Months are getting nearer where we will witness the new evolution of CPU combined with GPU to accelerate the performance of the CPU by offloading the video graphics from CPU front to GPU front on the same die.
I was thinking what would be the theoritical and compute performance of Llano CPU excluding GPU.
Since Intel has already jumped on the bandwagon of fusion (technically proposed by AMD from an eternity).
We all have seen the performance of the Sandy Bridge processor for example the performance of
Core i7 2500K is around 10-15% up and down against Core i7 950 and Core i7 2500K is costing around $220 around $70 less than Core i7 950.
AMD says that Llano has 500 GFlops of compute power with the GPU. Not sure the relevance of without it. But the numbers seem about the same as Champlain (no L3) plus maybe 10% for the faster RAM and HT. Plus, you have to examine the Turbo scores.
but I am talking about the cpu capability and since it is being developed on half node of 32nm that it might be around and 2.8+ ghz and taking it for clock to clock.
it would be 5-10% faster than the current generation Phenom II X4 945
or may be less. but will have a very good OverClocking headroom.
AMD has said that Llano would START at 3GHz. It will use DDR3-1866 whereas Champlain is using DDR3 1333. They say that the HT links are faster, it has revamped DIV unit and faster switch times so I stand by the 300MHz better prediction.
OCing will be interesting as it will depend on the nature of the links between GPU and CPU. I for one wouldn't buy a Llano for OC. But then my BE Deneb sits at STOCK.
Llano to bring new power gating thanks to the new approach enabled by SOI manufacturing process, AMD can completely disconnect cores from the power grid. According to AMD, usage of NFET power gating transistor reduces power leakage versus previous power gating solutions by 10 times. Besides, ground-gating can also use the much more conductive chip package for gate supply redistribution rather than a special thick metal layer on the die.
Sorry but that's a crap analysis. It doesn't base it on features, just how fast Nehalem is. If you start looking at features starting from Thuban
wider front end
Dual Prefetechers for DATA AND INSTR
the new RIP (relative instr pointer)
4 issue instead of 3
faster HT links
Full 2x128 bit SSE (no more combining 2x64)
2x128 bit FPU
Fully OoO LD\ST
A LOT MORE CACHE
To me that equals a helluvan increase over Thuban\Lisbon. This doesn't even include Turbo or FMAC. he did say JF mentioned a number but JF also said the single threaded perf would be higher than 17% (50%-33%).
When people start looking at the arch and not at Nehalem, it becomes clear that BD will live up to its name.