I made a similar observation using information from another site. It looks like AMD has now entered into a pattern of creating new designs for the high end of the market and moving older designs downmarket with maybe a few tweaks, new features or higher clocks to differentiate them from the cards they are based on rather than creating variants based on the newest technology to fill the different market tiers. The specs I have seen for the 7870 look suspiciously like the 6970, the 7850 looks like the 6950, the 77xx cards look like the 6870/6850, and on down the line. The only cards that are substantially different from the HD6000 series cards are the high end dual GPU card and the single GPU card just below it and the Lombok cards at the bottom of the market. Everything else in the lineup looks like it's being recycled from the previous generation. The Lombok cards look to be a new design with more rendering units than the cards they are directly replacing but not as many as the units that currently occupy the next higher rung on the ladder, so those aren't just renamed carryovers that got shunted downmarket. I think it was necessary for them to replace the cards that occupied that part of the market with new technology so they could do hybrid Crossfire with their APU's and onboard chipsets.
This.
I only expect to see the HD7900 cards with this new memory architecture because they are the only ones that will be able to saturate 256-bits of GDDR5. It would be an unnecessary expense to bring this tech lower down the market when cheaper solutions are still viable there.
According to the chart here,
http://www.neoseeker.com/forums/8/t1677781-amd-hd-7000-specs-roadmap-revealed/
We finally see the long overdue death of the 64-bit memory bus on low end cards with Lombok being 128-bit and everything above being 256-bit. Lombok are the ONLY 128-bit cards in the HD7000 generation so could we even see the end of that soon, as well?
An interesting side note, compare the specs of Cayman XT (HD6970) with those of Thames XT (HD7870). The number of rendering units is the same but Thames XT has slightly more memory bandwith and higher clocks but look at the power consumption at the bottom. Less than HALF that of Cayman XT.
And it's this that proves that simpler designs are better than more complex ones. Everyone who does Bitcoin mining seriously uses AMD cards because they have more horsepower for GPU computing than the competing nVidia designs. nVidias CUDA cores are a joke for Bitcoin mining compared to how AMD does it. AMD would be foolish to ignore this and try to make their cards more like nVidias'.
From the Bitcoin wiki
Why are AMD GPUs faster than Nvidia GPUs?
Firstly, AMD designs GPUs with many simple ALUs/shaders (VLIW design) that run at a relatively low frequency clock (typically 1120-3200 ALUs at 625-900 MHz), whereas Nvidia's microarchitecture consists of fewer more complex ALUs and tries to compensate with a higher shader clock (typically 448-1024 ALUs at 1150-1544 MHz). Because of this VLIW vs. non-VLIW difference, Nvidia uses up more square millimeters of die space per ALU, hence can pack fewer of them per chip, and they hit the frequency wall sooner than AMD which prevents them from increasing the clock high enough to match or surpass AMD's performance. This translates to a raw ALU performance advantage for AMD:
AMD Radeon HD 6990: 3072 ALUs x 830 MHz = 2550 billion 32-bit instruction per second
Nvidia GTX 590: 1024 ALUs x 1214 MHz = 1243 billion 32-bit instruction per second
This approximate 2x-3x performance difference exists across the entire range of AMD and Nvidia GPUs. It is very visible in all ALU-bound GPGPU workloads such as Bitcoin, password bruteforcers, etc.
Secondly, another difference favoring Bitcoin mining on AMD GPUs instead of Nvidia's is that the mining algorithm is based on SHA-256, which makes heavy use of the 32-bit integer right rotate operation. This operation can be implemented as a single hardware instruction on AMD GPUs (BIT_ALIGN_INT), but requires three separate hardware instructions to be emulated on Nvidia GPUs (2 shifts + 1 add). This alone gives AMD another 1.7x performance advantage (~1900 instructions instead of ~3250 to execute the SHA-256 compression function).
Combined together, these 2 factors make AMD GPUs overall 3x-5x faster when mining Bitcoins.
More complex cores are *WORSE* at GPGPU operations. See my post above.
