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

No,
there is always something good in all the bad things that happen.
Last period when prices had been relatively low was when ATI (or AMD) had to catch up performance wise. So, the relatively low performance can be a factor to keep prices in check, rather than go sky rocket.

Bulldozer shows negative performance scaling in many applications when you compare it to the old X6 1100T.
With Radeon HD 7000 we should see at least 20-30 % performance improvement (yeah, it might be relatively low increase, but it's better than none or negative) over Cayman in all applications. I mean there is no way in hell we will see Tahiti being slower than Cayman, in any application.
The good thing for AMD is that this time they will not have competition for several months and that should give them time to release a second generation GCN product just in time to counter.

the eyefinity users might benefit more due to the larger resolutions.
not to long to go now tho.
 
:oops:
So.. erm.. did this thing happen at London or not?
:D

Normally, you'd think someone would have created a presentation for such an occasion. And under normal circumstances some of the slides of that presentation would have been bound to leak even before the presentation was being held. Now, it's december 6th, and I have yet to come across such a leak.

Not even the infamous OBR (ovski) has more on his blog to offer than AMD bashing without any source. :)
 
London? Munich? Paris? It seems that AMD has a list of dates + places and gave a different combination to each editor to reveal leakers ;)
 
Can we expect any pefromance boost from shinked 69xx chips, or will be only get cooler and les power hungry cards with 78xx? Will they clock them any higher?
 
Last edited by a moderator:
It's now official: the HD7600M are rebadged HD6700M (Whistler Pro), and the HD7500M are practically rebadged HD6600M (Whistler LT, trading a 128bit DDR3 controller for a 64bit GDDR5 one).

So none of those new laptops are bringing new GPUs to the table.
That Asus X53TK is still a spectacular gaming machine for the price, especially if we can count on some +35% 3D gaming performance through hybrid Crossfire.
 
The 64bit GDDR5 gpu could be also a fused off juniper XT (16 ROP and 800 SP) core too. Maybe they want to get rid of the 5k radeon leftovers.
 
Do you conclude that from shader count?

I didn't conclude anything, I simply repeated what is told by anandtech, which seems pretty assured to me.
There's also the fact that with the rise of APUs for mid-to-low end graphics needs, there was little reason to develop a new chip for the same performance target.
 
With UVD3?

Why not :?: The difference betwen UVD2 and UVD3 sounds to me more like a bios update , than real hardware difference.
Support for DivX and Xvid via MPEG-4 Part 2 decoding and Blu-ray 3D via MVC is for 100 percent just driver and bios update :rolleyes:.
 
Why not :?: The difference betwen UVD2 and UVD3 sounds to me more like a bios update , than real hardware difference.
Support for DivX and Xvid via MPEG-4 Part 2 decoding and Blu-ray 3D via MVC is for 100 percent just driver and bios update :rolleyes:.

Unlikely, if it was HD6700 (Juniper) would have support for all those, yet it doesn't.
 
I didn't conclude anything, I simply repeated what is told by anandtech, which seems pretty assured to me.
There's also the fact that with the rise of APUs for mid-to-low end graphics needs, there was little reason to develop a new chip for the same performance target.

Sorry, my mistake. I was referring to this statement only:
„That Asus X53TK is still a spectacular gaming machine for the price, especially if we can count on some +35% 3D gaming performance through hybrid Crossfire“
 
and the HD7500M are practically rebadged HD6600M (Whistler LT, trading a 128bit DDR3 controller for a 64bit GDDR5 one).
Well it is still listed as 64bit ddr3/gddr5. One thing is for sure though, with ddr3 performance is going to be abysimal, right there on intel igp level. Of course I don't know if we'll see such parts, but if past is any indication we definitely will (did you know abominations like desktop HD6450 with low clocked 32bit (!!!) ddr3 exist?).
And, even with 64bit gddr5 I wonder if this makes sense. I thought part of the reason why all lower end nvidia chips are ddr3 only is because 128bit ddr3 is both cheaper and more power friendly than 64bit gddr5, so if the chip is 128bit anyway outfitting it with 64bit gddr5 seems kind of pointless.
 
And, even with 64bit gddr5 I wonder if this makes sense. I thought part of the reason why all lower end nvidia chips are ddr3 only is because 128bit ddr3 is both cheaper and more power friendly than 64bit gddr5, so if the chip is 128bit anyway outfitting it with 64bit gddr5 seems kind of pointless.
Who knows... maybe GDDR5 has become cheap enough that 64bit GDDR5 is more cost-effective than 128bit DDR3.
It might be beneficial for marketing reasons, too. GDDR5 @ 3600 MHz looks better than DDR3 @ 1800 MHz to uninformed consumers that don't read/understand the fine-printed stuff.

The 64bit DDR3 parts are a rip-off, of course.


One thing strikes me as odd, though. Our dear Charlie writes that the 7600M and 7500M are Thames parts, but other sites say they are Whistler Pro (Turks). So either Charlie is wrong (not impossible, of course), or Lombok/Thames is just a new code-name for Turks/Whistler (new stepping or 28nm shrink, maybe?).
It's somewhat confusing...
 
Don't bash on me, but the pattern of those solder points for the memory chips doesn't look like an ordinary GDDR5 device. :???:
 
Back
Top