Tuesday, July 7th 2020

"Zen 3" is On Track and Launching Later This Year: AMD CEO

In a video message posted on her Twitter timeline, AMD CEO Dr Lisa Su confirmed that the company's next-generation "Zen 3" microarchitecture is coming out "later this year." Speaking in context of 7/7 (a year since AMD debuted high-performance CPU- and GPU- architectures on the same day, leveraging 7 nm), and the Ryzen 3000XT series processor announcement, Dr Su stated "As you know with Ryzen, we're always on a journey, a journey to push the highest performance that we can for our users and our fans. So Zen 3 is exactly that. Zen 3 is looking great in the labs, we're on track to launch later this year, and I can't wait to tell you more about it." Watch the video in the source link below.
Source: Dr Lisa Su (Twitter)
Add your own comment

55 Comments on "Zen 3" is On Track and Launching Later This Year: AMD CEO

#1
mtcn77
Dr Lisa Su
I can't wait to tell you more about it.
Please do...
Posted on Reply
#2
steve360
Someone named their kid Ryzen...WTF?

Parents of this generation are clearly on brain killing drugs.

I feel sorry for that one kid in class who has a normal name.
Posted on Reply
#3
ARF
Where is Navi second generation which you much more desperately need than anything else ? ?
steve360
Someone named their kid Ryzen...WTF?

Parents of this generation are clearly on brain killing drugs.

I feel sorry for that one kid in class who has a normal name.
Yeah, and Zen is not even the right word from Japanese and Chinese.
It is Chen or Dzen.
Posted on Reply
#4
_Flare
and AMD has very early Zen4 in the Labs also. used TSMC N5 risk-process
Posted on Reply
#5
phanbuey
my body is ready

my wallet though... that's a different story
Posted on Reply
#6
mtcn77
_Flare
and AMD has very early Zen4 in the Labs also. used TSMC N5 risk-process

Does anybody have information on AMD's business relations? It seems AMD gets a good reception. Like a golden partner, or something.
I see no other reason why some more pronounced partners cannot lay waste to TSMC's output which might have happened with the 6nm thingy.
AMD both had access to DUV 7nm and EUV optimised DUV 7nm(6nm) first. Now this is happening all over again for optimised 7nm EUV(5nm)...
Now that is a precedent - AMD is picking high yield output nodes. How TSMC is letting them is beyond me.
Posted on Reply
#7
_Flare
N5 is the real successor of N7, with N7+ as an intermediate process
but N7P and its successor N6 are kind of cheaper upgrade path with very much common design rules
AMD did plan with the N7 --> N5 hard jump from the beginning
N5 entered risk production in march 2019, so last year
i think we will see Desktop Zen4 (YES Zen4) maybe 7-7-2021 because i think N5 high-perf will enter volume production end 2020 or early 2021

klick on the info tabs, for more info
www.tsmc.com/english/dedicatedFoundry/technology/logic.htm
Posted on Reply
#8
ARF
_Flare
N5 is the real successor of N7, with N7+ as an intermediate process
but N7P and its successor N6 are kind of cheaper upgrade path with very much common design rules
AMD did plan with the N7 --> N5 hard jump from the beginning
N5 entered risk production in march 2019, so last year
i think we will see Desktop Zen4 (YES Zen4) maybe 7-7-2021 because i think N5 high-perf will enter volume production end 2020 or early 2021

klick on the info tabs, for more info
www.tsmc.com/english/dedicatedFoundry/technology/logic.htm
Isn't 5-5-2021 a better date?
Posted on Reply
#9
_Flare
ARF
Isn't 5-5-2021 a better date?
you are absolutely right, i am very curious how early will launch Zen3, because the gap is even smaller, maybe Sept/Oct somewhere

no clue regarding the CCX size, but Zen3 ups it from 4 to 8 so maybe there is no need for a 12-core CCX with Zen4 ... who knows
Posted on Reply
#10
cucker tarlson
steve360
Someone named their kid Ryzen...WTF?

Parents of this generation are clearly on brain killing drugs.

I feel sorry for that one kid in class who has a normal name.
he has no daate of birth,he has a stepping.
and I hear his parents are anti-microcode updaters
Posted on Reply
#11
theoneandonlymrk
mtcn77
Does anybody have information on AMD's business relations? It seems AMD gets a good reception. Like a golden partner, or something.
I see no other reason why some more pronounced partners cannot lay waste to TSMC's output which might have happened with the 6nm thingy.
AMD both had access to DUV 7nm and EUV optimised DUV 7nm(6nm) first. Now this is happening all over again for optimised 7nm EUV(5nm)...
Now that is a precedent - AMD is picking high yield output nodes. How TSMC is letting them is beyond me.
Yes how dare Tsmc earn money, I doubt there's anything other than business involved , Intel are direct competition and Nvidia tried to goose them so AMD , Apple and Qualcomm it is.

End of 2021 for consumers IMHO because these chips are destined for a lot of servers and supercomputer first.(long day meant zen 4 here not 3)

Big push style, I'm worried for Intel.
Posted on Reply
#13
yeeeeman
Last year we got new parts, this year we get 100mhz.
Posted on Reply
#14
AnarchoPrimitiv
yeeeeman
Last year we got new parts, this year we get 100mhz.
Did you not read the article? Entitlement....
Posted on Reply
#15
Tomgang
My body is ready for Ryzen 9 4950X or what it will be called.

My bank account not so much. This coronavirus has cost me a good lump of my savings. But I hope i can manage to save up by then zen 3 comes out. So I am ready economic as well. It's time to put X58 to rest after over 11 years on it, I need more horsepower under bonnet and other new features.
Posted on Reply
#16
Mats
Yet another thread with clueless posts about how next-next generation will show up in no time at all, preferably guessing by superstition.

Developing a new CPU cost money. Lots of it.
Selling CPU's makes money.
In order to break even, AMD has to sell said CPU for a certain amount of time before moving on.
Judging by AMD's history, anything less than 12 months between launches isn't feasible.

The number of days between the three Ryzen desktop launches (not APU's) so far has been 400+ days, and I don't see any reason for AMD to speed this up given how well the current models are selling.
Posted on Reply
#17
mcraygsx
steve360
Someone named their kid Ryzen...WTF?

Parents of this generation are clearly on brain killing drugs.

I feel sorry for that one kid in class who has a normal name.
I do not find anything wrong with how people choose to name their kids.
I have twin Cousin's named Zen and Zane. People should not judge how others choose to live their lives.
Posted on Reply
#18
RandallFlagg
steve360
Someone named their kid Ryzen...WTF?

Parents of this generation are clearly on brain killing drugs.

I feel sorry for that one kid in class who has a normal name.
Yeah, would be kinda like if someone named their kids Atari and Coleco back in the 80s.
Posted on Reply
#19
ARF
Mats
Yet another thread with clueless posts about how next-next generation will show up in no time at all, preferably guessing by superstition.

Developing a new CPU cost money. Lots of it.
Selling CPU's makes money.
In order to break even, AMD has to sell said CPU for a certain amount of time before moving on.
Judging by AMD's history, anything less than 12 months between launches isn't feasible.

The number of days between the three Ryzen desktop launches (not APU's) so far has been 400+ days, and I don't see any reason for AMD to speed this up given how well the current models are selling.
I'd agree that the first version of Zen cost money. The latter versions, including Zen 4 are just evolutionary improvements, and mostly just not so significant optimisations of the slow parts in the design.
They don't cost money except the salaries and the masks, which I am quite sure AMD returns in no time.

Here, you can see an interview with Mark Papermaster who with his own words just confirms the AMD's "milking cows" strategy which I hinted above:
MP: I didn’t see the particular interview you’re referring to, but what I will say is that we’re not on a tick-tock model. What we’re doing is looking at each generation of CPU and marrying the best process variant that’s out there with the right set of IPC improvements, memory hierarchy, and all the things that we can put in there. We are committed to staying on the best possible pace of improvements each generation that we can. This is a formula that’s working well for us at AMD.


MP: There’s a natural cadence to where we are, and we have multiple design/implementation teams in play. As one team completes, they move onto the next project. You know I think it’s the right rate for us, because in some markets you have an increased expectation of a new product every year, like smartphones. For that market you have to catch the holiday season every year with pretty much a new core, and to a large extent they can do that. But that need to tie things off at the same time every year means that there might be performance still left on the table for that core, and that’s going to happen every generation. So we believe we’re on the right track to deliver for our high performance customers, and that’s what we’ve based AMD’s resurgence on.
www.anandtech.com/show/15268/an-interview-with-amds-cto-mark-papermaster-theres-more-room-at-the-top

Simply said - they know a lot of improvements but decide on political bases which version what improvement will get.
Posted on Reply
#20
Imsochobo
mtcn77
Does anybody have information on AMD's business relations? It seems AMD gets a good reception. Like a golden partner, or something.
I see no other reason why some more pronounced partners cannot lay waste to TSMC's output which might have happened with the 6nm thingy.
AMD both had access to DUV 7nm and EUV optimised DUV 7nm(6nm) first. Now this is happening all over again for optimised 7nm EUV(5nm)...
Now that is a precedent - AMD is picking high yield output nodes. How TSMC is letting them is beyond me.
Yes,

While my sources are thin but they generally say.
amd is the customer when it comes to the consoles, AMD and not sony.
Amd is a guaranteed revenue source on 7nm for Many many years to come, they were a source for many many years prior with Xbox One and Ps4, they were with ATI.
Many of the nodes are co-developed with AMD and AMD patents.
So yes, the partnership is deep.

Furthermore why wouldn't TSMC lets say.. give nvidia capacity over AMD ?
Also, having a strong AMD ensures they get a cut of the pie which Intel had a monopoly on, only AMD can access that market.
If they give Intel it over AMD they will squeeze AMD and Intel will go back to their own manufacturing and ditch TSMC.

So yes it's history partnership with ATI in particular, it's technological partnership and strategic.
Posted on Reply
#21
mtcn77
Imsochobo
Furthermore why wouldn't TSMC lets say.. give nvidia capacity over AMD ?
Cause Nvidia went with Samsung?
Posted on Reply
#22
Tom Yum
ARF
I'd agree that the first version of Zen cost money. The latter versions, including Zen 4 are just evolutionary improvements, and mostly just not so significant optimisations of the slow parts in the design.
They don't cost money except the salaries and the masks, which I am quite sure AMD returns in no time.

Here, you can see an interview with Mark Papermaster who with his own words just confirms the AMD's "milking cows" strategy which I hinted above:



www.anandtech.com/show/15268/an-interview-with-amds-cto-mark-papermaster-theres-more-room-at-the-top

Simply said - they know a lot of improvements but decide on political bases which version what improvement will get.
You think that even the yearly updates don't cost valuable R&D dollars? Those optimisations aren't just sitting there waiting to be 'turned on' when the yearly update comes around. Bottlenecks have to be analysed, solutions found and then reanalyzed to check what the second and third order effects of the change are to catch any reversions. Papermaster's comments are more about the need at some point to lock the design lest they suffer the Osborne effect. That always means some optimisations are left out, either because they aren't ready to commit to a configuration baseline or the 2nd/3rd order effects are getting worked out. Given AMD is still 'chasing' Intel in certain performance areas, it would be extremely unlikely that AMD are sitting on optimisations that could resolve that performance shortfall just to milk existing designs.

CPU's and architectures are incredibly complex systems, and you can never produce the optimal solution on the first past. That's why an architecture will typically only be redeveloped every 3-4 generations, with every other update being an optimisation (at the architectural level, not at the silicon level).
Posted on Reply
#23
mtcn77
Tom Yum
Papermaster's comments are more about the need at some point to lock the design lest they suffer the Osborne effect. That always means some optimisations are left out
Oh, I know it! Like the time they left Norman Osborne out to dry from the board of directors! He did not take it lightly, to say the least...
Don't ever leave out your optimisations, copy exactly!
Posted on Reply
#24
Mats
Tom Yum
You think that even the yearly updates don't cost valuable R&D dollars?
Oh don't bother, you're fighting dreams, gossip, wishful thinking, and lucky numbers, etc. :D You have my sympathy for trying, tho.
Posted on Reply
#25
mtcn77
Mats
Oh don't bother, you're fighting dreams, gossip, wishful thinking, and lucky numbers, etc. :D You have my sympathy for trying, tho.
I'm sure you know much of the EUV scaling has been on the backing of laser mask validation tools newly available to the market. Sources say the japanese company is booked for literally years at this time.
I haven't seen such straight-faced announcements, maybe you have more experience.
Posted on Reply
Add your own comment