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

[DylRicho]

Sorry, my mistake, i don't often use the code names so couldn't remember if TR used Colfax or PR, apologies.

No worries, man. I knew what you were getting at.

I guess it comes down to what we consider officially supported to be, personally I'd say not launching with it, adding it and then removing it, isn't what would be considered official support.

I would be in agreement with that as well. The issue I have is with AMD touting support in their materials, adding said support, and then apparently removing it (not able to confirm that), and all the while not saying anything. That's a pretty d**k move, in my opinion.

Yes and no, PBO doesn't identify the limitations of the VRM configuration for the board in use, it simply increase the limits imposed on PB, it's similar to increasing the power and thermal targets on their GPUs.

Our little conversation that we've had going here has been a lot more understandable than anything AMD has spewed out into its slides. If only AMD would define these technologies clearly and concisely and then confusion wouldn't be a thing. To the average person, it sounds like Ryzen comes equipped with three technologies that do the same thing.

Anyway yes. Now that I've had a couple of hours away, what you're saying makes absolute sense. PBO isn't a separate technology, it's an extension of PB2. And both work in conjunction with XFR to determine those hardware limitations.

[...] if i had to guess though i think you maybe drawing some equivalences in the wrong places, PB1 and PB2 are basically the same thing just with algorithm changes (PB2 is finer grained) and the changes from XFR1 to 2 basically extended how many of the cores to include in the data set (XFR1 used 1-2 cores, XFR2 uses all of them).

Sure, PB1 and PB2 both achieve the same thing, but the way they work changes, and that's where I was drawing the equivalences with Intel's Turbo Boost.

PB1 works with defined frequency tables, which is how TB1 and TB2 work, while PB2 only defines a maximum turbo frequency (the one advertised by AMD). Then, as you say, the changes to XFR as well, with XFR1 being the equivalent of TBM3.

Though as @Potatowithearsontheside pointed out to me, Intel has a new turbo feature that will work similarly to PB2 with XFR2.

 

[d_brennen]

Worlds First AMD 3000 Review:

The AMD 3950x and 3900x are fast.

Conclusion:

3950x and 3900x on x570 motherboards are faster than all previous non HDET AMD processors.....

You didn't open with "Alreet guys" Unsubscribing

 

[chrcoluk]

https://translate.google.com/translate?sl=auto&tl=en&u=https://www.computerbase.de/2019-05/reale-cpu-test-intel-cinebench-co-benchmarkreferenz/

Had a read and its weird, they not really advocating geekbench, but want benchmarking to change to just analyse specific target use case.

So yeah intel doing whats best for their marketing, but they effectively (seem to) want geekbench and other cpu benchmarks to not be used either, cinebench is the one mentioned as its by far the most famous one used now days. This is assuming the linked to article is saying the truth and isnt speculation.

 

[Potatowithearsontheside]

I think that the switch to one-click auto-overclocking is probably good for the industry as a whole. Leaving something on the table has led to effective stagnation, with successive generations simply being clock uplifts that utilise what was previously left on the table. Now I believe that AMD/Intel have to produce genuinely superior products from one generation to the next. For sure, it takes the fun away for those that like to tinker, but it does seem like a natural progression in their approach.
The downside is that pushing technical boundaries tends to come with higher price points.
The new norm is out-of-the-box best performance at inflated prices.

 

[DylRicho]

I think that the switch to one-click auto-overclocking is probably good for the industry as a whole. Leaving something on the table has led to effective stagnation, with successive generations simply being clock uplifts that utilise what was previously left on the table. Now I believe that AMD/Intel have to produce genuinely superior products from one generation to the next. For sure, it takes the fun away for those that like to tinker, but it does seem like a natural progression in their approach.
The downside is that pushing technical boundaries tends to come with higher price points.
The new norm is out-of-the-box best performance at inflated prices.

I absolutely agree.

I think AMD's goal was to turn overclocking into more of a mainstream concept. Ryzen has unlocked multipliers across the board, and though not officially supported, the Athlons are also overclockable as many BIOS vendors decided to ignore AMD on that one.

That decision allowed more and more people to gain better performance for no cost, as there are a lot of people that wouldn't touch overclocking by themselves.

But also, let's be realistic here. The other reason for it was because AMD needed absolutely every last drop of that 14/12 nm process in order to have performance parity with Intel.

 

[chroniclard]

Messed around loads with 2700X, eventually just let it do its own thing as overclocking(in my case) wasnt really worth it.

 

[Phate]

For anyone interested a Portuguese e-tailer has prices up:

Ryzen 3 3200G - €108,90
Ryzen 5 3400G - €163,90
Ryzen 5 3600 - €218,90
Ryzen 5 3600X - €273,90
Ryzen 7 3700X - €361,90
Ryzen 7 3800X - €438,90
Ryzen 9 3900X - €548,90

 

[tamzzy]

so from buildzoid's latest video...
1.2v 150A apparently needed for OC 12c = 180w
1.2v 200A for OC 16c = 240w

 

[Zeed]

so from buildzoid's latest video...
1.2v 150A apparently needed for OC 12c = 180w
1.2v 200A for OC 16c = 240w

But is 1.2 enough for overclocking ?? Remember that screenshot from 16 core ?? 1.425 same az zen 1

https://www.guru3d.com/news-story/a...erforms-roughly-similar-to-core-i9-9900k.html

 

[Ross Thomson]

It’s one game. Why are people making such an issue out of it? 99% of potential buyers probably couldn’t care less about arma 3.

never even heard of it till this thread.

 

[MG42Maniac]

I'm not sure what to do, I've got 2 x 2700X with 2 x Gigabyte B450 Aorus M. I've upgraded one of the displays to a 3440x1440 100Hz so keen to eek some more FPS out. I'm tempted to stick a 3900X into one of them, would the mobo take it okay? I've got a H115i Pro so would be looking to get a bit of an OC on the go.

 

[Ross Thomson]

so from buildzoid's latest video...
1.2v 150A apparently needed for OC 12c = 180w
1.2v 200A for OC 16c = 240w

the overclocked 16 core showed 245w and 151A at a stupid voltage.

 

[maltrabob]

Although MegaTroll's review on page 704 was second to none I tried a bit of my own guess game so let me know what you think.

Looking at the first slide from AMD presentation above we see a 1T score of 483 points for R5 3600 (this should be on stock settings with automatic boosting). Knowing a 1T score we can get a nT score by using a MP Ratio. We do not know what it is going to be on Ryzen 3000 CPUs but from OcUK online user database for 6c/12t Ryzen 2600/2600X we can safely use a MP Ratio of 7.4x.

Therefore R5 3600 results are: 1T - 483 points, nT - 3570 points

----------

Now, looking at the second slide above we know that R5 3600 is faster in 1T by 9% when compared to R7 2700X at stock (automatic boost). From what can be found online at OcUK user database, we know that stock OC can be anything up to 4.35 GHz for the best chips with a proper cooling solution. There are also other factors like memory speeds and timings but for our purpose we will use a 1T score of 440 points (this is valid for 4.3 GHz stock OC and memory running at 3000 Mhz at C14). Looking and comparing further we can use a MP Ratio for 8c/16t Ryzen 2700X CPUs of 9.4x.

Therefore R7 2700X results are: 1T - 440 points, nT - 4100 points

Now that we know how much faster the rest of new Ryzens are when compared to R7 2700X in 1T we can calculate our results as follows:

R5 3600 +9% - Single 480, Multi 3500 (MP Ratio 7.4x)
R5 3600X +13% - Single 500, Multi 3700
R7 3700X +15% - Single 506, Multi 4750 (MP Ratio 9.4x)
R7 3800X +19% - Single 524, Multi 4925
R9 3900X +21% - Single 530, Multi 7700 (MP Ratio 14.5x derived from 12-core TR 1920X)

Those results are derived from AMD presentation slide depicting R5 3600 performance. If those numbers provided by AMD are to be trusted then my results shouldn't be far off. Also, using the result achieved by PBO Automatic Overclocking by AMD we could add another 4% of performance for a R5 3600 (483 points vs 504 points).

----------

There is also this slide so maybe someone else can cross-compare against my results:

All slides here: https://www.techpowerup.com/256660/...ecks-ryzen-3000-zen-2-radeon-rx-5000-navi-etc

 

[GAC]

I think that the switch to one-click auto-overclocking is probably good for the industry as a whole. Leaving something on the table has led to effective stagnation, with successive generations simply being clock uplifts that utilise what was previously left on the table. Now I believe that AMD/Intel have to produce genuinely superior products from one generation to the next. For sure, it takes the fun away for those that like to tinker, but it does seem like a natural progression in their approach.
The downside is that pushing technical boundaries tends to come with higher price points.
The new norm is out-of-the-box best performance at inflated prices.

well lets be blunt its amd doing this which will force intel in to following so they cant drip drip drip faster cpus like they have been doing claiming they are a new generation when its just the same thing tweaked, remember 9th gen is more like 6th gen and possibly older depending how pedantic you want to get over them, 14nm+++++ what a shambles.

as for higher prices probably from intel but saying that look at the prices with their current act of drip feeding 100mhz bumps which half the time include a new motherboard socket, so in the long run could be cheaper. something the people bleating on about ryzen 2 being too much, which they arnt but if you are a x370 or 400 series owner its a bargain compared to the rubbish intels been pulling the past few years.

 

[Orange Nexus]

I'm not sure what to do, I've got 2 x 2700X with 2 x Gigabyte B450 Aorus M. I've upgraded one of the displays to a 3440x1440 100Hz so keen to eek some more FPS out. I'm tempted to stick a 3900X into one of them, would the mobo take it okay? I've got a H115i Pro so would be looking to get a bit of an OC on the go.

Might not make a difference if you keep the same gpu.

 

[AStaley]

Are we there yet?

 

[pandem0nium]

I hope the new APUs work with Playready 3 DRM for netflix UHD. And UHD blu ray playback to a lesser extent.

 

[Zeed]

the overclocked 16 core showed 245w and 151A at a stupid voltage.

1.425 You call stupid ?? Thats my DAILY but 12nm hhahaah

 

[DragonQ]

Yeah it'll be interesting to see the recommended maximum voltage for safe daily use for 7nm.

 

[4K8KW10]

People assume cinebench is some kind of god as all their favourite reviewers use it, they use it not because its meaningful and accurate but because it showcases new products well as it really favours logical threading and higher cores. Thus showcasing both new amd and intel flagship products well.

Yes, CB represents only potential, peak, best-case performance, and is not necessarily representative of real-world applications performance.
Just cherry-picking in order to highlight in the best possible light and mislead the customers, who will accept the fake results as something worth their trust.