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

[Necrotic]

Quote

I don't really know what to say, to be honest. The results is much worse than what I expected. Again, keep in mind what I said at the beginning, [in regards] to being a public survey what can be the negative influence about this but I am pretty sure that I was not wrong, that the CPU's in general cannot hit the boost and probably its just the silicone quality is maybe not good enough. I clearly want to state that I absolutely recommend buying those CPU's...

Although this doesn't change the overall value of Ryzen 3000 in my view and its more relevant on the higher chips (3900X in particular), one should be aware what you are buying. Up to this point the "boosts up to XXX" has been a fairly reliable numbers, but AMD is changing the equation here by not consistently reaching it and it seems unlikely that changes will be made (in bios and such) to correct it due to risk of chip failures in the field.

 

Source: 

Update: AMD has released a statement on twitter:

"AMD is pleased with the strong momentum of 3rd Gen AMD Ryzen ™ processors in the PC enthusiast and gaming communities. 3rd Gen AMD Ryzen users are reporting boost clock speeds below the expected processor boost frequency. While processor boost frequency is dependent on many variables including workload, system design, and cooling solution, we have identified the feedback from our customers and have identified it as having an impact on our firmware. We are in the process of preparing a BIOS update for our motherboard partners that addresses that issue and includes additional boost performance optimizations. We want to provide an update on September 10 to the community regarding the availability of the BIOS."

https://twitter.com/AMDRyzen/status/1168901636162539536

Edited September 3, 2019 by Necrotic

[Humbug]

AMD should have just advertised the clockspeeds as 100Mhz lower. People are happy with the performance which is what matters.

 

What is the point of this? Do they think that they will lose sales because some people don't unerstand that clockspeed is only one factor in performance.

[Humbug]

What I understand so far is that even though Ryzen has a fairly low TDP and runs cool on a stock cooler, the boost clocks are still dependent on cooling. So if you have good cooling they will boost a bit more.

[Necrotic]

1 minute ago, Humbug said:

AMD should have just advertised the clockspeeds as 100Mhz lower. People are happy with the performance which is what matters.

 

What is the point of this? Do they think that they will lose sales because some people don't unerstand that clockspeed is only one factor in performance.

This is probably a complicated answer. Perhaps they didn't have the right data of production lot performance at the time they decided the marketing issues. Alternatively, between Intel's traditional higher clock speeds and trying to make 3000 series look better than their own 2000 series and make a big splash, they may have simply over-committed on the marketing aspects of boost clocks.

[Bitter]

I smell another Bulldozer class action lawsuit. Heck, there was a class action (several actually) about lawn mowers not doing the advertised horsepower.

[Blademaster91]

6 minutes ago, Humbug said:

What I understand so far is that even though Ryzen has a fairly low TDP and runs cool on a stock cooler, the boost clocks are still dependent on cooling. So if you have good cooling they will boost a bit more.

So AMD should be advertising boost clocks based on averages using the stock cooler, otherwise its a bit misleading if boost clocks aren't consistent.

[Humbug]

9 minutes ago, Bitter said:

I smell another Bulldozer class action lawsuit. Heck, there was a class action (several actually) about lawn mowers not doing the advertised horsepower.

Unfortunately I think AMD will get away with this because the chips tend to hit the boost frequencies in certain ideal conditions... i.e. particular applications when running with good coolers.

[Humbug]

Just now, Blademaster91 said:

So AMD should be advertising boost clocks based on averages using the stock cooler, otherwise its a bit misleading if boost clocks aren't consistent.

absolutely!

[Necrotic]

1 minute ago, Humbug said:

Unfortunately I think AMD will get away with this because the chips tend to hit the boost frequencies in certain conditions... i.e. particular applications when running with good coolers.

I also suspect that all the marketing probably says "...up to XXX Ghz" giving themselves some room.

[Sauron]

7 minutes ago, Bitter said:

I smell another Bulldozer class action lawsuit. Heck, there was a class action (several actually) about lawn mowers not doing the advertised horsepower.

hope not, those are just free money for lawyers and people will get 5$ at most

[RejZoR]

44 minutes ago, Bitter said:

I smell another Bulldozer class action lawsuit. Heck, there was a class action (several actually) about lawn mowers not doing the advertised horsepower.

But no one has an issue with Intel's BS TDP ratings...

[DrMacintosh]

1 hour ago, Humbug said:

What is the point of this? Do they think that they will lose sales because some people don't unerstand that clockspeed is only one factor in performance

The average consumer is, uhh, “not very intelligent.” So yes, that probably could be a real concern at some level. 

[5x5]

The survey itself is flawed since people aren't running the same cooling, power delivery and firmware. A lawsuit is not possible. AMD have listed a given boost, not what users need to achieve it. If someone is complaining about getting 100MHz lower on a cheap 45$ B350 board, then that's their problem

[Energycore]

I think this one's easy. If AMD were careful to say "up to x GHz", they're 100% safe. If they didn't, they need to hire a new marketing manager.

[porina]

"up to" is one of the most intentionally misleading marketing phrases that is widely used, not just by AMD. It's a way of looking like you're giving useful information when you're really not, as it could be anything lower.

 

1 hour ago, RejZoR said:

But no one has an issue with Intel's BS TDP ratings...

It has a definition. That a few either choose to ignore it or want it to mean something else, is their problem, not Intel's.

[Curious Pineapple]

15 minutes ago, porina said:

"up to" is one of the most intentionally misleading marketing phrases that is widely used, not just by AMD. It's a way of looking like you're giving useful information when you're really not, as it could be anything lower.

So the better option is to lock the CPU frequency down to whatever it is rated at and remove any turbo or boost features?

[Jito463]

My personal opinion is that AMD should restrict their boost numbers to what can reasonably be achieved with stock cooling, then clarify higher boost available dependant on cooling solution used.

 

That said, these numbers are so close that I'm not going to hassle them over it.  Also, this is based on an informal, unscientific survey, so I take those results with a grain of salt.

[RejZoR]

AMD should post 4GHz as boost clock and then everyone would be shitting their pants when CPU would be hitting 4.4GHz... That's kinda like what NVIDIA is doing. Stating absolutely pointless boost clocks and then GPU always goes kilometers past that mark. Coz in this case, AMD's "Max boost clock" makes more sense. It's a number which will NEVER be exceeded by the CPU. Getting as close to it as possible falls down to many factors and none of which are made up by AMD. TDP, temperature, threads used for the workload, all this affects the clock.

 

It's kinda ironic how people always want MOAR CORES, AMD gives them MOAR CORES and then they start bitching it's not running at 4.6GHz at all times. Especially in this time and age when everyone wants everything as multithreaded as possible. Newsflash, you can't put load on all cores and expect them to run at 50GHz. But you'd also want that 105W TDP at the same time. You know, I'd love to break laws of thermodynamics too. Overclocking would be so much easier without these stupid restricting laws. What idiot made them up anyway?

[porina]

31 minutes ago, Curious Pineapple said:

So the better option is to lock the CPU frequency down to whatever it is rated at and remove any turbo or boost features?

Define it in a more meaningful way. Maybe something like "average CPU speed" which will no doubt lead to endless arguments as to what should go into that average. Minimum could be useful in some situations, but we have that already as base clock.

 

With Zen 2, it really feels like AMD went all out pushing the single core turbo so that can advertise higher speeds.

[Curious Pineapple]

1 minute ago, porina said:

Define it in a more meaningful way. Maybe something like "average CPU speed" which will no doubt lead to endless arguments as to what should go into that average. Minimum could be useful in some situations, but we have that already as base clock.

 

With Zen 2, it really feels like AMD went all out pushing the single core turbo so that can advertise higher speeds.

They sell the CPU with a specified base clock which all cores on all processors will run at on motherboards built to minimum specifications. The boost clock is dependant on so many factors that the "up to" term is used. If the bare minimum PSU is a 350W and a 350w is used, then it may sag under the extra loading of jacking up the speed and cause instability. The only way they can market guarenteed speeds is to lock the CPU and what you get is what you get.

[Eaglerino]

I have a 3700x and it'll boost to 4.3Ghz but not the advertised 4.4. I just thought it was my cooler not being able to keep up, but I doubt 100mhz would make a difference for what I use it for anyway

[Taf the Ghost]

4 hours ago, Bitter said:

I smell another Bulldozer class action lawsuit. Heck, there was a class action (several actually) about lawn mowers not doing the advertised horsepower.

AMD will be fine. Every chip can hit the "max boost", emphasis on "max" here. AMD never said you could all-core sustain at that boost, nor did they ever say anything beyond that the chip could hit those clocks. If you load up very specific types of single-core workloads, you'll hit the boost without much issue. Anything that hits AVX you won't see it, however. Plus, a lot of motherboards have less aggressive single-core boost effects in their power tables.

 

Currently, your max boost matters what Motherboard & BIOS you have. Chips will still hit the all-core speeds without much issue.

[Taf the Ghost]

1 hour ago, Eaglerino said:

I have a 3700x and it'll boost to 4.3Ghz but not the advertised 4.4. I just thought it was my cooler not being able to keep up, but I doubt 100mhz would make a difference for what I use it for anyway

It really doesn't effect anything. The core blipping to 4.4 Ghz for a microsecond doesn't change the actual processing capabilities. The longer it can hover at 4.2 Ghz the better off you are.

[Necrotic]

2 hours ago, 5x5 said:

The survey itself is flawed since people aren't running the same cooling, power delivery and firmware. A lawsuit is not possible. AMD have listed a given boost, not what users need to achieve it. If someone is complaining about getting 100MHz lower on a cheap 45$ B350 board, then that's their problem

If you look at the start of the video he discusses this, including the fact that his average viewer isn't a newbie or the fact that boost limitation isn't tied to specific cheaper boards. You also can't claim that "listed a given boost, not what users need to achieve it" as in legal terms that doesn't reasonably stand, its called false advertising and its illegal. When you promote something in marketing its something that has to be reasonably understood by the population, you can't just say that you have a different definition. Please  see the bulldozer settlement if you want to see an example of that. What will let them off the hook is that they probably say "up to" or "max" which doesn't necessarily guarantee it.

1 hour ago, Curious Pineapple said:

So the better option is to lock the CPU frequency down to whatever it is rated at and remove any turbo or boost features?

No, this hasn't been a issue before. They test all the CPU's, they just need to put a more reasonable/consistent max boost on "standard" equipment (ie their fan).

1 hour ago, RejZoR said:

AMD should post 4GHz as boost clock and then everyone would be shitting their pants when CPU would be hitting 4.4GHz... That's kinda like what NVIDIA is doing. Stating absolutely pointless boost clocks and then GPU always goes kilometers past that mark. Coz in this case, AMD's "Max boost clock" makes more sense. It's a number which will NEVER be exceeded by the CPU. Getting as close to it as possible falls down to many factors and none of which are made up by AMD. TDP, temperature, threads used for the workload, all this affects the clock.

 

It's kinda ironic how people always want MOAR CORES, AMD gives them MOAR CORES and then they start bitching it's not running at 4.6GHz at all times. Especially in this time and age when everyone wants everything as multithreaded as possible. Newsflash, you can't put load on all cores and expect them to run at 50GHz. But you'd also want that 105W TDP at the same time. You know, I'd love to break laws of thermodynamics too. Overclocking would be so much easier without these stupid restricting laws. What idiot made them up anyway?

Please see the video, this was tested on single core, not multi core...and on an application and monitoring program recommended by AMD.

8 minutes ago, Taf the Ghost said:

AMD will be fine. Every chip can hit the "max boost", emphasis on "max" here. AMD never said you could all-core sustain at that boost, nor did they ever say anything beyond that the chip could hit those clocks. If you load up very specific types of single-core workloads, you'll hit the boost without much issue. Anything that hits AVX you won't see it, however. Plus, a lot of motherboards have less aggressive single-core boost effects in their power tables.

 

Currently, your max boost matters what Motherboard & BIOS you have. Chips will still hit the all-core speeds without much issue.

Please see my previous comments. There is also no reliable consistency between motherboards and bios versions that lead to the issues, which highlight that it seem to be a chip issue.

[BroliviaWilde]

8 minutes ago, Taf the Ghost said:

AMD will be fine. Every chip can hit the "max boost", emphasis on "max" here. AMD never said you could all-core sustain at that boost, nor did they ever say anything beyond that the chip could hit those clocks. If you load up very specific types of single-core workloads, you'll hit the boost without much issue. Anything that hits AVX you won't see it, however. Plus, a lot of motherboards have less aggressive single-core boost effects in their power tables.

 

Currently, your max boost matters what Motherboard & BIOS you have. Chips will still hit the all-core speeds without much issue.

I went through 3 different bioses on my x470 taichi and teh 3900x, first one allowed the rare fleeting max boost clock to 4.6~4.65 on various cores only on one chiplett.  But that bios had issues with IF.

couple of bios versions later and IF and RAM compatibility works great, max boost clocks are now no higher than 4.5(still even those are fleeting), but the interesting thing is that results of doing 'sustained workloads' such as cinebench r15, rendering a video or running a game are slightly better despite the lower max boost clocks.  And I'm going to stay with this latest  bios, that slight perf bump while under a load is way more important than some split second boost that only seems to happen when idle at desktop with nothing open.