Slashdot Mirror
IBMs CMOS 9S
TigeR writes "I saw over at 3DNews that IBM has just unvieled some new chipmaking technology. " Its called CMOS 99:"copper wiring, silicon-on-insulator (SOI) transistors and improved, "low-k dielectric" " All this and 0.13 microns. Smaller chips with more punch using less electricity. Everybody wins. Gimme now.
I remember a while back IBM announcing that they'll be using SOI technology with Alpha procs. CMOS 9S (not CMOS 99) and Alpha? What a team. The press release from the horse's mouth is here. Thet whole "low-k dielectric" thing reminds me of an article I read some time ago (June 1998) in Discover.
"`Ford, you're turning into a penguin. Stop it.'" -Douglas Adams, THHGTTG
The article was guessing that 9 stood for 9 layers. I would think since the other feature of this chip is Silicon-on-Insulator its safe to say S is for Silicon!
They read like Guiness Book entries. The world's largest submarine sandwich, largest grilled cheeze, largest ________ sandwich.
They are all chips but our special feature in doing it is _________ and that means we can put out a Gee Whiz press release.
The rules say we should be impressed but pardon if I don't have enough "impressed" emotions to go around.
--
A feeling of having made the same mistake before: Deja Foobar
Those aren't chips, they're DIMMs. A DIMM consists of either 4 or 8 chips on one board (I forget which number).
I believe a 10Ghz G3 would toast a 500MHz G4.
You listnin' Motorola?
These are my friends, See how they glisten. See this one shine, how he smiles in the light.
I'm sure Motorola will have a benchmark that shows that the 500MHz G4 is faster. :-)
The next Cmdr Taco duplicate will be ready soon, but subscribers can beat the rush and see it early!
e= mc squared? nahh its about 2.718281828459045235602874713527
Nah, it is E=mc^2. It's case-sensitive, you know.
The young ones must be out of school for vacation. I think /. has a worse time with them than radio shows.
Yep, I never spell check.
More incorrect spellings can be found he
With the usual inverse proportionality between size of component and size of packaging, as we approach incredibly small chips, I forcast the need for individual 747s to transport each unit.
Daniel
And in other news, Water is Wet, Gravity Sucks, and Nader Still Lost.
--Joe--
Program Intellivision!
Program Intellivision!
This is nothing new. IBM has always had the best process technology in the industry, at least for the last 10 years. What IBM has been missing is a marketing department that announces vaporware (see Intel announcement today). If you look at it, IBM has invented just about every process technology that has enhanced IC performance in the last 10 years.
- shallow trench isolation (STI). People considered IBM crazy when they said they'd just polish the wafers essentially in a mud slurry and keep defect counts low. IBM pulled it off and everyone followed.
- Copper interconnect. Copper is a deep trap in silicon, so people said you'd be crazy using this stuff in a fab. If you do it incorrectly, you can contaminate the entire fab. IBM did it.
- SOI technology. Other companies cancelled their SOI programs in the last downturn, IBM kept it alive. They are now harvesting the fruits.
- low k1 (leading to lower capacitance, faster switching speeds). Admittedly, others are working on it too, but it seems IBM solved the remaining problems first.
In general, there are a whole bunch of areas where IBM is the undisputed leader (e.g. lithography - IC patterning, SOI, DRAM). They don't need to announce vaporware to push their stock, they just do it.
I like the Gimme Now comment... Sur, give it to him NOW, then watch him complain that it wasn't ready for market and that they released it too soon =)
"Intel topping the charts"? What news have you been reading? Can anyone name a real technology breakthrough like this that Intel has come up with in the last 5 years? The only thing Intel tops the charts in is announcements of lower-than-expected earnings and recalls of their newest processors. There's no "technology deathmatch"--IBM is winning, hands-down. Intel is just a crappy monopolistic giant trying to force second-rate technology like RAMBUS down everyone's throat. Even worse, I have to work at this dump... I'm going to be looking to move to IBM as soon as possible.
Well if you scroll Slashdot's homepage down just a tad bit further you'll notice the post announcing 30 nanometer transistor technology achieved by Intel. Don't be blinded by the hype. This company has made some bad decisions in the past but that doesn't mean they can't come up with nifty hardware tech anymore.
Just out of curiosity, what job do you have at Intel ?
With great power comes great electricity bills.
PowerPC. It's IBM...when's the last time IBM created an x86 architecture processor?
Blar.
SOI is significantly different from SOS, for one thing the insulator is not sapphire, and the advantages do look like they will be able to compensate for the higher cost. Chief among them are freedom from body effect and latch-up and ultra low-voltage operation. People have been interested for years in SOI technologies, including SOS, but have in the past only used them for radiation-hard military/space type things because of the cost. Today, with much lower defect densities in SOI wafers, the cost is decreasing dramatically.
I can't imagine a 0.13 micron chip. Packaging must be a nightmare.
Not really. All you need is a bag of your regular ~3-7 cm chips and a large mallet...
They've pretty much stripped the latency down, almost to the cell-level factors I spoke of. Plus those imbedded DRAMs are quite a bit smaller than the 256Mb stuff seen in commodity SDRAM today. Remember that imbedded DRAM doesn't need big off-chip drivers, doesn't need multiplexed addresses, doesn't even need multiplexed data input and output. Some amount of conventional access time comes from all the stuff added to make a DRAM talk to other chips over a small number of pins.
The living have better things to do than to continue hating the dead.
This is allow chip performance to be increased
:-]
Going to 0.13 circuitry get smaller
these guys must have went to the same grammer school
"My mother never saw the irony in calling me a son-of-a-bitch." - Jack Nicholson
Here
Actually, I'm using Windows 2000, and I could leave it on for months on end, but my parents don't like having me leave the computer on for so long. Also, it's their power bill, so I only leave it on when I'm using it.
"Ancillary does not mean you get to rule the world." --U.S. Circuit Judge Harry Edwards, speaking to the FCC's lawyer
IBM's announcement
The Register article concerning 10 GHz Power PC processors.
--
A feeling of having made the same mistake before: Deja Foobar
...the linux substrate registry won't correctly identify this manufacturing process.
Treatment, not tyranny. End the drug war and free our American POWs.
See my user info for links.
the technology is called CMOS9S not CMOS 99
"My mother never saw the irony in calling me a son-of-a-bitch." - Jack Nicholson
You can find this yourself by searching "silicon spheres integrated circuits" on Google.
Yeah they are called Ball Semiconductor. They want to process the balls in evacuated glass tubes to reduce the expense of manufacturing facilities. Try www.ballsemi.com
First guess at an URL was www.ballsemi.com.
There are several problems when doing 3D chip design, even using modular "beads"... most of the problems have to do with propogation. That is to say clock signal propogation and heat propogation.
It's impossible to make 3D "modules" that don't end up burying circut components *very* deeply (think about a 16x16x16 bit "processor": even if each junction has just three transistors (up, down, right) you end up with transistors that are 24 times their size away from heat dissipation -- in ANY direction. they don't have a substrate to wick away heat, nor a nice big surface of nice heat conductors (read: metal) only a few microns away (like in the case of your M1,M2... layers in traditional chip fab). Of course, this will typically be MUCH higher since at each element in the matrix you'll want to actually DO something instead of just switch...
-Chris
PS. I'm talking out of my ass so if you moderate me up, I'll cut off your balls.
...More Powerful than Otto Preminger...
We have just received word that Intel and Pringles are in league with one another. Apparently Pringles' motto: "Once you pop, you can't stop." and the x86 architecture assembly language command "pop [reg]" go hand in hand. There are reports from users that the new 386 instruction, "popad", actually displays Pringles ads on the display. We will have more at 11. :)
---
(due to the lack of time in writing this, I grabbed the first link containing the info I need from a search. It may or may not be the best ASM document, but that's not my point here
---
Apparently Ball Semiconductor is the company making this product. What I had believed to be 3d modular architecture is a stand-alone product. The 3d sphere itself has the whole chip etched on it. Ball Semiconductor states that they can produce chips at a much higher rate .
This could all be propaganda. I am interested in the heat dissipation on these bad mamajamas though, so I emailed them to get some tech specs on how they deal with this.
l8rs
In the distance you hear an ominous moo.
Boot time is such a worthless metric. And in the media, they're not only using it, but are *excited* about it. What gives? I reboot my computer as little as possible. Advances in power saving modes seem to be a superior trick, since not only does the computer come online in minimal seconds, but it's exactly in the state you left it in. Clamoring for reducing boot time on computers seems equivalent to clamoring for cup holders in cars -- nice, but when it comes down to it, should you really care?
the advantage of ball semiconductors is manufacturing. you can mass produce them at an incredible rate.
however, the draw back is that the transistor size is MUCH larger. (5 microns, i think) in other words, we won't see these 2 technologies combined for a long time.
i don't have time to look up the details but if i remember correctly, their transistors were on the order of 5.0um. Thats about 30 times larger than the 0.13um transistors that we can print on wafers.
do a search for "ball semiconductor" and you will find a lot of information.
http://www.ballsemi.com
86x or PPC ?
Is this a leap over Intel current capabilities.
An old college roommate had a theory that they simply put one Pringle's chip in the bottom of an empty can and sealed the can shut.
The Pringle's chips subsequently reproduce asexually. When all oxygen and space resources are consumed, they cease to reproduce.
Voila, a full can of eerie, perfectly-formed potato chips.
:)
If it ain't broke, it doesn't have enough features yet.
I still think Pringles lead in chip technology. How do they make them so perfect?
GONE FISHING
That would be very useful, especially for notebooks. I don't reboot my desktop PC much, so I don't care much on its behalf, but for a mobile device it would be wonderful. For them, instant on is essential. I would rarely, if at all, use my Palm if it took eight seconds to boot.
If you are modding me down because you disagree with me, use the "Flamebait" category, not the "Troll" one.
It's a shame that news like this so seldom gets people excited anymore. "They made a faster, smaller microchip!? Who would have thought it?" Leaps in technology like this, however, don't happen automatically. There are researchers busting their asses daily trying to squeeze every last drop of performance out of hardware.
Got Rhinos?
We've been using CMOS processor complexes for the better part of the last 2 or 3 years. Is this really new, or am I missing something?
- Despite popular opinion, I am not perfect.
We do have people working on it. Instead of Intel doing it though, we have some Linux hackers taking a stab at it. Check it out - Linux Bios project.
witty sig goes here
Everybody wins.
Mmmhhh, new stuff. New stuff = expensive stuff = evil(tm).
Will wait.
People replying to my sig annoy me. That's why I change it all the time.
CMOS 99? Isn't that marketing "retro?" Or should I think 2099? Ooh. Cool, far-off future tech.
Evan - needs to hit preview before submitting
I wonder how long before this process is used in mainstream applications? 12-24 months is my guess. Maybe we will see some small equipment, like lighter laptops and smaller CPUs.
Amigori
"The quality of life is determined by its activites."--Aristotle
It stands for layerS, of course. 9 layerS, 9S :)
"The question of whether a computer can think is no more interesting than that of whether a submarine can swim" -EWD
They already make 512mb chips, and as I'm looking on pricewatch here, they apparently have a 1gb pc100 chip. But I see what you mean, this can probably lead to, ya know, bigger stuff. gandalf18
I've got about as much eye-hand coordination as Christopher Reeve. thelung187
The IBMlogic and DRAM processes are somewhat separated because they are drive by different needs. DRAM is driven almost exclusively by density and cost. Logic is driven by performance and wirability. The big sharing point between the two is in the photolithography development.
Incidentally, DRAM is unlikely to move into SOI any time soon, because the raw wafers have too many defects. For ordinary circuits that class of defects doesn't really matter, but when you're trying to store fewer than 50,000 electrons for 64,000,000 nS, they can kill.
DRAM is much better off in bulk or epi silicon, rather than SOI. Besides, there's so much density and cost pressure that relatively crude, slow devices are used. Even if one wanted to pay for faster transistors, it wouldn't do you much good. The paramount need to shut off the switch into the DRAM cell (so it can hold that 0 or 1) means that particular transistor *can't* be optimized for performance, and that one link can quickly become the performance-dominating factor. In other words, it isn't terribly cost-effective for an ordinary DRAM to pay for fast transistors.
The living have better things to do than to continue hating the dead.
It has been tried before, but an insulating substrate is costly. Both RCA and HP tried "SOS" chips, silicon-on-saphire, but the advantages didn't compensate for the much higher cost.
IBM working on faster chips, Intel reported doing same
:)
Nothing I can say would be any funnier than that headline.
This sig is worse than my last.
Those researchers don't get the respect they deserve because normal people don't rate those breakthroughs as high, because the need for such technical progression, from their point of view, is simply not graspable for them. It's not as apparent anymore, because the applicaitons they use allready run.. on a functional level, computers have little more to offer, it can only get faster and easier to use, but that's about it.
So sure, in the good old days, people cared to go from 16 mhz to 40. Even though that was a small step, together with a few other enhancements to the system architecture, it made windows 3.11 a reality (I'm not trying to say windows is my criterium for progression here, it's just the OS I used back then).
Now clockspeeds jump from 1.1ghz to 1.5ghz or whatever, but John Doe doesn't care about this.. he cares about reading his email now. Windows 2000 runs fine on his 166 or 180, he doesn't really need "faster", that is just a convenience that 'happens automatically when you buy a new pc". The ones who need "faster" are the ones playing games, like his kids perhaps, but then you also see that todays games and gamesystems are shifting allmost completely into a dedicated market. PS2, DC, Xbox,..
I'm NOT saying chipmakers should stop getting on with their new designs and research. *I* WANT these fast things as much as the next guy who likes to game every now and then. But to most people, the difference between 1.1 and 1.5 means as much as the difference between windows 98 and 98OSR2 I think.
Still, *clap* *clap* for Big Blue!.. the one minute it's Intel topping the charts, the next it's IBM.. seems like technology deathmatch at times..
With great power comes great electricity bills.
--
It might be significantly harder to scale the G4 with that vector engine. RISC CPU's are much easier. I'm not saying it couldn't be done, but I believe IBM is more likely to apply it to their own G3 successor.
If you are modding me down because you disagree with me, use the "Flamebait" category, not the "Troll" one.
On behalf of os/390 system programmers everywhere, I'd like to say, "we love ya too". Now, get outta here, ya big lug...>sniff,sniff...
- Despite popular opinion, I am not perfect.
huh? How often do you boot? Oh, wait you're running windows... my bad!
___
___
If you think big enough, you'll never have to do it.
CMOS = Complimentary Metal Oxide Semiconductor
CCS = Compilementary Copper Semiconductor
But the complementary thing is a bit abitrary, so
CS = Copper Semiconductor
is perhaps more appropriate
"This new manufacturing technique will be used to produce future generations of the IBM POWER4"
The POWER4 is not a PowerPC processor, but they are related. The best way to define the two, POWER and PowerPC, is...
The POWER4 will be a 4 chips on a single die. Each chip in the POWER4 will consist of 2 processing cores with a shared L2 cache. Speed between cores in a chip will be >100 GB/s. Speed between chips will be >500 MHz (>35 GB/s).
More details can be found at http://www.chips.ibm.com/news/1999/microprocessor9 9.pdf
- George
"And now you shall learn the secret of boot to the head"
Shift to the Right!
Pop Up!
Push Down!
Byte! Byte! Byte
Michael D. Crawford
GoingWare Inc
-- Could you use my software consulting serv
You can find more information from IBM here and here.
now combine this with intel's nanotechnology, the 3-atoms-thick thing. of course, that would require a lot more cooperation than everyone would be willing to provide. but if this eventually happens, we may hit the 10GHz mark even faster than everyone thought.
...is to pump out 512 MB and 1 GB RAM chips. If the process can be done in a conventional fab facility, it shouldn't take more than year or two. RAM is, as solid state devices go, among the simpler things to make. (It's still complicated as hell, but not compared to processors or whatnot. That's why RAM has a lifetime warranty.) However, I have no clue as to whether or not most mobo's would support a RAM chip of that capacity.
And if I'm way off on the semiconductor stuff, I'd be as interested as anyone to find what's correct.
Evan
I'm not an IBM fanboy or anything but... Get on with your bad self IBM!!
Assuming this isn't vaporware, this is the 3rd time in a month IBM is in the news with a new tech. My point behind this post is that IBM was downgraded (IMO) to underdog status because of stagnation in the mid 90's. Now look at em, they're pushing ahead into the future as hard as AMD/Intel by revolutionizing and inventing instead of incrementing technology.
"Me Ted"
BOSTON SUCKS!
and then in a desperate attempt to get back on topic(!)
I think the S in CMOS 9S is in some kinda rare isotope of Silicon, Silicon-9, which is soooo rare they can scarcely manage enough to make a chip of 0.13 microns. No there's a low fat chip...
oooh, it was almost a good save, but that pun will surely be rewarded with a downward mod...
--
A feeling of having made the same mistake before: Deja Foobar
Yes, straying from the topic, but still semi-relevant. Recently, Intel set a goal to have a PC boot from the BIOS in less than eight seconds. Currently, that figure is at around 30 seconds with things like on-chip virus detection and ATA-66/100 adapters delaying the boot time. Why can't the companies start working on that?
"Ancillary does not mean you get to rule the world." --U.S. Circuit Judge Harry Edwards, speaking to the FCC's lawyer
Yes, IBM's 9S is more advanced than the current 8S2 process. However the current process (8S2) already uses copper wires and SOI, and the low-k dielectric is already in use in IBM's C-11 ASIC process. In addition C-11 has embedded DRAM on chip.
IBM seems to be fighting with Intel and AMD for the lead in process technology. Remember IBM is also fabing the next generation of Alpha processors as well as Transmeta's Crusoe.
Does anyone else find it interesting that Intel is making all these announcements right before they release their prospective earnings report? Just a thought.
everybody wins
:-)
Well, maybe Intel doesn't win.
I'll see your senator, and I'll raise you two judges.
Does anyone recall the company from Japan that is researching a high speed manufacturing process using spheres of silicon instead of waifers? Apparently they would each have specific functions and be stacked in a 3d matrix for connectivity. The product was ?microbead architecture? or something like that. If anyone could find information on this I would appreciate it.
Increase in production, combined with increase of chip speed= more chips we can use to design faster ways of producing faster chips. Sounds like natural log to me.
e= mc squared? nahh its about 2.718281828459045235602874713527
In the distance you hear an ominous moo.