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

[Murphy]

Most of what you said here is wrong AND your own links prove it.

Click your Samsung and global forge strat colab link, it straight up says this

I had a suspicion someone would point that out but i honestly couldn't be bothered to go back further in time when this sort of stuff is publicly available in the first place and mentioning what I'm about doesn't change the original point that 12nm was not a mature phone/tablet CPU node. If you want to get further into the origins of the non-Intel 14nm node then try using Google, I'll even start you off with this article that tell you how...

Ever since GlobalFoundries was spun off from AMD in 2009, Samsung, IBM, and GF have maintained what they called the “Common Platform.” Common Platform was an agreement on broad technology standards that would be deployed across all three companies. It was never entirely clear which technologies were governed by CP rules and which were independently developed, but the goal was to create an infrastructure that allowed for easy design porting across all three corporations.

Now how much of the 14 and 12nm that was eventually used in Ryzen came from each company in that common platform agreement, the strategic collaboration, the buying up of IBM's plant, process, and numerous semiconductor technology IPs is, i would say, subjective as that information is not publicly available for obvious reasons, we can only speculate.

Full stop, that's the end of it. Anyone not bs'ing their way through basic knowledge knows that IBMs 14nm was decent but had very different design parameters, it was designed primarily for IBM chips, that is screw power efficiency, max clock speeds were the only concern, it was more expensive and was focused on producing a relatively low number of extremely high cost server chips. It was basically unsuitable for creating more efficient higher yield cheaper chips. So straight off, and this was incredibly widely and publicly known, Global dumped IBM 14nm and licenced what was a fully Samsung process node. The original Zen was made on a purely Samsung node, developed by Samsung for mobile with the focus on efficiency and being high yielding and it went into production before Global did and it was licenced by Global. Suggesting this was originally an IBM node literally immediately highlights that you don't know what you're talking about here. Here's a hint, if one company licences a node to another company... they won't launch it before the company who actually finished it then licenced it.

Firstly I'm not bs'ing my way through with basic knowledge, secondly I'd appreciate if you could keep the ad homeni attacks to yourself.

Thirdly I'm fully aware of the characteristics of IBM's 14nm process but diverging into a long explanation as you've done above is completely irrelevant to the history of where the 14 and 12nm that's used in Zen and Zen+ came from, it was suffice to say IMO that IBM had the first working 14nm FinFet outside of Intel way back in 2014.

If however you want to get into a discussion on how much of that FinFet on a SOI substrate ended up in Zen and Zen+ I'm more than happy to have that discussion but perhaps a thread about Zen2 isn't the place to have it so feel free to start another thread.

Secondly, the defining characteristics on size and electrical performance come from in huge part the basic parameters of a node, the basic gate pitch, metal pitch, they pretty much define the 14nm Samsung node as what it is, 12nm makes no significant changes and ignoring the headlining numbers, 12nm is 'faster' not because it's denser, but because slightly thinner taller gates allow the chip to be less dense.

Keep in mind that Intel relaxed transistor gate pitch for subsequent 14nm nodes, that's how the clock speeds went up. 14nm had a transistor gate pitch of 70nm, 14nm + didn't officially raise it but supposedly did in the design rules, 14nm++ is a transistor gate pitch of 84nm.

Intel called all their node tweaks 14nm/+/++/+++, in the past they would have called them a new stepping on a shipping chip. With ostensibly the same cpu they wanted to make the changes sound bigger than just a stepping. With Global they wanted it to sound like a bigger improvement and they gave it a new name. It used the same equipment, it used the same patterning, it has the same physical limits, it just tweaked the design, nothing more or less. On any given set of equipment and choice of patterning you are really limited to certain physical size parameters, within those limits anyone can choose to optimise for density, voltage, performance, in fact many or most nodes do that upfront and have different designs on purpose, 14nm industry nodes were really the first where they only really bothered with one design largely because Samsung planned to go 10nm before 7nm and earlier so wasting time on extra versions of 14nm apparently didn't seem worth it to them.

https://en.wikichip.org/wiki/14_nm_lithography_process

Look at Intel, they started off, before launch, with an optimised high performance, low voltage and high density library. 12nm is nothing more than a tweaked 14nm with different design goals, using the same equipment, using the same physical limitations, using the same patterning. No one in the industry would call it a new node.

Also note the IBM 14nm, it is dramatically larger despite being 14nm, because lower density = higher performance in general. This is why Intel's nodes decreased density to raise clock speeds, IBMs node was completely unsuitable for a high volume, high yield, lower cost chip and still is, also a really really big deal here, IBMs 14nm is SOI, Samsung's is bulk, no Zen chip is produced on SOI, full stop.

https://www.globalfoundries.com/sites/default/files/product-briefs/pb-14lpp.pdf

You will notice here, particularly top of the second page, twin well CMOS bulk FinFET, with every single design parameter matching Samsung 14nm and nothing matching IBMs 14nm SOI node.

Jesus H, and you accused me of bs'ing my way through with basic knowledge, that's not even worth a response.

Although the "defining characteristics" part did give me a good laugh.

And what's with the rant about Intel got to do with anything?

 

[beany_bot]

Urgh I hate coming in for for Ryzen info and just stumbling into a bar Brawl. I don't care who is right and who is wrong. But respect to Humbug for (at least trying) to end this and let it go. Murphy you would do well to do the same.

 

[boomish]

Urgh I hate coming in for for Ryzen info and just stumbling into a bar Brawl. I don't care who is right and who is wrong. But respect to Humbug for (at least trying) to end this and let it go. Murphy you would do well to do the same.

Well said, please let it go, I’m here to read up on all the info and it’s quite sad to see, I’m sure if you both agreed to disagree and move on to something you both agree on..I’m looking at getting an AMD again for the first time in many years and it’s so exciting. Just reading this I’m thinking of getting a cheap 2600 build till the new chips come out My 3570 is really showing it’s age and the Vega 56 I got from OCK has been awesome.

 

[Murphy]

Urgh I hate coming in for for Ryzen info and just stumbling into a bar Brawl. I don't care who is right and who is wrong. But respect to Humbug for (at least trying) to end this and let it go. Murphy you would do well to do the same.

Yea apologies for that, i had let it go after what humbug said but then drunkenmaster wadded in, in my defense I've been trying to keep it related to Zen2 by explaining why IMO 5Ghz is wishful thinking and that Zen+ only achieved a 200Mhz increase despite the process being turned for performance, that Zen+ wasn't simply a mature phone/tablet CPU node...

Besides there's not really going to be much Zen2 related news for a few weeks yet so we've gotta talk about something.

 

[beany_bot]

Well said, please let it go, I’m here to read up on all the info and it’s quite sad to see, I’m sure if you both agreed to disagree and move on to something you both agree on..I’m looking at getting an AMD again for the first time in many years and it’s so exciting. Just reading this I’m thinking of getting a cheap 2600 build till the new chips come out My 3570 is really showing it’s age and the Vega 56 I got from OCK has been awesome.

I saw recently R5 1600 going for $110 new. That is amazing value.

 

[michty_me]

I'm considering getting myself a Ryzen 3550h laptop with RX560X graphics for using when I'm away at work to play some older games.
Seems a good value although noticed the Intel 8300h with a gtx1050 is the same prices now so it's a bit of a toss up.

 

[Wucked]

Getting excited for this now, my trusty 4770k has served me well but it’s time for an upgrade. And I haven’t had AMD in my rig since the glory days of the first Athlon chips. People seem to be blasé about 8 core now, but for me it’s going to be a big deal.

 

[orbitalwalsh]

23rd April came and has nearly gone.
Where are the leaks?

theres been plenty - just have to look hard



look forward to seeing the old mans stock told him to put a hold pensions worth into AMD before Ryzen 1*** details started leaking out. bit of a risk as the enigeering samples were s***

looking at the above 12 and 14nm talk. regardless of how its done etc, can see 300/400hz increase with 1700 to 2700 all core overclock , Ryzen 3000 APU 12nm+ has 300hz increase in overclock over 2000 APU 14nm ... so hopefully on a loose trend should see Ryzen 3000 hitting 5-5.3ghz all core overclock .

which would be lush and see why VRMs have been beefed up from x470/z390 as its the overclocking on 8 cores plus regardless of what node it is will push TDP quite high ! to keep costs down should see more cheaper mosfets being used - pretty much like Gigabyte UD and z390 Aorus boards ..

and I loaf for either team that can get the old mans pension fund increasing - also got him to drop a second pension pot on 9th gens intel before z390 boards and 8 cores came on the news..

 

[thebandit]

Getting excited for this now, my trusty 4770k has served me well but it’s time for an upgrade. And I haven’t had AMD in my rig since the glory days of the first Athlon chips. People seem to be blasé about 8 core now, but for me it’s going to be a big deal.

Lol, I'm going to do my first upgrade for years when Ryzen 3xxx is released, should represent a decent jump from my current intel i5 750, my son is literally asking everyday when these CPU's will be released.... he can't wait.

 

[humbug]

Yea apologies for that, i had let it go after what humbug said but then drunkenmaster wadded in, in my defense I've been trying to keep it related to Zen2 by explaining why IMO 5Ghz is wishful thinking and that Zen+ only achieved a 200Mhz increase despite the process being turned for performance, that Zen+ wasn't simply a mature phone/tablet CPU node...

Besides there's not really going to be much Zen2 related news for a few weeks yet so we've gotta talk about something.

Just to be clear, the vast majority, including me have already agreed 5Ghz Zen 2 is very unlikely and have done for some time.

Zen+ was never promised to be a high clocking god chip, AMD promised a 10% overall up lift in performance and that's what we got, a mixture of 3% IPC and 7% higher clocks.

 

[beany_bot]

Lol, I'm going to do my first upgrade for years when Ryzen 3xxx is released, should represent a decent jump from my current intel i5 750, my son is literally asking everyday when these CPU's will be released.... he can't wait.

There really are a lot of us! I hope there isnt a stock shortage at launch. But I suspect with their claimed 70% yields it will be OK.

 

[phrases]

Urgh I hate coming in for for Ryzen info and just stumbling into a bar Brawl. I don't care who is right and who is wrong. But respect to Humbug for (at least trying) to end this and let it go. Murphy you would do well to do the same.

Same i just sifted through about 10 pages just to try to see if anybody picked up some information along the way. Filter mode needed or at least longer pages 310 is a lot to get through.

Anyone know when the first media release/benchmark is going to be?

 

[humbug]

Same i just sifted through about 10 pages just to try to see if anybody picked up some information along the way. Filter mode needed or at least longer pages 310 is a lot to get through.

Anyone know when the first media release/benchmark is going to be?

The thinking is they are due out in June, so sometime around then.

 

[Murphy]

Just to be clear, the vast majority, including me have already agreed 5Ghz Zen 2 is very unlikely and have done for some time.

... so hopefully on a loose trend should see Ryzen 3000 hitting 5-5.3ghz all core overclock.

BTW that's not doubting that you're not expecting that, it's just me being a bit of a dick by pointing out there's plenty that are expecting 5Ghz.

 

[EsaT]

Yea apologies for that, i had let it go after what humbug said but then drunkenmaster wadded in, in my defense I've been trying to keep it related to Zen2 by explaining why IMO 5Ghz is wishful thinking and that Zen+ only achieved a 200Mhz increase despite the process being turned for performance, that Zen+ wasn't simply a mature phone/tablet CPU node...

Are you telling that Intel's 5GHz level boost clocks are imaginary?
Or is Intel somehow exempt from your logic?

IBM's SOI tech isn't used anywhere near Ryzens.

14nm High Performance (14HP) process has been optimized for different vectors than the low-power-optimized processes developed by companies such as Samsung and TSMC. In order to avoid any confusion, we want to point out that 14HP is very different from GlobalFoundries standard bulk CMOS process (e.g., 14LPP) used by products such as AMD’s Zen.
https://fuse.wikichip.org/news/956/globalfoundries-14hp-process-a-marriage-of-two-technologies/​

And already GloFo's own ad lines of "12nm" not needing new silicon level chip design work tells it's not really any new node, but just normal tweak.
Very limited clock speed increase shows it being shackled to same basic limitations.
While voltage dropped little, scaling of clocks per voltage is still crappy already before 4GHz:
https://www.gamersnexus.net/guides/3290-exponential-ryzen-voltage-frequency-curve

Like ~2,5% clock increase from 3,8 to 3,9GHz needing ~5% more volts.
And already ~7,5% more volts for next 2,5% clock increase!
That's signature of chips being pushed past what node was designed to do.


I'm not expecting any more real than Intel near 5GHz all core clocks.
(meaning if those are achieved power consumption/heat output is excessive)
But near 5GHz level boost clocks for few cores from actually designed for high performance node should be realistic.
Especially when AMD has more granular clock boosting system than Intel.
No doubt developed partially because of knowledge of very uncertain availability of high performance node with competitive clocks.


Can't avoid curious observation how in forums in different language speaking countries there are relatively new user accounts, whose proclamations are combined by single common theme:
That AMD can't ever catch Intel...
Bad upgradeability doesn't matter...
Intel's price levels are justified regardless of bad performance level retaining prospects...

 

[Potatowithearsontheside]

As long as the base clock is around 4.2GHz on 8c+ then it'll beat a 9900K running 24/7 at its TDP; 9900K clocks down to 4.2GHz itself when out of turbo (at 100% load)
Of course, at the bleeding edge it'll need to beat an overclocked 9900K for the die-hards to take note.

 

[beany_bot]

We already saw an 8 core Zen 2 (not at final clocks) beating a 9900K in cinebench. Now, yes, this isnt in games. but can we assume that as everything else is equal that Zen 2 would beat the 9900k in games if it's beating it core for core in cinebench?

 

[LePhuronn]

We already saw an 8 core Zen 2 (not at final clocks) beating a 9900K in cinebench. Now, yes, this isnt in games. but can we assume that as everything else is equal that Zen 2 would beat the 9900k in games if it's beating it core for core in cinebench?

Careful, you'll get DG frothing at the mouth again with asking that.

But despite his apparent rabies, he is correct in stating that Cinebench is not any real indicator of gaming performance because they're completely different workloads. Cinebench will eat up every core and thread you have available and schedule the work perfectly, games won't do the same so it'll still come down to per-core performance, and of course how the I/O die + chiplet design of Zen 2 adds latency and what the impact would be.

I think it's safe to say though that Zen 2 will have at least significantly reduced the per-core performance gap to Intel once again, perhaps even matched it now, at a good chunk less power. That bodes well for gaming.

 

[pandem0nium]

Yeah, that's a good point for anyone reading this, the Second M.2 slot at the bottom of the board is for SATA (6GB/s) Drives, you can't put an NVMe drive in there, that's the slot just under the CPU, if you want to raid 2 NVMe drives you need an X470.

However, that means you can only run your GPU at 8x PCIE 3.0 as any boards that offer two M.2 NVMe slots has to take lanes from the 16x GPU, still not as bad as some Intel configs.

@Journey Not quite. Most boards actually use 4 pcie gen 2 lanes from the chipset for the second nvme slot. It's only Asus who use CPU lanes for secondary m.2 slots.

@humbug It's doable on the B chipset boards as well, since the CPU lanes can be split to x8 x4 on those. The Asus B450-F for example has 2 full speed m.2 slots with the usual trade-off of dropping the GPU lanes down to X8.

Intel, yeah, NVMe raid is a complete no-go on the mainstream boards.

 

[Journey]

Not quite. Most boards actually use 4 pcie gen 2 lanes from the chipset for the second nvme slot. It's only Asus who use CPU lanes for secondary m.2 slots.

My comment was with regards to RAID NVMe M.2 as mentioned, so using a PCIE 2.0 slot would be utterly pointless unless you are using low end drives that are much slower, and I am well aware of who uses what, ta.

In other news, TR 2950X are coming down ~£630 to your door.