: Testing shows lots of Result Hash Match, must be good sign.

 

[drea.drechsler]

...
Also, im interested in your cpu, what voltage (bios setting) and ghz is it? Which cooling?
..

forgot about that...I'm not running that board any longer, my board is a B450m Mortar now. But I ran a 1700 on it with a cooler master LM-240 AIO. I had to run volts at about 1.412-1.425 to hold a 3.8Ghz OC in stress tests. That was terrible...and to boot the VRM temp was running around 110C in stress even with a 50mm fan blowing directly on it. It would cause the CPU to throttle occasionaly if it overheated to 115 or 120C.

BUT...that was my silicon's characteristics and not representative of what you should expect. It's not wise to extrapolate anything from it since everybody's hardware has unique characteristics.

And BTW...I'm now running that 1700 on my B450m Mortar at 3.95G, 1.38-1.4V with VRM temps never exceeding 82C. The difference is the VRM.

 

[winncootMKD]

forgot about that...I'm not running that board any longer, my board is a B450m Mortar now. But I ran a 1700 on it with a cooler master LM-240 AIO. I had to run volts at about 1.412-1.425 to hold a 3.8Ghz OC in stress tests. That was terrible...and to boot the VRM temp was running around 110C in stress even with a 50mm fan blowing directly on it. It would cause the CPU to throttle occasionaly if it overheated to 115 or 120C.

BUT...that was my silicon's characteristics and not representative of what you should expect. It's not wise to extrapolate anything from it since everybody's hardware has unique characteristics.

And BTW...I'm now running that 1700 on my B450m Mortar at 3.95G, 1.38-1.4V with VRM temps never exceeding 82C. The difference is the VRM.

Holy damn, 110C on the VRM that is a lot, so i should be happy with my 65C.

Im plannin on the 3rd gen Ryzen. Either the 3600 or 3700. So defenetly need new MB if I want to overclock that, or maybe BIOS update will do the work.

But thats crazy how different MB changed the VRM temps and also at lower voltage.

Stress news: 45mins in, everything good.
Cpu max temp: 70.3C
VRM max temp: 65C
Voltage: 1.332/1.344v
Will mention again, Stock cooler (Wraith Spire)

 

[drea.drechsler]

Holy damn, 110C on the VRM that is a lot, so i should be happy with my 65C.

Im plannin on the 3rd gen Ryzen. Either the 3600 or 3700. So defenetly need new MB if I want to overclock that, or maybe BIOS update will do the work.

But thats crazy how different MB changed the VRM temps and also at lower voltage.

Stress news: 45mins in, everything good.
Cpu max temp: 70.3C
VRM max temp: 65C
Voltage: 1.332/1.344v
Will mention again, Stock cooler (Wraith Spire)

I would completely not plan on running a Ryzen 3000 on that board. I even wonder if Gigabyte will release a BIOS with microcode updates for Zen2 on AB350m Gaming 3 boards.

I was reading this on a German enthusiast site: the Reason is B350 boards only have 16Mb BIOS chips and Zen2 will require 32Mb chips to hold microcode for Ryzen 1000, Ryzen 2000 and as well Ryzen 3000 chips. They're concerned about a strategy that kills support for a Ryzen 1000 chip when updated to a later BIOS, as they did for Bristol Ridge with the release of Ryzen 2000 microcode.

If someone sees a later BIOS and naively updates to it using their Ryzen 5 1600, for instance, it will leave them with a board that won't boot until they get a Ryzen 2000 or 3000 processor. Imagine the support problems then.

 

[winncootMKD]

I would completely not plan on running a Ryzen 3000 on that board. I even wonder if Gigabyte will release a BIOS with microcode updates for Zen2 on AB350m Gaming 3 boards.

I was reading this on a German enthusiast site: the Reason is B350 boards only have 16Mb BIOS chips and Zen2 will require 32Mb chips to hold microcode for Ryzen 1000, Ryzen 2000 and as well Ryzen 3000 chips. They're concerned about a strategy that kills support for a Ryzen 1000 chip when updated to a later BIOS, as they did for Bristol Ridge with the release of Ryzen 2000 microcode.

If someone sees a later BIOS and naively updates to it using their Ryzen 5 1600, for instance, it will leave them with a board that won't boot until they get a Ryzen 2000 or 3000 processor. Imagine the support problems then.

Damn. Well I completly understand those who dont know much about computers. But I ready the updates and of course would not update to that BIOS.

But will defenetly get a new board probably 370 or idk.

 

[winncootMKD]

Stress test News: Crashed at 58min mark.
Bumped the voltage up tiny bit.
Will see how it does now.