Question Tips for first time overclocking CPU

[ahmd3500]

+So hey there. only recently I bought the AOC C24G1 monitor which is a solid 1080p 144hz VA.

Basically, I used to have a 1366 x 768 60hz monitor so I didnt really care about frames that much.
And right now, I'd of course like to achieve 144 solid in almost any games.

Sometimes, my games stutter in the bigger resolution and ofc frames have dropped a lot, not sure if this is a gpu or cpu issue but yea here goes! So basically this is my kit: MOBO:B350 Mortar

Ram: (8gb) DDR4 2 x 4 crucial ballistix sport 2400mhz Ram

GPU: Zotac GeForce GTX 1060 AMP! Edition 3gb

CPU:Ryzen 3 1200

Monitor:AOC C24G1

HDD: Seagate BarraCuda Internal Hard Drive 2TB SATA 6Gb/s 64MB Cache 3.5-Inch (ST2000DM006)

PSU:Seasonic s12ii 520w 80 + bronze certified

I've mostly been able to achieve stable 144hz in games like Fortnite, league of legends without an issue. But apex legends and other games might actually drop to 120-100 and even lower.
Now 3 very small questions:

Will overclocking the cpu help at all, if so what benefits exactly will it bring?

What would be a better component for my build or what exactly should I improve next for gaming from what I stated, better mhz ram, 16gb perhaps or should I aim towards for a more stable / better pc?

Will overclocking with this PSU be okay?

Thanks a ton!

 

[DavidM012]

https://forums.tomshardware.com/threads/ryzen-3-1200-overclocking-help.3113729/

https://forums.tomshardware.com/threads/ryzen-3-1200-4ghz-1-2v-stable.3282477/

Others say they got to 4ghz which is the kind of averagely accessible overclock for that cpu

1st. download hwmonitor and monitor temps. Don't exceed 75-80c.

You will have to go one step at a time, with prime 95 version 26.6. only. not the latest version which includes avx instructions that will overheat cpu. If your cpu temps. exceed 75c stop the test at that point.

Each time you increase the clock frequency by 1000mhz only, 0r 0.1ghz, run the the torture test, 30mins to check for errors but really 2 hours to ascertain stability. There shouldn't be any at default speed. If you increase to 3.7 you might not get any errors and need not increase the vCore. Info above says a vCore of up to 1.4v is ok on ryzen. Others may recommend less. You can search plentiful guides on overclocking ryzen, be careful and don't start with the highest settings. Is the caution.

At some point maybe 3.8-3.9ghz you might start getting prime 95 errors and will have to increase vCore by a small increment until you find the lowest stable vCore for that clock frequency.

Things vary between chips due to silicon lottery.

Also you are using slower ddr4 2400 dimms. You might have been better off trying for 3000 or 3200 to aggregate the cpu overclock with memory overclock.

Anyway here's some blurb on fortnite benchmarks that might have some relevant info too.

https://www.tomshardware.com/reviews/fortnite-best-performance-benchmarks,5541-6.html

I would say don't go BLAM and try and hit the max overclock first time without testing the intermediate stages for the lowest vCore you can apply. Don't use auto overclock software. It will go BLAM and probably apply too much voltage the first time. A lot of auto tune software isn't all that well tuned. Don't imagine it will do anything clever or sophisticated.

So you will go through a few stages of stability testing and may eventually achieve a 9% boost of cpu if stable at 4ghz. It starts at 3.6ghz? And finishes at 4ghz that's 9%. quantity of cpu performance.

You might not hit 4ghz with the stock cooler.

Also seriously lagging hdd data transfer rate using a hdd rather than ssd.

 

[ahmd3500]

https://forums.tomshardware.com/threads/ryzen-3-1200-overclocking-help.3113729/

https://forums.tomshardware.com/threads/ryzen-3-1200-4ghz-1-2v-stable.3282477/

Others say they got to 4ghz which is the kind of averagely accessible overclock for that cpu

1st. download hwmonitor and monitor temps. Don't exceed 75-80c.

You will have to go one step at a time, with prime 95 version 26.6. only. not the latest version which includes avx instructions that will overheat cpu. If your cpu temps. exceed 75c stop the test at that point.

Each time you increase the clock frequency by 1000mhz only, 0r 0.1ghz, run the the torture test, 30mins to check for errors but really 2 hours to ascertain stability. There shouldn't be any at default speed. If you increase to 3.7 you might not get any errors and need not increase the vCore. Info above says a vCore of up to 1.4v is ok on ryzen. Others may recommend less. You can search plentiful guides on overclocking ryzen, be careful and don't start with the highest settings. Is the caution.

At some point maybe 3.8-3.9ghz you might start getting prime 95 errors and will have to increase vCore by a small increment until you find the lowest stable vCore for that clock frequency.

Things vary between chips due to silicon lottery.

Also you are using slower ddr4 2400 dimms. You might have been better off trying for 3000 or 3200 to aggregate the cpu overclock with memory overclock.

Anyway here's some blurb on fortnite benchmarks that might have some relevant info too.

https://www.tomshardware.com/reviews/fortnite-best-performance-benchmarks,5541-6.html

I would say don't go BLAM and try and hit the max overclock first time without testing the intermediate stages for the lowest vCore you can apply. Don't use auto overclock software. It will go BLAM and probably apply too much voltage the first time. A lot of auto tune software isn't all that well tuned. Don't imagine it will do anything clever or sophisticated.

So you will go through a few stages of stability testing and may eventually achieve a 9% boost of cpu if stable at 4ghz. It starts at 3.6ghz? And finishes at 4ghz that's 9%. quantity of cpu performance.

You might not hit 4ghz with the stock cooler.

Also seriously lagging hdd data transfer rate using a hdd rather than ssd.

Thanks a ton, almost everything I needed if not everything <3.

 

[Karadjgne]

With a gtx1060, even at 1080p, 144fps is questionable in many games, especially the seriously detailed graphics of BF5 and other such games. You may have to drop some details.

 

[drea.drechsler]

...
You will have to go one step at a time, with prime 95 version 26.6. only. not the latest version which includes avx instructions that will overheat cpu.
...

Why would you not want to know your overclocked system is thermally stable when tested using AVX instructions?

 

[Karadjgne]

Because 99.9% of games use extremely little to no AVX, AVX-2 or AVX-512, so testing for that kind of 130+% heat is pointless. P95 26.6 small fft uses close enough to 100% continuous load that it enables any cooler, air or AIO to generate the maximum amount of cpu generated heat that any gaming load would apply.

P95 is also not to be used for stability testing. It's purely for testing cpu output heat. Instructions like AVX only surpass heat loadage, they don't affect stability. To test for stability use Asus RealBench as it puts the entire pc through a workout, simultaneously using cpu, ram, USB, pcie, gpu etc. AR will generate heat, but it's main focus is bouncing the pc every which way looking for issues.

 

[drea.drechsler]

Because 99.9% of games use extremely little to no AVX, AVX-2 or AVX-512, so testing for that kind of 130+% heat is pointless. P95 26.6 small fft uses close enough to 100% continuous load that it enables any cooler, air or AIO to generate the maximum amount of cpu generated heat that any gaming load would apply.

P95 is also not to be used for stability testing. It's purely for testing cpu output heat. Instructions like AVX only surpass heat loadage, they don't affect stability. To test for stability use Asus RealBench as it puts the entire pc through a workout, simultaneously using cpu, ram, USB, pcie, gpu etc. AR will generate heat, but it's main focus is bouncing the pc every which way looking for issues.

It's not pointless if you like to de-rate an overclock. That may sound contradictory, but I do like to squeeze the manufacturing margin as much as anyone else and yet have a high-availability system. P95, small FFT's, latest iteration is one of the best ways to know that.

That said: I do understand how everyone gets those crazy insane overclocks. Just don't stress test it very strictly "if it boots Windows10 is great...but if it can play Fortnight for a couple hours it's gold". Yeah. Right.

 

[Karadjgne]

Well since you'll never see AVX-2 or especially AVX-512 outside of professional type production and content creation programs, there's no need for average Joe to worry about that on his gaming machine, and most new motherboard bios come with an AVX offset you can set that'll drop clocks a few 100MHz if the cpu determines its using those instructions. So getting 5.0GHz under normal gaming or 4.8GHz under AVX unrealistic temps really amounts to the same thing. Setting a 4.8GHz OC and using an oversized cooler just to stress temp test with AVX enabled just leaves you 200MHz shy at all times instead of the slim chance you might see a few seconds of AVX in a game.

 

[drea.drechsler]

.... Setting a 4.8GHz OC and using an oversized cooler just to stress temp test with AVX enabled just leaves you 200MHz shy at all times instead of the slim chance you might see a few seconds of AVX in a game.

It may come as a surprise but a few people do more than game with a system. It's telling when I can go all day long 'stress testing' with Realbench...and it won't make one pass at Cinebench20, or one run at my own Handbrake encoding benchmark (simultaneous encoding of 2 movies, about 5 minutes total).

I have to know there's operating margin so that as components age and dust accumulates it stays stable. Realbench just doesn't help with that, it would appear.

 

[Karadjgne]

Not sure. Asus RealBench uses handbrake as part of its testing.

 

[drea.drechsler]

Not sure. Asus RealBench uses handbrake as part of its testing.

As I understand, h.264 (the real encoding engine, Handbrake being the front end) uses AVX instructions for encoding (I'm not that sure though.) I also noticed that's when my CPU seems to get it's heaviest load during the RealBench benchmark script...at least based on fan speed that starts to increase because move the mouse and it stops.

But the way it works seems to be a series of small videos to encode before moving on to another task in the script. My 'real-world' test entails a couple of long movies encoded simultaneously to put maximum sustained load on the processor for an extended duration. These are also some older legacy cinema movies that have never been restored so they probably present a bit of a challenge to the encoding engine. But that's the kind of tasking I do, so stability is also important there.

 

[Karadjgne]

True. There's nothing really that's modern that gets along well with legacy stuff lol. I gotta go hunting for a DX9.0c sdk just to be able to play some older favorites..