AMD Updates Ryzen Product Pages to Elaborate on "Max Boost Clocks"

[einard]

the graph for 3700X doesnt seems right, in CB R20 all core I can see 3,92Ghz max.
the graph shows 16 threads 4,2Ghz-4,25ghz.
if this would be right you would see much more higher scores at auto clock.
when I manually OC to 4,2Ghz all core the scores are higher then auto.

I also find it strange. My 3700x hits 3.9 all core stock, 4.275 sc boost. When I got my noctua cooler mount, I hoped that the thermal headroom would help, but it didn't.
PBO enabled makes it go to 4.1 all core. SC seems to take a hit then though. Adding +100/200 auto-oc gives me nothing. In any case PBO or OC doesn't even get me close to techpowerups review unit.

 

[Domokun]

I would go as far as to call that false advertising...

Also, what's nominal conditions?

That's exactly what it is.

I don't know what AMD's definition of 'nominal' is, but it's clearly not a 240mm AIO water cooler within the winter season, because I can't achieve the rated boost frequencies on my Ryzen 9 3900X.

 

[hzy4]

I also find it strange. My 3700x hits 3.9 all core stock, 4.275 sc boost. When I got my noctua cooler mount, I hoped that the thermal headroom would help, but it didn't.
PBO enabled makes it go to 4.1 all core. SC seems to take a hit then though. Adding +100/200 auto-oc gives me nothing. In any case PBO or OC doesn't even get me close to techpowerups review unit.

What is your MB? Yeah PBO an autoOC does nothing, I am patiently waiting for the next BIOS revision, hopping for higher boosts.

 

[TheLostSwede]

Meh, I beg to differ. Its really just a problem with the definition of 'turbo' or 'boost'.

In fact, I think we can safely say only Nvidia has its GPU boost story in good order. As in: you always get an advertised clock, and most of the time, you get much higher clocks. Even if you nearly cook the GPU it will still run base clock or you'll be using it completely out of spec.

Perhaps what we need is an industry standard wrt boost clocks. I could imagine you'd set that at something like a total achieved frequency across all available threads divided by number of threads. And when you do it like that, suddenly AMD doesn't look all that bad. Another approach could be the total deviation from advertised clocks - again, if you'd put Intel versus AMD spec sheet and reality; AMD will be coming out much better.

Let's not go blind on that peak clock number and define something around thát. I really don't like Intel's Turbo and the spec sheet trickery they've deployed over the years (and how recent parts handily turn the 'headroom' into 'used room' and how CPUs royally boost beyond TDP on stock bioses). They're playing the game for marketing, AMD is just bad at it (once again... it never ends does it) and you're right they've left a gap here for a lawsuit. But AMD does deploy a much better type of boost.

I mean for this example here, what should AMD advertise, 4375 mhz on the box instead of 4.4? 25 mhz...

As for the 'need' to see what conditions are required to get the advertised boost clocks...maybe. Maybe not... isn't the actual defining factor for a CPU in the end performance? You cannot grasp CPU performance based on clocks alone, nor could you do it based on the advertised turbo/boost clock of a spec sheet, after all there are barely any use cases that only hit one core.

People buy hardware based on the manufacturers spec.
How about DDR4 4400MHz memory "only" managed DDR4 4375MHz, would people be happy? I think not.

The issue is as you say with the marketing, but the users have been "sold" a number and as such are expecting said number, not 200-300MHz less than said number.
How do you suggest that situation should be resolved, as it seems AMD is now trying to come up with some rather lame excuses as to why we're not quite getting what we paid for.

4.2-4.3 is not far off 4.5. You're only supposed to get 4.5 if you have load on a single core, while the others are literally shut off.

Am I? Where does it say so? I haven't seen anything that specifies this. And the 3700X gets the same clocks by the looks of it, so what did I pay extra for that I got with my 3800X?

I also find it strange. My 3700x hits 3.9 all core stock, 4.275 sc boost. When I got my noctua cooler mount, I hoped that the thermal headroom would help, but it didn't.
PBO enabled makes it go to 4.1 all core. SC seems to take a hit then though. Adding +100/200 auto-oc gives me nothing. In any case PBO or OC doesn't even get me close to techpowerups review unit.

And this is why we're here discussting this matter. A lot of people are not seeing the expected speeds, while others are getting better than the expected speeds form their CPUs.
There's been zero clarity from AMD as to why this is happening.

That's exactly what it is.

I don't know what AMD's definition of 'nominal' is, but it's clearly not a 240mm AIO water cooler within the winter season, because I can't achieve the rated boost frequencies on my Ryzen 9 3900X.

Nor a 280mm AIO liquid cooler on my 3800X...
Maybe 360 is the magical number?

 

[hzy4]

The issue with AMD is not the performance its the advertisement. They should advertise 4,3Ghz with 3700X and say it might reach more with PBO and all would be well, same goes for other SKUs. Also they could openly say, we are still working on our drivers and it should get only better from there.

 

[bug]

Am I? Where does it say so? I haven't seen anything that specifies this. And the 3700X gets the same clocks by the looks of it, so what did I pay extra for that I got with my 3800X?

It says so on their page now:

Max Boost Clock is the maximum single-core frequency at which the processor is capable of operating under nominal conditions.

When other cores are running, they are eating into the total TDP, so common sense dictates you can't expect the same clocks. But AMD was crafty enough to leave this out of their description.

Fwiw, Intel doesn't do a much better job either:

Max turbo frequency is the maximum single core frequency at which the processor is capable of operating using Intel® Turbo Boost Technology and, if present, Intel® Thermal Velocity Boost. Frequency is measured in gigahertz (GHz), or billion cycles per second.

Though for Intel we can find this: https://en.wikichip.org/wiki/intel/core_i9/i9-9900k
so it seems it can hold that boost when one full core is under load.

 

[TheLostSwede]

It says so on their page now:

When other cores are running, they are eating into the total TDP, so common sense dictates you can't expect the same clocks. But AMD was crafty enough to leave this out of their description.

Fwiw, Intel doesn't do a much better job either:

Though for Intel we can find this: https://en.wikichip.org/wiki/intel/core_i9/i9-9900k
so it seems it can hold that boost when one full core is under load.

But I didn't buy the product based on what it says now, I bought the product based on what they marketed it as. If they go and change the marketing language, that's not my problem.

Without enabled all the overclocking settings in my UEFI, I simply can't get about 4.4GHz, which is not what their specs claim.
On top of that, their PBO video claimed you can get an extra 200MHz or maybe even 300MHz out of the chips if you get the "right" motherboard, but I guess my board is simply too cheap...

Again, my system performs well, but the product I was sold, is not what I got.

 

[bug]

But I didn't buy the product based on what it says now, I bought the product based on what they marketed it as. If they go and change the marketing language, that's not my problem.

Without enabled all the overclocking settings in my UEFI, I simply can't get about 4.4GHz, which is not what their specs claim.
On top of that, their PBO video claimed you can get an extra 200MHz or maybe even 300MHz out of the chips if you get the "right" motherboard, but I guess my board is simply too cheap...

Again, my system performs well, but the product I was sold, is not what I got.

You also bought the product based on what benchmarks and reviews said about it and that's what you got. So there you have it.

 

[xorbe]

Maybe it's just me but, look at benchmarks, look at my wallet, buy the one that fits my needs. Use new PC to browse for next computer.

 

[TheLostSwede]

You also bought the product based on what benchmarks and reviews said about it and that's what you got. So there you have it.

Except AMD didn't seed any 3800X CPUs...
So no.

Say you bought a car with a certain fuel usage and a certain performance, but it turns out you're only getting 80% of the claimed fuel usage and performance, would you be happy?
It was the model that didn't get tested by any car magazine, but you bought it based on the performance of the nearest model.
Then the car manufacturer went out and stated it was during "nominal conditions" without explaining those conditions, would you be happy?
I think not.

 

[AMX85]

everybody is arguin by Clock Speed, nobody is talking about IPC, Ryzen2 having lower clocks at every core count, but the performance gap is lower than clock gap


greetings!

 

[Xuper]

It's simple.people forgot that Bus Clock's Ryzen is NOT 100mhz , Always below 100, mine is 99.76mhz but Multiplier is always a Fixed number.4375 / 44 = 99.5Mhz.

 

[Dragonsmonk]

Except AMD didn't seed any 3800X CPUs...
So no.

Say you bought a car with a certain fuel usage and a certain performance, but it turns out you're only getting 80% of the claimed fuel usage and performance, would you be happy?
It was the model that didn't get tested by any car magazine, but you bought it based on the performance of the nearest model.
Then the car manufacturer went out and stated it was during "nominal conditions" without explaining those conditions, would you be happy?
I think not.

You are aware that that is what is being done, correct?
The l/100km is also done under VERY specific conditions, with VERY specific behaviour that you'd never see in normal every day life.

Also, I think that this is being pushed out of proportion - I don't expect either camp to have a 100% up to scratch product (in relation to BIOS and drivers) within the first 3 months... too many variables that you cannot test for or expect...

 

[EarthDog]

Also, I think that this is being pushed out of proportion - I don't expect either camp to have a 100% up to scratch product (in relation to BIOS and drivers) within the first quarter... too many variables that you cannot test or expect...

I don''t think anyone expects perfection, but, some launches (from either camp) are worse than others. This has been better than Zen, but not as good as Zen+ as far as out of the box support/compatibility/stability...(and yes, well aware one was a refresh).

 

[TheLostSwede]

You are aware that that is what is being done, correct?
The l/100km is also done under VERY specific conditions, with VERY specific behaviour that you'd never see in normal every day life.

Also, I think that this is being pushed out of proportion - I don't expect either camp to have a 100% up to scratch product (in relation to BIOS and drivers) within the first 3 months... too many variables that you cannot test for or expect...

And the car companies get in legal trouble over it, no? They also have to provide multiple specifications based on different conditions.
Sadly, in today's world it feels like everyone's trying to cheat the consumers, one way or another.

everybody is arguin by Clock Speed, nobody is talking about IPC, Ryzen2 having lower clocks at every core count, but the performance gap is lower than clock gap


greetings!

Never said it performs bad, but AMD threw a number in the hat and isn't meeting that number...
I paid for something that hasn't been delivered.

Am I disappointed with the performance, no.
But I would like a damned good explanation from AMD as to what has gone wrong and why so many people are seeing side a widespread performance difference (not talking MHz here, but rather actual benchmark numbers).

It's simple.people forgot that Bus Clock's Ryzen is NOT 100mhz , Always below 100, mine is 99.76mhz but Multiplier is always a Fixed number.4375 / 44 = 99.5Mhz.

Same on most Intel boards though, so that's not the issue.

 

[mouacyk]

nominal as defined at https://www.collinsdictionary.com/us/dictionary/english/nominal

You use nominal to indicate that someone or something is supposed to have a particular identity or status, but in reality does not have it.

Using the potential of best silicon to sell all silicon...

 

[Vayra86]

Maybe it's just me but, look at benchmarks, look at my wallet, buy the one that fits my needs. Use new PC to browse for next computer.

You forgot a critical step: after purchase run a few benches and post on TPU, then browse for next computer.

Also, mandatory repaste due to being 1-2C below temps in Youtube review

People buy hardware based on the manufacturers spec.
How about DDR4 4400MHz memory "only" managed DDR4 4375MHz, would people be happy? I think not.

The issue is as you say with the marketing, but the users have been "sold" a number and as such are expecting said number, not 200-300MHz less than said number.
How do you suggest that situation should be resolved, as it seems AMD is now trying to come up with some rather lame excuses as to why we're not quite getting what we paid for.

I've already suggested what should be done with the 'boost' spec on a spec sheet to make it a more useful number to gauge relative performance (between CPUs of the same product line, because that's about as far as it goes anyway). I mean what else should it be other than a number you can use for at least something. If you can't, might as well remove it altogether. Its right up there with HD Ready, Watt RMS and 'dynamic contrast ratio' in bullshit stats.

As others and myself point out, you didn't buy a boost clock you bought an idea of performance and that performance is what you got. Don't get me wrong though, AMD is doing itself a service correcting this one way or the other but let's not overreact either.

Still is an interesting issue and like you I'm curious where the real problem is at. Still though if this is anything like Nvidia's GPU Boost it might just be as simple as 'losing a boost bin' just a tad earlier than most people would expect it to, which could translate to something silly as moving a temperature or voltage target up or down a notch. And maybe, just maybe, AMD concluded it can't guarantee stability if it does that. We already know these CPUs happily clock above boost if you really want them to.

And then I spotted this

wtf?

 

[yeeeeman]

They should put a clear explanation on the single core boost value, since as far as we see here, 99% of the guys here don't have a clue what it is and when it can be reached. You will reach max boost of single core when you do a single core cb run for example, since only one core is pushed and that also might not happen since you might have other apps that use other cores, so the power budget needed for max boost on one core is not available.
Tldr: you will reach max boost clock in limited situations.

 

[junglist724]

The 9900K graph 'looks flatter'... yeah, if you disregard the actual numbers they've used. Intel's turbo is an ancient piece of junk compared to XFR.

Intel 'looks flatter' ... with a 500mhz frequency gap versus AMD's 150 mhz.



Good point, because if they advertise 4.4 Ghz and hit 4375, that is really under

I'll still take XFR any day of the week though...

Yeah and you'll never see the 1 core boost clock on a 9900K without modifying your power plan. Windows wants to spread threads across all cores and even with no background applications open Windows + a hardware monitor will create load on more than 1 core. You have to enable core parking so that Windows won't assign threads to parked cores, but even then I only ever see the 2 core boost on my OC'd 9900k using custom turbo clocks. Of course by default core parking will actually decrease your performance too because it doesn't enable enough cores for some workloads, something you can only fix by modifying hidden PPT settings that you have to unlock with registry edits.

 

[yeeeeman]

They should put a clear explanation on the single core boost value, since as far as we see here, 99% of the guys here don't have a clue what it is and when it can be reached. You will reach max boost of single core when you do a single core cb run for example, since only one core is pushed and that also might not happen since you might have other apps that use other cores, so the power budget needed for max boost on one core is not available.
Tldr: you will reach max boost clock in limited situations.

Adding to my response...I believe the issue here is the following:
- as AMD stated, at least one core is able to to max boost; that means that many others might not be able to do so;
- when you launch a single core load like cb you cannot control the process (or maybe you can) to be loaded on that one core that can do max boost. So most of the time you end up getting lower boost clocks because you hit a slower core, hence the problem people complain about.
I do agree that this is a problem, but I do think it is hard to make a fix since it is hard to detect what threads will use a core to 100% and then maybe migrate it to the core that can go to max boost...
The only solution is to either lower the specs or improve the process tech with tsmc so that all core can achieve min freq = max boost.

 

[Vayra86]

Adding to my response...I believe the issue here is the following:
- as AMD stated, at least one core is able to to max boost; that means that many others might not be able to do so;
- when you launch a single core load like cb you cannot control the process (or maybe you can) to be loaded on that one core that can do max boost. So most of the time you end up getting lower boost clocks because you hit a slower core, hence the problem people complain about.
I do agree that this is a problem, but I do think it is hard to make a fix since it is hard to detect what threads will use a core to 100% and then maybe migrate it to the core that can go to max boost...

What? Where does that definition get clarified? Only one core on the package being even capable and its luck of the draw whether you might use it? That sounds like... well I don't know, a whole lot of WTF.

 

[Nordic]

I purchased a 3900x after looking at various reviews. I knew I wouldn't get as good of single threaded performance as Intels latest just given the clockspeed difference, but it is faster than the 4790k I upgraded from. Intel has been so boring lately and hasent really given us anything new in so long. AMD earned my purchase. My build performs pretty well despite some complaints I might have.

This is an enthusiast forum and I am not the average user. AMD sold me a 4.2ghz chip as a 4.6ghz chip. I average 4.2ghz when gaming or running single threaded benchmarks. I have caught the cpu running as fast as 4525mhz, but only at idle. During any kind of load no matter the amount of cores used, it does not boost that high. I knowingly purchased my 3900x knowing about these problems, but that doesn't mean they aren't problems.

AMD's XFR or boost does seem to work better than intels turbo boost. AMDs boost even seems magical given that it can overclock the cpu higher than us enthusiasts can manually do it. It is fascinating that AMD can do this. AMD marketed their cpus with this increase even. I think AMD should have sold these as 4.2ghz cpus that can boost higher given AMD's technology that you can't get in intel. This way we would be talking about how cool AMD is for giving us more performance potentially, even though the cpu doesn't really boost higher than 4.2ghz under load. The 4525mhz I get would be exciting even though it only happens at idle rather than under load when I could use that clockspeed increase.

Every Intel cpu I have purchased has been able to meet the turbo boost clock. I have also consistently been able to overclock the cpu by manually setting the cpu to run all cores at the boost speed. Intel has binned every cpu so that it can for sure run at the max boost speed. From an hardware enthusiast point of view this is valuable. We can trust Intel to have consistent as advertised performance but we can also run our cpus as fast as Intel advertised on all cores if we have the knowledge, cooling, and motherboard to do it. In contrast AMD has a theoretical boost clock as a marketing gimmick hiding behind some really cool technology.

Am I upset? Not perticularly. I knew what I was buying. I would buy it again. I still think AMD's boost clock is bad advertising and isn't worth it despite being better than intels. The problem isn't the boost algorithm but the silicone not being able to run as fast as marketed, especially when it is needed.

 

[biffzinker]

The problem isn't the boost algorithm but the silicone not being able to run as fast as marketed, especially when it is needed.

Even with the switch from GloF 12nm LP process to TSMC's 7nm process it's still targeting a low power process. TSMC not to long ago announced a revision to the 7nm process for high power draw. AMD could move Zen2 chiplets over for a clock speed boost.

 

[bug]

Except AMD didn't seed any 3800X CPUs...
So no.

Say you bought a car with a certain fuel usage and a certain performance, but it turns out you're only getting 80% of the claimed fuel usage and performance, would you be happy?
It was the model that didn't get tested by any car magazine, but you bought it based on the performance of the nearest model.
Then the car manufacturer went out and stated it was during "nominal conditions" without explaining those conditions, would you be happy?
I think not.

Ok, I get it, you're bitter.
Try this: fire up SuperPi (single threaded) and see whether you reach the advertised speeds. If you don't, then yes, AMD has shafted you.

 

[RH92]

They should put a clear explanation on the single core boost value, since as far as we see here, 99% of the guys here don't have a clue what it is and when it can be reached.
You will reach max boost of single core when you do a single core cb run for example, since only one core is pushed and that also might not happen since you might have other apps that use other cores, so the power budget needed for max boost on one core is not available.

It seems that you are a part of those 99% aswell !

Most guys don't ever get close to max advertised boost clocks even when running single core loads with everything else closed .
The only way to hit advertised boost clocks is to move around the mouse when PC is idle , this is what AMD means for '' optimal conditions '' and this is where the marketing BS comes to play !

The debate here is to know if it is ethical from AMD to advertise clocks you can hit only when you play with your mouse at idle instead of advertising clocks you can hit under '' normal '' single core loads ?
The answer is hopefully obvious , hence why many peoples do rightfully so complain about not getting what they paid for !