Monday, September 2nd 2019

Der8auer: Only Small Percentage of 3rd Gen Ryzen CPUs Hit Their Advertised Speeds

World famous overclocker Der8auer published his survey of boost clocks found on 3rd generation Ryzen CPUs. Collecting data from almost 3,000 entries from people around the world, he has found out that a majority of the 3000 series Ryzen CPUs are not hitting their advertised boost speeds. Perhaps one of the worst results from the entire survey are for the 12-core Ryzen 9 3900X, for which only 5.6% of entries reported have managed to reach the boost speeds AMD advertises. However, the situation is better for lower-end SKUs, with about half of the Ryzen 5 3600 results showing that their CPU is boosting correctly and within advertised numbers.

Der8auer carefully selected the results that went into the survey, where he discarded any numbers that used either specialized cooling like water chillers, Precision Boost Overdrive - PBO or the results which were submitted by "fanboys" who wanted to game the result. Testing was purely scientific using Cinebench R15 and clock speeds were recorded using HWinfo (which got recommendation from AMD), so he could get as precise data as possible.
Der8auer comments that he still recommends Ryzen 3000 series CPUs, as they present a good value and have good performance to back. He just finds it very odd that AMD didn't specify what you need to reach the advertised boost speeds.

If you would like to see the more in depth testing, here is the English version of the video:

Add your own comment

248 Comments on Der8auer: Only Small Percentage of 3rd Gen Ryzen CPUs Hit Their Advertised Speeds

#76
cadaveca
My name is Dave
EarthDog, post: 4108783, member: 79836"
Most of us will be waiting with bated breath. :)
It's not like that is unexpected... :p Today, my 1950X is fantastic. It took a while before it got there. A long while. I'm sure you saw that many boards early had the usual USB problems, even. Like it's all been way too familiar. So familiar that I kind of feel all warm and fuzzy inside, YKWIM? LOL.

TheLostSwede, post: 4108789, member: 3382"
But that's EXACTLY the point, it's NOT just down to cooling. My CPU couldn't go 1MHz past 4,400.25MHz at one point, along comes a new UEFI and now it boosts to 4,525.25MHz on multiple cores. So in other words, people not hitting advertised boost, is NOT related to cooling alone. Don't try to simplify this issue, as it's really a lot more complex than that.
Extremely more complex. This has been largely understated by the media for this launch, how complex this CPUs power handling really is.

That's AMD's fault for not emphasizing that aspect of these chips enough to the media.
Posted on Reply
#77
Totally
xkm1948, post: 4108649, member: 50521"
Man the amount of butt hurt AMD fans.
You're like the biggest one.

Irony much?
Posted on Reply
#78
Jism
TheLostSwede, post: 4108789, member: 3382"
But that's EXACTLY the point, it's NOT just down to cooling. My CPU couldn't go 1MHz past 4,400.25MHz at one point, along comes a new UEFI and now it boosts to 4,525.25MHz on multiple cores. So in other words, people not hitting advertised boost, is NOT related to cooling alone. Don't try to simplify this issue, as it's really a lot more complex than that.
CPU´s FIT dude. It s a small and hidden feature in Ryzen CPU´s that only The Stilth has shined some documentation on. With a new UEFI it proberly losened up some smalls things on where the FIT´s absolute values are a bit more widened. I wish AMD published a bit more on the workings of the FIT so the masses would finally come to an understanding on how the boost algorithm works.

Things that derbauer does not documentate about. And creates a hope FUD by spreading bullshit really.
Posted on Reply
#79
TheLostSwede
Darmok N Jalad, post: 4108756, member: 170588"
Just as an aside, I notice that with HWiNFO, there is a lag in readings when I apply a load. I’m not sure what the polling interval is, but it’s clear that it doesn’t capture the CPU state for every clock cycle. This is still why I would like to see what AMD says. I think there is silicon lottery, but there may also be very brief (millisecond) boost instances. I know mobile SOCs employ this strategy considerably, in order to stay within power and cooling budgets.
This is correct, BUT it's unlikely that you're going to miss the boost every single time if you run it for hours on end, especially as the boost can happen on any core. If you run it for a minute and call it a day, well, then you didn't really care.

This has been running for a few hours and at least three of my cores are 25MHz over advertised boosts, with an additional two hitting advertised boost clocks.
Every single core now boosts over 4,400MHz with some margin.

Posted on Reply
#80
Turmania
This is as bad as bulldozer core count fiasco.for some reason AMD is getting away with murder and people turn a blind eye when it comes to them.
Posted on Reply
#81
TheLostSwede
EarthDog, post: 4108771, member: 79836"
So, as I've said earlier... I've used at least 6 boards (more now I think, lol) with multiple bios and agesa's. Nada. They've been out for almost 3 months.. maybe this will all end with another update...sure. But it hasnt yet. ;)
I think you've been too busy testing stuff, it's five days to go until it's been two months since launch ;)
Posted on Reply
#82
cadaveca
My name is Dave
Turmania, post: 4108796, member: 182201"
This is as bad as bulldozer core count fiasco.for some reason AMD is getting away with murder and people turn a blind eye when it comes to them.
is that AMD's fault, or the media's?

I seen Microsoft, with that specific issue, treat AMD's Bulldozer CPU far differently than the rest of the world...

After being part of that media for pretty much a decade and now having left it, I can't really say that looking back that there was any real problems and more so misunderstandings. Its pretty common in this industry and some companies do take advantage of that fact more than others for sure, but they all do it.
Posted on Reply
#83
Jism
TheLostSwede, post: 4108795, member: 3382"
This is correct, BUT it's unlikely that you're going to miss the boost every single time if you run it for hours on end, especially as the boost can happen on any core. If you run it for a minute and call it a day, well, then you didn't really care.

This has been running for a few hours and at least three of my cores are 25MHz over advertised boosts, with an additional two hitting advertised boost clocks.
Every single core now boosts over 4,400MHz with some margin.
And beyond 60 degrees thermals, you should be feeling lucky then having a piece of silicon that is able to sustain those clocks (and beyond).

FIT ensures that the cpu simply does´nt boost beyond oblivion, i.e a unstable condition or pulling too much current through the silicon. If it would you would be degrading your CPU really fast. With a new AGESA update they proberly extended the XFR values which made your CPU boost a bit further compared to the previous AGESA. Congrats, AMD deliveres more then you wanted now.
Posted on Reply
#84
ZoneDymo
Turmania, post: 4108796, member: 182201"
This is as bad as bulldozer core count fiasco.for some reason AMD is getting away with murder and people turn a blind eye when it comes to them.
yeah, you can see from the amount of comments and the video Der8auer made they are just getting away with it and people turning a blind eye!
Dont you just get tired of yourself sometimes?
Posted on Reply
#85
TheLostSwede
cadaveca, post: 4108777, member: 25138"
I think AMD needs 4-6 months before they can suss out all the little niggles I see. I mean... do we even have odd CAS numbers working yet? ;)

So like, this is a real problem, there can be no doubt. I don't see it as that big of a problem (as is usual for me and this stuff), but it is something that consumers should be made aware of in a better way. I'm not sure that this is the correct approach, as we have to filter out the mail carrier's bias as always.

Lisa Su said that'd get better, and that's the one area that I think AMD could really improve upon. but there's too many large egos in the media that get in the way of that ever being an easy task. AMD jsut doesn't use what most consider as traditional avenues to reach their end users because of this, for example. How do you fix that? You just can't.
Apparently some people do. Posted this elsewhere earlier today.
https://lab501.ro/memorii-stocare/review-hyperx-fury-rgb-ddr4-3200-cl16-2x8gb-hynix-cjr-pe-platforma-amd/5
Not sure what kind of black magic Monstru is pulling, as those Kingston sticks can run CAS 17 and CAS 19 on his setup.

To be honest, I think most people would be ok with it, as long as it wasn't for the deafening silence from AMD on this issue. They simply say that they're aware of the issue and they're working on it. Not even a peep in terms of why it might be happening.

Well, in my case, it has gotten better, as all my parts are now running at better than spec, so although I was having a lot of issues to start with, it took about a month and a half or so to get everything working properly. Sadly, not everyone else has been as lucky. And yeah, I think AMD needs to post on their site about this, rather than hanging out on reddit...
Posted on Reply
#86
Xuper
I think It's better AMD adds Minimum boost clock, Problem solves.
Posted on Reply
#87
TheLostSwede
Vya Domus, post: 4108781, member: 169281"
And the default value on mine was a staggering 2000ms, good luck keeping a CPU under the right conditions for single core boost clock to work for two seconds.
So please explain my readings in the post above, HWinfo is sett to 2000ms in my case as well...
Posted on Reply
#88
Nkd
Turmania, post: 4108796, member: 182201"
This is as bad as bulldozer core count fiasco.for some reason AMD is getting away with murder and people turn a blind eye when it comes to them.
I am not going to lose my shit over 25mhz that can easily be optimized by bios update. In the meantime my cpu smokes everything out there in the 500 range. So no I am not losing my shit over it. I got more things to worry about in life.
Posted on Reply
#89
TheLostSwede
Jism, post: 4108793, member: 91255"
CPU´s FIT dude. It s a small and hidden feature in Ryzen CPU´s that only The Stilth has shined some documentation on. With a new UEFI it proberly losened up some smalls things on where the FIT´s absolute values are a bit more widened. I wish AMD published a bit more on the workings of the FIT so the masses would finally come to an understanding on how the boost algorithm works.

Things that derbauer does not documentate about. And creates a hope FUD by spreading bullshit really.
Link? As it's impossible to search any of the things you reference.

Turmania, post: 4108796, member: 182201"
This is as bad as bulldozer core count fiasco.for some reason AMD is getting away with murder and people turn a blind eye when it comes to them.
Far from it, as this is most likely going to be something that can be fixed with a simple UEFI update, as has already happened in my case.
Don't try to make this out to be something it isn't.
Posted on Reply
#90
Nkd
Xuper, post: 4108804, member: 83814"
I think It's better AMD adds Minimum boost clock, Problem solves.
That is not how AMD boost works. Their boost algorithm has always worked with cooling and temp in mind. Next thing you someone will be bitching using a low profile cooler in a sff case.
Posted on Reply
#92
Jism
TheLostSwede, post: 4108811, member: 3382"
Link? As it's impossible to search any of the things you reference.
The stilth has done some threads here and there about going in depth on the Ryzen 1 and 2 series. In the 2 series he is mentioning the FIT which is responsible for the silicon health in general. Now this FIT simply controls the maximum current a core or cores can use, without frying the chip from inside out while being in boost state. It also controls the voltage thats being supplied to a single core workload or a all core workload (hence why the variance in the two when going full load). It also prevent the CPU from consuming too much current which leads to degeneration of the CPU. At last it dictates the current boost state in combination with thermals and VRM threshold.

So if people are not having their advertised maximum boost states, even when tweaking such as undervolt or cooling, then its pretty much the FIT preventing the CPU core or cores to run into oblivion by going unstable or so. When i initially bought my Ryzen 2700x, i had to take a few months to learn about the tech behind it. Its not so simply as putting up a multiplier and call it a day. Or its not so simple to rev up the base clock while running into lowered latency dividers once you go beyond 104Mhz or so.

This is a new platform and this platform comes with alot of nice features, but also difficult features to understand. If you ask me what would be better to play with, XFR or manual OC? Then i would answer that investing in good cooling (i.e 240mm / 360mm rad) and some tweaking such as undervolting, and call it a day. Overclocking these days is´nt so sophisticated as it was before. Slap on a big cooler and let technology such as XFR do its thing.

If you wonder who the stilth is, he´s mentioned in every populair Asus board under tweaking settings. Its someone who knows more then Derbauer does related to AMD tech.
Posted on Reply
#93
TheLostSwede
Nkd, post: 4108815, member: 42675"
That is not how AMD boost works. Their boost algorithm has always worked with cooling and temp in mind. Next thing you someone will be bitching using a low profile cooler in a sff case.
I think you misunderstood what he meant. The idea is what AMD should've put something like the boost speed that they knew all their CPUs in a certain SKU could reach, not just some of them.

Jism, post: 4108817, member: 91255"
The stilth has done some threads here and there about going in depth on the Ryzen 1 and 2 series. In the 2 series he is mentioning the FIT which is responsible for the silicon health in general. Now this FIT simply controls the maximum current a core or cores can use, without frying the chip from inside out while being in boost state. It also controls the voltage thats being supplied to a single core workload or a all core workload (hence why the variance in the two when going full load). It also prevent the CPU from consuming too much current which leads to degeneration of the CPU. At last it dictates the current boost state in combination with thermals and VRM threshold.

So if people are not having their advertised maximum boost states, even when tweaking such as undervolt or cooling, then its pretty much the FIT preventing the CPU core or cores to run into oblivion by going unstable or so.

If you wonder who the stilth is, he´s mentioned in every populair Asus board under tweaking settings. Its someone who knows more then Derbauer does related to AMD tech.
So please provide a link or two then...
I have a 280mm liquid cooler. I have not touched any of the overclocking features and never asked about this.

Ferrum Master, post: 4108816, member: 90058"
Splitting hair over 100MHz.... WTH people?
It's more about how AMD marketed things than anything else. It's actually a bit like the storage companies counting in even 1,000's per byte, but operating systems doing 1,024 to a byte. It starts small and soon everyone is trying to get away with it. At least storage devices have disclaimers these days, AMD didn't have one and still doesn't. On top of that, they went and posted a video, claiming that under the right circumstances, with the right cooling and motherboard VRM, you can expect your shiny new (yet to be launched) Ryzen 3000 CPU to boost even higher, which so far it seems almost no-one has managed.
Posted on Reply
#94
Vya Domus
TheLostSwede, post: 4108806, member: 3382"
So please explain my readings in the post above, HWinfo is sett to 2000ms in my case as well...
I already said how HWinfo wouldn't report that my 1700X is running at it's boost clocks even though the benchmarks confirmed it was in fact running at those speeds. So I have experienced readings that didn't reflect reality. It would be perhaps more relevant for everyone to just run some single thread CPU tests, see the results and watch if they always match with what the readings say.

What's more important, performance or a pretty number ?
Posted on Reply
#95
TheLostSwede
Vya Domus, post: 4108826, member: 169281"
I already said how HWinfo wouldn't report that my 1700X is running at it's boost clocks even though the benchmarks confirmed it was in fact running at those speeds. So I have experienced readings that didn't reflect reality. It would be perhaps more relevant for everyone to just run some single thread CPU tests, see the results and watch if they always match what the readings say.
Wow, you've clearly followed all the other threads in the forums on this topic very closely... :banghead:

Just please don't comment any more on this topic, as you clearly have no hands on experience or understanding of the issues people are facing with the Ryzen 3000.
Posted on Reply
#96
Jism
TheLostSwede, post: 4108818, member: 3382"
So please provide a link or two then...
I have a 280mm liquid cooler. I have not touched any of the overclocking features and never asked about this.
I posted in this thread before, the behaviour of my boosting 2700x, that once it passes the 60 degrees mark, i increased my cooling capacity, slightly undervolted a bit, was enough to have it running at 4.2GHz all core.
Posted on Reply
#97
Vya Domus
TheLostSwede, post: 4108831, member: 3382"
Wow, you've clearly followed all the other threads in the forums on this topic very closely... :banghead:

Just please don't comment any more on this topic, as you clearly have no hands on experience or understanding of the issues people are facing with the Ryzen 3000.
I follow what's on this thread and I'll comment whenever I like, thanks.

Software readings are unreliable, in addition to that making random polls is as useful as asking people what's the size of their johnson.
Posted on Reply
#98
TheLostSwede
Vya Domus, post: 4108837, member: 169281"
I follow what's on this thread and I'll comment whenever I like, thanks.

Software readings are unreliable, in addition to that making random polls are as useful as asking people what's the size of their johnson.
Is making ill informed comments that adds nothing to a topic something you enjoy doing? Then please be my guest. Let me just add you to my ignore list so I don't have to waste my time reading them.

Jism, post: 4108834, member: 91255"
I posted in this thread before, the behaviour of my boosting 2700x, that once it passes the 60 degrees mark, i increased my cooling capacity, slightly undervolted a bit, was enough to have it running at 4.2GHz all core.
That's not a Ryzen 3000-series CPU though, right?

And I would still like a link to this AMD CPU magician of yours.
Posted on Reply
#99
vega22
cadaveca, post: 4108798, member: 25138"
is that AMD's fault, or the media's?

I seen Microsoft, with that specific issue, treat AMD's Bulldozer CPU far differently than the rest of the world...

After being part of that media for pretty much a decade and now having left it, I can't really say that looking back that there was any real problems and more so misunderstandings. Its pretty common in this industry and some companies do take advantage of that fact more than others for sure, but they all do it.
welcome back to the real world dave, can't wait to read your insight onto things in the future now you're out from under that weight :)

on topic

title is misleading when the % is really down to the chip/mobo combo. really seems like some mobo have polished bois while others are still quite immature.


it will be xmas before amd and the mobo makers finally get their shit together. this is like another totally new platform for them with these new features.
Posted on Reply
#100
EarthDog
Lol, faking shyte.

No links, referencing zen+ not zen2, a dude saying to run single threaded tests (we did already in the other monster thread), and more users adding 2 cents that isnt worth a penny.

Man do I hate forums, lol!
Posted on Reply
Add your own comment