Are you sure it is the CPU that is the bottleneck here? What I am getting from above, is you don't see a performance increase going from processor x to processor y. That tells me that possibly the GPU is probably holding you back, not the motherboard/CPU combination. Just a thought.

Sorry to clarify I'm not comparing one processor to another - I have a 3800x which is supposed to be able to boost to 4.5GHz as standard but it's not doing that it is just getting to about 4.4GHz on one core at a time. The original problem, now solved, was that after a BIOS update all cores were stuck at an even worse 4.2GHz fixed frequency.
My mentioning needing all the CPU I can get in games is because, while I know my 2080 Super's not the newest card, I still think that the CPU is working so it's worth it working as well as it can.
Now that I solved the annoying CPU not scaling issue I would be fairly happy with my system but for the fact, as mentioned, I could have spent about $200 less for a system that, according to AMD's figures at least, would have performed just as well.

If you are getting one or 2 cores to reach 4400 then it seems to be scaling as needed. Note that it was mentioned that most games are single threaded with the GPU heavily involved so I would not expect very many of the CPU threads to peak. One or 2 peaking seems right in my book.

Well, it is scaling, yes, but AMD have ripped me off since they state 4500 is possible.

Well just find the biggest most expensive CPU cooler you can find and maybe you can get there and beyond . You know, something like a liquid nitrogen setup . Ha!

I already have an NH-D15 and, according to k10temp, at least, I'm only hitting about 75C when running mprime on all cores and leaving it an hour or so.
Like I typed, I'm annoyed at having spent so much to get less than I should have. If I bought the 3700x and a cheap motherboard and used the stock cooler I got with the CPU I'd be happy but, as it stands, I've wasted over $200 thanks to AMD lies.

I wouldn't say 'lies' or 'ripoff' here - folks these days are too quick to jump down that rabbit hole imho. I would say that '4500mhz' figure is probably an 'up to' not a 'guranteed all-core' figure, which is pretty typical for modern CPU specs - it sounds like everything is working exactly as it should, to spec, and the performance gains you'd get from an extra 100MHz would be imperceptible (if not immeasurable). All of that said, yes, the diminishing returns curve for CPUs with gaming are absolutely at a 'low' point price-wise (an Intel i3-10100 is around 'more than good enough' point for the vast overwhelming majority of games), and AMD's marketing has done a great job pushing $400-800 chips as 'requirements' for most games. But it does not sound like your chip is running out of spec or not doing what its supposed to be doing. Note that '4.5GHz on all cores all the time' is a fairly hefty overclock for a modern multi-core chip.

If I were getting 4.5GHz on one, maybe two core at a time I'd say no harm done but disappointing - the fact it won't hit 4.5 on any core with adequate power and cooling is what makes me say it's a ripoff. Ideally, of course, I'd like to be able to hit, say, 4.5 on all cores for a minute or two, or whatever, but to me it's a ripoff when they advertise an up to, you put the chip in the optimum situation and it still doesn't do it. Perhaps the 3700x is equally below advertised, I have no way of knowing, but if it isn't then I was ripped off.

Remember one thing. The interval required for measurement may be slightly off the actual peak speed. CPU demands fluctuate in the millisecond range and most monitoring tools function in the seconds. There is a big difference in timing there and it is unlikely that you will see the actual peaks and valleys since monitoring tools only take a snapshot.

Accept that the CPU you have is "up to" 4500 MHz and far from "4500 MHz on all cores all the time". The fact that you have seen it at 4400 on at least one core says it is working as designed.

Your expectations are far out of line with reality.

Also note that on my system with an AMD Rhyzen 5 3600 the spec says yet the way I use it I frequently see all cores at 4400 or better. In fact, right now with the system running at ~60% load on all CPUs I get this: Have you even tried using something such as "inxi" to see what is actually happening?

On my laptop, again at ~60% load, I see:

I don't think it is wrong to expect a CPU rated to be able to boost up to 4.5GHz on a core to actually be seen to boost to 4.5GHz on a core. Besides, if you take a look here https://en.wikipedia.org/wiki/Ryzen#CPUs_4, you will see that I, literally, paid for that extra 100MHz extra turbo. That is my issue here and why I say AMD ripped me off not that I can't get 4.5GHz on all cores - that kind of thing may or may not be feasible (I feel sure I have seen at least one person state it has been done) but is besides the point that if AMD are to be believed I paid ~$100 for nothing at all.

Before buying or upgrading any kind of processors, may I recommend userbenchmark.com?

Just been through a number of searches for a "new" cpu and it's the same story as a decade ago... a myriad of cpus with 100-200-300 MHz differences... sigh.

For practical use a jump must be at least 15-20% in performance before you can feel the difference in daily work...


Now your two AMDs...

Actually, noone is seeing that increased performance... https://cpu.userbenchmark.com/Compar...00X/4043vs4047

If you had seen that, you clearly would have gone for the 3700X?

But to get to the magical 4.5 GHz there's probably a timing/voltage setting somewhere you've missed. My point is, you're wasting time as it will make no difference whatsoever...

That does sum it up, yes. I had put the machine together "in a hurry" with the idea that a higher binned part would be more likely to clock higher. I mean, it still might actually be a tiny bit more performant than the 3700x but, yes, had I realised I would have either gone for the 3700x (or even a 3600 on an x570 motherboard for future upgrade) or splurged on a 3900 for more cores. The idea being I'm not going to be able to afford a new PC for the forceable future.
Sadly AMDs lies, sorry marketing, got to me. Oh, well.

He, I've just bought a second-hand 3600 for $170. The 3600X model? all I get is +50% increase in powerconsumption...

https://cpu.userbenchmark.com/Compar...00X/4040vs4041


Now I upgraded from a core i7-930 and this is what I call an upgrade!

https://cpu.userbenchmark.com/Compar...-930/4040vsm79

Keywords are "much faster" and "hugely faster"... 48% and 71%...

Plus the power bill will drop like a stone!

Haha, my colleague made the same upgrade as you (I think exactly but if not very close) and was very, very happy with it. Don't get me wrong, this is a far better system than my old AMD Piledriver and it still wasn't as much as the Intel equivalent just sucks that I should have spent more or less to get better value.

Exactly this. Whatever tool you're measuring with (e.g. watch'ing /cat/proc or inxi or what-have-you) is probably only polling every 1000-2000ms, but the CPU is able to make adjustments far more frequently. And again, this is an 'up to' value (and even if it were 'hitting it on one core' it would make no perceptible difference to the system's performance, and the measurable difference would be exceedingly small (as in, within margin of error for most measurement techniques)). This truly is an excercise in trying to fit dancing angels on a pinhead. Further note that 'single core' boost speeds are always worthless, because they're unattainable in real systems (even with a 'single core application' the kernel will always schedule tasks to other threads if SMP is available so unless you want to disable SMP, you will never see 'single core boost' in real world). The only CPU family for which that doesn't appear to be uniformly true is the AMD 15h processors (e.g. FX, most A-series APUs, Opteron x200 and x300 series, etc) because the clocking is per-module (e.g. 2 threads) and it isn't uncommon to see one module sustain 'single core boost' while the others scale back. Of course the other option, if you feel you've been somehow denied 'performance that you paid for' is to disable all those slick power management and dynamic boost algos (which, I would note, you also paid for) and run the chip with no limits of any kind (e.g. socket/core power limits set to 4kW, current limits set to 999A, etc) - power draw will be obscene, temps will be obscene, but you'll get 'all the performance' (at least until it invariably overheats and throttles back, or the VRMs overheat and the board crashes/fails). If you've got a slick enough mainboard and sufficient cooling it can probably be run this way more or less sustianably (but who knows what long-term life-cycle may look like as the 'boost' may require voltage that promotes degradation of the chip - basically 'try it and see' and its probably not catastrophic but you'll know when you know), but generally the performance increases are still quite small relative to the power/thermal costs.

As far as the 3800X not being 'better' than the 3700X - my question would be, does the 3700X actually hit its advertised 'up to' clocks as you're demanding of the 3800X? My suspicion is not - so while there isn't a great deal of difference (again, because most chips these days aren't that far apart, and are all binned pretty aggressively), the 3800X *is* likely hitting higher clocks for whatever that may be worth. 'The law of diminishing returns' has always applied to higher-end CPU SKUs.