Ryzen 3000 boost clock controversy - der8auer publishes his survey results, not a good look for AMD

But much like you can't get a Bugatti Veryron to it's top speed on any old road, you can't expect a variety of motherboards and cooing solutions used for these survey results to give you a clear idea of what is actually limiting the boost clock, here.

Well, if you have the same fuel, same tires, same track, same conditions, you should hit that speed, right? Or at least within a really close margin. Same thing with the MPG on the sticker - you recognize there are situations where you aren't going to hit it, but if you are running the same conditions, you should expect to get the same results.

Sounds like that isn't the case here - people are using same motherboards, same BIOS, same cooling, and aren't getting the same results.

I think what we are seeing is just the auto-overclock (PBO) is really good at getting to the limits of the silicon, so you are getting the most out of your chip. It's just what is printed on the box is what most people expect to get, and AMD printed a rosy estimate on the box, not a minimum baseline case.

If AMD had printed 4.4Ghz Boost, and some people were seeing 4.5's occasionally, everyone would be overjoyed and estatic, and those that "only" get 4.4 wouldn't feel like they are getting cheated. Instead, they printed 4.6, because they got one sample in a lab to hit it and they were afraid of Intel's 5.0, and now everyone feels like they are getting gypped. Even though the product is no different, and the results are not different. It's a marketing plunder.
 
I think the 20+ hardware exploits for Intel CPUs from 1995-present are a bit more scandalous, and serious, than the lack of boost clocks.
Agreed. But Intel's old flaws don't excuse AMD's new actions. Whataboutism falls flat here.

Even without it, these are still damn good CPUs, especially for the cost.
Agreed. But that's not what's on trial here. Why lie when they had a solid product and didn't need to, just for a few extra percentage points?
 
Agreed. But that's not what's on trial here. Why lie when they had a solid product and didn't need to, just for a few extra percentage points?

At this point "what's on trial" here is the interpretation of the "max boost clock" definition. We're running circle trying to define that so let's wait for AMD's explanation if any or the typical lawsuits.
Your internet provider will say "up to 400Mb/s, will you sue if you get 380 ? What's the difference with "Max speed of 400Mb/s" ?
 
People are also not realizing that most of the ryzen motherboards, all though set at a bus speed of 100, are actually running at a bus speed of 99. Which means a 3700x for example will boost to 4356, or there about when hitting the boost multiplier of 44.

4.375GHz is actually pretty much hitting the advertised boost clocks. The reason you aren't seeing 4.4GHz is because the base clock is generally below 100MHz. If you add a slight increase to your base clock you should see 4.4GHz. (Theoretically, as I haven't tested this exactly.)

AMD-Ryzen-3900X-Review-Benchmark-1.jpg

9051_35_amd-ryzen-3900x-3700x-zen2-review_full.png

3-1080.82cbf7dd.png

f951f98b-eb34-446e-a6ab-7cde828c546d.png
big_3900x-cpu-z-1.jpg.jpg

3900X-stock-clocks.png

4350-benches.jpg

index.png
 
At this point "what's on trial" here is the interpretation of the "max boost clock" definition. We're running circle trying to define that so let's wait for AMD's explanation if any or the typical lawsuits.
Your internet provider will say "up to 400Mb/s, will you sue if you get 380 ? What's the difference with "Max speed of 400Mb/s" ?

This is true. Part of that will come down to precedent.

ISPs have never been able to deliver on Max Speed consistently, and most people understand it's a big complicated interconnected network with a lot of external influences, but you should get close on occasion.

CPU's, on the other hand, historically have been able to hit the number on the box. Consistently. Without fail. And often, by virtue of overclocking, go above and beyond what's printed on the box.

Your right in that it's no different really, and AMD hasn't really printed something that is false (the "up to" disclaimer), but precedent hasn't carried that way.

Legal decisions often hinge on precedent.
 
  • Like
Reactions: N4CR
like this
This is true. Part of that will come down to precedent.

ISPs have never been able to deliver on Max Speed consistently, and most people understand it's a big complicated interconnected network with a lot of external influences, but you should get close on occasion.

CPU's, on the other hand, historically have been able to hit the number on the box. Consistently. Without fail. And often, by virtue of overclocking, go above and beyond what's printed on the box.

Your right in that it's no different really, and AMD hasn't really printed something that is false (the "up to" disclaimer), but precedent hasn't carried that way.

Legal decisions often hinge on precedent.

I wonder if the current settlement by AMD about 8 cores can be counted as precedent in this case ? I didn't read about it but does accepting this settlement means they accept that they were wrong and deceived or only "end" the fiasco ?
The only number on the box that should be taken seriously is the base clock here (my point of view). Everyone that bought the product should of read the reviews and by then knew what they were getting.
 
I wonder if the current settlement by AMD about 8 cores can be counted as precedent in this case ? I didn't read about it but does accepting this settlement means they accept that they were wrong and deceived or only "end" the fiasco ?
The only number on the box that should be taken seriously is the base clock here (my point of view). Everyone that bought the product should of read the reviews and by then knew what they were getting.

You mean the reviews what were written before AMD lowered boost clocks for people with an AGESA update?
 

What are you trying to show us? What I see is you hit a single core boost of 4.525 Ghz which is only 75 below it's spec, which most likely is due to you running your memory at 4600 and not spec 3200 Mgz. Also your bus speed is not consistent and fluctuates when you are not manually overclocking. So.. what's your point?
 
You mean the reviews what were written before AMD lowered boost clocks for people with an AGESA update?

I may have missed that, could you please link something regarding AMD lowering max boost clock specification (I still think that's a characteristic) or changing the name after the release date ?
Read again your post, you're saying an AGESA update lowered the reported boost clock (Or at that point let's call it an overclock) after an update ? Updates are normally to fix issues and/or make a product better so maybe it was boosting too much OR the overall performance may be better even if the max boost frequency is lower ? Was this tested ? This may also be a new problem which will get fixed later, not sure what your point is here ? From the get go I painted myself as someone who thinks max boost clock is not a guaranteed spec but rater a characteristic that some chips will hit.

Why is this an issue now when the 2000 series was in the same boat ? How many 2700X hits 4.3GHz constantly ?
 
  • Like
Reactions: N4CR
like this
I think this is some of the most egregious car analogies I have seen in quite a while. The more obvious comparison to make would be a manufacturer hitting horsepower/torque ratings.

More easily measurable, repeatable than top speed and both ratings are achieved when peak performance is necessary, e.g you stomp on the accelerator.

Hell typically people even equate horsepower to their rigs as horsepower is a part of the equation to top speed, same as the CPU is one part to maximum FPS in a game.

As for the actual thread, this discussion reminds me of when AMD released the 290/290x and their poorly advertised base clocks and highly advertised boost clocks. Cards would throttle and not hit the rated boost. While Nvidia cards would almost always push well past their rated boost.

They make their advertisement without having everything bulletproof for the ensuing scrutiny. It’s the MO of an AMD release.
 
I think this is some of the most egregious car analogies I have seen in quite a while. The more obvious comparison to make would be a manufacturer hitting horsepower/torque ratings.

More easily measurable, repeatable than top speed and both ratings are achieved when peak performance is necessary, e.g you stomp on the accelerator.

Hell typically people even equate horsepower to their rigs as horsepower is a part of the equation to top speed, same as the CPU is one part to maximum FPS in a game.

As for the actual thread, this discussion reminds me of when AMD released the 290/290x and their poorly advertised base clocks and highly advertised boost clocks. Cards would throttle and not hit the rated boost. While Nvidia cards would almost always push well past their rated boost.

They make their advertisement without having everything bulletproof for the ensuing scrutiny. It’s the MO of an AMD release.

The 290/290x suffered from AMD's usual crap stock GPU coolers. Waterblock or AIB cards with custom coolers hit the max clock all day long. Bit different here where it seems like even with good cooling you're not guaranteed to see the max boost for more than a brief moment or two.
 
The 290/290x suffered from AMD's usual crap stock GPU coolers. Waterblock or AIB cards with custom coolers hit the max clock all day long. Bit different here where it seems like even with good cooling you're not guaranteed to see the max boost for more than a brief moment or two.
Yeah I do understand that the situations aren’t 100% comparable. I think it does show how I don’t think AMD properly plans these launches. They thought they could get away with not hitting the frequencies and that people wouldn’t put them under the microscope. Similar to the 290x coolers keeping them from hitting rates boosts.
 
Yeah I do understand that the situations aren’t 100% comparable. I think it does show how I don’t think AMD properly plans these launches. They thought they could get away with not hitting the frequencies and that people wouldn’t put them under the microscope. Similar to the 290x coolers keeping them from hitting rates boosts.

I think they choose those numbers because average joe shopping for a new computer will look at this and directly compare it to Intel offerings. There's a difference between educated customer and average joe... there's also a big difference in the market share between those two.
To the educated customer even if the max boost clock was 200MHz less they would rely on reviews and not what's written on the box (mind you the interpretation of it). My guess is they wanted to ensure sales from that fantasy number more than anything else. The CPU delivers the performance/$ hands on... average joe won't complain much.

This may impact company image in the long run but as long as they deliver performance... they may get away with it (Intel and NVidia comes to mind ;) )
 
and now everyone feels like they are getting gypped.

speak for yourself home slice. I am more than happy with my 4500-4550mhz

To me to get twin chiplets that are 7nm and extremely dense in transistor packed volume to\ run at these speeds is quite an achievement.

Ya'll want to sue AMD, I say give them a fucking award!
 
The performance is there.

Maximum does not mean Minimum.

Do the world a favor and sell your hardware and go back to the intel forum if 25MHz means that much to you. In reality, it means nothing so stop falling for clickbait that gets you all riled up.

I’m tired of reading all of this bullshit. Fkn crybabies.
 
It looks like the majority can hit 4.5 ghz, which is indistinguishable from 4.6 ghz in everyday use; 4.6/4.5 = TWO PERCENT. That's like complaining about 62 vs 60 fps.

Once again, this ia a big Nothingburger. The performance hit from Bulldozer's "cores" was more significant (20% below that of traditional CPUs in multi core integer workloads, and up to 50% lower in FPU-heavy code), but you'll still hear people here whining about it being a pointless lawsuit.

The reviews show the 3900x to still beat the 9900k in almost every benchmark, so it's not hurting anything noticeable. For perfectionists here, you will see the process improve and clocks get up to 4.6ghz on all parts in a few more months.


Hold up there. I picked up a 3900x and board week and have been playing with it, and no, it's not faster in gaming, I know that for a fact. I have the 9900K with 3733 memory, 2080 ti and 3200mhz, the 3900x and my 2nd 2080 Ti that I've been trying to sell, and playing with both systems. The Intel is just snappier on the desktop, gets my windows tasks done quicker and in gaming, there is no comparison. The Intel is just faster.

My 9900K is @ 5ghz across all cores. My boost speed on my 3900x is not at advertised speeds but now we know, so are most others.

At least this is my experience.

I'm not a fanboy of Intel or AMD or nVidia or AMD GPU ... I just want bleeding edge speed. That's what I go after. I'm a fan of performance.

I know a lot of you guys just want more cores, that's how you equate value and performance but, more cores does not mean "cpu, windows OS, chipset maturity" or "the best gaming performance"

I'm still iffy on this whole AMD "chiplet" design stuff. If Intel ever copied AMD with a chiplet design, that would be freaking incredible.

I can't wait for Intel to come out with their 10nm stuff. Intel uses smaller gates so 10nm Intel performance will match 7nm and from what I read, AMD's 7nm is not exactly ... well, I don't want to cause any issues.

I have both but I still go to my Intel system for gaming and general usage.
 
The performance is there.

Maximum does not mean Minimum.

Do the world a favor and sell your hardware and go back to the intel forum if 25MHz means that much to you. In reality, it means nothing so stop falling for clickbait that gets you all riled up.

I’m tired of reading all of this bullshit. Fkn crybabies.

you're so vain, you probably think this song is about you
 
  • Like
Reactions: Meeho
like this
This is on my 3600X. Max boost of 4.4 GHZ, and I'm getting 4.325, a difference of 75 MHz.

And this is with an old B350 motherboard I bought for Ryzen 1. Maybe with a newer system I would get the last 75 MHz.

Not a big deal.

Ryzen44.PNG


Also, there are many companies that advertise theoretical maximums that are not guaranteed.

Your ISP says you will get up to 400mbps but it happened once. Or "save up to 15% on your car insurance" is not a guarantee of any savings at all.
 
Last edited:
Hold up there. I picked up a 3900x and board week and have been playing with it, and no, it's not faster in gaming, I know that for a fact. I have the 9900K with 3733 memory, 2080 ti and 3200mhz, the 3900x and my 2nd 2080 Ti that I've been trying to sell, and playing with both systems. The Intel is just snappier on the desktop, gets my windows tasks done quicker and in gaming, there is no comparison. The Intel is just faster.

My 9900K is @ 5ghz across all cores. My boost speed on my 3900x is not at advertised speeds but now we know, so are most others.

At least this is my experience.

I'm not a fanboy of Intel or AMD or nVidia or AMD GPU ... I just want bleeding edge speed. That's what I go after. I'm a fan of performance.

I know a lot of you guys just want more cores, that's how you equate value and performance but, more cores does not mean "cpu, windows OS, chipset maturity" or "the best gaming performance"

I'm still iffy on this whole AMD "chiplet" design stuff. If Intel ever copied AMD with a chiplet design, that would be freaking incredible.

I can't wait for Intel to come out with their 10nm stuff. Intel uses smaller gates so 10nm Intel performance will match 7nm and from what I read, AMD's 7nm is not exactly ... well, I don't want to cause any issues.

I have both but I still go to my Intel system for gaming and general usage.

At what resolution do you game and care to provide numbers ?
 
Hold up there. I picked up a 3900x and board week and have been playing with it, and no, it's not faster in gaming, I know that for a fact. I have the 9900K with 3733 memory, 2080 ti and 3200mhz, the 3900x and my 2nd 2080 Ti that I've been trying to sell, and playing with both systems. The Intel is just snappier on the desktop, gets my windows tasks done quicker and in gaming, there is no comparison. The Intel is just faster.

My 9900K is @ 5ghz across all cores. My boost speed on my 3900x is not at advertised speeds but now we know, so are most others.

At least this is my experience.

I'm not a fanboy of Intel or AMD or nVidia or AMD GPU ... I just want bleeding edge speed. That's what I go after. I'm a fan of performance.

I know a lot of you guys just want more cores, that's how you equate value and performance but, more cores does not mean "cpu, windows OS, chipset maturity" or "the best gaming performance"

I'm still iffy on this whole AMD "chiplet" design stuff. If Intel ever copied AMD with a chiplet design, that would be freaking incredible.

I can't wait for Intel to come out with their 10nm stuff. Intel uses smaller gates so 10nm Intel performance will match 7nm and from what I read, AMD's 7nm is not exactly ... well, I don't want to cause any issues.

I have both but I still go to my Intel system for gaming and general usage.
Faster by 1%? 2%? You did not mention any numbers for any programs or games.
 
At what resolution do you game and care to provide numbers ?

Every benchmark I've seen shows 9900k as the fastest cpu for gaming. It's the fastest in all but a few games where it gets just barely beat by AMD while there are a few games where the 9900k absolutely thrashes AMD such as StarCraft 2.

Now for the most part it doesn't matter because you'll be GPU bottlenecked, but 9900k is the undisputed best gaming CPU.
 
By mentioning a maximum it sets a limit. If a 3700X can boost to a max of 4.4 Ghz then it should be able to be replicated. There has to be a condition that will push every 3700X to 4.4 Ghz. So far that is not the case. Not sure why people are finding it hard to understand this.

If I buy a car that can do 250 MPH max then it should be able to do 250 MPH. Yes, under certain conditions. But it should be able to be replicated under those conditions for every car produced.
 
What are you trying to show us? What I see is you hit a single core boost of 4.525 Ghz which is only 75 below it's spec, which most likely is due to you running your memory at 4600 and not spec 3200 Mgz. Also your bus speed is not consistent and fluctuates when you are not manually overclocking. So.. what's your point?
My point is the bus speed is 99.8 MHz at the lowest so it is not the reason for not reaching the advertised boost clocks.
 
I see both sides of this. 1

MHz isn't that much, considering it's a deviation between 4.5 to 4.6 GHz.

On the other hand, they were specifically spec'd to achieve boost clocks that they are not, so that puts it into potential judicial action territory.

nope because the fine print protects them.. you think they didn't double check this with the lawyers? it's the same thing as isp's saying "up to x speed".. even if you never hit that speed there's nothing you can do about it because it wasn't guaranteed..


also i hate you cybereality, you beat me to the ISP comparison.. GRR!! :p
 
Last edited:
My point is the bus speed is 99.8 MHz at the lowest so it is not the reason for not reaching the advertised boost clocks.

I doubt 99.8 is the lowest. In your screen shots yes.. but it bounces all over, which is normal.. All though I have a different mother board (Crosshair VII hero) in a 2 minute time frame, I have watched it go from 97.9 to 99.9 on my board using cpuz... This is the case for Most AMD ryzen compatible motherboards, including x570 series. I suspect your's drops lower, you are just not catching it. However, possibly you have a motherboard that doesn't fluctuate as much. Have you tested it using 3200 Mhz memory speed?
 
I doubt 99.8 is the lowest. In your screen shots yes.. but it bounces all over, which is normal.. All though I have a different mother board (Crosshair VII hero) in a 2 minute time frame, I have watched it go from 97.9 to 99.9 on my board using cpuz... This is the case for Most AMD ryzen compatible motherboards, including x570 series. I suspect your's drops lower, you are just not catching it. However, possibly you have a motherboard that doesn't fluctuate as much. Have you tested it using 3200 Mhz memory speed?
Those are screenshots from various reviews. I don't have a Ryzen to test it myself. The reported clocks are consistent with the CPU's 25 MHz increments, so while a valid theory, I don't believe the explanation is as simple as that. Even if it was the cause, it would mean 95% of AMD motherboards are either not in spec, or they are and the CPU never had a chance to reach advertised clocks. No, the issue is elsewhere. Possibly in the poor chiplet silicon quality and AMD's bad assessment of what they would be able to produce in quantity.
 
If the latest amd optimizations increased effective performance on average by over 3% vs initial reviews then they effectively could argue that they gave you the performance equivalent to the rated boost even in cases of hitting 4.275 out of a max 4.400 boost.

Just saying, I think that this will be extremely hard to argue.
 
If the latest amd optimizations increased effective performance on average by over 3% vs initial reviews then they effectively could argue that they gave you the performance equivalent to the rated boost even in cases of hitting 4.275 out of a max 4.400 boost.

Just saying, I think that this will be extremely hard to argue.
They aren't advertising performance on the box, though.
 
False advertising is false advertising. Just because you get triggered by people calling it out doesn't make it right.
MAX BOOST != base. It's not false advertising. Like saying a Camaro LS3 has 425hp. Sure, usually it does, sometimes it doesn't. It doesn't in Denver, I'd be quite sure of that. Or my ISP saying I have 200mbit, good luck with that.

AMD always said boost was dependent on many different factors. If AMD said 'rated boost' or something implying regular operation there then it would have merit.
Just because Intel can usually hit it (not in thermally constrained operation.. apple memebook cooling anyone?) and Nvidia often hits boost clocks, doesn't make them guaranteed clocks.

2700x was the same boat, yet heard nothing of it. Yet now AMD is on par or ahead in most workloads, it's a massive deal..

So yes Dan_D there is precedent for this; Zen+ not hitting 4.3, Nvidia/AMD GPU boost not always reaching max etc and the wording 'max boost' leaves them plenty of 'out' with this in mind.
I'd be quite confident this will go nowhere.

The other issue is this is early days, people are still figuring out max ram vs boost vs agesa vs sillicon quality vs cooling vs etc. So right now there are a lot of fingers pointing at AMD but I feel some should also be pointed at other parties, AMD is not the only variable in this mess as clearly demonstrated by trials of same CPU different motherboard.

and edit: don't think I'm making excuses for AMD. I am looking at it from a biz/legal POV. They said and did the right thing in my books. I will say though it's a bit low to not say 'maximum possible boost' but it fits better on the box... and yes as a businessman it's easier for me to read 'marketing speak', for normal people it's not so easy. So I can understand there is some unhappiness there and it sucks. I do think AMD could have been a little more conservative, even 50-100MHz less. Aka Over deliver and under promise works every time! Always a happy customer.
AMD could learn from that - systems I design and build or sell are usually 10-15% under spec for this reason.
 
Last edited:
MAX BOOST != base. It's not false advertising. Like saying a Camaro LS3 has 425hp. Sure, usually it does, sometimes it doesn't. It doesn't in Denver, I'd be quite sure of that. Or my ISP saying I have 200mbit, good luck with that.

AMD always said boost was dependent on many different factors. If AMD said 'rated boost' or something implying regular operation there then it would have merit.
Just because Intel can usually hit it (not in thermally constrained operation.. apple memebook cooling anyone?) and Nvidia often hits boost clocks, doesn't make them guaranteed clocks.

2700x was the same boat, yet heard nothing of it. Yet now AMD is on par or ahead in most workloads, it's a massive deal..
It's dishonest and, while not a massive deal, it is A deal. If for nothing then for the fact that Intel and Nvidia guarantee the stated boost clock (outside specific restrictions), but with AMD there is no guaranteed performance.You could get a proper advertised performance, but most of the time you'll fall short of it and your neighbour will have some random +/- performance compared to you.

It's AMD's surprise mechanic. You paid extra for that 3600X? Tough luck, you may end up with a lower boosting CPU than a 3600. Surprise!!!
 
It's dishonest and, while not a massive deal, it is A deal. If for nothing then for the fact that Intel and Nvidia guarantee the stated boost clock (outside specific restrictions), but with AMD there is no guaranteed performance.You could get a proper advertised performance, but most of the time you'll fall short of it and your neighbour will have some random +/- performance compared to you.

It's AMD's surprise mechanic. You paid extra for that 3600X? Tough luck, you may end up with a lower boosting CPU than a 3600. Surprise!!!

I agree it's definitely an issue they need to fix in future releases (3950) if not with AGESA if possible.
Reality of the issue is down to semantics. As you said the others do whilst AMD doesn't guarantee, they just say (as you said) depending on certain factors/restrictions it will reach a max boost of X.X. Others also do this - Nvidia doesn't always reach boost clocks - neither does Intel in AVX or thermal situations! At this point there are platform OEM factors also as clearly demonstrated by motherboard variations. For now there may still be some alleviation of this to within a very small margin with updates.
So for now i'd put the pitchforks down but definitely press AMD and OEMs for more updates to fix the issue on affected boards and also to make clear statements regarding it.

After semantics it's transparency and AMD has dropped the ball a little here. 2700X didn't reach boost usually either but in this case it's multiple SKUs and motherboards having issues.
 
I still feel like this is blown way out of proportion... I have a 3700x and it's rated at 4.4 boost.... Highest I see is 4375. What a travesty.

Only people who have room to complain are 3900x owners imo... And still 4.5 ghz is not that far off from 4.6 and we still don't know if it's the boards fault or some sort of issue with the reporting software.

Also I've found my cpu does better with Pbo off. How many people are running pbo vs not?
 
I'm one to buy the product, not the story. Marketing says X but I don't really care about the label. I care about the performance and value. What actual performance will I get in the applications I will use, and how much do I have to pay for that performance.

It's a moot point for me anyway, as the soonest I'll be able to afford anything new is next year's tax returns, if even then.
 
At the end of the day are they really missing out on performance? Do the people crying even own these $500 chips? My guess is no on both parts. This is off topic but I hate the culture of people complaining about shit on behalf of others.
 
Back
Top