AMD Trinity & Richland APU Discussion

[Skud]

re: AMD Trinity & Richland APU Discussion

AMD "Richland" Desktop APU Lineup Detailed | techPowerUp

The series will be led by AMD A10-6800K, which features every component on the "Richland" silicon unlocked, which includes two "Piledriver" CPU modules amounting to four x86-64 cores, and all stream processors on the iGPU unlocked, with the highest CPU and iGPU clock speeds enabled in the lineup. The iGPU model for this chip is Radeon HD 8670D. The A10-6800K features unlocked multipliers, making overclocking a breeze. Trailing it is the A10-6700, which features all physical components unlocked, but with slightly lower clock speeds, and locked BClk multipler. It features the same iGPU as its bigger sibling, the HD 8670D.
The AMD A8-6000 series consists of the A8-6600K and A8-6500, both of which feature four CPU cores, but slightly toned down iGPU cores, labeled Radeon HD 8570D. The A8-6600K features unlocked BClk multiplier. The next APU in the lineup is the A6-6400K, which is dual-core, features a further scaled down iGPU, bearing the model number Radeon HD 8470D, and unlocked BClk multiplier. Lastly, there's the A4-6300, an entry-level dual-core APU with Radeon HD 8370D graphics. All models with -K extension feature rated TDP of 100W, others 65W.

Moving on to the platform itself, it's known that "Richland" APUs will be built in the same packages as "Trinity," and as such existing A55, A75, and A85X chipset-based motherboards should be able to run them with BIOS updates, yet AMD plans to launch a trio of new FCH chipsets. Leading the pack is the A88X (eight SATA 6 Gb/s ports), followed by A78 (six SATA 6 Gb/s ports), and A68 (probably four SATA 6 Gb/s ports, entry-level).

PS: Changed the thread title to reflect the recent developments.

 

[topgear]

^6 Thanks for the news but with all those code names of new APUs and moreover 3 new chipsets the situation is going to get very messy and difficulty to recommend . We have not seen any Trinity APUs apart from 5800K and 5300 so really can't think of when the new APUs will see the limelight here.

 

[Skud]

*www.xbitlabs.com/images/news/2013-01/amd_richland_trinity_fusion_specs.png

That's the updated line-up, so finally a CPU running at 4+ GHz at stock. Graphics seems just a rename/re-badge though.

 

[Cilus]

Graphics is not renamed, Richland is going to use GCN based GPU, not the VLIW 4 design of Trinity. Also in Richland, you can pair the IGP with any Desktop class HD 7000 series GPU in Hybrid Crossfire.

 

[sukesh1090]

^^
nope bro if i read things correctly its going to be PD and 6XXX i.e., vliw4.even AMD told 2nd generation graphics which will point to vliw4(1-vliw5,2-vliw4,3-GCN) and it can't be GCN 2.0 as you know APU always lag by 1 generation but still gets their branding.if the GPU was GCN then i am sure that we would have been seeing minimum gpu improvement as 40% not 20% because we know the performance improvement from vliw4 to GCN.so it is basically PD and vliw4 on 32nm with some tweaks.the big guns SR and GCN are kept for kavery which will be on 28nm.

 

[Skud]

Re: AMD Trinity & Richland APU Discussion

@Cilus: this slide tells something else:-

*www.expreview.com/topic/img/review/News/201301/0124/AMD-2013-roadmap-635x355.jpg

28nm, Steamroller, GCN, HSA - all are expected in Kaberi, Richland is basically Trinity 2.0

 

[Cilus]

Skud, on the slide nowhere it is mentioned that Richland is having VLIW GPU. Instead it is termed as 2nd gen directx 11 class gpu. 2nd gen dx11 gpu means GCN as both hd 5000 & 6000 series are based on VLIW architecture.
I think you are referring to some old leaks/slides , have a look at the following:

AMD Richland APU Will Boost up to 4.4GHz

 

[sukesh1090]

^^
but no where else it is mentioned as GCN and if i am not wrong then 2nd gen dx 11 stands for vliw 4 because first was vliw5 and 2nd is vliw 4 and GCN should be 3rd and over that if it was GCN don't you think that they would have been saying GPU boost is more than 40 or 50% because that was the performance we saw from desktop 6000 to 7000.a member in fudzilla told that the boost was due to a trick which AMD didn't do in trinity so they are doing it in richland.
there is lots of confusion about richland,only AMD can clear it.some people even say it will be on 28nm
here a slide from ces2013,nowhere it is mentioned as GCN in the slide as it is mentioned for kavery,
*images.anandtech.com/doci/6567/AMD-016_575px.jpg

 

[Skud]

^^
but no where else it is mentioned as GCN and if i am not wrong then 2nd gen dx 11 stands for vliw 4 because first was vliw5 and 2nd is vliw 4 and GCN should be 3rd and over that if it was GCN don't you think that they would have been saying GPU boost is more than 40 or 50% because that was the performance we saw from desktop 6000 to 7000.a member in fudzilla told that the boost was due to a trick which AMD didn't do in trinity so they are doing it in richland.
there is lots of confusion about richland,only AMD can clear it.some people even say it will be on 28nm

This. And most of the sites are saying it will be based on HD6000 only. On the other hand, AMD is claiming some 30-40% improvement in graphics performance, which seems impossible with the older graphics.

BTW, that AMD slide is dated 7 Jan 2013, so not so old. Also it was part of their CES presentation afaik.

 

[sukesh1090]

AMD is again doing a mistake by allowing for the unnecessary build of hype on richland that it will be having GCN and will be on 28nm,they should better clear it or they will end up in the same situation as they were in during BD release..

 

[tkin]

^^
but no where else it is mentioned as GCN and if i am not wrong then 2nd gen dx 11 stands for vliw 4 because first was vliw5 and 2nd is vliw 4 and GCN should be 3rd and over that if it was GCN don't you think that they would have been saying GPU boost is more than 40 or 50% because that was the performance we saw from desktop 6000 to 7000.a member in fudzilla told that the boost was due to a trick which AMD didn't do in trinity so they are doing it in richland.
there is lots of confusion about richland,only AMD can clear it.some people even say it will be on 28nm
here a slide from ces2013,nowhere it is mentioned as GCN in the slide as it is mentioned for kavery,
*images.anandtech.com/doci/6567/AMD-016_575px.jpg

You're right, much of Richland is clouded in mystery, its impossible to tell for certain if it will have GCN or VLIW cores.

 

[sukesh1090]

@skud,
i think as AMD says the gpu improvement is 20-40% so it will be mostly 20-25% in most cases and we will be seeing 30-40% improvement in some few cases only.

 

[Skud]

^^Even that seems impossible with just 44 MHz speed bump unless there's some architectural changes.

 

[sukesh1090]

^^
thats what i said a reputed member in fudzilla said AMD is doing it by a trick but didn't say what it is.it could be even some small changes including driver optimisations which may end up in giving 20% improvement.so we have to wait and see.

 

[Cilus]

Re: AMD Trinity & Richland APU Discussion

^^
but no where else it is mentioned as GCN and if i am not wrong then 2nd gen dx 11 stands for vliw 4 because first was vliw5 and 2nd is vliw 4 and GCN should be 3rd and over that if it was GCN don't you think that they would have been saying GPU boost is more than 40 or 50% because that was the performance we saw from desktop 6000 to 7000.a member in fudzilla told that the boost was due to a trick which AMD didn't do in trinity so they are doing it in richland.
there is lots of confusion about richland,only AMD can clear it.some people even say it will be on 28nm
here a slide from ces2013,nowhere it is mentioned as GCN in the slide as it is mentioned for kavery,
*images.anandtech.com/doci/6567/AMD-016_575px.jpg

Search carefully. Let me quote from the link that i gave above:

Something else that's worth mentioning is that the GPU on the Richland APUs will be part of the HD8000 Series based on the GCN architecture. This will allow Dual Graphics support with the current HD7000 Series of dedicated GPUs.

I hope you got it clearly.

^^Even that seems impossible with just 44 MHz speed bump unless there's some architectural changes.

The architectural change is GCN.

 

[Skud]

Re: AMD Trinity & Richland APU Discussion

Search carefully. Let me quote from the link that i gave above:



I hope you got it clearly.



The architectural change is GCN.

Apparently, TweakTown thinks otherwise:-

LeakedTT: AMD's upcoming Richland APU specifications :: TweakTown USA Edition

The unlocked K-series processors are the only ones to feature the 100W TDP. Integrated graphics remain based upon the 6000-series GPU, even though they feature numbers in the 8000 range. Reports say that the Richland APUs are already in the hands of system builders, so we could be seeing a release of these parts soon.

If the last line is true though, we will know sooner.

 

[topgear]

HD 8670D and HD 8570D most probably based on Oland chip and it's based on GCN not GCN 2.0 though.

 

[Cilus]

Exctly. Kaveri will be based on Stream Roller CPU Core and GCN 2.0 Core whereas Richland is based on Tweaked PileDriver CPU Core and GCN Graphics Core. Even AMD has used renaming of Trinity's VLIW4 Graphics Core as HD 7000M series, they have never used exact same name for two different architecture based cards. For example, A10-5800K has HD 7660D, based on VLIW4 but AMD does not have any GCN based Mobile Graphics card with the same name. But in case of Richland APU, A10-6800K has HD 8670D IGP whereas it is confirmed that AMD is already having a GCN based HD 8670M Graphics card with 384 SIMD unit (6 Compute Engine, 16 ROPS). Here is the link: AMD Radeon HD 8670M - NotebookCheck.net Tech

So it is highly unlikely that Richland is going to have re branded VLIW4 IGP again.

 

[Skud]

Re: AMD Trinity & Richland APU Discussion

Thanks for the info Cilus, lets hope its GCN based.

Thanks for the info Cilus, lets hope its GCN based.

 

[mohiuddin]

when will the kaveri arrive?