AMD X399 !!!!!

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

french toast

Senior member
Feb 22, 2017
988
825
136
I'm not ignoring what Papermaster said, but people may be reading into what he said incorrectly.

Tock does not necessarily imply a new CPU core, and the roadmap that just leaked just showed it.

Process and circuit optimization is a totally and completely legitimate way to gain more performance. If 1800X could OC to 4.4GHz on all cores reliably, rather than the 4GHz-ish, then that'd be a solid improvement, no?
Not on 14nm lpp it wont.
It will a HP process for that, samsungs 14nm lpu would do the trick, BUT i cant see AMD doing that for a meagre upgrade, expensive so i think some low hanginging fruit will be fixed, netting 5-10% ipc, not sure if they would make the jump to 14nm lpu, be great if they could.
 

Ajay

Lifer
Jan 8, 2001
15,332
7,792
136
Not on 14nm lpp it wont.
It will a HP process for that, samsungs 14nm lpu would do the trick, BUT i cant see AMD doing that for a meagre upgrade, expensive so i think some low hanginging fruit will be fixed, netting 5-10% ipc, not sure if they would make the jump to 14nm lpu, be great if they could.

Well, two things:
1) GFL should be able to wring a bit more out of 14nm LPP (yes this process has been in use for a while, but this is the first time it is being used for a high clocked CPU - the GFL engineers might be learning a bit more about their process)
2) If Zeppelin has any clock limiting bottlenecks, then Zen+ might correct them.

So it is may be possible to to hit a 4GHz base clock.

I suppose, with Raven Ridge coming online soon, AMD might be able to move Zen2 over to LPU @ Samsung, that decision would have been made a year or two back, otherwise it would be coming out later than 2018. I think 14 HP will be volume limited (where is it being deployed? Fishkill/Burlington?).

There is just allot of stuff we don't know.
 

french toast

Senior member
Feb 22, 2017
988
825
136
Well, two things:
1) GFL should be able to wring a bit more out of 14nm LPP (yes this process has been in use for a while, but this is the first time it is being used for a high clocked CPU - the GFL engineers might be learning a bit more about their process)
2) If Zeppelin has any clock limiting bottlenecks, then Zen+ might correct them.

So it is may be possible to to hit a 4GHz base clock.

I suppose, with Raven Ridge coming online soon, AMD might be able to move Zen2 over to LPU @ Samsung, that decision would have been made a year or two back, otherwise it would be coming out later than 2018. I think 14 HP will be volume limited (where is it being deployed? Fishkill/Burlington?).

There is just allot of stuff we don't know.
Yea I predicted 'pinnacle ridge' using 14nm LPU, with some 5-10% IPC improvements, if those improvements included some latency improvements to the infinity fabric that would be awesome!.

Not sure about 14lpu, but I figured Samsung would need a large customer for it, why not amd?
 

iBoMbY

Member
Nov 23, 2016
175
103
86
AMD will not produce their main CPU at other manufacturers than GloFo. Only selected GPUs, and semi-custom stuff will go to Samsung, unless the GloFo fab burns to the ground.
 
  • Like
Reactions: Arachnotronic

french toast

Senior member
Feb 22, 2017
988
825
136
AMD will not produce their main CPU at other manufacturers than GloFo. Only selected GPUs, and semi-custom stuff will go to Samsung, unless the GloFo fab burns to the ground.
Is it possible that glofo can use this process?
 

lopri

Elite Member
Jul 27, 2002
13,209
594
126
Given the sensitivity to memory speeds - am I right in assuming that the L3 cache and RAM access are operating at the same frequency?

If they could de-couple that, and run the internal CPU memory (i.e. cache) at a faster frequency, then it might make the L3 latency problems go away.
Not a good idea since the current setup appears optimal. Basically the IMC/LC digests as much as data memory sends, there is no waste. Having it run faster when memory is not able to feed means wasted power, and having it run slower than memory can feed means a bottleneck. The sensitivity to memory speed in this case is a good thing, not a bad thing.
 

lopri

Elite Member
Jul 27, 2002
13,209
594
126
What I am more curious about is: How much credit is due for Samsung in Zen's creation/production? I know that GF (and by extension AMD) licensed Samsung's 14 nm, which was originally used for A9 production. Does anyone know how long this cooperation will last?
 

french toast

Senior member
Feb 22, 2017
988
825
136
Yea but they appear to be saying 2 die mcm? That looks odd to say the least.
Also low clocks for consumer, getting interesting though.
 

imported_jjj

Senior member
Feb 14, 2009
660
430
136
Yea but they appear to be saying 2 die mcm? That looks odd to say the least.
Also low clocks for consumer, getting interesting though.

It's a 16 cores Naples ofc it's MCM, AMD is not doing a native 4 or 16 C and 8 is all they have this year.
The clocks are needed to stay inside a sane TDP but maybe they are unlocked.
Ofc then we better hope that there are some water cooling solutions designed for it.
 

french toast

Senior member
Feb 22, 2017
988
825
136
It's a 16 cores Naples ofc it's MCM, AMD is not doing a native 4 or 16 C and 8 is all they have this year.
The clocks are needed to stay inside a sane TDP but maybe they are unlocked.
Ofc then we better hope that there are some water cooling solutions designed for it.
Yea ignore my previous post, I'm multitasking and had a total brain fart, for some reason my brain told me mcm = two separate dies ala dual socket, doh!
 

Ajay

Lifer
Jan 8, 2001
15,332
7,792
136
Yea but they appear to be saying 2 die mcm? That looks odd to say the least.
Also low clocks for consumer, getting interesting though.

This really looks like a workstation/entry level server platform. If the boards support ECC, that will confirm. Of course, it can still be dropped into a board w/o ECC if motherboard companies think it will sell well to consumers.

Not sure what Canard is saying about clocks: 2.4G base + 2.8G turbo or 2.4-2.8G base?
 

french toast

Senior member
Feb 22, 2017
988
825
136
This really looks like a workstation/entry level server platform. If the boards support ECC, that will confirm. Of course, it can still be dropped into a board w/o ECC if motherboard companies think it will sell well to consumers.

Not sure what Canard is saying about clocks: 2.4G base + 2.8G turbo or 2.4-2.8G base?
ryzen already supports ecc on some boards, yea unfortunately although I'm french toast, I'm not really french! :)
 

moonbogg

Lifer
Jan 8, 2011
10,635
3,095
136
Is this just a rumor or is there good reason to believe they will release this in consumer form?
 

imported_jjj

Senior member
Feb 14, 2009
660
430
136
So they just switched Snowy Owl from "opteron" (or naples) moniker to "ryzen" or there's something new here?

Snowy Owl might be APU but aside from that yes it should be a server SKU (more or less) brought to consumer just like Intel HEDT is. In the end even Summit Ridge is a server die.
The interesting bit would be if the x399 platform is unlocked and it better be or there isn't a point to it.
Chilled water anyone?:)

 
  • Like
Reactions: lightmanek

lopri

Elite Member
Jul 27, 2002
13,209
594
126
This really looks like a workstation/entry level server platform. If the boards support ECC, that will confirm. Of course, it can still be dropped into a board w/o ECC if motherboard companies think it will sell well to consumers.
Didn't almost all AM3/AM3+ boards support ECC? I cannot speak for budget boards (<$100), but several of them I used all had ECC support as long as the CPU does. All CPUs in AMD's case, that is.
 
  • Like
Reactions: Ajay