I think we are on the verge of a new era of partnership with government - Steve 'Understatement' Ballmer
|
|
|
Boffins create superchip
BOFFINS at the University of Idaho say they have created a single chip more powerful than 17,000 quad core processors that will run on less than 0.03 per cent of the power those chips would require.
The chip will be used on NASA's developing Geostationary Synthetic Thinned Aperture Radiometer (GeoSTAR) project, which will observe hurricanes and other severe storms in the US. It is the latest in a long series of microprocessors created for NASA by the University's Center for Advanced Microelectronics and Biomolecular Research (CAMBR) located in Post Falls.
Sterling Whitaker, who led the team of computer engineers, said that everyone was under pressure against the clock to get the chip made. This was because the team could only use an IBM 90nm fab facility twice in a year.
The chip is responsible for correlating 588 antennas in real-time. This meant that the team had to ensure it fit with the system of electronics featuring many inputs and outputs without crossing any data streams. It also had to run on 120 watts of power.
The system uses a packaging system designed to deliver power throughout the chip via a number of half-spheres spaced evenly across its surface. µ
Share this:
Share
Wow, why don't you just show the original article instead of making a bad copy?
http://www.idahobusiness.net/archive.htm/2009/09/29/Less-is-More-Computer-Chip-Faster-than-17000-Quad-Core-Processors
more powerful than 17,000 quad core processors that will run on less than 0.03 per cent of the power those chips would require.
if that were so then everybody would have one in their phone especially if it's a 750,000 transistor part.
and if that was possible Intel and AMD would have thought of it first. no matter what sort of chip it is you can always stick put x86 decoders on it. that's if it really was that fast.
heck if that was possible you could use it for software 3d rendering and still probably keep up with the last green goblin offering.
bunch of bollocks if you ask me...
I emailed Dr Whitaker and here is his reply:
It is always interesting to see how a reporter interprets what is said in an interview, especially seeing the 'quotes' which you supposedly gave.
We at CAMBR are a part of the over all GeoStar project. JPL is supervising the program and JPL engineers are developing the antennas for the instrument. The Space Physics Research Lab (SPRL) at the University of Michigan is developing the electronics for the proof of concept instrument which will utilize multiple copies (12) of the proof of concept chip CAMBR designed (the final version will utilize nx12 chips where n will be set by pin limitations). The chip was released to fab and will be processed through the IBM 90nm (9sf) bulk CMOS process on a TAPO multi-project wafer run. SPRL also has a test chip on the TAPO run which has a set of A/D converters which might eventually be merged with the CAMBR core design for the full final version (to reduce pin count and power). The CAMBR chip in fab has the core processor which will be tiled on the final version. 2-bit data from two sets of antenna are cross correlated by the CAMBR processor. Data is clocked into the chip at up to 1GHz. Data is integrated for up to 1 second. The processor will consume less than 250uW per correlation. The control circuitry is radiation hardened (the instrument will be on board a space craft and observing the formation of hurricanes in order to expand our understanding of that process). The correlated data is dumped from the chip on a 100MHz output bus. Statistics are also kept on each of the antenna inputs and dumped. The proof of concept chip contains about 750,000 transistors. The size of the core, which will be tiled on the larger final version, is 869um x 1280um. It is a full custom layout to minimize power. The proof of concept chip will be packaged in a Flip Chip Ball Grid Array and has 282 bump pads.
The press release concentrated on the wow factors of the final instrument, but I suppose the wow factors important to engineers would glaze the eyes of most readers and not make good press.
Ok, you say it's 17000x faster, but what score does this 'super chip' get in Linpack, the industry standard benchmark for dickwaving?
"X more powerful than Y general purpose processor at doing Z specialized tasks in parallel."
Ho. Hum. Maybe they'll come up with some cool name like "stream processors" too.
You know what, why don't we up the stakes? We'll build a super fast, "massively parallel," 2-input NAND-operator "CPU". It can compute some odd gazillion NAND operations--in parallel--as well as being so sophisticated that it doesn't even use a "clock." EAT THAT INTEL. Paste that ALL over the BBC's "Technology" section, and on the front right next to the day's "groundbreaking" survey-based psychology research paper about kids using Twitter. Don't forget the stock photo of a silicon wafer.
A gpu is hundreds of times faster at rendering than using software on an x86 but a GPU sux at doing non 3D things
"Anyway I'm sure this chip is more powerful than 17,000 QCcpus." - At doing a particular task, a task that it was built to do.
... "Will it run Crysis?" I will shoot my computer and yours.
Crysis is yesterday, ArmaII is today.
Anyway I'm sure this chip is more powerful than 17,000 QCcpus.
The QC is a general purpose cpu that has many instructions to do a whole bunch of different things.
Now this new cpu probably only needs a few instructions to do only a couple of tasks; As it was built to focus on those few tasks then yes it would be more powerful than a QC.
Simple and general, but I'm sure you get the point.
Also; I'm sure it's super cooled and has a multitude of co-cpus.
Again - ENOUGH WITH THE CRYSIS BS!
... for the specific application that the chip was designed for. Application-specific hardware is always faster than generic compute hardware.
Non-news IMO
Dickens and ALL the poets!
Get reality ppl~!
This is a "CPU". HA!
We all know: that the CPU is dead.
17000 times deader than quad.
Do Uself a Faviour! Get Fermi - he don't need no Katrina pr0n.
The drashional of some empaths is beyond me.
Do not llik the msgr.
It's probably rubbish, but what's actually important is the generalised functionality it delivers.
If it's a single function device with custom hardware, it could well be faster than lots of quad core chips but absolutely useless at anything other than its primary purpose.
In what way is this faster than 17,000 Intel Quad CPUs?
Smells like horse manure.
"The system uses a packaging system designed to deliver power throughout the chip via a number of half-spheres spaced evenly across its surface"
...So it's BGA. Big whoop.
http://www.uidaho.edu/newsevents/item.aspx?id=40174
No additional details, but it's obviously the source (might as well have been the article, really!)
Any sources to back the article up? How have they got such an amazing boost in processing power with only a little bit of time at the IBM facility to create the chip?
Any further info on how the chip works? is it x86? does it have 10000 cores?
does it run at 500Ghz?
How about some info about where Nick got the information? Where can I find out more?
how does it work? And why cant I have one in my desktop?
WOW!!I've got more than me eyebrows raised,lol
thats all well and good but....will it allow me to play Crysis at 9000 FPS? if not the "super chip" is worthless LOL HAH HARR
...they are raised.