[SOLVED] Confused on 5900x OC

[victortsoi]

Hey everyone- so I'm completely lost regarding what the best course of action is for the 5900x.Coming from an 8700k that I've been running fixed at 4.65 for 4 years straight, I:

1. I Don't even know if the chip benefits from OC in the way intel chips did.
2. Is it the case that PBO2/undervolting is the most efficient and safe way to optimize the chip, instead of forcing a single clock across all cores?
3. Is this best done through bios, or a program like Clock Tuner Ryzen (CTR), or Ryzen Master? (I guess I shouldnt use the asus intelligent processors program that comes with the motherboard).

Thanks in advance. I just want the best for my 5900x!

 

[Phaaze88]

I can chime in just with number 1... hopefully, you get assistance with the other 2.

I Don't even know if the chip benefits from OC in the way intel chips did.

NONE of the modern chips benefit from it as much as the old ones did, including Intel.
Older Core i K SKUs could be all core OC'ed to match their single core turbo - pretty much guaranteed, even with a crappy bin that couldn't be pushed further than that. [Those Sandy Bridge cpus were OC beasts.]
Since the 9900K, that has not been the case; 5.0ghz single core turbo, and according to Silicon Lottery's data, not all samples could hold that across all cores.

Now, it's sacrificial overclocking, with the tradeoff of all core OC being worse the higher one goes up a particular product stack.
The single core performance is being dropped in favor for more multicore - this action does not benefit all applications; some applications still benefit from higher single core, thus the sacrifice.
Ryzen 3600, 3600X, and 5600X are the only models among the last 2 Zen lineups I believe that avoid SO'ing.
3700X, 3800X, and 5800X: If you get a good bin, great. If not, it's a sacrifice of 100-200mhz.
3900X, 3950X, 5900X, and 5950X: Good luck not throwing 300-500mhz out the window...


Short answer: No, and Intel K SKUs after the 9700K aren't immune either.

 

[Schlachtwolf]

I have a 5900x since last November and have had a bundle of fun OC'ing, undervolting, fiddling with RAM timings, memory latency etc. etc. and seeing apart from a few extra points in Cinebench absolutly no real world benefit whatsoever LOL !!!
Your cooling solution is more important than a wee undervolt or any other setting to get the best out of a 12 core CPU.

So my answer is

1. No real benefit, just runs hotter and therefore louder.
2. My Ryzen on STOCK settings gets the same score as an 4.45 ALL core OC. Undervolting keeps the voltages down a tad lowering temps but you lose a few % in performance.
3. What I have done..... Ryzen Master-Áuto OC- leave it alone and enjoy your PC. Let it do it's thing and boost up to 5gh like mine can.

Congrats on getting one as the 5900x is a fantastic CPU.

 

[drea.drechsler]

....

Now, it's sacrificial overclocking, with the tradeoff of all core OC being worse the higher one goes up a particular product stack.
...

I just wanted to add one other point to your 'sacrificial overclocking' analogy for Ryzen 3000 in particular. Many people completely ignore light threaded performance loss as Ryzen 3000...and 5000 in particular...will boost super eagerly in light threaded work loads. That's the kind of work load that benefits gaming.

So what they do is dial in a fixed overclock that gets them a performance uplift in an extreme all-core benchmark (Cinebench, timed X.264 transcode, whatever) and call it great. But they ignore the single thread benches because they take too long to process. If they would they'll often find a fall-off in performance (sometimes pretty significant) and that's what gaming most needs.

 

[Phaaze88]

I just wanted to add one other point to your 'sacrificial overclocking' analogy for Ryzen 3000 in particular. Many people completely ignore light threaded performance loss as Ryzen 3000...and 5000 in particular...will boost super eagerly in light threaded work loads. That's the kind of work load that benefits gaming.

So what they do is dial in a fixed overclock that gets them a performance uplift in an extreme all-core benchmark (Cinebench, timed X.264 transcode, whatever) and call it great. But they ignore the single thread benches because they take too long to process. If they would they'll often find a fall-off in performance (sometimes pretty significant) and that's what gaming most needs.

Aye, it's because of the Primary Thread(s), which all game commands are run, are still dependent on single core performance.
Games may be using more cores now, but the commands/code are still being run through a single core.

If folks want to try their hand at overclocking out of curiosity, that's fine.
It's the ones pursuing it for more performance, like it's guaranteed to be better...