Richland ES now out in the wild.

[inf64]

Kabini is really a honey badger .

 

[Ajay]

It's just me or did anyone else find strange a 50% drop in TDP for the same frequency coming from a half-node SOI-less shrink?

Bogus. That's all I can say. Unless GFL's 32nm is absolutely FUBAR, then a 50% decrease on a 1/2 node just isn't possible. Unless we get collaborating info from another source, the info from that IRC chat is basically useless.

There are allot of us looking for AMD to pull a rabbit out of it's hat, myself included. So it's real easy to get pumped up on tiny little scraps of unverified info. It would be very sad to see AMD implode.

 

[Arachnotronic]

I hope Richland is good on the iGPU front...

 

[ViRGE]

If we wanted to hear your opinions on each other, this would be the Politics & News forum, not CPUs. So watch it
-ViRGE

 

[inf64]

I have no clue why my post was gone,it had nothing personal in it(Was about Kabini and Richland)... But hey,we know how moderation works here,don't we

 

[NTMBK]

I have no clue why my post was gone,it had nothing personal in it(Was about Kabini and Richland)... But hey,we know how moderation works here,don't we

Weren't there some personal comments about ShintaiDK in there? (And going round throwing out accusations of mod bias is just sad.)

Yes. I wiped the slate clean for the last dozen posts or so rather than infract you guys.
-ViRGE

 

[inf64]

No wasn't anything personal and about Shintai in it.

PS I have nothing against good moderation on any forums.

It was in direct reply to Shintai's post, hence its inclusion. Now I ask that you either drop this subject or take it to PMs so that this thread doesn't get derailed.
-ViRGE

 

[Idontcare]

Kabini is really a honey badger .

What do we know of AMD's launch plans/timeline for Kabini?

CTho9305 was kind enough to point out that AMD announced it had taped out in mid-Oct, and it appears the tapeout went so well that AMD has functional ES from first-silicon.

So realistically they could be looking at launch in Q213. But does that jive with AMD's plans? They don't want to create another Llano-type inventory buildup which Kabini could create for them if they don't get the channels cleared out first.

 

[inf64]

What do we know of AMD's launch plans/timeline for Kabini?

CTho9305 was kind enough to point out that AMD announced it had taped out in mid-Oct, and it appears the tapeout went so well that AMD has functional ES from first-silicon.

So realistically they could be looking at launch in Q213. But does that jive with AMD's plans? They don't want to create another Llano-type inventory buildup which Kabini could create for them if they don't get the channels cleared out first.

I agree,maybe late Q2 as best case scenario. It's good to have working early ES but it's far from final product. But it looks like a very good product and nice successor to Bobcat core.

 

[jpiniero]

I agree with your market analysis . But AMD won't abandon the "big core" yet,maybe after 2015 they will reevaluate their startegy

It would make sense if Richland is the last of the "big cores", just so they can cut costs and focus on Kabini and the ARM stuff. It would be dead already if not for the PS4 contract. AMD doesn't want to admit it yet (& the layoffs) because they need the engineers to finish it.

 

[NTMBK]

It would make sense if Richland is the last of the "big cores", just so they can cut costs and focus on Kabini and the ARM stuff. It would be dead already if not for the PS4 contract. AMD doesn't want to admit it yet (& the layoffs) because they need the engineers to finish it.

That's assuming that the PS4 is using Richland and not Kabini, of course.

 

[Idontcare]

I agree,maybe late Q2 as best case scenario. It's good to have working early ES but it's far from final product. But it looks like a very good product and nice successor to Bobcat core.

Did you see this xbitlabs article on SR and Richland?

The company, according to media reports, cancelled its Steamroller-based performance-mainstream Kaveri APU and substituted it with Richland APU (which features Piledriver x86 cores, but comes with latest-generation Radeon HD GCN cores) made using 32nm SOI process technology.

Richland is 32nm SOI HKMG and not 28nm? Is that true or did xbitlabs mess up?

 

[inf64]

Who knows,anything is possible. The x264 dev who supposedly got Richland ES might have gotten the 2.3Ghz 35W chip which has 3.2Ghz Turbo(and not a stock 3.2Ghz clock). I also doubt that bulk 28nm can yield such a massive perf./watt jump.

 

[NostaSeronx]

That's assuming that the PS4 is using Richland and not Kabini, of course.

Playstation 4 is using Thebe-J which hasn't finished yet nor is it related to the Jaguar or Trinity or Kaveri architectures. The only one that is showing any signs of finalization is Xbox's Kryptos which is a >450 mm² chip. To get back on Thebe-J it was delayed from Trinity side-by-side development to Kaveri side-by-side development.

 

[Idontcare]

Playstation 4 is using Thebe-J which hasn't finished yet nor is it related to the Jaguar or Trinity or Kaveri architectures. The only one that is showing any signs of finalization is Xbox's Kryptos which is a >450 mm² chip. To get back on Thebe-J it was delayed from Trinity side-by-side development to Kaveri side-by-side development.

How is Thebe Jaguar not related to Jaguar? Did they abort using the jaguar core at some point?

 

[NostaSeronx]

How is Thebe Jaguar not related to Jaguar? Did they abort using the jaguar core at some point?

I assume if they are going to use Jaguar it is going to be in a big.LITTLE formation. Which will have them in a configuration where the Jaguar portion will control all of the system, os, etc stuff that generally isn't compute intensive. While the Thebe portion will control all of the gaming, hpc, etc. stuff that is generally compute intensive. Since, each year the performance part of the Playstation Orbis was upgraded it is safe to assume that they are going for an APU with the specs.


First:
A8-3850 + HD 7670
400 VLIW5 + 480 VLIW5 => 880 VLIW5 -> VLIW4 => 704

Second:
A10-5700 + HD-7670
384 VLIW4 + 480 VLIW5 => 864 VLIW4/5 -> VLIW4 => 768

I have heard that the third generation of the test Orbis uses an APU with GCN2.
Unknown APU + HD8770
384 GCN2 + 768 GCN2 -> 1152 GCN2

It is assumed that the APU only has four cores because AMD doesn't plan to increase the core count other than the GPU cores from now on.

 

[Ajay]

Did you see this xbitlabs article on SR and Richland?



Richland is 32nm SOI HKMG and not 28nm? Is that true or did xbitlabs mess up?

I've read both now. I've also read that SR is due out in 2014 now on 28nm, which would be ridiculous. Excavator in 2014 on 22/20nm might have a snowballs chance in hell of being somewhat competitive with Intel, but SR?! People are going mad with all the speculation.