AMD Zen - Key Dates and Information

[superstition]

Since there hasn't been any leaks for motherboards (which always has happened in the past for both camps), I am not seeing any kind of a real launch during CES, unless it is for OEM machines.

I would love to see an end to the practice of having dramatically different boards with the same name and just different version numbers.

Instead of, for instance, 990FX–LB2A with version 1.0, 1.2, 2.2, 3.0, 4.0. and 5.0 — with big differences in specs, like different audio, a different level of VRM robustness, power regulation parts removed or added, etc. Why not this:

AM4 Goldie v1, AM4 Goldie v2, AM4 Goldie v3, AM4 Goldie v4, AM4 Goldie v5

Do we really need long cumbersome strings of nonsense letters and numbers like 970A–UD3P 2.2? What does the "A" stand for? Does anyone care? UD is ultra-durable, as if anyone cares about that. 970 is the chipset which is at least useful information. 3 means a lower tier than their higher tiers (4 and 5), but that's arcane brand-specific knowledge and thus poor-quality naming convention. What is "P"? Does anyone care? I don't. I own the board, have done a ton of testing with it, and have never bothered to check.

 

[Atari2600]

Indeed. Pity AMD couldn't "persuade" their board partners to release a common naming convention.

A320_T1-- // B350_T3-- // X370_T5--
A320_T3-- // B350_T5-- // X370_T7--
A320_T5-- // B350_T7-- // X370_T9--

T1/T3/T5/T7/T9 corresponding to spec tiers. Of course, more tiers can be added from the start with greater fidelity.

If possible, add minimum specs that must be met for a board to be a tier. So, VRM quality, # USB ports, # SATA ports, # PCIe slots, # NVMe slots etc etc

 

[superstition]

in this case I'm more interested in VRM capability more than anything else at this point, 8c16t SR will need a beefy VRM at >4GHz (more than likely it can cross the 4GHz mark), let's hope properly overengineered high end boards aren't too expensive.

relevant:

AM4 requires much beefier VRMs (all CPU / APU planes), higher quality PCB (due higher currents, faster signaling), etc. The design guidelines for all Zen based platforms are quite demanding, so you cannot get away with the same garbage quality as one could with "AM1" or FM2+. Unless you want to break the platform cross compatibility of course. All FM2+ boards < 60$ (CSP) are generally garbage and in most cases cannot operate properly with all the APUs available for the platform.

AMD parts are not factory overvolted any more than Intel products are. The load-line specification for both AM3+ and FM2+ is extremely loose (1.3 mOhm & 2.1mOhm) and because of that the voltage droop by the specification is very large. To ensure that the operation parameters remain within a spec, an AM3+ part which draws 100A of current and requires 1.3000V to be fully stable must have at least 1.4300V default voltage (130mV droop @ 100A). Some of the motherboards are built to have lower Rll (< 1.3mOhm / 2.1mOhm) than the specification dictates, or the end-user might adjust it to be lower than the default value. If the droop is lower or non-existing for either reason, it appears that the parts are extremely overvolted from the factory. The truth of course being that they are configured for perfectly right default voltage, which complies with the specifications (droop) and contains some standard safety margins.

On AM4 the situation is significantly better, since the load-line spec. is less than half of what it was on AM3+.

For a 8C/16T Zeppelin I wouldn't even consider a motherboard with a 4+2 phase config. While a making a board with sufficient 4+2 phase VRM is entirely possible (as previously said), none of the manufacturers will use high enough quality components in their boards to make a 4+2 phase configuration good enough. A native 6+2 phase configuration will be used on the high quality mainstream / high-end boards. There will most definitely be doubled (4+1 to 8+2, etc) solutions too, especially in the cheaper "enthusiast" / "high-end" boards.

The VRM requirements on Zeppelin & Raven are not demanding due the high power draw, but due the high current draw.

95W TDP Zeppelin parts should draw around ~ 80W from the primary (VDDC, CPU core) power plane. The rest (~15W) is for the northbridge, IO, FCH, etc and will be drawn from the secondary VDD_SoC plane and other minor, low-current planes.

When AMD moved from 32nm SHP SOI to 28nm BULK the voltage stability became extremely important. Despite the platforms using parts made with different processes (e.g AM3+ and FM2+) had exactly the same load-line specification (1.3mOhms) in reality the smaller and othewise inferior 28nm process was significantly more sensitive to voltage variations / fluctuations. Achieving a stable voltage supply through proper (load dependent) load-line calibration can result in hundreds of MHz additional headroom when close to Fmax, even on the more recent 28nm (Godavari) chips.

For Zen the load-line appears to be (based on the existing VRM designs) significantly tighter than it was with previous AMD designs and much tighter than the Intel VR12 spec (which is already strict) specifies...

FX-9K series SKUs use dies with so extreme leakage characteristics, that they could never be sold as FX-8K series SKUs. Even if they are downclocked to same specifications as FX-8K series their power consumption and current draw would violate the infrastructure specifications (95 / 125W TDP). If AMD hadn´t create the 220W TDP infrastructure for AM3+, the parts now sold as FX-9K parts would have been sent to the crusher.

Relative leakage:

FX-9590 = 11.9 (avg)
FX-8370 = ~16.6 (avg)
FX-8370E = ~ 21.5 (avg)

The first FX-8350s had significantly higher leakage, around 14 on average.

 

[zinfamous]

The point I made is that if it works really well there's no good reason to manually overclock.

This. I'm not one to spend the effort/time/money to do such things. I guess this isn't a hobby for me (I much prefer spending hours on preparing a meal, for example).

But if this works as suggested, I'd be intrigued. I'll remain skeptical, though, until release.

 

[superstition]

This. I'm not one to spend the effort/time/money to do such things. I guess this isn't a hobby for me (I much prefer spending hours on preparing a meal, for example).

But if this works as suggested, I'd be intrigued. I'll remain skeptical, though, until release.

I enjoy overclocking when it's not painful with no gain. That has been my experience with trying to tune RAM with a 6700K recently. Every time I thought I had stability I would fail Memtest after a huge amount of coverage. I finally got fed up and put the RAM back at stock.

 

[Atari2600]

(I much prefer spending hours on preparing a meal, for example).

Screw that!!

10 mins cooking the meal, 10 mins eating it and 5 hours drinking with mates after it.

 

[AtenRa]

Screw that!!

10 mins cooking the meal, 10 mins eating it and 5 hours drinking with mates after it.

You can always drink in between as well, just for continuity

 

[zinfamous]

Screw that!!

10 mins cooking the meal, 10 mins eating it and 5 hours drinking with mates after it.

I tend to drink while cooking and with friends as a bonus.

And 10 minutes preparing a meal? I'm not talking about boiling a box of Kraft Mac 'n "Cheez," here.

 

[DrMrLordX]

I tend to drink while cooking and with friends as a bonus.

And 10 minutes preparing a meal? I'm not talking about boiling a box of Kraft Mac 'n "Cheez," here.

Pssh, Easy Mac is quicker anyway.

 

[zinfamous]

Pssh, Easy Mac is quicker anyway.

You people make me sad with your expensive computer whatsits but terrible eating habits.

 

[SlowSpyder]

Thanks for the dates and other info, looking forward to this launch. Definitely has a different feel to it than the Bulldozer launch.

 

[Rifter]

I just want some IPC results already.

 

[Vesperan]

Couple of days ago Fortune.com had an interview with Lisa Su where Lisa says Zen in first half of 2017.

http://fortune.com/video/2016/12/21/the-game-changer-amd-needed/

While I don't expect Fortune to have their ears closer to the ground than a tech focused outfit, they are a big outfit and it is interesting that the date seems to bounce between Q1 and H1.

 

[jpiniero]

While I don't expect Fortune to have their ears closer to the ground than a tech focused outfit, they are a big outfit and it is interesting that the date seems to bounce between Q1 and H1.

I still think it's more like a paper launch of the top model in February ($$$) and the full launch in end March-April. Which kind of fits what that is saying. Honestly, as long as it's widely available before Skylake-X they can probably get away with it now.

 

[kalmquist]

"***Supposed*** public launch of consumer Zen at CES -- 5th-8th Jan"
according to http://www.digitimes.com/news/a20161003PD200.html

A few things to note:

1) That article is dated October 3. As far as I know, no other reporters have gotten this information from their sources. The lack of confirmation from other sources, this close to CES, is reason to wonder if the report is correct.

2) The article also states that, "Zen-based products are unlikely to start mass shipments until after February...according to sources from the upstream supply chain." So while one source is telling the writer about a CES launch, other sources are telling the writer not to expect Zen processors to be available until March or later.

3) The writer also predicts that Zen is, "unlikely to start contributing profits until the second quarter 2017," which, if correct, means product availability very near the end of March at the earliest.

Also worth considering is Vesperan's report of a Dec. 21 article where Lisa Su referred to a release in the first half of 2017, rather than in the first quarter. My guess is that the target date for Zen is in late March, and could slip into April if problems crop up.

 

[coercitiv]

Also worth considering is Vesperan's report of a Dec. 21 article where Lisa Su referred to a release in the first half of 2017, rather than in the first quarter. My guess is that the target date for Zen is in late March, and could slip into April if problems crop up.

Based on reinforced commitment from Lisa Su for Q1 2017 I think it's safe to say they're on track for March. However, it's also safe to expect low initial availability.

 

[Arachnotronic]

"***Supposed*** public launch of consumer Zen at CES -- 5th-8th Jan"
according to http://www.digitimes.com/news/a20161003PD200.html

A few things to note:

1) That article is dated October 3. As far as I know, no other reporters have gotten this information from their sources. The lack of confirmation from other sources, this close to CES, is reason to wonder if the report is correct.

2) The article also states that, "Zen-based products are unlikely to start mass shipments until after February...according to sources from the upstream supply chain." So while one source is telling the writer about a CES launch, other sources are telling the writer not to expect Zen processors to be available until March or later.

3) The writer also predicts that Zen is, "unlikely to start contributing profits until the second quarter 2017," which, if correct, means product availability very near the end of March at the earliest.

Also worth considering is Vesperan's report of a Dec. 21 article where Lisa Su referred to a release in the first half of 2017, rather than in the first quarter. My guess is that the target date for Zen is in late March, and could slip into April if problems crop up.

They could announce the products/SKUs/pricing ahead of product available. By CES, they should know what frequencies these things will run at out of the box and the chips should be running thru the fabs.

They could hard launch AM4 boards and Bristol Ridge CPUs at CES.

 

[Soulkeeper]

Is it not unreasonable to think they are already stockpiling them ?
I mean don't this usually start weeks or months ahead of a launch, they need hundreds of thousands afterall.

 

[Arachnotronic]

Is it not unreasonable to think they are already stockpiling them ?
I mean don't this usually start weeks or months ahead of a launch, they need hundreds of thousands afterall.

It is unreasonable. If they were already stockpiling them, they wouldn't have been going on about needing to do further work, getting turbo enabled, etc. The chip would have been qualified for production and they'd be able to give us exact details.

I suspect it will be qualified by the time CES rolls around so they can paper launch the lineup.

 

[blublub]

Based on reinforced commitment from Lisa Su for Q1 2017 I think it's safe to say they're on track for March. However, it's also safe to expect low initial availability.

Since AMD can now easier FAB at Samsung (since new WSA) I actually expect a good availability

 

[blublub]

It is unreasonable. If they were already stockpiling them, they wouldn't have been going on about needing to do further work, getting turbo enabled, etc. The chip would have been qualified for production and they'd be able to give us exact details.

I suspect it will be qualified by the time CES rolls around so they can paper launch the lineup.

If they are still working on the silicon on the 13th of December a Q1 launch is out if question as fabing, binning, packaging and delivery will take longer than April - just fabing needs close to 90d!

The reason for not stating the turbo can simply mean they don't know of many CPUs will clock to XGhz - so in other words binning isn't done yet.

Second possibility is they are still working on the microcode which influences the turbo

3rd ulis that turbo is so good they won't tell ppl

 

[KTE]

Since AMD can now easier FAB at Samsung (since new WSA) I actually expect a good availability

Depends on yields for the parts they require. The 6 month delay already is almost always bugs/yields. If they miss January, they either have poor yields or higher power@clocks than required. Really no other way around it.

K10 when produced was matching/beating Intels best on the Server side, had it not been delayed so much so that it became a relic on release.

There is so much AMD has to get right with a new product launch... Its not easy.

Sent from HTC 10
(Opinions are own)

 

[blublub]

I highly doubt there are yield issues because at least at Samsung fabs it is a mature process and Su said 3.4Ghz+ - so if quality wasn't OK they could sell chips with 3.2Ghz

 

[AtenRa]

They already producing 232mm2 Polaris dies at 14nm LPP GloFo, yields shouldnt be a concern at this point.

 

[itsmydamnation]

It is unreasonable. If they were already stockpiling them, they wouldn't have been going on about needing to do further work, getting turbo enabled, etc. The chip would have been qualified for production and they'd be able to give us exact details.

I suspect it will be qualified by the time CES rolls around so they can paper launch the lineup.

I very much doubt the work they are doing has anything to do with silicon. So yes they very much could be stocking up on them. Then its just either a microcode update or maybe even just a Motherboard side configuration set.

If they did have to do another metal layer ( very much doubt it ) then we could easily see another few hundred mhz on the base clock as well.