• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

AMD Ryzen 9 3900X

I have a question regarding Precision Boost Overdrive.

In Computex and E3 Amd showed these slides:

hx9446bddm331.png


1a5b6mFmkZD0KjjJ.jpg


And couple of days ago AMD showed this video on Precision Boost Overdrive:

Updates to Precision Boost Overdrive for the AMD Ryzen 3000 Series

In video AMD representative clearly gives example of single processor that boosts to 4.55 GHz, and after PBO it can reach 4.75 GHz.

In all those cases AMD shows PBO lifting single core performance and frequency (+200 MHz Auto Overclock for Cinebench R20 single thread on first slide).

I haven't really seen any Ryzen 3000 on any review that boosts past their stated single core boost clock, except when using liquid nitrogen. Why is that?

Maybe the reviewers weren't testing PBO properly? Early UEFI? Wrong drivers?
AMD just launched new drivers today, there are new UEFIs out in the past day or two, so it might be worth taking these first tests as just that.
Unlike Intel, AMD seems to be a bit sloppy with coordinating launches and the boards makers have to do a lot of the heavy lifting.

Note that this is apparently tied to the power budget allocated to the chip, in addition to having sufficiently good cooling.
The power budget (as I pointed out in a different thread) is 87W on the 65W TDP parts and 143W on the 105W TDP parts.
It's clearly working, although unfortunately, these guys don't mention the exact frequency their CPUs hit. It's seemingly working best in productivity related software, as it doesn't seem to have a huge impact in games. Also, you might need Google translate to read that properly.

Say anything in these forums and you'll eventually get called a fanboy I find.

Still, I can confirm what you say. I wrote the infamous hypetrain article a while back saying exactly that... and next to the cries of "no politics" (because I used a political analogy to... Mugabe of all people) there were the cries of "Intel fanboy!"

I am neither an AMD supporter nor an AMD one. I just like actually reviewing with realistic expectations.

So you're neither/nor a supporter of AMD? :laugh:
 
Low quality post by TheLostSwede
Or intel... but yeah. Fixed.
Sorry, it was too funny a slip up not to take advantage of...

oh my goat, the fanless 570x cost 1100 $ in denmark
Maybe you should sell the goat, it might fetch you enough for a motherboard?

Also, can you please change your profile picture, seeing you with your hand down your speedos is disconcerting...
 
Low quality post by R-T-B
Sorry, it was too funny a slip up not to take advantage of...

I was editing it seconds ago thinking "lol there is no way everyone missed that, even if it was only up for a minute"

Thanks for not disappointing.
 
Low quality post by Eskimonster
Sorry, it was too funny a slip up not to take advantage of...


Maybe you should sell the goat, it might fetch you enough for a motherboard?

Also, can you please change your profile picture, seeing you with your hand down your speedos is disconcerting...
It aint speedoes, i just pulled up the shorts to get sun, and my hand is tucking them. I did not even know i was beeing pictured b4 it was to late :)
 
I have a question regarding Precision Boost Overdrive.

In Computex and E3 Amd showed these slides:

hx9446bddm331.png


1a5b6mFmkZD0KjjJ.jpg


And couple of days ago AMD showed this video on Precision Boost Overdrive:

Updates to Precision Boost Overdrive for the AMD Ryzen 3000 Series

In video AMD representative clearly gives example of single processor that boosts to 4.55 GHz, and after PBO it can reach 4.75 GHz.

In all those cases AMD shows PBO lifting single core performance and frequency (+200 MHz Auto Overclock for Cinebench R20 single thread on first slide).

I haven't really seen any Ryzen 3000 on any review that boosts past their stated single core boost clock, except when using liquid nitrogen. Why is that?
Non-x cpu
 
Non-x cpu

The second slide shows PBO increases across most of the product range, and in AMD Youtube PBO explanation an example of CPU that boosts from 4.55 GHz to 4.75 GHz is clearly not 3600 non X (since it's boost is much lower).

I think it's fair to assume that AMD showed PBO overclock of single core boost above stock maximum boost for all new Ryzen 3000 CPUs. And reviews simply don't show that.
 
I think it's fair to assume that AMD showed PBO overclock of single core boost above stock maximum boost for all new Ryzen 3000 CPUs. And reviews simply don't show that.

I think It's fair to say that this is textbook illustration of misleading marketing. Advertising at its core.
 
But why didn't a single reviewer point that out? The only one pointing that out was Der8auer in his short pre-NDA comment on overclocking, who also told that all of his processors don't even achieve advertised boost in single core loads.
 
oh my goat, the fanless 570x cost 1100 $ in denmark
Look, we get it, you love intel stuff, but you're just constantly pointing out irrelevant talking points in some sort of attempt to try and make these Zen 2 CPU's look bad(which they quite obviously are not).

First it was 'you need an expensive x570 board' and which removes the better value they have over their intel counterparts. We've all pointed out is not the case, so now, completely off topic without any outside influence, you mention what I assume can only be the Asrock x570 Aqua, a 100% waterblock motherboard is $1100(which doesn't even seem to be in stores yet?). We can do that same game with the z390 motherboards: Gigabyte Z390 Aorus Waterforce.

These Zen 2 CPU's are good. They're good value(especially the 3700x/3600x), and even if you're playing at 1080p with a $1,300 2080 Ti, which is highly unlikely, there is roughly a 5-8% difference, which is to say, a completely unnoticeable difference.
 
So much is missing from your review... your graphics test for 1080p, you don't say at all what graphics settings you're using. Ultra? medium? low? AA on? AA off? You don't tell us at all. You stick to only 3200 mhz memory as well. The 2000 series were certified up to 3600 Mhz memory for most motherboards for X470 but you don't even try 4000 or higher? Why not? Do you people not even have access to ram at 4000 mhz or faster? You're also only testing an AIO.. So you claim the chips run hot and that limits the OC. Okay, so try a custom water loop instead? Do you not know how to set up a custom water loop? *Sigh* This is the major problem with tech review sites. Most of them don't even try to actually push the chips properly to a normal usage situation, custom water, fast ram, etc. Most people have a decent custom water loop these days. Back to the reddit list to find a proper review site.
 
But why didn't a single reviewer point that out? The only one pointing that out was Der8auer in his short pre-NDA comment on overclocking, who also told that all of his processors don't even achieve advertised boost in single core loads.

1. Reviewers are biased. Not necessarily biased to a particular brand (although that certainly exist), but biased to private power in general. In this case, the vendors that ship them products beforehand for review, and that they are dependent upon for their livelihood. This is simple political economic understanding that is well-understood. That is what the whole PR industry is for, after all; to manufacture consent. And reviewing products is a part of that; it's a form of indirect advertising.

2. To avoid backlash from a hostile AMD fanbase (which is very real). Just look at all the criticism various sites, like Techspot and Gamersnexus, recieved when Zen and Zen+ launched. These sites would prefer to stay away from unnecessary drama, if possible.

*Sigh* This is the major problem with tech review sites. Most of them don't even try to actually push the chips properly to a normal usage situation, custom water, fast ram, etc.

Custom water loop is not a "normal usage situation", it's a niche. AIO and air coolers are what define as a "normal usage situation". Same is true in terms of RAM speeds; tinkering with RAM timings is something only a small minority do, whereas the majority of users either leave it at stock or just choose a simple XMP profile and leave it at that.

Maybe there's an argument to be had that they should test in the instances you talked about. But don't call it normal when it isn't.

Honestly, the biggest criticism of this review is the lack of minimal FPS metrics, as well as lack of more games tested. I would have liked a greater pool of titles, including the most popular titles out there: PUBG, Overwatch, Dota 2, CS: Go, Apex Legends.
 
Last edited by a moderator:
This is going to be an amazing upgrade from my 1700X
 
Same is true in terms of RAM speeds; tinkering with RAM timings is something only a small minority do, whereas the majority of users either leave it at stock or just choose a simple XMP profile and leave it at that.
All X470 motherboards have certified ram kits to certified up to 3600 mhz ram for all Ryzen 2000 series chips. And that is not "tweaking with settings". That is "Buy the listed memory kit, stick it in and it's guranteed to work out of the box at the specified speeds". They didn't even try anything above 3200. Ryzen is supposed to scale with ram speed and they're just sticking to stock ram speed.

EDIT: I'm even learning from seeing other websites that AMD provides ram kits with the reviewer kit for the new chips, Trident-Z Royal DDR4-3600. So it's actually provided with the new chips and guranteed to work with the new chips and Techpowerup here isn't even using the provided ram. Instead they used their own 3200 mhz kit. What the heck?
 
3900x is a beast, but for a gaming CPU seems like they better choice is the 3700x due to price and performance comparison.

Intel has been cornered into a single thread, gaming performance category at this point.
 
Last edited:
If you can use all the cores and threads, no doubt!
I already manage to max both my CPU and RAM out

So the 3900X and 64GB of RAM is a great upgrade
 
So much is missing from your review... your graphics test for 1080p, you don't say at all what graphics settings you're using. Ultra? medium? low? AA on? AA off? You don't tell us at all. You stick to only 3200 mhz memory as well. The 2000 series were certified up to 3600 Mhz memory for most motherboards for X470 but you don't even try 4000 or higher? Why not? Do you people not even have access to ram at 4000 mhz or faster? You're also only testing an AIO.. So you claim the chips run hot and that limits the OC. Okay, so try a custom water loop instead? Do you not know how to set up a custom water loop? *Sigh* This is the major problem with tech review sites. Most of them don't even try to actually push the chips properly to a normal usage situation, custom water, fast ram, etc. Most people have a decent custom water loop these days. Back to the reddit list to find a proper review site.

Did you not see how much content just went up? Again, as I suggested to someone else, please contact @W1zzard and ask for an internship, I'm sure he'd love to have your expertise and opinion at hand to help him test things. How much hardware can you test in 24h? I presume you don't need to eat, sleep or use the little boys/girls room, right?
It's easy to criticise from the outside, but once you've done some serious last minute testing, you know what goes into something like this and you appreciate the hard work that has gone into it.

What you might also not know, is that AMD has seeded different hardware to different sites. They've seeded at least three different motherboards and not all sites were given boards by AMD, some were seeded directly from the board makers. So if not everyone got boards, do you think everyone got RAM? AMD has a list of priority media that gets the full kit and some don't even get a CPU to test. But hey, you know best, right?
 
Low quality post by Dyatlov A
When a Core i5 is faster than this Core monster, I do not see why is this AMD hype? Make fewer fast Cores, rather than many slow ones!
 
3900x is a beast, but for a gaming CPU seems like they better choice is the 3700x due to price and performance comparison.

Intel has been cornered into a single thread, gaming performance category at this point.
Even for gaming, it only makes a marginal difference if you are using a 2080ti.

The 3900x is an absolute monster, not quite sure I can justify upgrading from my 5820K though. Retail price on Amazon is £480, as usual we overpay, though for 12 cores that's still astonishingly low.

If this were Intel that 3900x would cost upwards of £1k, easy.
The i9-9920X 12-core does indeed costs $1200.
And it is Skylake-X, which doesn't game as well as CoffeeLake either.
 
The 3900x is an absolute monster, not quite sure I can justify upgrading from my 5820K though. Retail price on Amazon is £480, as usual we overpay, though for 12 cores that's still astonishingly low.

If this were Intel that 3900x would cost upwards of £1k, easy.
 
Did you not see how much content just went up? Again, as I suggested to someone else, please contact @W1zzard and ask for an internship, I'm sure he'd love to have your expertise and opinion at hand to help him test things. How much hardware can you test in 24h? I presume you don't need to eat, sleep or use the little boys/girls room, right?
It's easy to criticise from the outside, but once you've done some serious last minute testing, you know what goes into something like this and you appreciate the hard work that has gone into it.
Most (almost all) review sites have had their review samples for weeks and they've been testing and reviewing the chips for a while now, but just under NDA and not allowed to publish the results until this morning, today. They definitely do not have "just 24 hours to do all testing". Where did you ever get that?

I don't know everything but I'm just looking for an in-depth review that lists everything and really tries to push these chips as fast as they're capable of going. It's not hard to at the very least write into the review "Ultra" for the 1080p tests.. to let us know what graphics settings they used. They didn't even do that much here at techpowerup. It's well known by now that All ryzen chips scale with ram speed, yet despite this, the reviewers here stick to 3200 Mhz. Why? That's so confusing. I've so far seen 10 reviews from the reddit thread and only one site used ram above 3200 mhz when testing these chips and it was the 3600 mhz kit provided with their review kit. Out of 10 so far.. only one even tried a custom water loop to see what the chips can really do.
It's just frustrating to try and find information on these chips. I'm considering buying one but if I can't see how it performs maxed out on a CWC loop with fast ram then I'll have to wait a while. I'm still hunting to find a review that even remotely tries to max out the new chips. Almost everyone is testing either stock speed with stock ram or just an AIO.

Maybe one of my friends will actually buy one and I can talk to them 1:1 to give me the information about the chips that reviewers won't tell us.

EDIT: So at least one motherboard, the X570 Godlike has QVL-Certified ram kits listed up to 4800 Mhz DDR4 for the ryzen 3000 series. Yet still no one is trying above 3600 that I can find. Sad.. depressing even. These chips may have untapped potential but no one is even trying.
 
Last edited:
Back
Top