Der8auer: "Good number of Ryzen 3000 chips does not reach boost clock"

You still don't seem to get my point: "virtually" the same as in Hilbert's review is probably an understatement. If those numbers we are discussing in this thread are something to go by, they are NOT as in Hilbert's review. That's the whole point of this thread...
Also, you should not be surprised, since I know that FW updates can change the performance. But that doesn't mean I have to like it. Or did you think it was good that Intel's mitigations, especially on non-last gen products crippled their performance beyond the margin of error, so that you could measure it, and at times, feel it? For me personally, and I understand that this is subjective, it takes out the confidence of a product. I am aware that this is highly subjective in nature, and you don't share that sentiment. But that doesn't mean I am wrong to think so, nor are you to just trust that thing and not think it's a bother.
Sure we are talking about a new CPU node, architectural improvements, chipset signalling (PCIe 4), probably soon a new IMC design etc. etc. but that again, doesn't say I have to like that one has to go through this when spending hundreds of $/€ on your hardware. "Plug and play", "fire and forget" comes to my mind... which is not installing a quaterly security mitigation with Intel, or a monthy AGESA version with AMD tbh. But, I guess it comes with being innovative... or carrying over faulty design for decades in the case of Intel.

Yes, your argument with more cores and multithreaded applications is valid. With AMD's growing market share, if so, and even more now that Intel gets off their butt and sells more cores. But... we have been talking about this for years. Why do you think did I buy a 5930K back in the day, for 6/12, instead of another quad core? But... sadly, it still doesn't feel like we're there. 5 years later. And I wouldn't want to bet money on seeing real performance gains between 4/8 and 6/12 in today's games. I already bet money on that and probably lost when I invested extra in a hexa core CPU those years ago.

The comparison with a car is good, but not 100% (since interior might be better built, more extras etc.). But I see what you mean, and I have to agree.

Yes that's true... I guess Zen 2+ (Ryzen 4K) will still be on AM4, but beyond that I doubt that AMD will be able to keep their platform. As for Intel, they could just name their sockets by the year, enthusiast platform "year+" and server wise "year-S". Would also be easier to compare for people not reading into it.

 

Not sure if something inside the CPU actually sets this, but instead some of it's behavior is being watched by AGESA/BIOS and the clocks are being adjusted accordingly, e.g. "this cpu needs 1.45v to go up to 4.3ghz and is above 70ºC, so let's not push much farther"; "this cpu got to 4.4ghz with only 1.4v and is averaging 70ºC, so let's use that"; etc. So, a kind of automatic binning of the chip, setting an overclock accordingly, regardless of the model.

I didn't tested each core, so I just let the stock configuration figure out the best core (I think in my case it's core #4).

You just described "boost clock" and that's exactly how it's supposed to work: higher boost clocks for single/dual-core workloads. It doesn't really matter how many cores can do the maximum boost - as long as two of them can, the performance will be the same.

All-core frequency is closer to the base clock, which is the absolute minimum guaranteed to work, though I think most Ryzen 3000 CPUs hover at 4ghz all-core 100% load in stock config (mine does). So, no issue there.

 

What you describe means that AMD either did not test that given CPUs can achieve rated clock and just randomly sold them under given name.
Or that they did bin those chips and ignored binning.

 

or they screwed up the SMU somehow

 

Or they binned the chips the best they could but realized after that they needed to adjust things?

I'm not sure why we're still having this discussion when both AMD and ASUS said this is what's happening. ASUS said AMD adjusted the limits for reliability in AGESA update and HWInfo apparently reports this as well:

https://www.overclock.net/forum/28085580-post549.html

 

I am not discussing that. I am discussing people who indirectly or directly suggested that AMD sold those chips knowing they are not physically able to reach given clock while remain stable. (Which did not look like the case based on people who did manual "OC".)

Is that only about temperatures through different boards and BIOSes? Apparently, there is huge difference between under and above 80°C. But CPU was set to keep given clock till 95°C.
Instead it should have had more steps, and actually lower clock as it gets closer to 95°C limit. Which is likely what AMD is tuning up now.

Interestingly enough 2700X has limit temperature of 85°C.

 

@fantaskarsef
And why i bought an 8C AMD back then too.
But the big difference between then and now?
Both 5 gen consoles will run 6 or 8C with two T,
and since lots of games are either ported from console
to win, or at least made with that in mind (for pc version).
Knowing that they will make use of the additional power,
i expect the games (from at least end 2020 on) to be more tweaked for C/T, rather than clocks.

And in 4 to 5y i might just swap board/cpu completely,
but with old stuff sold i HAD to get something, and nothing on Intels side interested me, especially with the higher cost on cpu/board (paid 120$ for a decent 570 board).
And i said virtually, as i take the word "identical" literally, meaning if he'd get lets say 3198, that i would get that as well.
Running a hydro 100i (weird enough with a v2 rad) im getting min 3.6 without load (balanced power), 4.2 max boost, 4.02 to 4.07 all core boost, and 4ghz all core boost with prime95.
All stock/norm/reg settings in bios, any boost/oc settings turned off (to not see board maker mess with stuff),
and only ram man set to 3533, no xmp selected.
So without man tweak/oc, nor exotic cooling, i get advertised/expected numbers.
Of course i can only speak for my 3600....

 

Yet another one: Ticked Off: Clock Bug Leads to False Ryzen Benchmark Scores: https://www.tomshardware.com/reviews/amd-ryzen-clock-bug-benchmark-scores,6312.html

 

thats not even a real bug, its because those fools turned off HPET in the bios.

HPET needs to be enabled bios side, not forced OS side.

Toms is yet again, absolutely fkn wrong.

 

Like what?

Cinebench? isn't it one benchmark, that half of internet knows how to alter to change score without detection?

Secondly, article talks about intentional cheating. Not basic behavior of any chip or board.

 

In the end, the best tests are with real applications and games.

 

Yeah i got a 3600X and the max advertised boost clock is 4.4 GHz (4400).At most i'm getting 4.25 GHz (4250).I'm missing 150 MHz from this.If i knew that CPU's were working in the same way as storage say 1TB, but you get 930 GB's instead of 1000 GB's, then i wouldn't mind.This however is a different case.I'am a bit angry but also it's not something that's gonna make any big difference in anything.It's just that i payed for more and got less.

 

Did you pay for the clock frequency or the performance?

 

Yeah Toms will do ANYTHING to make click bait, the thing is TOMS is the arm pit of computing period! But they make them selves seem so important...........

 

Back when Thomas Pabst ran the place, sure.

even he is not happy with the quality of content these days...

 

@jose2016
But you cant compare between different rigs the same way you can with a bench.
E.g. If i want to see the difference between 2 gpus.
You wont have to try to be in the same spot at the same exact etc, just so load is identical.
Even harder with something like shooters where stuff isnt happening the same way on multiple runs..

@Megastitious
ST or MT? Pbo on?

 

Actually, I have a TR 2950X and it doesn't technically hit it's boost clock - which is supposed to be 4.4Ghz.

Running a Rog Zenith Extreme, with $1000 of water cooling equipment strapped to it. It will sit at 4.1Ghz+ on all cores, all day while encoding video and sit around 51-56c at 100% CPU use (ambient 28c) - but it won't hit 4.4GHz on any core.

 

What's the closet it gets with only a single thread loaded?