Slashdot Mirror
AMD Allies with Transmeta
respect sent us an article about AMD aligning with Transmeta, which isn't that surprising since everybody who is not Intel should be busying aligning with anyone who is also not Intel.
My favorite quote "The industry has been gradually moving toward a 64-bit architecture [From 32 Bits], which multiplies the amount of data the processor can access by four". Rock on CNN!
So are we going to wind up with a laptop processor that gets really hot or a desktop processor that needs software upgrades?
The obvious reason is Itanium compatiblity. AMD has a 64 bit processor that's likely much faster than Intel's 64 bit offering, but still doesn't have the market clout to push it's own instruction set. Transmeta has a hardware/software combination that can emulate other processors reasonably well.
Put the two together and you get a 64 bit processor that runs 32 bit apps and AMD 64 bit apps faster than Itanium, and might even be able to emulate the Itanium with decent results.
It is tempting, if the only tool you have is a hammer, to treat everything as if it were a nail. - Abraham Maslow
Linus who? Don't try to tell me that Transmeta has a Peanuts character working for them, because I won't believe it. My mother told me that cartoons were imaginary.
Oh, Linus Torvalds! Yeah, I suppose he's sort of famous too.
The industry has been gradually moving toward a 64-bit architecture [From 32 Bits], which multiplies the amount of data the processor can access by four". Rock on CNN!
Normally, I don't point out the journalistic errors of Slashdot, but I can't help myself here. At least CNN knows how to spell Transmeta without the D.
He who lives in glass houses should not throw stones, Taco.
First of all, AMD isn't making a fuss about this. Slashdot is making a fuss. AMD just made an announcement, and anyone who follows knows that they make several such announcements a week.
But the reason that teaming up with Transmeta is appealing is that Transmeta can offer them a processor that acts like a Sledgehammer well before they have silicon for the part. Software simulation of modern microprocessors is ridiculously slow, particularly when you are upping the word length from 32 to 64. However, writing a code-morphing layer on top of Transmeta's chips means that they can have a near-native speed chip before the design is even finished.
Being able to run tests at near-native speed, pre-silicon means that AMD can overlap software development with the hardware development, which could buy them months of development time. In this business, time to market is of overriding importance. And a shorter time to market is what AMD wants out of this alliance with Transmeta.
--Lenny
Well I did start off on that topic, agreeing that slashdot appears inconsistant, and then saying maybe in this case it isn't really. Then I wondered off and commented about the story, and then couldn't resist poking fun at the iTanic, and how utterly lame the x86-64'x REX prefix is.
A bit, yes. Couldn't you tell?
Can't we get back onto topic though? When is AMD going to do something about those orbital mind control lasers?
Has transmeta licensing a new CPU (or peripherals) bus reduced competition? Or has it increased it slightly because now you could more easily swap out the two CPUs? (or maybe just use the same high speed peripherals)
Has transmeta licensing the x86-64 instruction set reduced competition, or does it give you a second source for CPUs that can run 64bit x86 code?
I don't think so. The way AMD implements x86-64 isn't low power. As far as I know it isn't working either (not as in it is behind schedule, but as in it hasn't taped out). Transmeta last year got (as far as I know) one of their existing CPUs to implement the x86-64 instruction set so AMD could start working on system level code (reference system BIOSes, and one hopes GCC, and open source OSes).
As far as I know transmeta has only licensed the x86-64 instruction set (and I would assume any applicable patents). I could be wrong, since the article doesn't say one way or the other. I'm guessing because doing it the other way would make transmeta a 2nd source for the x86-64, and there would have been a press release about that.
Well SlashDot is pretty inconsistent (for example the far greater the normal number of Patent defenders for TiVo -- I love the product, but I still don't like the patent system).
However this may not be all that inconsistent. The AOL/TW merger was arguably the biggest dial-up ISP merging with a very large media conglomerate. Microsoft and anyone is the biggest software company plus, well, a little bug :-) In this case it is the second biggest x86 CPU seller (AMD has what, 30% of the market?) plus one of the smallest (I would guess less then 1%).
Plus this isn't even a merger, it is just "we will license an instruction set and bus". Nobody bitched when the PCI bus was wildly adopted (nobody I noticed at least), but slashdot wasn't around then. Nobody seemed to bitch when the clones adopted MMX either...
Now I think x86-64 is a huge kludge. Maybe not as bad as the iTanic, but pretty grimly crufty. I would much rather see a migration to the Alpha, or SPARC, but that doesn't seem likely. Transmets's adoption makes it more likely that x86-64 will take off, which isn't something I'm thrilled by. Unless the only other choice is really the IA64.
Who really wants an instruction prefix to switch 64bitness on and off and select a register bank for each instruction? Well not a register bank, one for the source and another for the dest. Feh.
No, it's new XML addressing. The four new 4Gig banks are addressed by putting the ascii strings "zero", "one", "two", and "thre" into the top 4 bytes. MicroSoft and Intel engineers patented the new spelling of 3 that allows them to save 8 bits per address!
The *real* problem is that there is only ONE independant implementation of Intel's IA32 architecture, and it's Transmeta's.
Yes, believe it or not, AMD is a licencee of Intel technology, and therefore the Athlon is somewhat dependant on Intel's goodwill.* Fortunately for them, Intel has a large number of contracts (government and otherwise) which require multiple sources for their tech.
Now, if Intel decides to be a bad boy and pulls AMD's x86 licence, Transmeta's IP portfolio starts to become very helpful. Namely, AMD will be able to stay in business.
*Source: AMD quartery report:
RECENT DEVELOPMENTS
On May 4, 2001, we, along with Intel Corporation, announced the renewal of the
patent cross-license agreement between the companies.
--
Business. Numbers. Money. People. Computer World.
...at El Reg
Well, well, well; three holes in the ground...
I recall an older Slashdot article that mentioned that Transmeta was working on adding AMD's 64-bit instructions to their code morphing, and that they'd be helping AMD test software, as AMD's processor simulator was very slow.
--
Hmmm... When I read that statement, it seemed like a rather ludicrous point of view, so I read it as sarcasm, and it made a lot more sense.
Next time folks should read the article before flaming CNN.
The CNN article originally said "multiplies... by four", not "four billion". They fixed it some time after the story was posted (kinda like how Taco fixed "Transmeda". If you want confirmation of this, an earlier poster has put up the text of the article, and it still contains the error. Generally, with typos in online articles, journalists just quietly fix the error. Retratctions and apologies are saved for serious mistakes. While "four" versus "four billion" may seem serious, nobody who would actually be basing important decisions on the article would be confused by that, and it's apparent to any knowledgable person that it's a typo.
--
Patrick Doyle
I mod down every jackass who puts his moderation policy in his sig. Oh, wait a sec....
Even with your own math, your conclusion is wrong. (Get a calculator and punch in the numbers.) Plus, x86s are byte-addressable, so no, it wouldn't "figure that 64 Bit processors would accesss data in 8 byte increments".
--
Patrick Doyle
I mod down every jackass who puts his moderation policy in his sig. Oh, wait a sec....
(Submitted to CNN's feedback address. There has been much debate here over this math, so I thought I should post it here too.)
Dear CNN,
I'm sure someone has told you this already, but in an article on Transmeta and AMD, you said this:
"Current 32-bit processors can address 4 gigabytes of data. The industry has been gradually moving toward a 64-bit architecture, which multiplies the amount of data the processor can access by four."
Actually, it multiplies the amount of data the processor can access by four BILLION.
Every byte of data has an "address" which is represented by a certain bit pattern, and every bit in that pattern can be on or off. If you have one bit, you only have two possible states, hence two possible addresses, hence you can only access two bytes of data. However, if you have two bits, you have four possible states (off&off, off&on, on&off, on&on), so you can access four bytes of data.
With 32 bits, you have 4.3 billion possible states, so you can access 4 gigabytes of data. With 64 bits, you can access 17 billion gigabytes (which is 16 exabytes).
--
Patrick Doyle
I mod down every jackass who puts his moderation policy in his sig. Oh, wait a sec....
Against Intel's market share and marketing engine (such as the Intel Inside campaign), AMD is doing the prudent thing by forging strategic alliances, rather than waiting for the computing public to recognize them as a superior provider.
.9Ghz Athlon at home and love it - I have never had better performance from a machine)
When AMD produces a superior product, they will be recognized by the market. Today they don't have that claim. You can't walk into a Circuit City these days without seeing that a plurality of the processors for off-the-shelf machines are AMD, so IMHO Intel's vaunted monopoly is a bust top hat. AMD runs in second place because it actually lags behind Intel in several key spots.
(Disclaimer: I run a
The fact is that AMD has beaten Intel only in two spaces: price/performance, and mindshare among the tech savvy. Apart from those advantages, there are still numerous failing in their product line:
1) Power consumption. They've kept themselves well out of the laptop and server farm market by eating power like a pig and blowing it in the form of heat. I have heard that AMD is reducing power in their Clawhammer chips though.
2) No non-vapor multiproc motherboards for the Athlon. Granted this is not really their fault, although I have heard that the Althon has a hard time achieving greater than 2-way SMP. Anyway, you can't enter the lucrative server market until you can at least support 2-way on a range of platforms and in reality you aren't viable until you have support for at least 8-way. Looking forward to the day when motherboard makers open their eyes to this opportunity.
3) Incompatibilities with gaming hardware. If you aren't in the server business you'd better be in the home user market and nothing uses a CPU like games. Once again this isn't exactly their fault, but at least in my case several pieces of my hardware had to have their drivers patched up before they would work with AMD and I even briefly had to crank down my AGP to 2x instead of 4x in order to prevent my games from freezing. All of these problems can be solved (enough other people have noticed them to have populated with web with helpful tips) but this is not the way to earn the trust of home users. The worst part is, it was most likely the makers of the other hardware items that failed to perform adequate QA, but fingerpointing never fixed a box.
4) Brand name. Where are the AMD ads to counterpoise the Intel blue guys? This company, despite its great product line, doesn't seem to have any message for ordinary consumers beyond "We're almost the same as Intel!" If you read through the last two problems above, you might agree with me that brand recognition is the root of both. So why don't they do something about it?
There are lots of advantages to buying AMD, but equally strong disadvantages at the present time. For certain applications AMD is the way to go. For others, regrettably the best choice is Intel.
-konstant
Yes! We are all individuals! I'm not!
-konstant
Yes! We are all individuals! I'm not!
I am absolutely thrilled to see companies agreeing on standard technology. I never want to by another VESA local bus machine...
Competition is good. So are standards, and the fewer of them out there the better.
Er, um, the "amount of data you can access" is the total amount of addressable memory, that is, not the total number of memory addresses but the total number of addressable bits. Since the word length changes to 64, you've gone from (2^32)*32 addressable bits on a 32-bit machine to (2^64)*64 bits on a theoretically maxed out 64 bit machine (may I live to see the day).
To follow this to it's 100% anal conclusion,
(2^64)/(2^32) = 4294967296.
((2^64)*64)/((2^32)*32) = 8589934592.
So technically, CNN was wrong by a factor of
(8589934592 - 4) or 8589934588 times. Their percent error can be calculated as (8589934588/8589934592) or 99.99999995343387126922607421875% (ok, yes, I rounded).
In short, CNN was nearly 100% wrong.
But you already knew that, right?
--
What happens when you outlaw guns
It's not strictly true that x86's are limited to 32bits of addressable memory. With VM and segment selectors (which I personally diplore) you can achieve a much higher number (don't remember if it was 36, 42, 46 bits or what). I'm not sure, but I was also under the impression that more than 32 address pins were available from the CPU.. Can someone confirm or refute this? (I'm too lazy to look it up).
With this, a sufficiently designed database / scientific app could handle very large memory/ disk mappings without too much difficulty.
Of course, things get a lot easier when all the data is 64 bits. It's kind of like the comparison of the old 8086 with 16bit segment selectors which could address 20bits of memory. It was done, but it wasn't pretty, plus each contiguous block was limited in size.
-Michael
-Michael
...Hypertransport is going to be an awesome technology when it hits the mainstream. And with little or no royalties, it should hit the mainstream pretty fast.
Whats interesting here is perhaps the biggest issue surrounding the computer: I/O. Intel, Sun, HP, Compaq, plus few dozen others are pushing Infiniband. However, AMD is pushing HyperTransport. Hypertrans seems like more of a local I/O solution (replacement for PCI-X), while Infiniband is more of a remote I/O replacement (or competator) to Ethernet. What are other peoples read on this issue?
Someone you trust is one of us.
Um..please forgive me, but you are stupid.
2^64 / 2^62 would be 4.
2^64 / 2^32 = 2 ^32 = 4,294,967,296.
Hardly 4.
"We see this is as a strategic deal, not a revenue deal," Weber said.
Translation: "We ain't getting much cash for this here deal."
But, believe it or not, I think that's the right move in this case. AMD realizes that revenue generating partnerships at this stage of the game isn't what's important -- AMD is not going to win by taking a few extra dollars here and there. AMD is going to win by taking market share and growing its brand.
I do fear, however, that we're going to see another Cyrix 486 problem (i.e. the rumors of "slightly different x86 architecture, not quite compatible with Intel/Windows, not gonna buy it.")
Ferrari and other exotic car rentals in New York
- Rampability. The pentium 4 is expected to reach 4Ghz, possibly 5Ghz by the end of 2002.
- Volume. Currently, AMD only has 2 fabs. The one in Austin is currently only producing Durons, and AMD is planning to turn it into a flash memory only fab. (Of course with
.13 micron and below you get more volume for the same wafer capacity.).
My estimate for the future: AMD will be the performance chip of choice for gamers and low-end server manufacturers. AMD will increase rampability with the Hammer line, due to stretching out the pipeline and other tweaks. The X86-64 instruction set is a great strategic move. It will give 64bits addressable, and double the number of registers, not to mention getting rid of some nasty ia32 quirks when in x86-64 mode.Actually this is not strictly true.
Although a 64 bit processor has 2^32 times more virtual address space than a 32 bit processor, x86-64 is not a 64 bit architecture, and IA-32 is not a 32 bit architecture.
X86-64's linear addresses are 48 bits (not 64 bits), and its physical addresses are 40 bits (again, not 64 bits), and IA-32's linear addresses are 32 bits, and its physical addresses are 36 bits.
So in reality, X86-64 can address 2^16 (65,536) times more virtual space than IA-32, and 2^4 (16) times more physical space than IA-32. Both are a far cry from 4 billion (or 4, for that matter).
I think this might even deserve a "!!"
Rich...
Ignore Alien Orders
Is polarisation always in everyone's best interests? Let's suppose we have one Big And Scary player in a field, and lots of little guys. The little guys decide to band together in order to be a match for Big And Scary, and what have we got? Two big and scary players instead of one, and a lot less diversity of choice. (Think of how often you've heard people say that Gnome and KDE should unify, because if they combined efforts they'd be able to be a strong contender against Windows: an attempt to increase consumer choice by killing off diversity.)
And I know "aligning" with someone, as Transmeta are doing with AMD, doesn't make you exactly the same as them. And I'm not saying standards are evil. I'm just saying that "them and us" thinking leads nowhere but multiple "them"s.
my plan
GROGGS: alive and well and living in
32 bits == 4 bytes, 64 bits == 8 bytes. Increase by 4 bytes. while bytes != times, cnn != programmers. ;p
I am !amused.
Transmeta, on the other hand, was up $0.75 on a basis of $14.10 at the open, from after hours trading. Since then, it fell steadily throughout the morning, but turned around at 10:30, and has since climbed back to $14.64, down $0.25 from the open, but up $0.50 from the previous day's close.
Isn't working for a brokerage firm fun? Disclaimer, for anyone who may be monitoring text transmittal: all quotes are delayed at least 20 minutes, and are approximate. This post is not intended as investment advice, and anyone who uses the information contained herein for any purpose is an idiot.
The only "intuitive" interface is the nipple. After that, it's all learned.
"The question of whether a computer can think is no more interesting than that of whether a submarine can swim" -EWD
From an analyst by way of CBS Market Watch
Brian Alger, an analyst at Pacific Growth Equities, said the licensing deals will attract attention, but may not amount to much. "People are going to see Transmeta and AMD working together against Intel (INTC: news, msgs, alerts) ," he said. "That's a nice perception, but it's not real. They're not going in side-by-side with their sales force saying 'Don't buy Intel, buy one of our two."'
This makes it sound like Slashdot "took the bait". The market opens in 10 minutes. Let's see how many traders react the same way.
For all intensive purposes, "whom" is no longer a word. That begs the question, "who cares"?
You forget that I was criticizing the statement, not necessarily the story at hand. Hell, you even quoted my post, not even realizing what I was saying! Are you daft?
Friends don't let friends use multiple inheritance.
I don't know about you, but from what I've seen in every other instance, Slashdot vehemently opposes industry consolidation and mergers (witness AOL/TW and MS). I guess less competition is only bad when you don't like the company.
Oh well, I guess the enemy of my enemy is my friend. That one always works. Just look how it's helped the U.S. Gov't in the last half a century.
Friends don't let friends use multiple inheritance.
NEXGEN. Do you remember what AMD was doing in the early 1990's? They were playing a continual game of catch-up with Intel. If memory serves me correctly, there were going to be a LOT of barriers to adopting the original Pentium technology to clone. Intel was going to squeeze them out.
Fortunately, there was a little company called NexGen. Simply put, they devised a way to hook up an instruction translator to a RISC core. I bought one of those motherboards in 1993 and it works well to this day - using Linux, of course. AMD bought the company and used the technology to to develop the K5 and the K6 series processors. I don't remember NexGen making a lot of money, either, but then they weren't public for very long before AMD bought them out. AMD had the brand name recognition as well as the funding to push the new technology.
Transmeta seems to fit this mold. Like NexGen, they've developed a disruptive technology. AMD is interested in them. I think there's a buyout in Transmeta's future.
Comment removed based on user account deletion
Comment removed based on user account deletion
"Normally you might look to Intel for leadership in this area, but Intel really hadn't come up with much of anything that looked like a good replacement yet," said Ditzel.
The reason why Transmeta is looking to AMD is NOT because they hate the evil Intel (AMD is just as much a competetor as Intel), but because AMD has innovated a useful technology.
Thankfully, competition continues to be the driving force in innovation, and in the end, the customers benefit.
Your reality is lies and balderdash and I'm delighted to say that I have no grasp of it whatsoever. - Baron Munchausen
Dave
Whilst I admit most of my industry experiance is in the software field as a dedicated IT professional I've been following the processor market for a while now, and I've got to wonder why on Earth AMD would pull this move.
Because despite the glowing hype to typical of public stories in the field, Transmeta are hardly in a strong market position. The huge hype that got them rolling turned out to be mostly marketing wind. Sure they had a great product, but it wasn't the industry-changing breakthrough they'd been attempting to make it look like, and so they've done no better than any other start up producing components for low-power devices.
Okay, not strictly true. They've done somewhat worse actually.
On the other hand AMD is going from strength to strength in a way that would have amazed insiders a few years ago. They've got a lot of well-deserved respect now, and people are really starting to take them seriously on their own rights, rather than as Intel's annoying little competitor.
But making a fuss over "strategic deals" with underperforming overhyped companies is not the way to go. They should concentrate more on their core strengths and carry on producing excellent processors.
Jon Erikson, IT guru
They were just accounting for the bloat in the next version of Windows :)
Bryguy
microsoftword.mp3 - it doesn't care that they're not words...
I like the article despite the misspelling. I have nothing against Intel, but I really don't care for thier monopoly on desktop computer processors. Perhaps someday soon, AMD will become a household name just like Pentium. Let's just hope that they don't have to resort to anything as stupid as those blue guys are the "bunny people."
In other news, Rambus flunked court in Italy:
EBN Online article
Next on the block:
Infineon in Mannheim, Germany
Oct 29, 2001, Wilmington, DE, Federal Court, Micron will ask that Rambus' patents be invalidated (due to that dirty business in JEDEC, no doubt.)
Maybe with an open processor workshop, they could show up and not tell anyone they're planning to patent adamantium interconnects...
-- .sig are belong to us!
All your
A feeling of having made the same mistake before: Deja Foobar
Proof 1:
Proof 2 (shorter version):
-Marcel
PS: Sorry, it's not that funny, but I just couldn't resist.
... who was wrong, but at the moment CNN spells it right:
The industry has been gradually moving toward a 64-bit architecture, which multiplies the amount of data the processor can access by four billion
Now is it CNN was nearly 100% wrong or somebody just need to actually read an article? We will never know.
The page now says " which multiplies the amount of data the processor can access by four billion" and it has a good explenation why.
LedgerSMB: Open source Accounting/ERP
Gee, I think my mini which uses powerpc chips has been doing 66bit computing for nearly 5 years! (and it is essentially one code change away from 128 bit processing - and guess what - no code changes required - code will be recompiled on the fly)
Gee, AMD/Intel grow up, the industry lauds up and coming 64bit while ignoring the fact the IT industry has been there done that.
* Winners compare their achievements to their goals, losers compare theirs to that of others.
I just wanted to thank you for letting us all know. I'm immediately liquidating all of my stock and mutual fund holdings related to either of those firms.
Oh, and thanks for also giving us such insightful explanations of why you have chosen to boycott AMD and Transmeta. It's that kind of technical commentary and expertise that makes Slashdot worthwhile.
People on slashdat, and rightly so, get a little worried about what happens when two super giants in related industries get together. If Intel and Microsoft merged into a single entity, well, after than Wintel would have a whole new sinister meaning. If AMD and Transmeta get together, damn skippy. Why, more choice, not less. At the end of the day we're all consumers, and we all want to be able to make an informed choice from a selection of products.
--Jimmy has fancy plans; and pants to match.
Nah, it's just CNN used a Pentium to (FDIV) divide 64 by 32...
--
KMSMA (WWBD?)
AMD isn't hyping this Transmeta deal any more than they have any of the other more major alliances (not alignments) they've developed. CNNfn isn't even really hyping this deal, even though they seem to immensely enjoy referring to Transmeta as an 'upstart'. In reality, the only things at work here are one CNNfn reporter's obsession with Transmeta (Seriously, try a search for 'Richard Richtmyer' at CNNfn. He's the author of this article, and 4 of his 6 articles this month so far are about Transmeta.) and the /. crowd's obsession with all things Linux. (If you don't know the connection between Transmeta and Linux, well, I can't help you.)
Anyways, nothing bad on AMD's part, they had no control over this.
-Jade E.
I disagree with the statement that the non-intels of the world need to band together for protection, AMD is, in my opinion, every bit as good if not better then our precious pentium cranking friends.
I agree that AMD has an excellent product. However, the ashbin of history is filled to the brim with products that had superior technology at the time, but were defeated by a competitor with better marketing (take Betamax, OS/2, and CP/M for instance).
Against Intel's market share and marketing engine (such as the Intel Inside campaign), AMD is doing the prudent thing by forging strategic alliances, rather than waiting for the computing public to recognize them as a superior provider.
Toronto-area transit rider? Rate your ride.
First off, I disagree with the statement that the non-intels of the world need to band together for protection, AMD is, in my opinion, every bit as good if not better then our precious pentium cranking friends.
So I wouldn't call this an alliance so much as a backing up of what deserves to be backed, Transmeta seems to need a step into the 64 bit world and AMD is the perfect company to help them into it. There's something in it for AMD and they don't have to give up anything, they earned this.
I love my chip! You love my chip too!
spacefem.com