The Transmeta Conspiracy Part V
Floydian Slip sent us a link to Time Digital which has an amusing 20 Questions with Transmeta. It doesn't say much, but it seems to debunk a few theories. I'm hearing 'RSN' rumors now, but nothing concrete yet. I need moles in Transmeta dammit. I know you're reading, why won't you squeal?!
Just to clear things up. That is, that there isn't a single piece of Transmeta hype out there right now, though admittedly the buzz is being generated specifically due to a lack of hype.
---
"'Is not a quine' is not a quine" is a quine.
"'Is not a quine' is not a quine" is a quine.
Quine "quine?
There are already asynchronous chips. University of Manchester developed the Amulet2e, a ARM7 derivate, from 1993 to 1995. It is compatible to ARM-v4G instruction set.
Comparison data (I wish I could use tables):
CPU: ARM 710 -- Amulet2e fab process: 600nm -- 500nm transistor count: 570,000 -- 454,000 cache size: 8kB -- 4kB speed: 23 MIPS -- 38 MIPS
Even though the cache is only half as big it's faster.
Another one is the TITAC-2 developed in Japan. It is based on the MIPS R2000 but is not binary compatible.
Comparison for that:
CPU: TITAC-2 -- MIPS R2000 transistor count: 496,000 -- 100,000 core voltage: 3.3V -- 5V power consumption / w/o cache: 2.11W / 1.02W -- - / 2W performance / w/o cache: 54.1 MIPS / 26.5 MIPS -- - / 12 MIPS
The TITAC-2 worked with 1.5V to 6V in environment temperatures of -196C to 100C (-320F to 212F, if I calculated correctly).
Clocked chips have to be clocked slow enough for the worst case (a set of commands that take a really long time before the chip enters a stable state), that's why overclocked / overheated CPUs work fine most of the time but only sometimes crash. These async CPUs get faster when they get cooler. They always run as fast as is possible given the temparature and commands to execute.
Rumors tell that German company Hagenuk will employ the Amulet3 in commercial products this year.
If you want to know more about micro pipelines, null convention logic and dual rail encoding, look it up in c't 17/99 (if you have that issue or want to reorder it and can read German of course ;-)
My wild guess at what Transmeta is up to some form of asynchronous logic design. A lot of the fabless chip companies have been making similar promises with that area of research. Theseus Logic and Cogency are two that come to mind. Clocked logic has its limits and the Intels and Motorolas of the world are going to hit those limits sooner or later. And barring any sudden developments in, say, quantum comptuing, asynchronous logic is the next, ahem, logical step. It's much quicker, much more flexible (didn't surprise me about the emulation speeds transmeta boasted of)and should consume less power, if done right. Transmeta, or their followers, is touting all three benefits. Asynchronous logic design actually has been around for awhile, though its always required a complexity of design that makes it more expensive to pull off. But as clocked logic is getting so complex, it might soon be the time where they'll be equivalent, cost-wise. Then again, Transmeta could be up to something *completely* different. joab
This was burried, but valuable. Thanks Chexum.
:)
Subject: Re: OpenGL mini drivers?
Date: 1999/03/31
Author: Dave Taylor
I think that Sean Baggaley pitched in with Russ Williams' pet anti-OpenGL sentiments because they're both British. It's obvious. I mean, come on. Evil empire. Elitist snobs. Still bitter about the "colonials" on the other side of the "puddle." Duh.
I love the D3D vs OpenGL struggle. I love the consistently inconsistent visuals you get from playing 3D games on today's PC's. I wish Microsoft and SGI would reduce each other to charred cinders with patent infringement lawsuits.
John uses OpenGL because he can afford to. He can lose hundreds of thousands of sell-throughs and be perfectly happy. Christ, he could throw every copy of Quake 3 in a wood chipper and sell the chaff to the government for use in hideous anti-personel weapons against the Serbs, and he would still make a killing on the royalty advances. This has always been John's modus operandi, whether he was turning a monster dime or a modest dime, and we should be thankful. Others follow suit because John's products become the IHV's benchmarks, so the IHV's improve their drivers and hardware to work with John's games, and so other game developers can eventually switch over. I think this is a better system than following Microsoft's lead just because one of their over-worked engineers or megalomaniac managers decided to arbitrate a new standard.
There's an alternative. You don't see anyone hotly debating whether to use DirectDraw. You don't see video glitches on games that use software renderers. They just work. They just deliver exactly what you specified and never crash. Sure, you have to come up with innovative gameplay and/or excellent artwork, but the technique has been a solid technique for hit-making from Myst to Heroes 3.
I am quite happy for the 3D cards and API's to beat each other black and blue while the real engineers either deal with what's available or quietly work on the solution to this whole mess.
Here's a fun multiple-choice quiz. What do you suppose that long-term solution is?
- A. Waiting for Nvidia to come out with the TNT N (as N gets large).
- B. Waiting for 3DFX to come out with the Voodoo N (as N gets large).
- C. Waiting for Matrox to come out with the GN00 (as N gets large).
- D. Waiting for ATI to come out with the Rage * (as * gets goofy).
- E. Waiting for OpenGL to get caps bits.
- F. Waiting for D3D code to be easier to read.
- G. Waiting for someone really brave to come out with a general-purpose processor w/ an open architecture that is suitable for high-performance parallel processing so that we have complete control over every pixel and get a consistent, high-quality, fast, innovative graphic experience.
I leave it to you, noble reader. (but if you pick anything but G., then you have gooey fluff where your brains should be, nyah)
-- What you do today will cost you a day of your life.
As a designer who worked for Transmeta until last week, I am risking my very life to bring this information to the /. readers.... I can safely say that the primary product shipped will be Warp Coil Drives that will allow man to travel faster than light. Transmeta's secondary product (the one I was involved with) will be Nachos that never go stale...
01101100 01101001 01101110 01110101 01111000 01110010 01110101 01101100 01100101 01110011
If you rearrange the letters in "Transmeta, Inc", you get "Intranet Scam." Sounds like a good investment.
Other possible letter rearrangements
Martian Scent
Instant Cream
Semantic Rant (sounds like a good name for a band)
Manic Rat Nest
Transient Mac
Men In Scat Art (ugh)
_______
2B1ASK1
Dr Mr Ditzel
I would like to work for you on the basis that there is absolutely no evidence your company is currently developing anything.
Because I am very smart, I have deduced that the average daily work load per employee must be nill.
Assuming this low-stress slow-paced work environment would give me unlimited time to post to slashdot, I would be willing to spend a period of one year at your company before moving on to a future life of playing video games and eating corn nuts while living in the basement of my mothers house.
Regards
Che G.
P.S. I apologise to all posters of slashdot as I know they may not want to read this but I find getting jobs using my resume and a telephone very frustrating. I hope posting application letters to message boards will eventually land me "the Big One".