Slashdot Mirror
Does Dvorak really know what Transmeta is Doing?
Asdren writes "Dvorak supposedly sheds some light on
what Linus and Transmeta are up to. Check out his opinion
column here"
Low power x86 chips. Never heard that before *grin*.
← Back to Stories (view on slashdot.org)
He's up there with Berst and Pournell for cluelessness.
It would be cool if these can be put to use in an SMP board. Even if they were slow as hell as long as they were cheap I wouldn't mind about 500 of them in my comp. Let see. I'll have about 300 doing x86, about 100 emulating voodoo2 in sli (he he), and the remaining 100 are custom distributed.net codebreaking cores.
Here's another prediction. The chip must be 64-bit. There's no way that they would be producing a brand new 32-bit chip. Especially when they're are rumours of Compaq developing a 128-bit alpha. Although, with their arcatecture (sp) could they do as many bit as they want?
except Dvorak getting something right.
hes an idiot and always writes crap. dont listen to his shit. he sucks, and he uses windows. fuck him.
Will we be able to create our own chip instruction sets, just for the fun of it?
It's seems to me that the author has not read Transmeta's first patent. They are working on VLIW chip without microcode. Emulation of another chips (not only x86!) is performed by software, which translates target instructions into host ones, just like Java JIT compilers.
Dvorak is not bad for a guy whose weekly opinions are archived. I respect him. Of course, he's not the Pope.
Well put! Stuff that grouchy little woman!
Divorak is usually wrong... But sometimes he's at least partially right.
Linus knows x86-i386 instructions really well. Linux takes more advantage of the i386 instruction set then most other operating systems. There has been much speculation that Linus will be doing some sort of x86 emulation of an SMP RISC processor while working at Transmeta. This has been brought up several times already. It's still speculation based on leaked reports from some employees as well as a list of the employees and their known specialties.
Whatever Transmeta is doing, it is going to blow everyone away. You don't hire some of the best and brightest chip designers and then hire Linus Torvolds, unless you are about to redefine everything we know in computing! This company is very ambitious and they have put together one hell of a team.
I can see Linus porting Linux to a new RISC processor natively. Remember when he ported to the Alpha (after someone donated one to him) within just a few weeks.
> host neosilicon.transmeta.comm has address 206.184.214.14
neosilicon.transmeta.co
hmm
Maybe he can write an article featuring some sort of boldface to get his point across. Even tho it manages to piss everbody off and makes his stuff more unreadable than it already is. Still, apparently boldface is a substitute for coherant writing ability and strong arguments.
perhaps, slashdot readers will again prove to the world how generally clueless they are.
First off, Dvorak can screw off.
Secondly, whatever Linus is a part of, I'm buying it. I don't care if it is too expensive, et cetera. I'm buying it.
Finally, hey Dvorak screw off!
CM
It makes sence that it would be 64 bit (or have a 64bit mode) as Linus seems to have something against 32bit. :)
128bit? You misread something.. Of course, you never can tell with the marketing people.. I'm supprised that the PII isn't labled 64bit (memory bus) or 80bit (fpu regs)..
Some loony at The Register wrote an article that had a one line mention of a 128-bit Alpha. It didn't say anything else, no details at all. It doesn't make any sense but then it is perfectly in line w/ the register's tabloidy lack of substance reporting style.
In short just ignore it because my 1,024 bit CPU will kick everyone else's ass.
*publically* traded? No wonder why I posted this as an Anonymous Coward. -publicly- :)
Laying aside the knee-jerk Dvorak bashing, the article sheds a little light at least on Transmeta's probable marketing strategy. Instead of marketing their chip as a "CPU for all platforms," they'll market it as an x86 clone that can keep up with Intel's "upgrades" to the instruction set. Not a mind-blowing revelation, but mildly interesting.
about some of it. The low-power-x86 is what they are working on. As for reprogrammability, I wasn't told, and didn't ask.
BTW - I was told there were going to be two primary OSes for the chip. WinNT5 and Linux. I was told in that order.
What's the advantage of 64 bit over 32 bit (or 32 bit over 16 bit for that matter), other then more memory? It will be a while until we reach the 4 gig memory barrier, so I don't think 64 bit architecture is necessary. What are other advantages?
Flames away!
./ poster is rapidly declining!
Don't feed me, I'm a troll!
The average intelligence of the
Transmeta would have to be stupid stupid stupid to try to take on the desktop market, especially with a very low power chip. Think a bit here people. M$ can be killed by Linux because it doesn't take multibillion dollar fabrication plants to make an OS. Hardware is a different world.
The market for very low power chips is in small devices. Low power means small batteries, low heat and therefore no fan. It also means underpowered. Still, many small devices don't need much processing power if they're implemented efficiently.
Imagine your Snap! network server, PalmPilot, auto GPS/navigation gadget, smart cellphone and 8.5x11x1/2" web pad all based around the same chip. What would that take?
Imagine learning to develop for one environment and being able to create software for all these devices (and more). What would that take?
It would take a small, fast, efficient and modular bare-bones operating system that could be easily expanded for a wide variety of devices. I wonder why they hired Linus?
Dvorak is the Walter Winchell of the computer press. He doesn't pretend to offer more than rumor and innuendo.
Oh, the boldface three-dot style has a long history in American journalism. It's just about dead though.
He does not know what's he speaking about; the whole article is nonsense directly from his head.
He did write for MacUser, he had the back-page column, entitled "The Devil's Advocate". That is, he was a PC fan being published in a Mac mag for the sake of both points of view.
Does anyone know of a site similar to Slashdot
that hasn't been inundated with all of the 2
line posting 14 year olds?
I have stopped reading slashdot due to its
general decline and this recent trip back only
worsens my opinion of it.
I believe I read somewhere that there will be no advantage to having more than one processor.
I actually read this article IN PRINT on PC Magazine (OK, OK, I know it was a mistake, please but the sticks down :) last Friday.
On another point, while Dvorak has written little about Linux, it has usually been good. First time he did (AFAIK) was about Yggdrasil, back in the 1.2/1.3 days, and he was quite impressed with it.
...for the vocabulary impared /.ers
if you want to get your point across owns all
That's what i look for, too. All the mac/be/nt loosers hanging around and degrading the formerly cool site to the point of womiting.
Come on, little dorks, weren't you happy advocating you favorite os's on usenet?
Here's where I found the rumours about 128 bit
http://www.theregister.co.uk/990203-000014.html
I couldn't have said it better myself. I wonder how much money these 3 make a year pretending they know something about computers.
Unix dead? I assume you don't get to play with many "real" servers.
But as I recall, he was a regular writer for Mac User magazine.
Well, he's an idiot, so why not writing for Mac User magazine while using Windows. Sounds good to me.
Well, I don't know about other 64 bit processors, but Alphas have 64 bit registers, which makes life a lot easier if you want huge integers.
Posted by Mike@ABC:
...this isn't such a bad move. If Transmeta can create x86 processors that flexible, the applications are limitless. Computers and computing devices are getting dumbed down every day. Simple flexible devices with nimble processors could fill all sorts of consumer niches, like AOL TV-boxtops or mobile news/data readers. If that is indeed the case -- and I've heard nothing else to back up the columnist's claims -- then it's not a bad move.
Posted by [naden]:
I dont know if anyone remembers, but IBM and Motorola actually created a chip named PowerPC 615 that was capable of executing x86 code natively whilst still being usable in Mac environments. There were so many problems that they ended up abandoning it. Why ?
1) Every time a x86 instruction was parsed a context switch had to occur. The number of these ended up being so high that the chip ran extremely slow.
2) The price of the chip was ridiculous when compared to the dismal performance. They both agreed that Intel/Cyrix/AMD would already make chips that were much faster and considerably cheaper.
3) The chip had problems over-heating in some cases.
So what I was wondering was if this chip is true, and it can load microcode instruction sets, then surely it would be much, much slower than simply a context switch. It also begs noting how Transmeta could come up with something that Intel/Motorola/IBM/Cyrix/AMD (all with vast amounts of money and R&D) couldnt think of.
Why is Transmeta so secret and closed? I do not think this is a 'cool' company. Some people will object that hardware companies cannot open up. In order to make profits they have to keep things as secret as possible, or else someone else who has the time and money will release the ideas first in a usable product.. Is this bad for us customers? No of course not, it would be *great*. I do not think Transmeta earns respect by having open-source folk inside. It's just another company, and the arrogant way in which they act disgusts me. Sorry Transmeta, your commercial does not work for me.
They are involved in a secret government project to make a mind control processor using alien and soviet technology. Their purpose is to control the human race through a simple solitaire game.
Or is that Microsoft? Either way, we're in for some hard times ahead. Our only hope is that they run NT on it and it crashes before the subjugate us.
It's far easier to forgive your enemy after you get even with him.
Cmon,
This is just Dvoraks opinion, and his opinions are usually wrong. He thinks he know the most closed kept secret in the known universe. RISC chip? Every newer x86 chip is internally RISC.
x86 is dead. Even Intel knows that, thats why they are working with HP.
We all know the truth. Transmeta is run by Aliens, and what they are actually doing is choosing which is the next Alien technology they will bring us. (The first was Linux)
As far as Transmeta RISC x86 clones, interesting, believe it when I see it.
As far as day dreams go, Transmeta/Linus T. do a Socket 7 RISC chip, Linux writes some LINUX code that drops all the "X86 Emulation" crap and runs the chip "native" for a 25% preformance boost, Transmeta sells the 500MHz version for $100/ea, and Boom.... Linux rocks the world. /dream off
Anyone else notice his comment on the URU fingerprint device? "It appears to be foolproof. I couldn't circumvent its security."
Doesn't he belong to L0PHT or something? Or was it CotDC? And he calls himself a cracker!
Bad Dan. Blathering at 7:04AM and not even using the Preview Button? Bad, bad, bad.
Once you pull the pin, Mr. Grenade is no longer your friend.
Easy to design electronics? Cool.
Everything I use my computer for split into 89437598375 different random devices? Look ma, it's Push 2!
I don't want one box that's a web browser, another that's a music player, another that's a word processor, another that's a network router, another that plays Quake, another that organizes my schedule, another that checks my email, and so on ad infinitum. Jini? Universal Plug and Play? I prefer all my wires *INSIDE* my box(which may or may not be half open). As for wireless, oh yes I can't wait for my wireless Jini enabled stereo system to be hacked from scannerboy and for me to go deaf at 4AM.
You just can't beat security through impossibility.
Every once in a while I get the feeling that half of the fads in Silicon Valley are started when one or two venture capitalists nods their head in understanding. Like a lightning bolt, "Bob Richguy Inc. just dropped a million bucks on that stupid concept" turns into "This is finally how we'll make money on the net".
Well, that million bucks had to go somewhere, so somebody's right.
In this case, I think the VC meeting was some poor schmuck sputtering trying to save his product(who knows what) until he finally--very accidentally--opened his mouth and let the magic words "More types of gadgets = More Units Sold = More Items For The Profit Margin To Multiply = $$$".
Don't believe me? You *KNOW* Push came from "We'll solve the problem of users not wanting to download $500,000 sites by doing it while they sleep, and we'll get a chunk of that $500,000 by finally getting BigCorp some hits!" Nod. Fad. Blink.
This isn't to say that I think Transmeta is going to fail. They won't--I'll flat out go wayyyy out on a limb and say that Transmeta is playing the industry with just a little too much skill to not truly have something up their sleeve. (Not even Microsoft has managed so much good press with so little product.) My honest guess is that they're planning to replace all those tiny custom chips that the semiconductor industry spurts out in mass quantities and replace them with one chip, in various die sizes, that just does everything.
God, Transmeta's good, they've even got me postulating. They've got something. I just hope they don't think that they're going to sell ten gadgets for each app my one PC manages quite well. Beyond every other argument I've already given, nine times out of ten three buttons and an LCD display does not a good user interface make.
Hurm. I'm going to start archiving these posts on my home page.
Yours Truly,
Dan Kaminsky
DoxPara Research
effugas@best.com
Once you pull the pin, Mr. Grenade is no longer your friend.
I dont believe that their secrecy is merely marketing tactics (wouldnt be unheard of however) as I have yet to see any mainstream coverage which would be the only case where such market frenzy would really be effective.
As it is, theyve got a small part of the general computing audience drooling in anticipation like the pavlovian blobs we are for a product that is not likely to change the average persons life immediately upon introduction, therefore few people really care.
Unless the reverse-engineered "facts" (read: rumors) are nothing but cover for a more novel application of their patents and collective technologies, in which case the tease wouldnt be very effective either since they still wouldnt have a huge market waiting to happen.
Actually it would seem quite a good reason to watch any company... after all what could posses and entice someone who by all accounts could get a job nearly anywhere in the industry (if for no other reason than for his name). Linus doesnt have a whole lot to gain in the end if it turns out he's cleverly manipulated his minions of loving Linux users into purchasing Transmeta technology because he gets stock options, or even if he just turns out to have been used as such.
Yes it make for good marketing, but any market niche they could go for would be one currently controlled or at least coveted by intel.
Intel has far more resources. If they had a good idea what Transmeta was up to, they could possibly get a competing product to market first or at least make things hairy for them.
"Reactionaries must be deprived of the right to voice their opinions; only the people have that right." - Mao
DVORAK> The U.are.U appears to be foolproof; nothing I could do would defeat it
whoa!!!! Dvorak-proof. must be way secure...
Dvorak needs to lay off of whatever he's smoking. that whole section on Neon is such a strech...
I wish someone could keep track a la Brill's Content of the computer pundits and how often their predictions come true. Dvorak would be batting like a .050. Anyone remember the "Unix is dead" piece? Also, anyone take time to read his other column in this month's issue, "Adobe is in shambles"? What the hell was that? I really have a hard time seeing where stories of Adobe's founders being taken hostage by terrorists fit in the rest of the content of the magazine. That column belongs in a tabloid, not a computer publication. The best part of the article is when he touts the "amazing" results he got from Adobe ImageReady - he was able to crank out this "amazing" page "in literally 30 minutes." What a disgrace. I can't believe PC Magazine lets him run one of his gossip columns each month, let alone two like he did this issue.
I think there is a world market for maybe five personal web logs.
Really, if he knows this is fact (right and I'm Bill Gates long lost son), it's very cool in idea. I'd love to have RISC chip like this, and with Linux installed on a 100GB drive (though I don't know if I'd completely trust a 100GB drive, maybe just a little paranoid after to many "big" drive failures).
if my wife asks a question, and I'm not paying attention, and then I answer wrong does it still count
I think the next major move in computer platforms is in wearables/ultraportables, capable of web browsing and email with radio links (such as Ricochet).
Full access from everywhere is the next trend. The limited Palms won't be enough.
I'd drop as much money into that company as I could even though I know NOTHING about what the hell they are up to...
Actually, AltaVista has been running on 8G machines for awhile (couple of years) now.
Just junk food for thought...
This is the guy who wrote the "Unix is Dead" article - "IBM's VM is the happening OS," so how much can we really trust this pundit.
Advanced 6-Issue RISC86® Superscalar Microarchitecture
- Seven parallel specialized execution units
- Multiple sophistocated x86-to-RISC86 instruction decoders
- Advanced two-level branch prediction
- Speculative execution
- Out-of-order execution
- Register renaming and data forwarding
- Issues up to six RISC86 instructions per clock
Large On-Chip Split 64-Kbyte Level-One (L1) Cache- 32-Kbyte instruction cache with additional predecode cache
- 32-Kbyte writeback dual-ported data cache
- MESI protocol support
High-Performance IEEE 754-Compatable and 854-Compatable Floating-Point UnitHigh-Performance Industry-Standard MMX(TM) Instructions
321-Pin Ceramic Pin Grid Array (CPGA) Package (Socket 7 Compatable)
Industry-Standard System Management Mode (SMM) IEEE 1149.1 Boundary Scan
Full x86 Binary Software Compatability
On top of that, this technique goes as far back as the K5, as documented in the Great Microprocessors of the Past and Present, and may even predate that.
If Dvorak thinks this is new, he's smoking something. If Transmeta supports more than just the x86 instructions, then they may have a viable product that is new. To be able to execute code compiled and written for other platforms would be a blessing. Programmers would no longer be limited to one platform, and one CPU could pretend to be another type. To acomplish this task is a daunting feat, and if they can pull it off with excellent performance, I'll be very impressed. To be honest, I'm not sure such an endeavor is possible. The K5, AMD's first attempt at such a task for the x86 resulted in slower performance. While code could be written specifically for the RISC86 instructions, the translation from x86 was horriblly slow. AMD bought NextGen and used their technologies to create the K6, a much improved design, which can actually outperform the Intel x86's, FPU not withstanding.
There is of course only one emulated platform for that CPU. If I understand the supposed ability of the Transmeta vaporware, it will be able to support even more, but I don't see how it will be able to outperform with out being "specialized". Maybe I'm thinking too hard, but I don't see the Transmeta chip, blowing away it's competition. Aside from a developers platform, A slower more capable CPU for the consumer isn't going to be a big selling point. If the user wanted a 68k processor, he'd have bought one...
Just my 2,
Baggio
Time flies like an arrow;
Time flies like an arrow;
Fruit flies like a bananna
Don't forget that Dvorak wrote that 1998's MacWorld would be the last one ever because Apple was definitely on the way out. Jobs, he wrote, was just a return to the past. Maybe so, but the 1999 MacWorld was definitely superior to 1998's.
:)
Nonetheless, however, I still enjoy Dvorak. Even if he's always wrong.
John Dvorak is as reliable as a Yugo. His articles are good as semi-fictional entertainment.
Ummmmm. One short word for you here. WRONG. All Intel Chips, including the Xeon you forgot to name are CISC. Intel will never use RISC, instead they plan to use a different instruction compression method in the IA64 chips, aka Merced.
As for the K6, I haven't examined AMD's chip details for a while, but I highly doubt it.
Here at work I use a one-processor machine with 1 Gig of memory. And on supercomputers I use a whole lot more than that. I topped 8 gig used just by my program for a large calculation. So it really just depends who you are and what you're doing.
--Artemisia
Hmm, for some reason I always got the impression that he was some closet-Mac abuser. Now, I haven't read any of his drivel since the mid-late 80's, so I could have a few leaks in the old memory. But as I recall, he was a regular writer for Mac User magazine.
About the only article he wrote that I actually agreed with was his tirade against 5.25 floppies. This was an article in the DEC Professional magazine sometime around 88-89.
The average intelligence of the ./ poster is rapidly declining!
Or the average age....
*--BigMan--- Time flies like an arrow.. but personally I prefer a nice glass of wine!
I think the title of this should have been aimed a little better at avoiding a Dvorak flame-fest. When I use to read Usenet I remember enough people thinking he's a dork, but do we really need to specifically AIM at him like this?
The signal-to-noise ratio on slashdot is already low enough as it is.
At least he's covering Transmeta.
Very cool. I hope the performance is there from the start. If so, I know what my 6th home computer will at its heart.
Did a search on quote.com anyway & came up with nothing.
=moJ
- - - - - -
Member in Good Standing,
I wanna hear prices...
That's all I want.. prices... and based on that I then want to know if they're naming it with a name to make fun of the Xeon (an over priced chip, but a good chip I think)... then, does this mean the performance will be similar too?
If this chip came out for $200 for a 500mhz version (or higher... lower power, means cooler, means faster in my head).. I'd get one even if it was a proprietary slot!
- Hugh Buchanan
- Userfriendly.com
Don't blame the author until you know what the editor changed.
I can't imagine Linus accepted this position without the belief that it was an opportunity to do something significantly beyond providing for Tove and the kids. I'm patiently waiting for the announcements to commence and expect to be pleased with the concepts and potential.
i can't remember which, but i'm pretty sure that i read somewhere that either the k7 or k8 will be a return to raw cisc. apparently they figured emulation was a waste of logic.