Trying to understand Ryzen 3000 series boost speed variations

[mtcn77]

I'm not enjoying proving you wrong, nor you can rile me up. You assume too much.

I would highly regard that if it were to be true, AMD would have carried along the hypetrain with a little bit more presence. Yes, it is there, but does it do anything? I'm not here to mince words, sometimes you have to call the voice of reason and I hold an esteemed viewpoint as a fact checker. I would expect the same prudence from yourself to disregard the empty promises.

 

[Zach_01]

I would highly regard that if it were to be true, AMD would have carried along the hypetrain with a little bit more presence. Yes, it is there, but does it do anything? I'm not here to mince words, sometimes you have to call the voice of reason and I hold an esteemed viewpoint as a fact checker. I would expect the same prudence from yourself to disregard the empty promises.

So, we are back at this:

Of course AM4 ZEN2 CPUs have it, but maybe he ment to say that they shouldn't...

And that is clearly a personal opinion. To have an opinion is respected, and me or anyone else who does not agree with it, can respectfully disagree.
That is a different matter.
But an opinion you tried to state as a fact, like an AMD statement, covered under old and obsolete articles and states, is a whole other one...

 

[mtcn77]

So, we are back at this:

And that is clearly a personal opinion. To have an opinion is respected, and me or anyone else who does not agree with it, can respectfully disagree.
That is a different matter.
But an opinion you tried to state as a fact, like an AMD statement, covered under old and obsolete articles and states, is a whole other one...

With all respect, do you think our assumed foregone conclusions makes any difference outside of our frames of rationality? It is, or it isn't. The difference is not for us to decide.
I would highly debate neither is for AMD marketting to decide, so thus we are at a fray. Why make any rationalizations when AMD blurts out the obvious elephant in the room. At this point you are picking between AMD's assigned market associations which is another form of spin doctoring, imo. Don't fall for it.
PS: had it been obsolete, it would be missing from the amd central.

I don't have to fight tooth and nail, no pun intended, to point out pbo is only for ctdp customisation in accordance with motherboard overspecification and not related to mainstream pcs, but have your side of the argument. I'll leave it to rest by pointing out just how much threadripper would use unless AMD prevented a meltdown like how Arctic runs asterixes on the bios bug that can cause intel 7980x use up to 270w unless boot restrained.

 

[Zach_01]

Well done. You've manage to drive the discussion to elsewhere from what I orinaly saying...
I'm out...

 

[mtcn77]

Well done. You've manage to drive the discussion to elsewhere from what I orinaly saying...
I'm out...

I might have difficulty of expression and perception. Don't hold it personal, but it was my argument to make. Whatever you say, don't neglect that I always keep a personal narrative, so I pull all arguments towards my end goal. Which is pretty simple, pbo by itself doesn't mean anything unless we find it an explanation to spend man hours. AMD isn't in a lofty idealist position to add fuel to the flames. They have to meet the field right then and there and they have to deliver on this persistently. Intel can bear these royal cluster fails. The hard line is AMD can't.

 

[Zach_01]

No worries... i definately dont take it personally.

 

[oxrufiioxo]

Speaking of PBO pretty interesting video although it probably only applies to dual CCX chips.

 

[Waldorf]

@mtcn77
posting a wall of text, without a single argument supporting what you claimed (TR only)?
Ok...

unwatched this anyway

 

[mtcn77]

@mtcn77
posting a wall of text, without a single argument supporting what you claimed (TR only)?
Ok...

unwatched this anyway

It would be a dissertation from the product placement strategy that tr brings.
You wouldn't wish me bring up the thread where I was promoting high operating temperature in mitigating ryzen heat load? It would only serve you to see the chips heat up by two folds by operating at 95°C in comparison to 65°C. Which is coincidentally the default case tr operates under normal circumstances.

Spoiler: Thread I cautioned you about:

65°C it draws 80Watts and sits at 4150MHz all cores
@94°C+ almost 170watts, throttling rules and it sits at 3900MHz all cores.

How often does it seem it would overheat with two dies than one is my question.
Don't forget I tried to sound a warning about overheating below maximum operating temperatures. It is contrary-wise to common intuition which is why we have chips both cold and throttled rather for the complete opposite of useful scale at max boost.
TR has 2x more chips and twice more capacity to overheat. You do the math whether leaving ctdp low prior to launch at pbo privation is a wise thing to do.
Why Intel CPU's run at 95°C and why AMD's should, also

 

[Zach_01]

Speaking of PBO pretty interesting video although it probably only applies to dual CCX chips.

Oh it does not I assure you...

Any way, he used scalar X10 and that is extremely agressive with voltages, hence the ridiculus power draw and temp. I tried it (limits 230) with X10, X5, X4 and lower is much better, with higher clock (on P95).

 

[mtcn77]

Let's leave it to Arctic to demonstrate us just what would happen without such a bios switch. Same as before with Intel automatic overclockery;
** c’t magazin has shown in issue 17/2017 on page 96 that Intel i9 CPUs on socket 2066 may draw up to 270 Watts since wrongly programmed BIOS by mainboard manufacturers fail to limit the TDP of the CPU to the specified 140 Watts. This may lead to thermal throttling of the CPU.

 

[TheoneandonlyMrK]

Oh it does not I assure you...

Any way, he used scalar X10 and that is extremely agressive with voltages, hence the ridiculus power draw and temp. I tried it (limits 230) with X10, X5, X4 and lower is much better, with higher clock (on P95).

setting PBO on for me set TDE EDC PPT to(auto which equates to) 255 and while crunching i got all cores to 4.025 all cores sustained boost at 100% load

following Buildzoids Theory I maually setup PBO to TDC168 EDC 168 PPT 168 im now at 4.135 all core boost at the same load and its stable , been on a day so far as it is.


worthy of note i have a manual temp limit of 80 set and it works well.

 

[Zach_01]

setting PBO on for me set TDE EDC PPT to(auto which equates to) 255 and while crunching i got all cores to 4.025 all cores sustained boost at 100% load

following Buildzoids Theory I maually setup PBO to TDC168 EDC 168 PPT 168 im now at 4.135 all core boost at the same load and its stable , been on a day so far as it is.


worthy of note i have a manual temp limit of 80 set and it works well.

I did follow my self this theory, and I restarted the system over 30 times the last night from testing different settings.
While in P95 the clock went from 3650 to 3950, the CB-R20 score gone down the drain. My temp never passed 72C(P95) or 64C(R20) with the default limit.
This kind of PBO usage is extremely aggresive (for voltage and current =power draw) and does not apply to all kinds of workloads. One must do some testing on the interested workload, otherwise performance could be crippled...
PBO needs a more refined and fine tune approach.

 

[mtcn77]

PBO needs a more refined and fine tune approach.

Maybe it is only for power gated hardware, no? We are stuffing words in AMD's mouth at this point...

 

[Zach_01]

Im not stuffing anything to anyone's mouth. This is all me...

 

[TheoneandonlyMrK]

I did follow my self this theory, and I restarted the system over 30 times the last night from testing different settings.
While in P95 the clock went from 3650 to 3950, the CB-R20 score gone down the drain. My temp never passed 72C(P95) or 64C(R20) with the default limit.
This kind of PBO usage is extremely aggresive (for voltage and current =power draw) and does not apply to all kinds of workloads. One must do some testing on the interested workload, otherwise performance could be crippled...
PBO needs a more refined and fine tune approach.

try with a manual temp limit, it will not exceed it , I went with a safe 80

 

[Zach_01]

try with a manual temp limit, it will not exceed it , I went with a safe 80

My temp never passed 72C I said...

 

[TheoneandonlyMrK]

My temp never passed 72C I said...

That's not my point , your also adjusting the temp it will boost upto it kept mine boosting more longer.

 

[moproblems99]

I figured I would stop posting in my thread and just post in here to try and consolidate.

I have been doing some testing over the last few says and I have almost got my 3900x into the 7500s on CB20. So far, I think the two most important changes I have made are setting my LLC to 5 to keep the voltage lower (and temp and power) and then limiting my EDC to 150A. I am not exactly sure why limiting EDC is effective unless it is limiting my current draw and reducing temps which is keeping boosts up. 7469 is the best I have made it to with all-core boost running between 4110mhz and 4140mhz. I think I could tweak EDC a little more for some room but if you limit EDC too much then you can't boost well and you are also unable to hit TDC targets either. I have been playing with the scalar but I can't actually see it having much of an impact at the moment.

Still testing so will update as it happens.

 

[lorry]

Finally got it watercooled! Took forEver it feels like. Went with the Optimus block as you know, skipped the Alphacool AIO (it's sitting here unused, lol, I'll use the rad at Least when I do the GPU most likely)
Went with Alphacool NexXxoS XT45 Full Copper 420mm Radiator V.2, 3 Noctua 140 industrial 2000, EKWB EK-Quantum Kinetic TBE 2r0 D5 PWM D-RGB Pump Reservoir - Plexi
Highest CPU temp is around 79C under a full load (Aida64, Prime95 etc) Water temp sits at around 24/25. I ended up using their block mounting in the end, If they had made their bolts 5mm longer you could use a Noctua backplate with ease, no idea why they don't though, cost?

Oh yeah, almost forgot, ambient is usually around 23C

http://imgur.com/2xrNtqg

 

[Waldorf]

@moproblems99
not sure, but i think i saw recommendation to limit EDC to about 140.
tho cant remember what chip (3700/3800?), as i rebuild 2 rigs, build a new one, and swapped parts on all incl mine in the past 2 weeks,
and read up on lots of stuff.

@lorry
nice to see it finally coming together

@All
finally changed proc for 3700x, but im getting same low SC/ST clocks i had with the 3600.
CB20 ST basically runs at 3.6ghz all the way, except for about a dozen bursts that boost to 4.35.
ST gets 508, MT4830, so not too bad.
stock bios except tweaks and optimization for hw/1usmus plan, xfr/pbo off, ram 3600@16/19/19/36/1T@1.35v
anyone knows how i see no boosting past 3.6 on ST stuff?