DerBauer hardware survey highlights issues with max boost clock on Ryzen 3000 "It's worse than I thought"

[Crowbar]

Just now, porina said:

Following link is best write up of it I've seen. Read it. Move on.

https://www.anandtech.com/show/14582/talking-tdp-turbo-and-overclocking-an-interview-with-intel-fellow-guy-therien

 

Intel spin doctoring . I'm not amused or surprised.

[RejZoR]

14 minutes ago, porina said:

Following link is best write up of it I've seen. Read it. Move on.

https://www.anandtech.com/show/14582/talking-tdp-turbo-and-overclocking-an-interview-with-intel-fellow-guy-therien

 

Advertise TDP at base clock. And then run CPU's WAY past that pretty much at all times. Yeah, sounds plausible lol... But hey, no outrage there since people don't even know what it is, but go absolutely mad because their CPU never hits max boost clock. Something which is the theoretic top of Ryzen CPU's. You can't ever go past that under normal conditions and during normal conditions, you probably also won't hit it because conditions are probably never met. Either thread count that most likely prevents it, total ACTUAL TDP, voltages, temperature and bunch of other things. And I'd sort of understand if Ryzen CPU's were shit. But they are excellent piece of hardware despite not even hitting the max boost clocks most of the time. It's why this thread is such a bizarre thing to see.

[Flying Sausages]

1 hour ago, RejZoR said:

That's just a spec decision. They could say it has 30 HP and then everyone would lose their shit if it actually had 10X the HP on a dyno. Coz wow it goes so far beyond the stated spec, it must be insane! But we all know it would be BS. Just like it is whining about not reaching max boost clock. It has fuck tons of cores, auto clocks as high as it can get based on several parameters and people still bitch about it. Sigh.

( Not owning Ryzen 7 3700x atm) But I would be still a little bit piss not to see my Ryzen 7 3700x hit 4.4GHz max turbo. Some people reported either 4.2 GHz or 4.375 GHz max turbo.

[5x5]

22 minutes ago, OlympicAssEater said:

( Not owning Ryzen 7 3700x atm) But I would be still a little bit piss not to see my Ryzen 7 3700x hit 4.4GHz max turbo. Some people reported either 4.2 GHz or 4.375 GHz max turbo.

Because 25MHz or 0.07% will affect you in any way shape or form. You want the 25MHz? Get a better board and cooler. Done.

 

Same thing with intel. Put a 9700K on a cheap B360 board and will struggle to maintain absolute max turbo.thats why the turbo boost description for every vendor states GIVEN POWER, THERMAL AND VOLTAGE REQUIREMENTS ARE MET, THE CHIP WILL BOOST UP TO XXXGHz.

[Flying Sausages]

19 minutes ago, 5x5 said:

Because 25MHz or 0.07% will affect you in any way shape or form. You want the 25MHz? Get a better board and cooler. Done.

 

Same thing with intel. Put a 9700K on a cheap B360 board and will struggle to maintain absolute max turbo.thats why the turbo boost description for every vendor states GIVEN POWER, THERMAL AND VOLTAGE REQUIREMENTS ARE MET, THE CHIP WILL BOOST UP TO XXXGHz.

 

1st comment with 360mm AIO and a hella expensive ASUS ROG board unable to get into the max turbo territory. Explain.

[5x5]

19 minutes ago, OlympicAssEater said:

 

1st comment with 360mm AIO and a hella expensive ASUS ROG board unable to get into the max turbo territory. Explain.

Easy, until he provides proof of ownership, he might as well be claiming to have a magical elephant that grants wishes. I know a guy whose entire job is to look up social media posts and videos of their competitors and pretend to have problems with their products from multiple accounts..it's aggressive corporate attack marketing in its purest form. Naturally, I can't disclose his specific employer or competitors but I can just suggest looking out for eerily similar posts across various sites.

[Curious Pineapple]

12 hours ago, Necrotic said:

Ignoring anything you don't agree with by claiming they are an intel fanboy just shows that you are an AMD fanboy...

I've not ignored anything, but I've been around here long enough to see the general attitude towards one brand over another. Also my history when it comes to choosing hardware would show no real preferance to either brand, however currently I have 3 Xeon workstations, an old Centrino notebook, a Core 2 Quad ITX board for embedded devlelopment and an AMD tablet so if anything I seem to be a bit of a fan of Intel hardware. My FX machine with a Radeon got replaced with a dual Xeon (with worse single thread performance) and a GTX650 because it suited my needs better.

[cj09beira]

3 minutes ago, 5x5 said:

Easy, until he provides proof of ownership, he might as well be claiming to have a magical elephant that grants wishes. I know a guy whose entire job is to look up social media posts and videos of their competitors and pretend to have problems with their products from multiple accounts..it's aggressive corporate attack marketing in its purest form. Naturally, I can't disclose his specific employer or competitors but I can just suggest looking out for eerily similar posts across various sites.

and its not like we know anything about his setup order than those two things, 

 

many things seem to be happening here, for one due to how new this is its obvious that software isn't quite ready for how fast zen can change frequencies, which leads to those high voltages at idle, then you have the fact that max boost is core dependent which means performance is in the hands of the scheduler, thus windows version chipset version and power profile all come into play,

and finally amd was quite aggressive with their boosts / binning.

 

so the result is that most people seem to hit 

3600 50% of them hit the target

3600x 18% hit it, with most being 75-50mhz off

3700x 50% hit it (or within 25mhz)

3800x 57% hit it (or within 25mhz)

3900x 18% hit it (or within 25mhz) with most of  the others hitting 50-100mhz of

 

25mhz is close enough to be considered hitting the clocks, as spread spectrum can easily affect base clock that much, even on intel many times bclk isn't set at straight 100 which leads to slightly less clockspeed, 

 

so it seems to me like there is here, specially with the 3900x and 3600x, but it doesn't seem like a deliberate choice by amd, 

i think its worse how boost is turning into the gpu boost that we never see on sustained workloads even if it does reach it from time to time

 

 

[RejZoR]

1 hour ago, OlympicAssEater said:

 

1st comment with 360mm AIO and a hella expensive ASUS ROG board unable to get into the max turbo territory. Explain.

Easy. Was it really an absolute single thread load under which it can hit max clocks? Was TDP under the limit? What was the temperature? All this affects clock and you can be assured you won't hit the max. Not to mention 4.6GHz on 3900X is never a round number. It never is. The bus clock or IF or whatever Ryzens run as core clock, multiplying that with the multiplier and depending on the internal clock, the end frequency will fluctuate. My 5820K is overclocked to max boost of 4.5GHz, but if I look at actual real-time clock, it's usually 4.45GHz or 4.49GHz, it's rarely 4.5GHz exactly.

[ravenshrike]

2 hours ago, OlympicAssEater said:

1st comment with 360mm AIO and a hella expensive ASUS ROG board unable to get into the max turbo territory. Explain.

What was ambient, his thermal paste, and his paste application method?

 

 

[7nationarmy]

10 hours ago, 5x5 said:

GIVEN POWER, THERMAL AND VOLTAGE REQUIREMENTS ARE MET, THE CHIP WILL BOOST UP TO XXXGHz.

I am genuinely curious, did AMD mention this anywhere in their packaging, its contents, or anywhere in their Ryzen product page? If there isn't any, they are just setting themselves up for a lawsuit. This isn't about the complexities and factors of achieving boost clocks, which I am pretty sure der8auer is quite knowledgeable in. This is more about the potentially misleading marketing of the 'boost clock' aspect of the CPU, which depending on the answer to my initial question, could have legal weight.

 

10 hours ago, 5x5 said:

Because 25MHz or 0.07% will affect you in any way shape or form. You want the 25MHz? Get a better board and cooler. Done.

Yes the performance difference is minuscule and won't affect your day to day usage of the PC. Der8auer said that himself and he still recommends the 3600/3900X over the competition. Again, the potential issue is pure marketing. People will find every bit of opening to sue. Not saying AMD is like evil or something if they did not elaborate on how the boost clocks can be achieved, but they can do better in terms of consumer orientation especially when it comes to the average consumer. And also protecting themselves from potential lawsuits.

 

On another note, I was reading a local forum and saw some people trying to sell their Ryzen 3xxx because of this news and said they should have stuck with Intel. Pretty sure some are just trolling but I think some also aren't. LOL

[Derangel]

13 hours ago, Crowbar said:

So how should intel be setting their boosts then since not all of their CPU's come with a stock cooler despite being priced higher?

 

As much as I enjoy a lot of derbauer's content this one is completely brain dead. You can't blame AMD when everyone is running a different system and I'm willing to bet a lot with completely shit airflow cases with solid front panels since trend whores prioritize rainbow lights and glass over actual function.

I highly doubt der8auer's audience consists of average users. I'd imagine the vast majority of it are people that are informed enough on how stuff works not to stuff their parts into cases that will cause it to overheat.

[7nationarmy]

1 hour ago, Derangel said:

I highly doubt der8auer's audience consists of average users. I'd imagine the vast majority of it are people that are informed enough on how stuff works not to stuff their parts into cases that will cause it to overheat.

But what about my SFF PC with the 3900X and Noctua L9?

[ravenshrike]

1 hour ago, Derangel said:

I highly doubt der8auer's audience consists of average users. I'd imagine the vast majority of it are people that are informed enough on how stuff works not to stuff their parts into cases that will cause it to overheat.

As the vid I posted from L1 shows, Ryzen is finicky as all hell when it comes to something as simple as thermal paste application. Also, the fact that the users aren't average doesn't mean they don't fuck up or have an inflated sense of their own ability when it comes to building PCs.

[RejZoR]

17 minutes ago, ravenshrike said:

As the vid I posted from L1 shows, Ryzen is finicky as all hell when it comes to something as simple as thermal paste application. Also, the fact that the users aren't average doesn't mean they don't fuck up or have an inflated sense of their own ability when it comes to building PCs.

How in hell do you screw thermal paste application? Oh right, doing stupid cross or pea size dot and have absolutely no clue if it has actually covered anything...

[Derangel]

3 minutes ago, RejZoR said:

How in hell do you screw thermal paste application? Oh right, doing stupid cross or pea size dot and have absolutely no clue if it has actually covered anything...

You're one of those people that bitch at Youtubers for applying thermal past "wrong" despite numerous tests proving there is no wrong way to do is as long as the CPU is covered, aren't you?

[RejZoR]

7 minutes ago, Derangel said:

You're one of those people that bitch at Youtubers for applying thermal past "wrong" despite numerous tests proving there is no wrong way to do is as long as the CPU is covered, aren't you?

Numerous tests prove you can see through a copper block to verify it (and if you say you can check if you remove it, well, how do you know reapplication will do the same for sure? Didn't know you're all Supermans who can do that. With pea sized poop or numerous variations of crosses, you're NEVER sure it's applied properly unless you apply it so much it goes out the sides. And even then there can be small portions that don't get covered. I'll still defend that only proper way is thin layer of paste across entire IHS. Fight me. And yes, the "air bubbles" myth is total BS. It just doesn't happen. This stands especially true for massive Threadripper IHS's.

[Derangel]

15 minutes ago, RejZoR said:

Numerous tests prove you can see through a copper block to verify it (and if you say you can check if you remove it, well, how do you know reapplication will do the same for sure? Didn't know you're all Supermans who can do that. With pea sized poop or numerous variations of crosses, you're NEVER sure it's applied properly unless you apply it so much it goes out the sides. And even then there can be small portions that don't get covered. I'll still defend that only proper way is thin layer of paste across entire IHS. Fight me. And yes, the "air bubbles" myth is total BS. It just doesn't happen. This stands especially true for massive Threadripper IHS's.

So you're just going to ignore that EVERY test on application method shows it does not matter at all just so you don't have to admit that you're wrong? If spreading a thin layer makes you feel secure that's fine, but it doesn't matter and you are willfully ignoring data to say otherwise. It shows you are only interested in what you believe is right, not what the facts show.

[leadeater]

1 hour ago, RejZoR said:

you're NEVER sure it's applied properly unless you apply it so much it goes out the sides.

So the correct way 

[Dabombinable]

On 9/2/2019 at 6:11 AM, Kisai said:

Internet speed "up to XMbps" is the worst misleading statement ever seen, and when seen in the context of "horsepower" eg CPU speed or Engine speed, it's very misleading because you can create a scenario where that number is true, but only if certain common situations are also true.

 

Like I'd ask how many people were using a stock cooler, and are in an air conditioned room. Because that very likely changes the maximum boost speed. If you are in a warmer climate that means you hit the the thermal throttle much sooner.

Considering just the ambient temperature of my room in winter makes my tablets core M run faster than an i3 7100U (in summer it's far slower).

[ThreePinkApples]

21 hours ago, porina said:

The two older generation mobos I own, X370 mobo went from 1.0.0.2 at Zen 2 launch to 1.0.0.3AB currently. Similarly B450 mobo went from 1.0.0.1 to 1.0.0.3ABB. There are multiple versions of 1.0.0.3 going around which might further confuse matters.

 

I can't say I looked at boost clocks too closely with either. I do recall initial testing with 3600 on 1.0.0.1 I never directly saw a single thread task go to rated boost, but if I left hwinfo64 logging, it did detect it at some point. It may be changing faster than the monitoring sample rate so get missed. Or maybe it is a measurement effect, that when doing a measurement you effectively have a 2nd active thread so you don't get single core boost. At least ram support got a LOT better from 1.0.0.1 to 1.0.0.3x.

The different 1.0.0.3x are just bug fixes on top of 1.0.0.3. 1.0.0.3A fixed some bugs, then 1.0.0.3AB fixed more bugs, and finally 1.0.0.3ABB fixed the random number instruction that caused newer Linux distros, and Destiny 2, to not work + WHEA errors/log spam. 
Basically the numbers go up when AMD has done changes in behaviour, or added new features, while the letters are for bugfixes

[Paddo]

On the topic of thermals: I did the test with my 3900x and a Noctua NH-D15. ~75°C max on the CPU, ambient around 28°C, no voltage drops, nothing. Still "only" 4.5GHz during testing. It did hit 4.6GHz once shortly after booting but never again. Even with an 360mm AiO I didn't get to 4.6GHz. I don't think that thermals are the problem for most people. Sure, if you use the stock cooler or a cheap aftermarket cooler your temps might be too high but I also think that most people participating in der8auers survey do have AiOs or good aftermarket coolers.
Maybe we'll get more infos on the topic when he releases a follow up video where we can get a deeper look into what board+AGESA combo did achieve the correct boostclocks.

[leadeater]

2 minutes ago, VegetableStu said:

official response (in form of an image because AIN'T NOBODY GOT TIME FOR 280 CHARACTERS)

Prepare for the 20 videos testing this . Better actually fix it because waaaayyy more people will be looking now lol

[leadeater]

1 minute ago, VegetableStu said:

(you know there's an angry thread over at GD... )

No? *foot steps*

 

Also I like how people are praising AMD on the twitter post, seems a bit odd you'd do that for having a bug, well fixing it but still amusing.

[leadeater]

5 minutes ago, VegetableStu said:

hey while we're all here (?), anyone knows if it's reasonable to expect all the BIOS fine-tuning work on Ryzen 3000 to reduce early-adopter issues on TR3K? o_o

 

Somehow I wouldn't be surprised if there were more. Different IMC, more chiplets, more power, more fun.