AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***

[Twinz]

And stress test? One of the recent changes was introducing a lower temp at which it throttles of 75c

I ran P95 default(mixed) for about half an hour (right after I finished the build)and I don't think it got anywhere near 4.5 under that load. I'm on a 280mm AIO, so my use case will put the chip in a better light than out of the box.

And the new window AC unit is really going to skew benchmarks. My 1080ti was running around 1987 before the extra AC, but I just did a qual+ race in Project Cars 2, and it's running between 2012 and 2025!
Just for the heck of it, I ran Time Spy, and my CPU physics score was 10753. Cooler is definitely better.

 

[Drollic]

Now we know why AMD lies about boost clocks after insights from ASUS development team

The chips can boost higher, but were software lowered to avoid damaging the chips. Then recently they were lowered again.

AMD jumped the gun when their marketing put 4.7ghz boost clock on the packaging. The chips were not properly tested, the whole Ryzen 3000 launch has been a mess and AMD clearly Jumped the gun and noW everyone is just a beta tester

I can understand some of the frustration if buying the 3800x or 3900x but for the rest of the chips, who honestly cares as they're such good value. I don't see why anyone would bother with 12/24 now anyway, unless it's for content creation or the like. Complete overkill with extra cost and heat.

 

[Murphy]

They haven't given an "up to" speed if you're never seeing the advertised clocks on full custom watercooling, on an X570 Taichi, with more than enough power available. If they've changed the way boosting works with AGESA 1003, then they have changed the way the chips behave post release and printing of the boxes. That in itself is enough of an issue.

Yes they have, whether you're getting it or not is irrelevant as it's the technical maximum speed, however like i said before IMO it's wrong of them to do so if, like you say, it's all but impossible to achieve that speed.

And no they didn't change it post release, 1.0.0.3 was released (afaik) 2-3 days before the processors were available for sale to the public, like i said though (again IMO) that was deceptive as most reviews were done on the older AGESA that boosted higher, i think only two reviewers had the time to redo all their test when AMD dropped the bomb on them that they should be testing with the newer AGESA.

 

[spoffle]

Yes they have, whether you're getting or not is irrelevant as it's the technical maximum speed, however like i said before IMO it's wrong of them to do so if, like you say, it's all but impossible to achieve that speed.

And no they didn't change it post release, 1.0.0.3 was released (afaik) 2-3 days before the processors were available for sale to the public, like i said though (again IMO) was slightly deceptive as most reviews were done on the older AGESA that boosted higher, i think only two reviewers had the time to redo all their test when AMD dropped the bomb on them that they should be testing with the newer AGESA.

If they change the behavior post release, then they have changed things. It's not an up to if it never will hit it based on the changes they made.

Some boards were shipping with 1.0.0.2, and again, AGESA changes made it behave differently. What's on the box is based on boost behavior that isn't present in AGESA now.

 

[Murphy]

If they changed the behavior post release and told reviewers to test with the AGESA code that customer would get then they have not changed anything as the only people who were ever meant to use the older AGESA were people who were testing it before release, how is that so hard to understand?

And it is an up to as it will hit the boost clocks, as seen in what others have posted in this very thread and various other places on the internet, yes maybe only 5% of the processors sold are achieving their maximum boost clocks but that's irrelevant, even if you had to use LN2 to reach the maximum boost clocks they could still claim that's the maximum boost clock, because it is.

It's also irrelevant if some boards were shipping with 1.0.0.2, that's not the responsibility of AMD, it's the responsibility of board partners.

Like i said it's deceptive to say the least and IDK what country you're from but we, in the UK, had this same BS from ISP for years, quoting up to speed that hardly anyone ever received, after 20+ years or more our advertising standards agency finally ruled that they were being deceptive and tighted up the rules, they still quote up to speed but now something like 50% of customers have to be able to reach that speed.

 

[spoffle]

If they changed the behavior post release and told reviewers to test with the AGESA code that customer would get then they have not changed anything as the only people who were ever meant to use the older AGESA were people who were testing it before release, how is that so hard to understand?

And it is an up to as it will hit the boost clocks, as seen in what others have posted in this very thread and various other places on the internet, yes maybe only 5% of the processors sold are achieving their maximum boost clocks but that's irrelevant, even if you had to use LN2 to reach the maximum boost clocks they could still claim that's the maximum boost clock, because it is.

It's also irrelevant if some boards were shipping with 1.0.0.2, that's not the responsibility of AMD, it's the responsibility of board partners.

Like i said it's deceptive to say the least and IDK what country you're from but we, in the UK, had this same BS from ISP for years, quoting up to speed that hardly anyone ever received, after 20+ years or more our advertising standards agency finally ruled that they were being deceptive and tighted up the rules, they still quote up to speed but now something like 50% of customers have to be able to reach that speed.

So they changed the boost behavior of the chips so that most of them won't hit the clocks on the box, and you're performing mental gymnastics to explain why that's okay.

If most chips aren't reaching speeds listed on the box because it's set by AGESA, then they've changed things.

"Upto" shouldn't be on the box if you need to win the silicon lottery to actually get those speeds.

 

[Grim5]

Considering ryzen chips kind of self overclock - it would be like Intel advertising the 9900k with a 5.3ghz boost clock which only 2 or 3% of all chips can hit using a water cooler.

 

[Rofflay]

I would expect a class action lawsuit, You basically leave yourself open to mass returns if something differs from what is on the box. You need a disclaimer about any up to look the BT infinity small print on every advertisement. It needs to be on the cpu box or really it is a big problem. The EU would give you a pretty big fine for that i reckon.

 

[Murphy]

So they changed the boost behavior of the chips so that most of them won't hit the clocks on the box, and you're performing mental gymnastics to explain why that's okay.

If most chips aren't reaching speeds listed on the box because it's set by AGESA, then they've changed things.

"Upto" shouldn't be on the box if you need to win the silicon lottery to actually get those speeds.

There's no mental gymnastics and I'm not explaining why that's OK, in fact I've said on numerous occasions why IMO it's not OK, there is however a lack of comprehension skills on your part.

Yes they changed the boost behavior but as you've been told and admitted yourself that was before they released the product to the general public, before it was available for purchase, before reviews were made public, how is that not getting through to you? You yourself say that "if most chips" which mean some are reaching the speeds listed on the box, you're the only one performing mental gymnastics here.

 

[kitfit1]

I would expect a class action lawsuit, You basically leave yourself open to mass returns if something differs from what is on the box. You need a disclaimer about any up to look the BT infinity small print on every advertisement. It needs to be on the cpu box or really it is a big problem. The EU would give you a pretty big fine for that i reckon.

I can't disagree with that statement. I've pretty much not posted in this thread since the Zen2 release, for a couple of reasons. The first being that i intended to wait for the 3950 release, the second is because of the dire state of bioses with the consequent inability of parts to reach advertised clock speeds.
This has to be laid fairly and squarely at the feet of AMD. I already know that AMD has borked the bios for my CH6, because 7106 simply will not clock my 2700X past 4.13Ghz (since it's release my 2700X has always all cored to 4.25Ghz).
How on earth AMD has managed to get to this position is anyone's guess. As it stands at the moment though there is no way i'm going to fork out over £700 for a 3950X unless all these issues are 100% sorted out and are seen by the community as having been sorted out.

 

[Yaayuh!]

It's also irrelevant if some boards were shipping with 1.0.0.2, that's not the responsibility of AMD, it's the responsibility of board partners.
.

1.0.0.3 wasn’t available for implementation across the board before release. Like all Ryzen releases, very little time and information was given up to the launch and the board partners were screwed for time.

It’s AMD’s mess through and through.

 

[Potatowithearsontheside]

I have a different take on all of this.
The individual cores on my 3600X can all hit their clocks listed on the box; 4.4GHz. In fact, they can all be clocked at 4.5GHz for me.
However, none of the cores will hit the clocks listed on the box by themselves, i.e. automatically by themselves whether under default settings, PBO, or Auto-Overclocking. I think that is an issue; you can only hit the clocks by breaching your warranty.

On another note, how do we properly define when a clock is being hit?
I stated above that each of my cores can hit 4.5GHz. This is true if the workload is a Ryzen Master 3 minute stress test. However, I couldn't get my fastest core to run CB R20 at 4.4GHz. A touch bizarre because I had no issues with all 6 cores at 4.35GHz.
If I run CB R20 in stock mode then it'll hit about 4.1GHz on all cores, or 4.3GHz on a single core, and using significantly more voltage than I could run those speeds at manually.

I think the big problem here is that with Intel you do get the advertised boosts, whether it be for 2 cores are all cores. Even at stock, with MCE etc disabled, a 9900K will hit 5.0GHz on 2 cores for 24-28 seconds, or hit 4.7GHz on all cores for the same amount of time, followed by it dropping to 4.2GHz indefinitely under load at 95w.
If you simply ask me whether I'd prefer 4.1GHz all core with AMD, or 4.2GHz all core with Intel, then the choice is simple; AMD all day. However, that isn't the choice being made, and it certainly isn't what the marketting clocks say. Intel are normally the ones playing tricks, but here on these numbers you do get exactly what you're expecting (and more if you don't disable MCE).
It is a shame that the clock issue muddies the waters for an excellently performing CPU.

 

[Murphy]

1.0.0.3 wasn’t available for implementation across the board before release. Like all Ryzen releases, very little time and information was given up to the launch and the board partners were screwed for time.

It’s AMD’s mess through and through.

Only it was, 1.0.0.3 was released something like 2-3 days before launch. Like i said AMD is not, and can not be, responsible for what board partners and reviewers do, was it a dick move to release an AGESA update that reduced the thermal throttling limit by 5°C 2-3 days before launch? Sure, they would've known that board partners wouldn't be able to update board that had already shipped, they also would've known that very few review sites would go to the bother of retesting, however it's entirely wrong to say that this changed happened after release, it simply didn't.

Like i said previously though i agree with the sentiment that this mess is entirely of AMD's own making, however to say this change was introduced after release is simply wrong and to say Zen 2 is not reaching its on the box maximum boost clocks is also wrong. As I've repeatedly said in my opinion AMD have mislead people and been deceptive so I'm in no way attempting to defend such a move, releasing an update that changes your boost algorithms 2-3 days before public release is a dick move.

What I'm saying is that it's wrong to say this change happened after release and it's wrong to say processors are not reaching the maximum boost clocks written on the boxes, if AMD bother to respond, and i hope they do, it's important to understand the issue and how it came about.

 

[Orange Nexus]

I have a different take on all of this.
The individual cores on my 3600X can all hit their clocks listed on the box; 4.4GHz. In fact, they can all be clocked at 4.5GHz for me.
However, none of the cores will hit the clocks listed on the box by themselves, i.e. automatically by themselves whether under default settings, PBO, or Auto-Overclocking. I think that is an issue; you can only hit the clocks by breaching your warranty.

On another note, how do we properly define when a clock is being hit?
I stated above that each of my cores can hit 4.5GHz. This is true if the workload is a Ryzen Master 3 minute stress test. However, I couldn't get my fastest core to run CB R20 at 4.4GHz. A touch bizarre because I had no issues with all 6 cores at 4.35GHz.
If I run CB R20 in stock mode then it'll hit about 4.1GHz on all cores, or 4.3GHz on a single core, and using significantly more voltage than I could run those speeds at manually.

In Cinebench 15 or 20, Ryzen has always run all cores at a lower speed in Multi-thread bench. It applies to all Ryzen gens. My R7 2700, at stock, runs at 3.3GHz on these benches.

My R5 3600 hit clocks listed on two cores using Agesa 1002. Less than 70MHz on newer Agesa. All cores, though, can run 4.2GHz stably with lower power and temp draws.

 

[triss]

Only it was, 1.0.0.3 was released something like 2-3 days before launch. Like i said AMD is not, and can not be, responsible for what board partners and reviewers do, was it a dick move to release an AGESA update that reduced the thermal throttling limit by 5°C 2-3 days before launch? Sure, they would've known that board partners wouldn't be able to update board that had already shipped, they also would've known that very few review sites would go to the bother of retesting, however it's entirely wrong to say that this changed happened after release, it simply didn't.

Like i said previously though i agree with the sentiment that this mess is entirely of AMD's own making, however to say this change was introduced after release is simply wrong and to say Zen 2 is not reaching its on the box maximum boost clocks is also wrong. As I've repeatedly said in my opinion AMD have mislead people and been deceptive so I'm in no way attempting to defend such a move, releasing an update that changes your boost algorithms 2-3 days before public release is a dick move.

What I'm saying is that it's wrong to say this change happened after release and it's wrong to say processors are not reaching the maximum boost clocks written on the boxes, if AMD bother to respond, and i hope they do, it's important to understand the issue and how it came about.

I think it will remain to be seen if the CPU's with all the latest updates do continue to hit the box speeds. TBH the reviews dont matter as those arent what AMD is stating it can hit.

 

[z10m]

Considering ryzen chips kind of self overclock - it would be like Intel advertising the 9900k with a 5.3ghz boost clock which only 2 or 3% of all chips can hit using a water cooler.

under those circumstances intel chips would probably hit 6Ghz as Ryzen 3rd gen only bosts to max clocks for small minority and only one core for a split second using very high voltage...

 

[Lolik]

Last few pages like


p.s: Mine boosts over advertised, but still AMD should get the *********.

 

[Tomuks]

I think main problem is with 3900x chips as those are the ones which cant reach their max boost clocks.. but now I am worried not even about those clock speeds, but how long CPU will hold before sad death if they cant even hold their advertised clocks before degrading, what even to say about PBO then

Also have a bad feeling abou 3950 achieving 4,7ghz

 

[4K8KW10]

I think main problem is with 3900x chips as those are the ones which cant reach their max boost clocks.. but now I am worried not even about those clock speeds, but how long CPU will hold before sad death if they cant even hold their advertised clocks before degrading, what even to say about PBO then

Also have a bad feeling abou 3950 achieving 4,7ghz

The Marketing department is desperate for some 'on-paper' victories, even when the Engineering department has no solutions...

They should have started with a 45W Ryzen7 8-core and 65W Ryzen9 12-core, much smarter approach of pushing the new 7nm technology.

 

[Zeed]

I think main problem is with 3900x chips as those are the ones which cant reach their max boost clocks.. but now I am worried not even about those clock speeds, but how long CPU will hold before sad death if they cant even hold their advertised clocks before degrading, what even to say about PBO then

Also have a bad feeling abou 3950 achieving 4,7ghz

IT will happen when you pour LN2 over it