AMD: Southern Islands (7*** series) Speculation/ Rumour Thread

CarstenS said:
Ok, so after doing a bit more reading (and thinking), it probably makes sense what Jawed and others explained with regard to the SIMD structures not being of much use in order to gather experience for GCN. Taken that for granted, however, does not make it easier to evaluate the role of Cayman and it's revamped albeit obviously short-lived VLIW4 approach. Any further insights into that?


On another, but probably related matter: Has it somewhere been confirmed that or if GCN was the first architecture which was started from the drawing board after the merger of ATI and AMD?
Click to expand...

I asked Eric Demers the same thing regarding Cayman, he said he thought it [Cayman] was.
 
Regarding the waitcnt instructions and function calls/DLLs, I assume the ABI will have to include a waitcnt on all writes as part of a function epilogue.
 
Does anyone have any idea how close the 7xxx series are in terms of release date? Also does anyone know what's happened to all the 6970's? Most online retailers either don't have any, or the ones that do have them for really expensive prices. Heck even Newegg only seems to stock 2 different 6970's

Everyone seems to have tons and tons of 6950's though if various different flavours.
 
caveman-jim said:
I asked Eric Demers the same thing regarding Cayman, he said he thought it [Cayman] was.
Click to expand...

Ah, thanks. I was under the impression that Cayman still was very close to the ATI stuff, but GCN seemed like someone flipped a switch. :)
 
CarstenS said:
Ah, thanks. I was under the impression that Cayman still was very close to the ATI stuff, but GCN seemed like someone flipped a switch. :)
Click to expand...

Cayman design might have started after the acquisition but it's still descendant from R600. GCN will be cleanest break from that lineage.
 
Rootax said:
Yes, in fact I was thinking, good luck maintaining the VLIW 4-5 drivers and the new one.
Click to expand...

Why would that be complex, if the former simply gets locked to the state it's in now?
 
Last edited by a moderator:
AlexV said:
Why would that be complex, if the former simply gets locked to the state it's in now?
Click to expand...



If the former is locked, yes it would be simple. But is it realistic (it's a serious question, I don't know) ?
I mean, you have always new apps, new games, new OS with Win8,etc, and I don't see AMD told 5xxx and 6xxx users to gtfo.

I'm not naive, I know the old generations are always on the back burner development wise, but still, you have to support them. My english is not good so I've hard time explaining myself, sorry :-/
 
Last edited by a moderator:
They could move the (first?) DX10 generation to a legacy support structure. Last time this happened, the oldest card affected was about 6.5 years old, the youngest about 2.5 years (9700 Pro Q3/2002 and X1950 XTX Q3 2006 respectively).

HD2000 series has turned four years a few months ago, HD3000 series is about 3-3.5 years old.
 
if i'm not wrong win8 will use the same driver model as win7 so that it will not be a necessity to update fast the drivers
and some page ago someone said that shifting from vliw5 to 4 is relatively simple
 
CarstenS said:
They could move the (first?) DX10 generation to a legacy support structure. Last time this happened, the oldest card affected was about 6.5 years old, the youngest about 2.5 years (9700 Pro Q3/2002 and X1950 XTX Q3 2006 respectively).

HD2000 series has turned four years a few months ago, HD3000 series is about 3-3.5 years old.
Click to expand...
Note though r3xx and r5xx were abandoned together, even though the differences especially in the pixel shader alus were huge.
HD2000/3000 is mostly the same, and even HD4000 is very similar from a programming point of view. The opensource driver indeed mostly splits the code (when necessary) along r600/evergreen/cayman, r600 here including r700 too. So I think it wouldn't make sense to abandon only HD2000/3000 but not 4000. Now HD4000 is also quite old (2.5 - 3 years with the exception of rv740) so that may be possible (though AMD still currently sells HD4000 based IGPs) but it probably doesn't really simplify things that much if they still have to maintain HD5xxx based cards. And obviously AMD can't abandon VLIW5 completely yet.
 
They had one thing in common, which made their differences seemingly small IMHO: They weren't unified.
 
TBH I'm expecting drivers to be a big factor for GCN. Lots of performance unlocked over the first 6-8months of release. If the APU's are using VLIW-4 and VLIW-5, and the new series is split arch with VLIW-4 entry/mainstream/performance cards and only GCN for the top end, there is lots of reasons for the driver team to focus first on the money segments and not the halo segment, except for specific applications ('games' as they are known sometimes).

I hope (and suspect) there has been a influx of talent for drivers and software, to boost responsiveness on many fronts (lots of new OS's supported, needing to be tested and verified) so that the APU's get the attention they need without starving the new gen GPU's.
 
You must log in or register to reply here.

Similar threads

I
AMD: Pirate Islands (R* 3** series) Speculation/Rumor Thread
140 141 142
Replies
3K
Views
616K
kalelovil
K
N
AMD: Volcanic Islands R1100/1200 (8***/9*** series) Speculation/ Rumour Thread
141 142 143
Replies
3K
Views
675K
pMax
pMax
S
AMD: Sea Islands R1100 (8*** series) Speculation/ Rumour Thread
69 70 71
Replies
1K
Views
389K
AnarchX
A
Back
Top