The Cellar

The Cellar (http://cellar.org/index.php)
-   Image of the Day (http://cellar.org/forumdisplay.php?f=10)
-   -   February 6, 2008: Two Billion Transistors (http://cellar.org/showthread.php?t=16575)

monster 02-05-2008 08:55 PM

February 6, 2008: Two Billion Transistors
 
1 Attachment(s)
Intel has launched Tukwila, the first chip with over two billion transistors. You won't be seeing this in your home PCs -this quad-core chip will be used in high-end servers. Operating at speeds of up to 2Ghz, this chip is not the fastest kid on the block, but it sure is pretty.

info from BBC, ProductReviews

Sheldonrs 02-05-2008 09:20 PM

Looks like a sattelite view of Manhattan.

xoxoxoBruce 02-05-2008 09:27 PM

Go Monster, we've been all waiting with abated breath.
It's a shame, one pair of leather soles and a wool rug, could fry that sucker.

btw... That BBC link has a great, simple, explanation of a transistor and how they work together on a chip.

linknoid 02-05-2008 09:42 PM

It's interesting to see that chip layout. In most pictures of processors I've seen, you can see different areas, but it's not so obvious what's what.

In this case, the large black areas are most likely cache RAM. All those lines running through it are probably busses to communicate with the rest of the chip.

Also notice the symmetry, you can see each of the 4 cores separated by large areas of RAM, and in the center is the controller which coordinates all the cores... or something like that.

I might also guess (all of this is guesswork) based on the layout that the two cores on the right are more closely tied, as are the two cores on the left, and for the right side and left side to communicate it has to go through the middle.

I'm really curious how accurate my guesses are, but I doubt I'll ever find out.

While 2 billion transistors sounds like a lot, it's not quite as intimidating when you consider that that's only about 31000 transistors wide. Kind of like how you can cross a 2500 square mile area in under an hour, because it's really only 50 miles by 50 miles.

classicman 02-05-2008 10:13 PM

Quote:

Originally Posted by linknoid (Post 430101)
While 2 billion transistors sounds like a lot, it's not quite as intimidating when you consider that that's only about 31000 transistors wide. Kind of like how you can cross a 2500 square mile area in under an hour, because it's really only 50 miles by 50 miles.

Interesting comparison. never looked at it that way.

xoxoxoBruce 02-05-2008 10:38 PM

Yes, but the difference comes in the 50 mph roads, instead of the 10 mph ruts they started with, not that long ago.

monster 02-05-2008 10:50 PM

Quote:

Originally Posted by xoxoxoBruce (Post 430099)
Go Monster, we've been all waiting with abated breath.
It's a shame, one pair of leather soles and a wool rug, could fry that sucker.

btw... That BBC link has a great, simple, explanation of a transistor and how they work together on a chip.

:lol:

I was also interested to read about Moore's Law:
Quote:


The chip industry is driven by Moore's Law, originally articulated by Intel co-founder Gordon Moore in 1965.

The industry axiom states that the number of transistors it is possible to squeeze in to a chip for a fixed cost doubles every two years.
But mostly I just thought it was pretty :D

lumberjim 02-05-2008 10:54 PM

Quote:

Originally Posted by xoxoxoBruce (Post 430099)
Go Monster, we've been all waiting with abated breath.

you and John Madden.

xoxoxoBruce 02-05-2008 10:56 PM

tw refers to Moore's law frequently, in the tech section, but many people avoid that, so this is a more likely place for people to grasp it.

monster 02-05-2008 11:01 PM

Well I've heard of it before too, despite not dipping into the tech forum very often. Apparently, my bachelor's degree is in computer science, but my degree is so old, it was still considered an "interesting statement" then, rather than an "axiom of the industry" :lol: My degree is so old, I was one of the first students to get a university email account as standard issue! (1989)

xoxoxoBruce 02-05-2008 11:06 PM

It's an industry thing for engineers and MBAs.

tw 02-06-2008 02:20 AM

Moore's Law is more of an industry benchmark to determine if the industry is still innovating.

What makes that chip possible is high-K dielectric that Intel took a great risk on in Jan 2006. Hafnium (with some other secret materials) made it possible to replace glass (silicon dioxide) with a gate material that does not leak so many electrons. Less leakage means less power consumption and less heat. This risk and its resulting rewards have left AMD in trouble both financially and in competitive products. AMD is suffering from processors that are either too hot or too slow. The buzz word for this new Intel technology is 45 nm transistors. AMD is just getting to 65 nm transistors.

Intel produces annually more transistors than there are stars in the universe.

Meanwhile, the next generation of processing is multicore. But as Sony's Playstation Three demonstrates, we have a serious problem making software that can utilize multicore processors. Where would be a good place to invest? Who can solve this software problem now that Intel has solved the Moore Law challenge using multicore processors?

nil_orally 02-06-2008 03:10 AM

Is it just me that sees a cubist representation of a robotic goatse in this image?

monster 02-06-2008 07:01 AM

Quote:

Originally Posted by nil_orally (Post 430183)
Is it just me that sees a cubist representation of a robotic goatse in this image?

thanks, now I do too :rolleyes:

glatt 02-06-2008 07:46 AM

Quote:

Originally Posted by monster (Post 430134)
I was one of the first students to get a university email account as standard issue! (1989)

I was too! I had to ask around, talking to a few professors before I found out who I needed to talk to to get an e-mail account. There were (I think) three terminals on campus I could use to check my e-mail, but only one was in a public location. The others were in physics labs, etc.

I was the only student I knew of with an e-mail account, but I heard from the professors that there were a couple others. This was 1988.

dar512 02-06-2008 09:09 AM

There are a number of "laws" to go along with Moore's law. Programmers and techies of all sorts have a fondness for them.

Here are a few:

Ellison's Law: The two most common elements in the universe are hydrogen and stupidity.

Hanlon's Law: Never attribute to malice that which can be adequately explained by stupidity.

Hoare's Law: Inside every large problem is a small problem struggling to get out.

Occam's Razor: The explanation requiring the fewest assumptions is most likely to be correct.

Cole's Law: Thinly sliced cabbage

Flint 02-06-2008 09:10 AM

But what is that sauce that goes on it? That's where the magic is.

torgut9 02-06-2008 09:54 AM

Is this fake-proof picture??? Over 2 billion transistors..??!!
*starts to count every single of them*

(wish me luck)

Gravdigr 02-06-2008 01:01 PM

Quote:

Originally Posted by nil_orally (Post 430183)
Is it just me that sees a cubist representation of a robotic goatse in this image?

Goatse. Tubgirl. Two girls one cup. Two girls one finger. There, we never have to hear about any of them EVER AGAIN!!!!:headshake

Shawnee123 02-07-2008 11:47 AM

1 Attachment(s)
I see this:

deadbeater 02-07-2008 10:52 PM

In ten years the Tukwilla will be part of a 16-core CPU/motherboard that will fit about the size of that current chip. The board will be used to play iTunes.

monster 02-07-2008 11:16 PM

In 20 years, the tukwila will be worn as jewelry in a junior high near you.

monster 02-07-2008 11:17 PM

Quote:

Originally Posted by dar512 (Post 430230)
There are a number of "laws" to go along with Moore's law. Programmers and techies of all sorts have a fondness for them.

Here are a few:

Ellison's Law: The two most common elements in the universe are hydrogen and stupidity.

Hanlon's Law: Never attribute to malice that which can be adequately explained by stupidity.

Hoare's Law: Inside every large problem is a small problem struggling to get out.

Occam's Razor: The explanation requiring the fewest assumptions is most likely to be correct.

Cole's Law: Thinly sliced cabbage

You missed Murphy

tw 04-30-2008 12:34 AM

Quote:

Originally Posted by tw (Post 430181)
Meanwhile, the next generation of processing is multicore. But as Sony's Playstation Three demonstrates, we have a serious problem making software that can utilize multicore processors. Where would be a good place to invest? Who can solve this software problem now that Intel has solved the Moore Law challenge using multicore processors?

Multiprocessing, multicore processors, parallel processing, and the associated software appears to be a next bottleneck for the computer industry.

From the NY Times of 30 April 2008:
Quote:

Race Is on to Advance Software for Chips
Stanford University and six computer and chip makers plan to announce Friday the creation of the Pervasive Parallelism Lab. Besides Stanford, the backers are Sun Microsystems, Advanced Micro Devices, Nvidia, I.B.M., Hewlett-Packard and Intel.

Last month, Intel and Microsoft announced they were jointly financing new labs at the University of California, Berkeley and the University of Illinois at Urbana-Champaign to tackle the same problem.

All three efforts are in response to a growing awareness that the software industry is not ready for the coming availability of microprocessors with 8 or 16 or more cores, or processing units, on a single chip. Computer and chip makers are concerned that if software cannot use the new hardware efficiently, customers will have little reason to upgrade.

dar512 04-30-2008 09:04 AM

I think the answer will be computer languages (and to a certain extent operating systems) that help the programmer facilitate multi-processing.

Coding for multiprocessing can be incredibly easy - if none of the tasks depend on the other or share resources. But that's not what people are going to want. Most people can't multitask past a few tasks. They want the one or two tasks they're working on to go faster.

In order to take advantage of multiple cpus in that scenario, a lot of interdependence and sharing of resources has to happen. This is typically pretty klunky to do in current languages. I suspect the next big thing in computer languages will be the one that sorts this stuff out.

Imigo Jones 04-30-2008 09:26 AM

Quote:

Originally Posted by linknoid (Post 430101)
I might also guess (all of this is guesswork) based on the layout that the two cores on the right are more closely tied, as are the two cores on the left, and for the right side and left side to communicate it has to go through the middle.

Link, the middle is called the corepus calculosa.



:p :headshake

tw 04-30-2008 10:45 PM

Quote:

Originally Posted by Imigo Jones (Post 449765)
Link, the middle is called the corepus calculosa.

Will you want a left handed or right handed computer?

Imigo Jones 05-02-2008 12:56 AM

Quote:

Originally Posted by tw (Post 450005)
Will you want a left handed or right handed computer?

Left-handed, please, tw. All hemispheric things being equal, and for what it's worth, the cor(e)pus cal(cu)losum is generally a bit more substantial in left-handers.

http://www.carleton.ca/49.663/1.jpg

"Okay, now stretch. . . . That's is. A little more. . . . Str-r-r-e-t-ch. . . . Okay, hold. Good, . . . a-n-d release." Sproing! :3_eyes:

Torrere 05-04-2008 11:55 PM

All that, and they named it after an airport town...

Agent-G 05-05-2008 01:57 AM

these chips must be fun to build......

Bitman 05-06-2008 01:31 PM

Quote:

Originally Posted by linknoid (Post 430101)
that's only about 31000 transistors wide.

"Only"?? The original 8086 that kicked off the personal computer revolution had 29000 transistors. This quad Itanium is like an 8086 squared.

tw 05-07-2008 10:17 PM

Quote:

Originally Posted by Bitman (Post 451436)
"Only"?? The original 8086 that kicked off the personal computer revolution had 29000 transistors. This quad Itanium is like an 8086 squared.

Non-editor's note: the 8086 was not used in PCs. 8088 was used. The personal computer revolution was pioneered by the 8080 in products such as the Altair 8080. Microprocessors starting with the 4004. 80186 was supposed to appear in the PC Jr but quashed by ignorant IBM management. 80286 - the first 16 bit IO personal computer. Then many who are now in the Cellar were born.

xoxoxoBruce 05-07-2008 10:23 PM

Uh, do you mean were not born?

classicman 11-28-2016 01:32 PM

Update ...

Gravdigr 11-28-2016 02:10 PM

I understood "Washington". The rest may as well be Klingon...


All times are GMT -5. The time now is 04:31 PM.

Powered by: vBulletin Version 3.8.1
Copyright ©2000 - 2025, Jelsoft Enterprises Ltd.