Intel Demos Sandy Bridge, Shows off Video Transcode Engine
Quote:...Sandy Bridge has around a billion transistors and that in 10 - 11 years we’ll have chips with 100 billion transistors...
So we are approaching picometer manufacturing in another 10 years or so.. Sounds good.. Looks like video transcoding is a very important aspect associated with CPU performance nowadays or else why would Intel care to involve dedicated circuitory for it.! However, transcoding just a 30 MB file is no way close to satisfaction.. Guess I'll hold my excitement till the chips actually get launched and put through some real value tests..
P.S. - That logo could've been better.. May be blue instead of yellow and white instead of blue..
Did you guys see this?
Dont care for Fuad, but it quotes Paul Otellini.
So it looks just as I thought. When there is a GPU on die, it has direct acces to the 'Last level" Cache, eg L3 cache where the CPU stores a copy of all the instructions. So it can grab it from there which is faster than asking the CPU.
Plus they are using a Ring based bus. ATI started that in the HD2K series, was a 512bit ring bus memory controller. Gave outrageous memory bandwidth, about the same as GDDR4 on a 256bit bus.
The details are starting to leak slowly. Soon we will see full fledge test on THG and Anand with ES samples and near finalized hardware/software.
@BadTrip.... holy hell..... "This simply proves that Intel was somewhat late with Sandy Bridge microarchitecture and it will get back on track with 22nm processors."
This guy must be a bit twisted. As far as I could tell from the first time that Intel released their 'Tick/Tock' strategy, 22nm was always slated for Q4 2011. He thinks SB is late even though it should be Q4 2010. Every even year is a new uArch, every odd number year is a new process on the same uArch.
LOL - remember when jennyh used to go on about how lousy Intel engineers were, and that AMD/GF's process engineers were superb, blah-blah?? That is, in-between declaring Thuban as the world's fastest desktop processor and other fantansies :P.
Anyway, if Intel does ship 22nm by Q4 of next year they will be back on track, as 32nm really didn't start shipping until Q1 of this year, instead of Q4 of 2009.
I note that the AT articles on Llano state that it won't ship until next summer apparently, so if it and BD are delayed even more then GF could find itself nearly 2 years behind Intel on process.