Slashdot Mirror
AMD takes a big hit & IDT exits x86 clone biz
About one billion of you wrote with the news that AMD took a operating loss this past quarter, and the COO and heir apparent to the CEO quit. In related news, IDT has declared that they quitting the x86 clone business. Wow-despite lower then expected earnings, Intel has to be pleased by this turn of events.
I read somewhere that IDT is letting the x86 part of their company go off on its own. So it will no longer be an IDT winchip, but rather and Centaur Winchip XXX.
http://www.jc-news.com/pc
"...here's too much news today! IDT posted some really good quarterly numbers but spontaneously decided to give up on x86 and make Centaur (the x86 wing of IDT) total autonomy. It is now the Centaur WinChip 4..."
he later reports that IDT has a buyer for their x86 division.
> incorporating 3DNow! instructions (K6-2) which have been a collosal flop for everything other than Quake2 on a 3dfx card
FWIW, the CVS versions of Mesa and the G200/TNT driver both support 3dnow builds. It doesn't look like 3dnow is going to be a lost cause for Linuxers.
Also, they may have felt like they didn't have any choice, after Intel came out w/ MMX.
> on-chip L2 cache (K6-3) which pumped up their integer performance at the time when even casual users had come to learn about the FP issue.
Just possibly there exist blighted souls who would rather have their kernel recompile faster than have Quake update the screen faster.
Sheesh, evil *and* a jerk. -- Jade
1) 3DNow! is hardly a flop, at least on windows platforms. It is supported by many games and graphics accelerator drivers now. Check out the review on Tom's Hardware comparing the performance of various 3d cards with 3DNow! enhanced drivers. Granted, it took a while to get there, and on linux the options are limited (but all we can use is 3dfx, and play quake, so it goes to figure), but it's hardly a flop.
/. readers and gamers know, but I wouldn't describe either as 'casual'.
2) most casual users don't know what FP performance is, much less associate AMD with a lack of it. Maybe most
The enemies of Democracy are
I think Rise will be gone by the end of the years. Their product isn't up to what the people want now adays. Instead of trying to beat out the major competition at the moment, the smart move would be to come together as one company. IDT, Rise, via and whom ever comming together as one company to actaully have some footage in the market and hopefully come out with a better product.
I ate my tag line.
I ate my tag line.
-=Ellis (D)25=-
This all sorta points in the direction that I have thought intel was going in anyways. I pretty much figure that IF intel can kill off AMD then I'm pretty much sure that the celeron will be killed off for a repackaged PII/III in a socket370 config or something similar. By doing this they can continue to segment the market and elminate the celeron which is eating into PII/PIII sales.
"An Ye Harm None, Do What Ye Will" Wicca Rede
1) I don't know about Wall Street, but AMD stated that their losses for this quarter could be as high as $200M
;).
2) AMD has stated that they plan on releasing 1GH processors in the first half of '00, not the first quarter. This however doesn't mean that Kryotech won't come out with a K7 system that is able to go this fast before then
3) Agreed
4) Don't forget the various issues Merced is having, and that it will be dirt slow on code that the compiler doesn't optimize right. In addition, old x86 code is supposedly dirt slow on it as well.
Few other things of note:
* AMD claims to be making their memory on the copper process, yeilding as well as their aluminum process
* AMD claims to have produced K6's using the copper process, and is getting decent yields
Re: Performance of K7 -- we'll just have to wait until it's out to declair that :). I personally think they can charge more, if reports of it's FP performance are true. But I'm glad they're not :).
.25um process. AMD doesn't NEED to shrink to a smaller immediately if this is true.
Re: Intel able to win the Mhz race -- the K7 is shipping at 600mhz, and reports are that it doesn't run hot at this speed. It is rumoured that AMD will be releasing 700mhz/750mhz versions before the new year. On a
Several sites are already projecting what the clockrate of the PIII would need to be, in order to beat the K7. Most of the figures say that the PIII needs to hit somewhere between 750-800mhz to beat a K7/600.
Food for thought.
I'm still not buying the part about the Athlon's selling well. Nobody cares if it beats the PIII in the SPEC benchmarks run in-house by AMD. If the discussion after Thresh's original article and its follow-up are even close to the truth, the Athlon will likely look very poor in most real world benchmarks until they can be optimized for the Athlon architecture. I'm worried that good Athlon optimized code will be just as slow in coming as 3DNow! optimized code. AMD has been trying to twist arms to get better 3DNow! support and still 3dfx is the only company who seems to be serious about supporting it. If the Athlon does turn out to require as much optimization as thought to show an advantage over the PIII, AMD is going to have a hard time selling them.
If *everyone* waits for the K7 to come down, the effect will be the same as AMD's previous problems with delays, and they'll go out of business.
There's no real bad news here... IDT was never a player, and AMD is better off than previously expected. AMD's fortunes depend entirely on whether or not high-end buyers take an interest in them.
Incorporate PS2 technologies in an x86-compatible CPU? What possible benefit is there? What does the PS2 have that would be even slightly useful?
You can already do this with the BSD's... 'make world'. Although, this won't be as effective since for say FreeBSD since the current release version still uses GCC 2.7.2.1 for stability reasons - you need to upgrade to current.
Also, Mandrake is pre-compiled for either P5 or P6 march...
Depends on what you are doing. Increasing RAM will only speed things up to a point. If you aren't experiencing a lot of swapping with your current RAM, adding more isn't going to make much difference. Most current computers are being sold with 64MB or 128MB of RAM, which is enough for most people.
Also, if you spend time using things that are CPU intensive, then SCSI obviously won't make any difference. I personally have a K6/300 w/ 64MB and Celeron OC'd to 450 w/ 128MB. Both are running off of standard 5400 rpm IDE disks. Both machines are more than fast enough for most tasks I encounter. However, when running Gimp filters to enhance large scanned images (3000x2000 or so), I find the Celeron system to be significantly faster. Also, when playing most 3D games (not just FPS) I get more than double the frame rate from the Celeron system from the same 3D card even when the Celery is clocked to the same speed (300). 3DNow! is great if you only play Quake2 on a 3dfx card, but it's practically useless with most games and all non-3dfx cards because developers and vendors won't properly optimize for it.
Maybe if I was doing different things with my machines, I would care more about RAM and disk speed. However, the way I see it the K6 FP issue is a big one for AMD, which hurts them in the most lucrative market segments (high end users and serious games). They should have fixed it by adding a second FP pipeline in the K6-2 instead of 3DNow!
I wonder though, whether the computer world has changed enough from those days for this not to happen. IIRC, Osborne produced *the* "portable" computer, there really wasn't any serious competition to them (well, Compaq had its models, but they were pretty far behindin terms of sales and future technology weren't they?). The market for sub $1k (hell, sub $.5K) computers looks to be picking up (this is off the top of my head, so corrections with backing joyfully accepted), and AMD could easily sell any backlog of K6-2 and -3's it might have to keep in business until the K7 sales pick up.
I suppose that with Intel being the market leader and producing chips that aren't that much more expensive than AMD's offerings, things could actually be worse for AMD than they were for Osborne, too.
OK, so I'm no market analyst.
"Oh, I hope he doesn't give us halyatchkies," said Heinrich.
Let's buy the Winchip design and put it in a GPL license! Anybody know how many people paying $20 is needed to buy it?
The retreivable serial number is excellent for at least one function: copy protection.
Here's where Linus and hist buddies at Transmeta earn their pay. Where's the 1GHz pentium++++ clone from Transmeta ???? We know they can code like bandits on speed; but can they hack hardware ???
Why would I care about anybody knowing where I bought my stereo? (and which vacuum tube does the serial number get stored in in the first place? Or would they store it in an array of diode tubes?)
I don't think that Intel is too happy about AMD's losses. For a couple of reasons:
.18 micron grade chips, is now online & AMD reports that they are on track to produce 1 GHz chips by first quarter of 2000.
1) Wall Street expected losses of 250 Million, AMD only lost 162 Million.
2) AMD's copper plant, a facility capable of producing
3) AMD's biggest problem has always been trying to get their average chip price above $100.00. With the new Athlon processor made to compete with the Pentium III, at same (or lower) prices with better performance, they will reach this goal, and also have the fastest processor on the market.
4) All of this doesn't even take into account that the Intel Merced has been delayed AGAIN, and is now not even slated to be a large production processor, its predecessor will be, but its not scheduled to come out until 2001.
--Remove chicken to e-mail
The K6-2 350 chips are obscenely cheap now. In the neighborhood of $55 apiece. It's tempting to pitch the motherboards in some of my older systems at home and get nice FIC motherboards and these chips. Of course, then I'd be tempted to also buy PC-100 SIMMs for the new systems. Suddently it gets a bit more expensive...
I would put in my $20 to buy the chip. Think about what we could do with the gpl'd x86.........
I ate my tag line.
I ate my tag line.
-=Ellis (D)25=-
Humm... I wouldn't be suprised if Acer wanted their own CPU company. They already make almost everything else. :)
Umm ... no. Actually, AMD is working on the following (and says as much, unofficially):
.18, and then move to copper. They will be positioned at the embedded market and anywhere that low power is key. AMD already has portable and "embedded" K6-2s and K6-3s. The improvements to the K6-3 will include making the cache optional, cutting the speed way back to save power if needed, and then freezing the design so that Intel's little market for embedded 486s dies, as Socket 7 stuff is dirt cheap now. So, the K6-3 will stick around for a long time to come. Also, the K6-3/500 isn't a no-show, just not out in larger yeilds yet. It will be generally available in normal quantities for about $150 by Christmas, with 450s for $100 and 400s for $65-70. The only slower K6s will be the portable ones, and only the portable K6-3s will be in production by end of the year. There will continue to be a lot of slower K6s around (including the portables) because of overproduction -- AMD hasn't sold them as fast as they should have, but if they can dump these on the secondary market they will make Celerons less attractive, especially if Intel kills that market by locking down overclocking and preventing SMP. So the K6 is doing fine, really.
.18.
.18 and smaller and with copper (possibly SOI, but I am told that this is still in negotiation), when the tweaks to make it modular (no cache, 128k, 256k; 133MHz at 1-1.5 watt to 500MHz at 6 watts), when the design is frozen as is forever and the tooling is paid for, you will be able to get a K6 core for about $25. Intel still makes a lot on 486s. This will make AMD a lot of cash. I think that this will do fine long run.
1. The K6-2s will be dropped as the K6-3s get cheaper, go
2. The issue with the K6 inventory is that most of the chips are at lower speeds. AMD doesn't want to remark for all the obvious reasons (they are getting too close to problem speeds as it is as a lot of K6s are still failing within the first 90 days of use because they are leaving no margin for later minor failures), but the issue was not the potential of faster chips, but that they were actually out there (and the Celerons).
3. The K7 is very nice. Yes, I have friends with samples. Yes, it beats the PIII, and yes, by more than the unofficial benchmarks suggest and there is room to improve. I would suspect that it will do very well, especially with faster cache at
4. I am buying AMD stock. Everyone I know who works there is too. This is the first time in a long time for all of us, and we are a pretty cynical bunch.
When the problems with the K6 are ironed out (packaging, mostly, and this will be solved soon), when it goes
Anyway, relax about AMD. Yes, they can still pull defeat from the jaws of victory, but it would be hard.
The WinChip deal is a pity, though. I have rechipped a lot of old Pentiums with those and they were always an improvement and ran very cool.
Well, for those "blighted souls" the K6-3 is perfect. However, I'm willing to bet there are a lot more games being played at any given time than there are kernel compiles in progress.
The Firing Squad review was incorrect. I have a prerelease K7-550 and it utterly kicks the PIII's ass, in every parameter. Rest assured, Coppermine will not be enough to close the gap.
"I work at Intel so obviously I like Intel chips better."
That's a shame. That kind of thinking has brought us the Crusades, the genocide of Native Americans, Communism, the Holocaust, and now, with the advent of the K7, obsolete processors.
I'd suggest a few introductory philosophy texts next time you're browsing Amazon.
Historically AMDs management have put their sel interest before anything else. Atiq Raza was an exception. My guess is that he quit because of frustration with the others.
AMD can design good chips but it takes solid management to ship a successful product. I would never buy an AMD chip. I don't believe in their ideology(or the lack of it).
Geeks like performance. But who drives home the big bucks? Businesses. You can bet your ass that even if the K7 doubles the PIII-Xeon, you still aren't going to see too many IS managers risking their neck by saying, "we're buying 300 K7-servers this next year". Hell no. Think about what the K7 competes with - Xeon class processors. Who buys Xeon class processors? I certainly don't. I find better value in a celeron. So in the end, you've got a chip competing in a market where it's almost impossible to give away the damn things. What if they bring the K7 down to the consumer level? Well, then they take losses up the wazoo.
You don't have to have the best products to win the game, you just have to have the better brand name. And AMD's brand is going down the toilet..
Pepsi is bigger than Coca Cola. (Yes, really!)
And I hate Pepsi...
--
I really like my cheap winchips. despite the name they run linux just fine. It's ashame so much hype is put on what should just be another piece of silicon. Ah well...
AMD need to survive this. One by one, Intels
rivals are dropping out, first Cyrix, now IDT. If AMD can't survive the price war with Intel, then Intel will be under far less pressure to reduce prices (and may raise them), and chip development will no longer be pushed forward by competition. This will be very Bad Thing for all x86 users.
First post? Probably not. Who cares anyway?
Yellow tigers crouched in jungles in her dark eyes.
She's just dressing, goodbye windows, tired starlings.
Cyrix.
IBM was also producing Cyrix chips for a while, I doubt they still do.
And is Transmeta working on an x86 clone?
There's a broken link (to IDT) in the standfirst
Intel has tried to keep the Celeron's performance a secret -- except at the "low end" where it has been pitted against AMD chips. I was a little surprised at how far they were willing to go to prevent cannibalizing sales of the PII when I first read the benchmarks on their site and found that they measured performance of the Celerons and the PII's using different benchmarks! They have since added several other benchmarks including SPECint 95, allowing comparisons. I guess with all the articles about Celeron performance they just gave up. However, my C366 came in a box clearly labeled "For basic computing" with no bragging about performance whatsoever.
Meanwhile, in many niches the Celeron is eating away at the PII/III market share now. If AMD loses their competitiveness, I fear that the days of cheapo, high performance processors will be over.
AMD's only chance is quality of manufacturing and Yield, Yield, Yield!
Geeky modern art T-shirts
I love AMD chips, lately they have been great (k6-3 and from what I hear about the k7). I am one of the firm believers that competition is good, it will force intel and amd to drop prices, develope better chips and what not.
A small problem that I see with intel vs. amd is that intel doesnt just sell chips. So if they were smart, they could lower chip prices and raise the price on a different product. Until amd couldnt stay with them.
I do not want to see this happen, but you never know these days. Business is war. Anyway, I have said my $00.02, now time to go get coffee.
- "My name is Legion, for we are many" -Mark 5:9
how many are there?
i can only think of three x86 clones that still exist (and only one of them is in production)
AMD
RISE
Transmeta
am i missing anyone here?
------- Oh damn.... the Sigfile escaped... -Great OM
click here
--
http://www.beroute.tzo.com
"Science will win because it works." - Stephen Hawking
I really doubt this is much of a surprise to anyone. I avoided upgrading to a K6-3 last quarter because I am waiting on K7 to come down enough to be affordable. I am certain I am not the only one in this boat. The forthcoming (at the time) K7 probably also put most of that dent in Intel sales.
But they beat the estimates wallstreet had for them. In that respect they are a little strong then what everyone thinks. Never count a company with a loss out.. look at Amazon... Barnes&nobles.com or basically and .com stock except ebay and yahoo.
"Imagination is more important than knowledge" -- Albert Einstein
Transmeta ?
I don`t think you could call that an x86 clone.
Allegedly it is much more advanced than current CPU`s and will provide x86 emulation at performance level equivalent to todays x86.
Though none of that has been comfirmed let alone been put into production
AMD is great...really great....I mean, their
chips become better and better, and they are
CHEAP ! LONG LIVE AMD !
>IIRC, Osborne produced *the* "portable" computer,
>there really wasn't any serious competition to
>them
Try Kaypro. It stacke the drives on one side rather than one on each side of the screen, allowing a 9" (?) monitor that displayed a full 80 columns at a time rather than Osborne's 5" with 52(?) columns that scrolled with the cursor to cover 80 columns. ANd it's been a few years, but it seems to me that they had the same price.
rick
I work at Intel so obviously I like Intel chips better. I've never had any problems with them and they are very well made. I know several people that have celeron 300's overclocked to 500 Mhz. The reason that AMD has trouble taking on Intel is because they don't have the ability to mass produce what they make. Intel can refit their fabs while they are still running, so there is no downtime between different processors. I think AMD is good though. Its a love hate relationship. Without them there is no competition. If they were not around the government would be all over Intel because of such a high market share. Anything that keeps us engineers on our toes is good.
People, their what's for dinner.
Intel is threatening to remove the SMP capabilities from their Celerons.
Hmmm... I wonder why, could it be becasue they don't want people building power system with their, 'low-end' processor for pennies.
/* CDM */
Btw, for those that have taken an interest in Microworkz from reading Slashdot, be sure to read the warrenty (anti-warrenty?) VERY CAREFULLY. The Microworkz sales staff will explain up and down that their machines are Y2k Compliant but when it comes to providing things in writting they state that they do NOT COVER "damages occurring to hardware and software as a result of the manufacturer's failure to comply with 'Year 2000' requirements." I am sure there are plenty of Comtrade customers that will testify to the fact that it doesn't matter what a sales person says over the phone, it is what is in writting that matters. Companies like Penguin Computing do have a customer friendly Year 2000 statement. But there is always the issue of if you have the time to deal with getting RMAs. And if you think that just because a company says they have tested their systems as being Y2K compliant that means it is a non-issue, think again. Gateway 2000 sales has made several promises yet several computers they shipped May of 1999 tested postive for the Crouch-Echlin Effect (and it effects Linux). GW2k still has yet to get back to us with the BIOS upgrade they promised over a *MONTH* ago! Over 12 employee hours lost sitting on the phone (mostly on hold) with GW2k just so they could save a couple cents by using an unbuffered real time clock on the motherboard. Unfortantly, several of the AMD resellers aren't making wise choices in what motherboards to use either.
I keep missing where the FP is so horrible in AMD chips. I've used AMD's since a 486 DX4/120; I have yet to find a game that doesn't work or works so horribly on my AMD CPUs that I rush out for an Intel stickered CPU.
If it's FP benchmark performance that everyone relies on, then consult the recent stories on the performance of Linux vs NT. NT may be faster on the quad-Xeon machines, but it would take many, many T3s and millions of hits a second for 24 hours every day for it to be too much for lowly Linux to handle. I don't care how fast a Quake demo runs, just how the game plays. With a 3d card, K6's play Quake, etc just fine. I have issues with the dull game style of all these 3d FPS games, but not with the drawing/calculating of graphics or anything.
Come to think of it, when I got my DX4/120 was about the time the huge 60MB Diablo demo came out. It put up a warning that it required an Intel Pentium to run, but worked just fine without. I just consider these sort of games, pushing for a one-vendor world evil and dislike them even more.
It would be really cool if Sony bought the rights to the WinChip. Incorporating Playstation II technologies into the next generation of x86 chips. And doing what the Japanese have done in
the past, mass produce and dump it below cost to gain market share, that would definitely benefit
the consumers and at the same time jab at Intel.
A Korean chipmaker buying the WinChip would be good too. I doubt the Taiwanese would step up to the plate again this time, as they already have
Rise and Via/Cyrix. What about to Mainland China?
Thousands of cheap (free) prison laborer.
Just to give you an idea of what Intel does with this extra money ... they pay companies like mine (we do web sites often with heavy Java content) to add extra bloat and unnecessary features to our code that we tweak to run better on Intel processors, then Intel gives us money to put a "runs better on a Pentium II" logo or somesuch on the site.
...
There are alot of small companies who are in a situation like ours who can't afford not to do this, when Intel will give us so much money to do it.
BTW if I was the one making the business decisions I probably wouldn't go along with it, but I'm just a coder
You buy something because of ideology?
Well, I *don't* buy something because of ideology (Eg. Microsoft products, something from intel etc.).
From what I have read and seen about the K6-3 is that for int operations is was faster than the PIII, but that the FP calculations weren't quite up to par with the PIII.
Now, we all know that benchmarks aren't everything so don't go raving at me about this but the one benchmark that I have seen for a BETA K7 was that the FP calculations weren't as fast as the PIII, nor as fast as a supercooled K6-3. That was over a month ago though and it was a BETA chip on a BETA motherboard.
My guess is that in August when some 3rd party benchmarks and reviews come out that the K7 will show itself to be the top dog. I say guess because I have read some articles discussing how the PIII FP calculations were set up in such a way to be more robust, not faster, but more robust then that of the K7.
Before anyone starts spewing out how the K7 has three FPU pipelines, and how it kicks the living ?hi? out of the PIII, please take these points into consideration.
1) We are all working on speculation. Unless you somehow have your own personal K7 to play with we are all relying on what other people have said, and posted.
2) A lot of the articles I read compairing the K7 to the PIII with actual performance data are now at least a month old, some up to three months old. If anyone has information from a NEUTRAL 3rd party with some actual performance data, please post a url to it.
3) I am planning on buying a K6-2 within a week. I am a supporter of AMD, but I also understand that Intel is starting to head down other roads. We don't hear about AMD starting to design a 64 bit processor, althought considering they are working with the Alpha group... . Intel has split its resources into several projects. With them working on upgrading the PIII, the Merced, etc etc etc, of course they won't be able to fight AMD in the 32 bit market as effectively.
Disclamer - Opinion of Person
It seems like a lot of people here just associate FP with games. I do pro audio -- lots of intensive algo processing. And I assure you, the K6/2's performance is really, really bad in comparison.
Not to mention a rather inapplicable analogy.
Did he say AMD chips should go away? No. Did he say he wished AMD's shrunken marketshare was lower, or that there was no competition? Again, no.
I use AMD chips. I use intel chips. They each have their uses and fall into different catagories based on price and performance, respectively. Maybe he gets an employee discount or they're a fine company to work for. Hence, he works for intel and thus likes them better.
The suggesting that he aquaint himself with introductory philosophical texts also appears to be a parsed insult that was uncalled for. While I would urge *everyone* to read up on a highly interesting subject, you're implying that there's something wrong with his thinking from an ethical standpoint. Which just ain't friendly, dig?
I see no flaws in logic or form without further evidence.
I am in a position to choose what computers we will buy at work and I would love to buy some boxes with AMD chips. Unfortunately all the amd machines I see are "home/game" computers. I need machines from a reputable dealer that are certified for/bundled with Win NT, MS Office SBE/Pro, etc. Until then I'm stuck with Intel by default.
> I'd like to see a disto that shipped a binary kernel and development kit (autoconf, compiler,
> etc.) and source for everything, so that your installation kit booted up, prompted
> you for what you wanted to install, and then compiled everything (including replacements for
> the kernel and development kit) while you took a day or so away from the terminal (or went on
> vacation, if you're running a slow processor).
If you don't have to use Linux get one of those BSDs. You can rebuilt the complete OS with a single command (make build (NetBSD (and OpenBSD?)) or make world (FreeBSD IIRC)).
I have read more info from "neutral" sources lately, but this is the only one I know offhand (I don't bookmark every cool tidbit I find about the K7 :)
http://unreal.epicgames.com/
The interesting quote is (by "Tim"):
The AMD Athlon Rocks!
My new 550 MHz AMD Athlon (K7) just clocked a jaw-dropping 68.5 Unreal timedemo at 1024x768,
running on a Voodoo3 3000 card. Even more telling, at no point did the frame rate ever drop below 38.0 fps. That's astounding, considering the intense lightmap and geometry usage in the timedemo level. Even while playing Unreal Tournament's most texture and polygon intensive level (Shane Caudle's DmGothic), the frame rate hardly ever went below 60 fps.
The Athlon's 128K L1 cache is awesome for memory-intensive games like Unreal. Operations like visibility determination, which thrash on the Pentium III's 32K cache, now run at full speed on the Athlon. This CPU truly shows a generational performance improvement, like going from my old 486 to my first Pentium.
When I saw AMD's K7 spec, I was pretty skeptical. The K6 had been hyped up, but in reality it was
slower for Unreal than a Pentium II of comparable clock rate, due to its poor non-SIMD floating point performance. The K7 claimed to fix all of that, and debut a new architecture with 3 execution pipelines. I decided to wait and see, without getting my hopes up.
Bottom line: I waited, and now I have seen! The Athlon is clearly the fastest x86 CPU at any clock
speed.
Congratulations go to AMD.
K7 isn't competing with Xenon processors. Where I work quad xenon servers power a database engine with serves/updates/does stuff with a database which requires an 80gig raid to serve (One freeking server is as big as my desk...yipes).
:)
... [which is also bs but I digress]).
;)
This is just one of the BO servers.
Those IS managers couldn't give a crap about what runs in their servers. They do care about what company they buy from. Sears buys ALL of their stuff from IBM. Sears doesn't care what's in the machine, just as long as it's from IBM. Hell, most of their machines still run OS/2 (not even OS/2 warp)... (the other biggie I notice is HP; they make some nice server boxes worth drooling over...).
K7 is targetted and the mid to high end of the consumer market. If it were targetted for high end business use, AMD wouldn't have bothered with FP performance (hello, most businesses don't care about FP -- that's why BUSINESS benchess stress the int performance of a processor). On top of that, if AMD were really targetting the high end server market, they would NOT lauch it without an SMP motherboard/chipset.
Celeron for you, is probably better. But you'll have to buy something new next year to play the new games. Don't play games? Then why did you even bother with the Celery? Get a $50 k6. Hell, my k6/233 does everything well except play the new games (and it isn't THAT bad either, if you like average framerates of 15fps (in Q3), which some people around here claim near the max the human eye can see
In addition, last time I checked, intel was making a profit by dumping the Celerons and charging $600/pop for the 450+ PII/PIII's.
...some people... I mean come on! Do you HONESTLY think that the rest of the world's opinion is confined to what YOU see? If that's true, we're all in trouble... then again, I have been noticing an increasing number of complete dolts during my morning commute lately...
That's just what Intel did. They dropped prices on the PIII and Xeon a shitload to undercut the prices announced by AMD for the K7. It looks like Intel is starting to get serious about using predatory pricing tactics, and may even be willing to start taking losses on some of their CPUs to drive up their market share.
AMD is going to be forced to drop their price on the K7 to get it into OEM boxes, and if their yields aren't good they are going to loose so much money they may really be forced out of the market for good this time. Already AMD is loosing money on the K6-2,3 lines despite good yields because they have been forced to match Intel's price cuts.
Unfortunately, the performance of the K7 isn't sufficiently better than the PIII to allow AMD to charge higher prices for similar clock speed chips. Also, it looks like Intel will be able to shrink the PIII to a smaller process before AMD and thus beat the K7 with higher clock speeds.
I think the future of AMD depends on whether Intel decides they are willing to loose money for a while to drive AMD out of business.
Here on News.com
RB
Most people have known for about two years that the K6 was fundamentally lacking in FP performance and a lot of people avoided the chip because of it. Instead of addressing the issue, they spent their money on: 1) incorporating 3DNow! instructions (K6-2) which have been a collosal flop for everything other than Quake2 on a 3dfx card, and 2) on-chip L2 cache (K6-3) which pumped up their integer performance at the time when even casual users had come to learn about the FP issue. As a result, they lost out in both the high end and gaming market segments. It is unfortunate that now that AMD finally has a processor that beats Intel on the high end (at least temporarily), most buyers have already come to associate AMD with low end processors.
I sure hope AMD makes it through ...
... yah for AMD!
They are the Pepsi to Intel's Coca-Cola
They are the Compaq to early IBM
They are the Linux to Microsoft's Windows
They are the alternate choice for those unwilling to put up with Intel's deceptions regarding their low end processors.
Computers @ The Manor, our retail store features AMD and Intel products, yet 75-80% of the systems we sell are AMD based.
AMD has forced Intel to speed up development, and drop prices, they are operating on record low profit margins, barely paying for their upgraded fabs before they have to drop prices
I have some questions. Had AMD done as much as intel has for Linux? Has AMD donated compiler technology? Had AMD donated machines and know-how to kernel dvelopers? Have AMD invested in Linux startups? I beleive Intel has done all that and they keep getting panned, but I have never heard any news about AMD doing anything for Linux...
Hard drives will return a serial number upon request
I believe there are certain video cards (high end ones albeit, but nontheless...) which do this too.
Get over it. Serial #s on processors are nothing. Nothing at ALL.
Andrew
According to AMD's benchmarks, the Athlon is 146% faster in SPEC_fp95 and 118% faster in SPEC_int95 than a PIII clocked to the same speed. However, AMD didn't publish the actual benchmark numbers, so we have no idea yet whether it was a fair comparison. To confuse things further, Thresh's Firing Squad did a bunch of benchmarks on a pre-release K7 and found that it lagged the PIII (and even the K6-3 in most cases) in every benchmark except CPU WinMark (an integer benchmark). Now the K7 used was probably not final silicon, and the benchmarks used are probably optimized for the PII/PIII architecture not the K7, so we'll have to wait for reviews of the actual production parts before deciding just how fast it is. But you can bet that the AMD SPEC benchmarks are going to look better than the real world benchmarks. 118% is definately not in the "blowing away" category. 146% is, but it is a best case number (3 FP pipelines / 2 FP pipelines = 150%) that is unlikely to be reached in real world apps & games. Furthermore, code is going to have to be optimized for the Athlon in order to take full advantage of it. Everything you are currently running isn't going to optimized for the Athlon, and AMD is going to face an uphill battle to get people to optimize for the Athlon just like they did with 3DNow!
.18u process. Intel is already planning to start production on the new process later this year, while AMD isn't planning the switch until the first half of next year. Combine that with AMD's track record of terrible yields every time they switch processes, who do you think is going to get to 1 GHz first? I'll bet that Intel stays ahead in the MHz race for a while to come.
For purposes of comparison, a 21264 at 500MHz matches a PIII 600 in integer performance, so the Athlon is probably going to be right about the same as the 21264 in integer performance. However, the 21264/500 already more than TRIPLES the FP performance of the PIII/600. That's what I call "blowing away". The brand new Athlon is still only going to produce less than half the FP performance of the 21264.
BTW, your comment about the PIII having a full speed L2 cache is wrong. Both the Athlon and the PIII (which is what it was compared against) use a 512K half speed cache. The Xeon processors have the full speed cache. The PIII is basically just a PII with the added FP SIMD ops.
Finally, neither AMD nor Intel is going to get to 1GHz before switching to a
I do think AMD is doing the right thing by using the EV6 bus and pushing RAM technology. It does mean that the Athlon has more room to grow than the current PIII/BX combo. It may give them a small head start since Intel will need the new PIII variant and support chipset to take advantage of increasing memory bandwidth.
Despite what I say, I'm no fan of Intel. I've owned 3 AMD processors myself. However, I'm very worried that AMD is going to go belly-up due to abysmal Athlon sales since there are going to be a lot of apps around that run like shit on it due to lack of optimization.
Please take a moment to check out this URL. AMD is in production at the Dresden FAB (ahead of schedule by a couple of months) with copper K6s. I may very well have been wrong abou the PIII cache, but I DO know that Xeons have been used in some of these performance comparisions, and got beat. Second: Thrash was using a performance impaired beta release of the processor on a performance impaired beta board. I'd take any all numbers from that test with a pound of salt.
re: Intel beating AMD to 1Ghz.
Only if Intel is putting them in the fabs by Thanksgiving. The Dresden FAB staff seem to have gotten it into their heads that 1Ghz will be rolling off their lines before 1/1/00, putting them in the market EARLY in 1Q/2000. This is a potentially devestating blow to Intel's server market, particluarly since AMD should also be running 8-way SMP about the same time. It'll be a VERY interest 1H 2k, that's for sure. Oh, and I think the Athlon is going to sell fairly well for one, and only one, reason: it's faster than any other x86 processor currently on the market. If AMD plays it right, in 5 years people will be saying "Intel used to do WHAT??"
Unfortunately, being a much larger and more diversified company, Intel can stand to loose money on their CPUs for a lot longer than AMD can. Unless of course they choose not to in order make bigger short term profits and keep their share price high, or unless they fear anti-trust action.
We're not talking about high end workstations here. Or at least I'm not. I'm talking about my home machine. Honestly I don't care if anything I use at work is subject to anyone's prying. But would you want anyone to be able to retrieve the serial number of your stereo, and then determine where you bought it? Admitedly, if my PC was ever stolen, then the serial number could be used to trace it, but that's the only use as a consumer I can see for it other than to tailor advertisements towards me. You're right; on a high end workstation that's not going to be the goal, but it IS the eventual goal of the same technology on a home PC. And I want none of it.
:-) That's the important bit, right?
And for what it's worth, though that's what gave me a bad taste in my mouth for Intel, that's not why I chose an AMD. I chose an AMD because every spec I'd read showed it outperforming equal or higher clock speed pentiums. And I'm happy with mine
Bad things often happen to good people,
It is up to them to see that they remain good.
You wonder why the K6-3 500MHz is a no-show? I think that probably the entire K6 production has been ceased so that any and all chip fabrication resources that AMD has at it's disposal are being concentrated to the effort to crank out full production of K7 Athlons only right now. We'll probably not ever see a K6-3/500. If the K7 is a flop, either technological or marketing flop, then AMD surely has pulled an "Osborne" and will be in unsurvivable trouble. You DO know the story of the Osborne computer don't you? ....and how that company's failure was caused by blabbing too much about a future, yet-to-be-released product that it caused all sales of it's current products to come to a halt, therefore ending their ability to even produce their new product.
If/when Linux and free software on it gets more dominant, we would be less dependent on the x86 architecture, and Intel wouldn't be so powerful. Open source allows easy porting, after all.
The only thing that doesn't suck about the x86 architecture is that it's mass produced and cheap. Imagine a world where you could choose between using an x86, Alpha, MIPS, PPC, ARM, etc., solely on the basis of price/performance. The demise of clone makers such as Cyrix or IDT would not then carry the risk of Intel re-taking its stranglehold on the market.
That would be cool, in more ways than one: CPUs wouldn't get as hot, because there would be no more need to be backwards compatible with some crappy 20 year old instruction set.