Blogs Has AMD made overclocking Ryzen CPUs pointless?

The chip must be giving itself more juice if its boosting more than a manual overclock maybe?

 

It definitely gives itself more juice. Mine will go as high as 1.57v for a second or two before normalizing. I manually did my OC, though. I much prefer knowing what my CPU clocks and volts are doing.

 

They may have taken our CPU overclocking but they'll never take our RAM.

 

So a similar situation to Intel chips, and GPUs from both vendors: the chips overclock themselves (AKA Turbo, GPU boost, etc) as hard as is possible with internal knowledge of sub-component voltage levels, power draws, clocks, local leakage, local power, etc. The only time manual overclocking is going to have an edge over that is when operating well outside the chips normal operating conditions (e.g. LN2 cooling, dangerously high voltages with an acceptance of reduced chip lifetime, etc).

 

With respect to Ryzen 1 it has pretty much been the case that overclocking is pointless for me, my chips all core is 3.9 stock (8 core TR1900X) in light threads often 4.15-4.2Ghz and I could only get 4.1Ghz manual overclocking with a level of volts I was comfortable with, just wasn't worth it for the negligible performance bump.

I'm hoping to drop in 12 or 16 core next but am only interested if PB2/XFR2/PBO etc allow me to run all core >4Ghz most of the time with out me having to dick about. I don't want to sacrifice core speed for more threads.

 

I would say so yes. My 2700x boosts itself to 4GHz on all cores when they're at 100% without any fiddling from myself. And that's running both synthetic tests like Prime95 and also when doing real world tasks that load all threads to 100% like exporting photos out of lightroom etc. I suppose I could fiddle in the BIOS for an extra 100-200MHz but it hardly seems worth it when it introduces the possibility of instability.

 

1800x 3.925ghz 1.41v peak (to get 4ghz I need 1.45v+ just not worth it for 75mhz+heat)

The 2700x has interested me but my money sense says wait for 3700x or alike model (1800x vs 2700x just means I don't have to overclock it as it does 4ghz out of the box all cores and peaks of 4.3ghz on 1-2 cores when load is only on 1-2 good cores)

 

So let's be real careful to not present these AMD developments as a negative, somehow*.

You overclock because either (a) you want to see how fast the technology can be pushed, or (b) to demonstrate your prowess in this, or (c) you want maximum heat-be-damned CPU performance 24/7. These are all fine reasons. And back in the day when intelligent boosting wasn't on the table, those of us who wanted "best daily performance" also had to choose (c).

But AMD (and possibly Intel; don't know) has advanced things. Sure, we can still choose (c), but we no longer have to. Now we can get almost the same thing - no noticeable difference - without overclocking, because AMD makes that headroom available automatically. Best of all, to me, is that we needn't dissipate an extra 50W or 100W continuously.

This is wonderful. And it's one of the reasons I bought Ryzen as soon as it came out last summer. I'm happy to not be overclocking which, as noted, may soon have no relevance beyond sport. Thank you AMD!


*E.g. as a "problem for overclockers". Problem?? You're begging the question that overclocking is inherently a good thing.

 

The problem is, AMD use unlocked multipliers and overclocking across the range as a specific marketing point against Intel - if, in reality, that ability actually results in lower performance, then that is a negative.

 

Doesn't it just strength that marketing point? Overclocking used to be a way of getting close to top end performance from chips lower down the stack until Intel made it the reserve of their high-end products.

Isn't this a return to that, enthusiasts and the technologically savvy customer can save themselves a few quid by getting a cheaper CPU and overclocking it while others can pay more for the X version and have it all done for them.

 

Why doesn't an Intel overclock plus enabling Speedstep give the same result as a Precision Boost selective overclock?