Overclocking Basics?

[Zonal]

I’m a bit confused, I was researching overclocking a ryzen 5 3600 and when they were done the MHz was only about 100 or sometimes none above the boost clock. I don’t understand why they did it when it’s so close. Thanks for all your help! :) 

[Guest]

1 minute ago, ZonalThrone said:

I’m a bit confused, I was researching overclocking a ryzen 5 3600 and when they were done the MHz was only about 100 or sometimes none above the boost clock. I don’t understand why they did it when it’s so close. Thanks for all your help!

You want to focus on the ALL core boosts and overclock that. The single core boost clocks are sometimes reachable with all core OC, but requires adequate cooling to do so.

[SFFDesigns]

Just now, ZonalThrone said:

none above the boost clock. I don’t understand why they did it when it’s so close. Thanks for all your help!

Boost clock is single core boost. My 3700X's technical boost clock is 4.4 GHz (never reaches it tho under stock or PBO), and my OC is 4.2 GHz all core, which is higher than the all core sustained boost (during CB20) of 4 GHz on my chip.

 

It rarely boosted above 4.2 GHz anyway

[SFFDesigns]

1 minute ago, ShrimpBrime said:

The single core boost clocks are sometimes reachable with all core OC

Keep in mind that on Ryzen it's very hard to reach single core boosts with an all core OC (even with per CCX, in my case my better CCX has one terrible core which brings it down)

[Guest]

2 minutes ago, Hymenopus_Coronatus said:

Keep in mind that on Ryzen it's very hard to reach single core boosts with an all core OC (even with per CCX, in my case my better CCX has one terrible core which brings it down)

I know.... But.... There's always exceptions...

There's a lot of variables. One I addressed in the sentence you only quoted half of.... 

 

 

[Cyracus]

ryzen stock pulls a crapload of voltage for the performance it yields, if you're limited on cooling potential it throttles like mad. 1usmus released a great OC tool for ryzen, https://www.guru3d.com/articles-pages/clocktuner-for-ryzen-v1-1-ctr-guide-by-1usmus,1.html  Exceeding stock single core boost speeds with an all core OC are a matter of silicon lottery and cooling

[SFFDesigns]

25 minutes ago, Cyracus said:

https://www.guru3d.com/articles-pages/clocktuner-for-ryzen-v1-1-ctr-guide-by-1usmus,1.html

Make sure to test your OC in small fft after using CTR tho, you may have to drop clocks a bit for stability like I did

[SFFDesigns]

32 minutes ago, ShrimpBrime said:

I know.... But.... There's always exceptions...

There's a lot of variables. One I addressed in the sentence you only quoted half of.... 

 

 

Yeah sorry, cooling definitely is a big issue for such OC’s, i should have quoted the whole sentence

 

I wish my chip was one of those exceptions  

[Guest]

29 minutes ago, Hymenopus_Coronatus said:

Yeah sorry, cooling definitely is a big issue for such OC’s, i should have quoted the whole sentence

 

I wish my chip was one of those exceptions  

I've found with Zen+ chips (didn't buy into Zen2, 2700X is plenty power house for me) 

That the processors respond very very well to sub ambient temps. 

The CPUs acted as AMD promised, would boost higher the colder you go.

 

When I experimented with TEC chilling, I was able to idle down to about -30c.

This resulted in all core boosts sustained at idle of 4300mhz.

However, the TEC could only maintain a load temp of roughly 25c depending on the water delta cooling the TEC.

 

For example, 4ghz All core/threads TEC chilled at 4ghz only required 1.188v. Testing was not done outside of benchmarks. 

This shows a pretty large decrease in leakage. However the chip's design prevents it from achieving higher clocks, somethings included in example the pipe lines. Longer looses IPC, and shorter increases it. One of the major differences between FX and ZEN. (There's more to it, but that is one example Die size and density is another).

 

But yeah, high voltage may not be necessary to run a higher clock, it's just most people don't go for sub ambient cooling where these chips actually do pretty well.

[Cyracus]

51 minutes ago, Hymenopus_Coronatus said:

Make sure to test your OC in small fft after using CTR tho, you may have to drop clocks a bit for stability like I did

Well, ctr gave me slightly lower than my presumed stable manual OC (I mean it never crashed but there were things I didn't run like p95 ffts because temps) granted I'm thermally limited so I'm not at the limits of what the silicon can do.

 

ctr 1.1 doesn't appear to be too friendly though (update notes mention changes in stress tests, can't even run the tests because temps are out of control even with very low reference voltage and frequency)
 

31 minutes ago, ShrimpBrime said:

The CPUs acted as AMD promised, would boost higher the colder you go.

to an extent, very minimal gains below 55C, though I'm quite curious how voltage scaling would plot out... https://www.gamersnexus.net/news-pc/3492-ryzen-cpu-thermals-matter-coolers-and-cases

[Guest]

13 minutes ago, Cyracus said:

 

to an extent, very minimal gains below 55C, though I'm quite curious how voltage scaling would plot out... https://www.gamersnexus.net/news-pc/3492-ryzen-cpu-thermals-matter-coolers-and-cases

Agreed. Just like most overclocking goes. It's called diminishing returns.