they could produce slow IC's with low yields now for low power. making working SRAM is much much easier than a cpu or gpu. intel demoed 22nm sram at idf and its not going to be ramped until 2H 2011.
Printable View
they could produce slow IC's with low yields now for low power. making working SRAM is much much easier than a cpu or gpu. intel demoed 22nm sram at idf and its not going to be ramped until 2H 2011.
Obviously not SRAM, most likely an FPGA from experimental customers like Altera or Xilinx.
Should probably be go for GPU in 6-9 months, depending on how long they've come. GF/AMD seem to be promising extremely loose schedules for whatever reason lately...
On the one hand you have GlobalFoundaries never having done such a small scale process before for a gpu and on the other hand TSMC screwed up 40mm.
So which way does ATi jump?
Andy
just cause GF has never built chips, does not mean they hired people who never have
are the people working for GF full of experience?
Technically yes. But it was just a tweaked 65nm process used on the 2600 and 2400 series not a genuine. This is really more of a way of one upping Intel's 65, 45 and 32nm with 55nm, 40nm and 28nm processes than a genuine node change from improved tech. It's marketing really
its price, die size and power consumption are typical of a midrange gpu. AMD clearly stated that they compete in the high end with their dual gpu card. this strategy became hard to notice when rv770 was so close to gt200.
Vrrrrrrrrrrroooooooooooooooooooooooooooooooooooooo ooooooooooooooooooooooooooooooooooooooommmmmmmmmmm mmm
im having problem seeing amd doing a half node on a cpu.... so this might be a test waffer for ic's ..... remember they need to be ahead of the tech to make sure their future cpu or gpu design actually works ok etc..... bring on the old amd :D
Good news, but sounds a bit a bit too late in my calender. Does this mean that ATI is not going to counter the Fermi lunch?
I suspect this could become a good excuse to just look at Fermi without even a refresh?. I hope not.
I don't know what's going on, really. But in good old days, ATI and Nvidia had a good fight, with countering each other's GPU almost simultaneity by lunching close to each other.
Nvidia had it's own "excuses" and didn't counter the 5x-series. I hope ATI is not going to use this as a "excuse" to jump over countering the Fermi (at least with a refresh). In case it is gonna to be bad news for Fermi-prices.
If ATI doesn't release a refresh to counter NVIDIA's Fermi, I just hope they get their next generation out of the door later on this year. Because a 28nm card does sound nice.
A 28nm card does sound nice, for sure. But that sounds a bit too late for making any news now, or be interesting for the purpose that I had in mind.
Be realistic. AMD will probably release new line this year/begining of 2011, but still on 40 nm. 28nm process is too way ahead, if something, we can expect another card like HD 4770 for testing. AMD will roll out on 28nm only if Fermi sucked too much thus no hurry needed and than we would see 28nm cards between H1 and H2 2011.
AMD is most likely going to have some form of refresh on 40nm to counter Fermi (depending on performance) and then make a test chip on 28nm before moving on.
The R800-series is the last of the R600-derived architecture supposedly, and so whatever comes next will be a new architecture, which gives AMD even more incentive to test out the 28nm process first on an existing architecture before moving on. RV740 probably saved AMD a ton of money on the transition to 40nm
am i the only one who see's this as only news that ati is allready doing test on a 28nm waffer..... and btw amd/ati will never do the 28nm without doing 32nm before .... the smaller the fab process the more painful they are... so it wouldnt make sense to forget a full node and go straight to a half node process etc......
GloFo's 32nm SOI tech is only for CPUs. For GPUs they'll have 28nm bulk:
http://img.hexus.net/v2/channel/news...slide1-big.jpg
how fast in general can they go from risk production to early production silicon... like say first test phase before they finalise the tweaking of a process ????
6 months ??? or even more