Thursday, July 26th 2018

No 16-core AMD Ryzen AM4 Until After 7nm EPYC Launch (2019)

AMD in its Q2-2018 investors conference call dropped more hints at when it plans to launch its 3rd generation Ryzen processors, based on its "Zen2" architecture. CEO Lisa Su stated in the Q&A session that rollout of 7 nm Ryzen processors will only follow that of 7 nm EPYC (unlike 1st generation Ryzen preceding 1st generation EPYC). What this effectively means is that the fabled 16-core die with 8 cores per CCX won't make it to the desktop platform any time soon (at least not in the next three quarters, certainly not within 2018).

AMD CEO touched upon the development of the company's 7 nm "Rome" silicon, which will be at the heart of the company's 2nd generation EPYC processor family. 2nd generation EPYC, as you'd recall from our older article, is based on 7 nm "Zen2" architecture, and not 12 nm "Zen+." 3rd generation Ryzen is expected to be based on "Zen2." As of now, the company is said to have completed tape-out of "Rome," and is sending samples out to its industry partners for further testing and validation. The first EPYC products based on this will begin rolling out in 2019. The 7 nm process is also being used for a new "Vega" based GPU, which has taped out, and will see its first enterprise-segment product launch within 2018.
With "Zen 2" based client-segment product being slated for 2019, there doesn't appear to be any immediate counter from AMD to Intel's 14 nm 8-core "Whiskey Lake" silicon; beyond maybe launching existing 12 nm "Pinnacle Ridge" based products with higher clocks, or lowering prices. The company will be launching its 2nd generation Ryzen Threadripper processors within 2018, but those are HEDT products not designed to compete with "Whiskey Lake." Any interim new 12 nm "Zen+" silicon with >8 cores would go against what Dr. Su stated in her Q&A, and is hence unlikely. "Rome" was the only CPU development mentioned by Dr. Su. Source: AMD
Add your own comment

29 Comments on No 16-core AMD Ryzen AM4 Until After 7nm EPYC Launch (2019)

#1
R0H1T
The 12 core AM4 chip is also rumored to be a part of the lineup, could be 6 core CCX or 6 (enabled) core per CCX, with 2 cores disabled for better yield? The 16 core chip will not be released anytime soon, not unless AMD absolutely have to. I'm personally more curious about TR3 or TR4, can AMD push 64c/128t in HEDT o_O
Posted on Reply
#2
Gungar
R0H1T said:
The 12 core AM4 chip is also rumored to be a part of the lineup, could be 6 core CCX or 6 (enabled) core per CCX, with 2 cores disabled for better yield? The 16 core chip will not be released anytime soon, not unless AMD absolutely have to. I'm personally more curious about TR3 or TR4, can AMD push 64c/128t in HEDT o_O
Dont expect 12 cores if there isn't 16 cores available.
Posted on Reply
#3
gmn 17
Here’s hoping for a ryzen 2800x in the mean time or some price cuts
Posted on Reply
#4
R0H1T
Gungar said:
Dont expect 12 cores if there isn't 16 cores available.
What do you mean? I doubt they'll go straight from 8 cores, mainstream, to 16 cores.
AMD's biggest USP is (more) cores, not the only one but the most obvious. So there will definitely be a 12 core part released before 16 one, IMO anyway.
So if they're gonna sell Ryzen & it's derivatives over the next half a decade, then it makes sense to pace the increase in core count to a more sustainable level.
Posted on Reply
#5
kings
No surprise, actually. 16 cores on a mainstream platform does not make much sense for now.

The vast majority of people is better served with CPUs with less cores and more aggressive clocks.

For those who need lots of cores for very specific tasks, that's what HEDT is for.
Posted on Reply
#6
Joss
there doesn't appear to be any immediate counter from AMD to Intel's 14 nm 8-core "Whiskey Lake" silicon
I find R5 2600X and R7 2700X to be perfect "counters" to whatever Intel is bringing, maybe an eventual 2800X.
This "Intel wins" because overall performance is x.1111% better is becoming nauseating.
Posted on Reply
#7
Hugh Mungus
kings said:
No surprise, actually. 16 cores on a mainstream platform does not make much sense for now.

The vast majority of people is better served with CPUs with less cores and more aggressive clocks.

For those who need lots of cores for very specific tasks, that's what HEDT is for.
My thoughts exactly. In fact, 8-core mainstream Intel is more exciting to me than 16-cores mainstream AMD! 12-cores could be nice though if single-core performance doesn't suffer. Would likely cost more though, like 400-500 euros or roughly as much as the 9700k and 9900k depending on version.

I might actually stick with Intel on this one. :fear:

Definitely'll be swapping the 1180 for an AMD equivelant if that arrives though, and if AMD can improve per-core performance enough AND increase corecount, I might have a team red pc before the end of next year!
Posted on Reply
#8
londiste
Why do we think there might be 16 cores on AM4?
If that 8-core CCX is true, AMD will more than likely just reap benefits from lower manufacturing costs as well as avoiding all the cross-CCX latency issues.

Joss said:
I find R5 2600X and R7 2700X to be perfect "counters" to whatever Intel is bringing, maybe an eventual 2800X.
This "Intel wins" because overall performance is x.1111% better is becoming nauseating.
Cost is where AMD currently wins. 2600X is not doing very well against 8700K.
Posted on Reply
#9
Vya Domus
londiste said:
Why do we think there might be 16 cores on AM4?
The density leap will make 16 core parts just as easy and cheap to manufacture as the current 8 core parts. It will simply be a matter of choice on their part if they think it's worth doing that against what the competitions has.
Posted on Reply
#10
Hugh Mungus
londiste said:
Why do we think there might be 16 cores on AM4?
If that 8-core CCX is true, AMD will more than likely just reap benefits from lower manufacturing costs as well as avoiding all the cross-CCX latency issues.

Cost is where AMD currently wins. 2600X is not doing very well against 8700K.
One 8-core cluster could certainly be better for the average consumer than two of 6/8. Threadripper and Epyc will get a corecount increase for sure though, which will help differentiate HEDT and consumer lineups again!
Posted on Reply
#11
Joss
londiste said:
If that 8-core CCX is true, AMD will more than likely just reap benefits from lower manufacturing costs as well as avoiding all the cross-CCX latency issues
Well reasoned; a single CCX 8-core may well make more sense (on mainstream platforms) than 2x8-core CCXs.
Posted on Reply
#12
R0H1T
Joss said:
Well reasoned; a single CCX 8-core may well make more sense (on mainstream platforms) than 2x8-core CCXs.
The single CCX designs will go to the APU, with Navi(?) based IGP.
Posted on Reply
#13
Joss
londiste said:
2600X is not doing very well against 8700K
If you're talking performance I think it does well indeed.
And if we consider price then there's no match.
Posted on Reply
#14
londiste
Joss said:
If you're talking performance I think it does well indeed.
And if we consider price then there's no match.
I know we all assume Intel continues to be greedy and price technically similar CPUs way higher. But like I said cost is where AMD currently wins. Technically, both these CPUs are very close.
2600X is 6c/12t @ 3.6/4.2
8700K is 6c/12t @ 3.7/4.7
Depends on your definition of well, I guess.
Even in best case scenario tests for Ryzen like Cinebench R15, 8700K consistently does 5+% better. More in most other productivity tests.
Posted on Reply
#15
GorbazTheDragon
I think intel can make a killing off their 8 core chips providing the TIM and the motherboards are not barriers to pushing around 5GHz on AIO cooling.

They should have a substantial MT advantage (even a 8c8t will) in overall performance and at more modest clocks the efficiency differences are insignificant to most users (desktop power efficiency is not as critical as server).

Intel's current IMC has a good track record while AMD's struggles to get much further than 3000MHz so intel has plenty of headroom in the memory bandwidth.

AMD also will need to start offering significantly better ST performance to really get a stronghold on the desktop market, and the current indications hint that they will at best be closing the gap to in the order of 10%.

AMD can slash prices all they want but the fact that intel will continue to hold the best overall desktop performance means AMD is cut out of what intel has shown to be a very profitable product bracket.

That said, for now, intel has to get whatever sales they can in the desktop market because that is the only place where their current architecture has a real advantage. They are getting destroyed on the server side by TR and AMD is only going to build a bigger lead as it stands.
Posted on Reply
#16
ppn
8/t/8/c should become mainstream, at 199$. ASAP
Posted on Reply
#17
MDDB
Just as until now, with Zen 1, AMD had 2 dies (one with 2 CCXx with 4 cores each used throughout server, HEDT and consumer segments, and another with 1 CCX and a iGPU), it would make sense in terms of costs to keep 2 dies for everything with Zen 2: one with 2 CCXs with 8 cores each to reach 64c/128t at the server level, and another with just one 8 cores CCX and an iGPU.

For the time being, topping at 8 cores on the desktop seems to be perfectly fine for all but hardcore users, maybe 99% of desktop users (not just us enthusiasts, but everybody). Adding an iGPU to it would make the proposition great for businesses that just need a working computer, no add-in cards needed (most office PCs are like that, very few workers need a powerfull GPU). It's what AMD needs to enter the volume market. At 7 nm, they should be able to cram 8 cores in a CCX (as the rumours circulating imply) plus a more than decent iGPU.

Then they would have the second die, with two 8 cores CCXs: that would allow the desktop to reach 16 cores on the AM4 socket, although without iGPU, the HEDT (Threadripper) to get to 32 cores (or even 64 if they use 4 working dies as is rumoured already for TR2, maybe limiting other features to segment the products and avoid cannibalizing the Epyc line, probably keeping 4 channel memory), and Epyc to reach 64c/128t with 8 channel memory and all the server grade features.
Posted on Reply
#18
brian111
I don't think anyone was expecting anything other than a Zen 2 launch in 2019. This isn't surprising or news. Only then will with a six or eight core CCX make higher core Ryzen cpus possible.

GorbazTheDragon said:


AMD also will need to start offering significantly better ST performance to really get a stronghold on the desktop market, and the current indications hint that they will at best be closing the gap to in the order of 10%.
As it stands right now IPC is already within 10%. What they really need is clockspeed.
Posted on Reply
#19
GorbazTheDragon
brian111 said:

As it stands right now IPC is already within 10%. What they really need is clockspeed.
Uh even in the more AMD favoured metrics they will have at least 10% deficit. CB should be around 2200 at 4.9 GHz (remember that 5.1-5.2 is very attainable on the 8700k) while a 4.2 GHz 2700x will be around the low 1900s.

You have to consider the attainable clocks on an arch if you want to look at practical performance, just stating per clock performance is irrelevant. At the moment ryzen isn't showing that it will push 10% extra clock speed out of the blue, even on a new process, intel is already at those clocks, I already mentioned that all they need to get right is the TIM and motherboard manufacturers need to actually ship good VRMs on sub $200 boards.
Posted on Reply
#20
Fx
londiste said:
Why do we think there might be 16 cores on AM4?
If that 8-core CCX is true, AMD will more than likely just reap benefits from lower manufacturing costs as well as avoiding all the cross-CCX latency issues.

Cost is where AMD currently wins. 2600X is not doing very well against 8700K.
It is in my book and I just bought one. It easily performs well enough for the gaming/productivity that I do. More importantly, it is on a platform (AM4) that isn't going anywhere any time soon. So what this means, is I bought this 2600x now and upgrade to the 3700x/3800x when it is released.

I only spent $220 for it and will give someone a good deal when I sell it to grab the latest and greatest and wont have lost much money compared to if I bought a 2700x (~$320) now.
Posted on Reply
#21
JRid
I will go for AMD just because of Intel think they are better and sell theyre shit for hi prices, just changing sockets, scaming costimers
Posted on Reply
#22
ssdpro
JRid said:
I will go for AMD just because of Intel think they are better and sell theyre shit for hi prices, just changing sockets, scaming costimers
Welcome to the forum! That is the best Pro-AMD argument ever made on the Techpowerup. Perfectly sums up the AMD movement in many ways.
Posted on Reply
#23
brian111
GorbazTheDragon said:
Uh even in the more AMD favoured metrics they will have at least 10% deficit. CB should be around 2200 at 4.9 GHz (remember that 5.1-5.2 is very attainable on the 8700k) while a 4.2 GHz 2700x will be around the low 1900s.

You have to consider the attainable clocks on an arch if you want to look at practical performance, just stating per clock performance is irrelevant. At the moment ryzen isn't showing that it will push 10% extra clock speed out of the blue, even on a new process, intel is already at those clocks, I already mentioned that all they need to get right is the TIM and motherboard manufacturers need to actually ship good VRMs on sub $200 boards.
My comment about AMD being within 10% is in regards to IPC, (which it is https://www.techspot.com/article/1616-4ghz-ryzen-2nd-gen-vs-core-8th-gen/)

Of course you have to consider attainable clocks when judging performance, that's why I specified AMD needs to increase clock speeds. If AMD could clock even close to Intel they would be in great shape.

I don't know how you can claim that a new process won't help.
Posted on Reply
#24
Sandbo
That's what I can call a fast advancement - my less than 1-year-old Threadripper 1950X will soon be obsolete,
where my X470 ITX can rock the same number of cores (with improvement) within maybe a year.
Posted on Reply
#25
GorbazTheDragon
brian111 said:


Of course you have to consider attainable clocks when judging performance, that's why I specified AMD needs to increase clock speeds. If AMD could clock even close to Intel they would be in great shape.

I don't know how you can claim that a new process won't help.
There are plenty of examples of process upgrades being downgrades in peak performance. These nodes are not optimised for peak clocks most of the time, they have to work well in ranges required by the majority of the market which is servers in the case of zen and laptops/workstations/servers in the case of intel's offerings. In all of these configurations power efficiency is key, so what they get at the high end is really just a bonus most of the time.

Look at SB to IVB and HSW to BDW, both cases the new node lost somewhere around 300MHz. Not to mention that the newer skylake iterations are faster by nature of being on the same node after several refinements.
Posted on Reply
Add your own comment