6 or 8 core Steamroller based AMD CPU likely?

[zlatan]

So is there any reason to not release a similar Steamroller based version?

Yes. The nextgen games will have several tasks that can run on the iGPU, and it will be much faster than a normal CPU. Even Intel want to push OpenCL for games.

 

[AtenRa]

Ok, so what are the numbers(or what do you think they are) are for Piledriver after 3 more quarters?

I dont have any numbers but, Vishera is faster, cheaper and in larger capacity one year after Bulldozer Q3 2012 sales. That should help you make an educated guess

 

[Abwx]

But FM2+ still doesn't use Hypertransport links, so...

HT for internal cross chip communication in the MCM itself ,
i wonder if they implemented one in Kaveri since it will be used
in low cost servers.

 

[Phynaz]

FX is dead. Long live FX

(At least until the rebrand something else as FX)

 

[inf64]

I think it's safe to say FX will not die as a brand. Sooner or later they will launch FM2+/FM3? SR/EX based FX product. That will require at least 3 modules though so we won't be seeing such a product in near future. My guess is Q2/Q3 next year the 3M SR based FX will arrive on FM2+ platform (with iGPU of course).

Latest Kaveri slides:

http://www.computerbase.de/news/2013-10/amds-kaveri-vorstellung-am-5.-dezember-erwartet/

 

[AtenRa]

New "SteamrollerB" Core

This slide is for Mobile but Desktop should be the same. Now, whats new on B version ???

 

[inf64]

I think "B" is Kaveri2.0. What happened to Kaveri1.0 is a good question .
From the article itself it looks like embargo lifts on 5th of December? Or my gtranslate has failed me duh

 

[Exophase]

Maybe "B" is a hint for why it's taking so long for Kaveri to be released..

 

[mrmt]

I think it's safe to say FX will not die as a brand. Sooner or later they will launch FM2+/FM3? SR/EX based FX product. That will require at least 3 modules though so we won't be seeing such a product in near future. My guess is Q2/Q3 next year the 3M SR based FX will arrive on FM2+ platform (with iGPU of course).

Given that AMD is trying to push into the entry-level workstation business, and this product you are describing would address one of the weakness of the current line up, don't you think that if this product would see the light of the day they would announce that product first for that more lucrative segment?

 

[inf64]

Do you see any desktop roadmaps featuring actual Kaveri models yet. At least one? No?
There is your answer.

PS One module adds ~28-30mm^2 to the die. Accompanying logic would add maybe ~10mm^2 more (just my guess). The Kaveri 2M die is smaller than Richland.

 

[ShintaiDK]

Do you see any desktop roadmaps featuring actual Kaveri models yet. At least one? No?
There is your answer.

PS One module adds ~28-30mm^2 to the die. Accompanying logic would add maybe ~10mm^2 more (just my guess). The Kaveri 2M die is smaller than Richland.

So you are saying they will sell workstation and server users 2M/4T. While for some odd reason sell desktop users with 3M/6T?

 

[mrmt]

Do you see any desktop roadmaps featuring actual Kaveri models yet. At least one? No?
There is your answer.

No, but we did see actual server roadmaps featuring Berlin and explicitly mentioning "4 Steamroller cores", and that roadmap goes up until the end of 2014.

I don't think AMD would launch a 3M Steamroller on the time frame you mentioned in consumer market without launching it on the same time frame on the server market. The opportunity would be too big to pass on, especially on the entry level workstation market where they are/were investing.

 

[inf64]

I didn't say they will launch it 100%, I said it makes sense and I expect them to launch product like that. That's a big difference.

 

[Exophase]

I can't find any AMD desktop roadmap that mentions beyond 2013 at all. But there are server and embedded roadmaps that go through 2014 that list Kaveri spinoffs but no 3M or 4M Steamroller non-APUs. Given that in the past AMD made the FX dies with the intention of being server dies, which followed closely in release time, this absence seems conspicuous.

 

[mrmt]

I can't find any AMD desktop roadmap that mentions beyond 2013 at all. But there are server and embedded roadmaps that go through 2014 that list Kaveri spinoffs but no 3M or 4M Steamroller non-APUs. Given that in the past AMD made the FX dies with the intention of being server dies, which followed closely in release time, this absence seems conspicuous.

I have a feeling that the reason of that is to not crater the sales of AM3+ systems. They had a fairly significant share in 2012 and they should have a significant share in 2013. Had AMD announced the death of the platform, in the next day AMD would have nowhere to dump their server rejects.

 

[pipm1]

I hope Steamroller B took in some performance-enhancing features that were originally planned for Excavator.

 

[inf64]

According to German article in my post above the press NDA embargo lifts on 5th of Decmeber and then we will know approximate performance. The desktop should be coming after CES,in February time and mobile is unknown. Judging from the embargo date maybe mobile will come first, right around CES?

PS Kaveri will have TrueAudio tech in iGPU meaning it is from Hawaii generation. Also the dedicated pcie x4 lanes for SSD are new info (I haven't seen it before).

 

[Yarn]

"AMD StartNow 3.0"

Am I the only one that thinks a lot of AMD's marketing names just aren't that good?

 

[Vesku]

From the comments of their rehired big core design leader, it's my guess that for various possible reasons (GF 28nm wafer supply, smaller than desired performance delta between piledriver and steamroller at higher TDPs, possibly GF abandoning anything 28nm but their low power node, etc.) AMD is waiting until 20nm Excavator to refresh the performance PC and 6+ thread server line.

Lack of or poor performing GF 28nm HPP, performance 28nm, could be a big part of all this. Anyone know of a chip in the wild made on that process. Only 28nm GF chips I know of are SLP, low power 28nm, Rockchip ARM SoCs.

 

[rtsurfer]

From the comments of their rehired big core design leader, it's my guess that for various possible reasons (GF 28nm wafer supply, smaller than desired performance delta between piledriver and steamroller at higher TDPs, possibly GF abandoning anything 28nm but their low power node, etc.) AMD is waiting until 20nm Excavator to refresh the performance PC and 6+ thread server line.

Lack of or poor performing GF 28nm HPP, performance 28nm, could be a big part of all this. Anyone know of a chip in the wild made on that process. Only 28nm GF chips I know of are SLP, low power 28nm, Rockchip ARM SoCs.

Even though this seems likely.
I wish this doesn't happen.
If AMD waits till EX to refresh the FX line, then Intel will decimate AMD(already half way there).

AMD won't have high performing 20nm till Q4 2015 (My guess).
Till then Intel will have launched the successor to the alleged Haswell Refresh.
People will have completely forgotten of AMD making anything more than APUs and Graphic cards.

 

[Vesku]

The tech industry is actually one of the more forgiving markets for releasing better products after a hiatus. Intel itself recovered from P4 mis-steps. Apple was barely surviving until the return of Steve Jobs.

So if Excavator is good compared to whatever Intel has at the time, there won't be much difference between refreshing the performance sector with Steamroller or not. Only reason to release performance/big server Steamroller chips is if they are decent. Which may not be the case if they had to make Kaveri on GF 28nm low power node or the high power node is not meeting expectations.

 

[AtenRa]

Even though this seems likely.
I wish this doesn't happen.
If AMD waits till EX to refresh the FX line, then Intel will decimate AMD(already half way there).

AMD won't have high performing 20nm till Q4 2015 (My guess).
Till then Intel will have launched the successor to the alleged Haswell Refresh.
People will have completely forgotten of AMD making anything more than APUs and Graphic cards.

1: AMD could have 22nm SR in Q4 2014 or Q1 2015.

2: 28nm Kaveri should be close to 15-20%(CPU) faster in games vs Vishera/Ritchland.

3: AM3+ market share is to small(like Socket 2011), whatever Intel introduce will not make a big difference even if AMD will not release another AM3+ CPU until Q1 2015.

4: AMD sells more APUs than AM3+ CPUs. Also it seams to me they have focused Kaveri more for mobile, like they should have.

ps: I have said it before, Kaveri may not be manufactured at GloFo. Havent confirmed it yet, just a hunch.

 

[rtsurfer]

I wish AMD would comment on the future of FX line.

I am trying to put together a new build and as usual torn on the CPU decision.

I could wait till Q2 2014 for an FX steamroller, not later than that.

But they need to clarify if its they plan to release one or not.

 

[mrmt]

I am trying to put together a new build and as usual torn on the CPU decision.

I could wait till Q2 2014 for an FX steamroller, not later than that.

If this is your deadline it is better start thinking about the backup plan. AMD is giving every indication that there will be no Steamroller with more than 4 cores for the remainder of 2014.

 

[NTMBK]

ps: I have said it before, Kaveri may not be manufactured at GloFo. Havent confirmed it yet, just a hunch.

Only way Kaveri could be not made at GloFo is if the PS4 APU was instead- AMD need to meet that WSA, or the payments will kill them.