Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Register for free to join our community of investors and share your ideas. You will also get access to streaming quotes, interactive charts, trades, portfolio, live options flow and more tools.
Re: It seems any Intel announcement, is lame, any AMD future annoucements are - well - announcemnts of new and exciting products
The thing is, AMD announcements have some validity. They are often late in delivering, (hey, it's tech) but AMD has an excellent track record of actually doing what they said they'd do in their announcements.
Back in 1998/1999 AMD said it was going to produce binary compatible 64-bit AMD64. AMD said it would move to a copper process. AMD said it would use dual cores in its chips. AMD said it would develop chips with direct-connect on die memory controllers. AMD said it would incorporate on die Hypertransport.
A little later AMD said it would move to DDR Ram. AMD said it would use SOI.
AMD has followed through on pretty much every aspect of every roadmap and announcement.
Now look at Intel. They said - actually more like shouted - that they'd never "extend" X86 to 64-bits. IA64 (Intel Architecture 64) was the only Intel 64 anyone would ever see from Intel. Intel said they'd never use DDR and Intel announced that Rambus would be in nearly 100% of their systems by now. Intel said they'd ship 5ghz parts in 2005 and 10ghz parts around 2008. Intel said that SOI was a waste of time and that bulk silicon had no serious problems with leakage.
This is why we scoff at Intel announcements and listen carefully when announcements come from AMD.
Re: falling further behind in process technology (IBM notwithstanding) at 65nm and 45nm
Have you seen how much lower power consumption is for AMD servers vs. Intel servers? While performing better than Intel servers.
And to see the effects of that, look at 80+% per quarter growth rate AMD is seeing in server share.
AMD's process may or may not improve relative to Intel, in the future, but to use the term "further behind" for the present is simply wrong.
and potentially an over-capacity scenario with Fab36 and Chartered coming online next year.
AMD overcapacity might be an issue for AMD - it would be a nightmare for Intel. As the smaller player in the market, having just signed up to take depreciation on a huge new FAB, AMD will be a "price taker" in some segments - AMD will sell many of its parts for whatever the market will pay, and that will drive prices down for each of the 2 to 3 times as many parts Intel sells (since that overcapacity and "sell it at any price" will result in a market share for AMD of 25% to 33%, maybe higher.
So AMD will see lower prices, on much higher volume, but Intel will see lower prices on lower volume.
You said it, not me.
Re: Leakage problem solved
Sure it is. Just like strained silicon solved let them get to 5ghz on 90nm.
Intel's big at promising huge gains 2 to 4 years down the road.
How about now? How about next quarter? How about 2 quarters from now? What about all those promises they made 2 years ago that were supposed to have come true by now?
Re: New Quad Xeon MP tries to match Dual Opteron
That's just sad - what's the cost penalty for the quad Xeon, 300%?
Let's test it against a 16 core 8 socket Opteron box, since cost and number of sockets don't seem to be criteria in this particular "comparison."
http://www.wsm.com/amd/servers/fusiona8/index.html
Re: The way some on this thread speak, you'd think anyone buying a Pentium 4 is getting a real piece of crap, while anyone buying an Athlon is getting a pearl. Kind of like the difference between buying a Yugo and a Ferrari.
It's more like sirloin steak vs. ground chuck - either one will make an adequate meal for most purposes, but the grocery stores are all selling ground chuck for $4 per pound and sirloin steak for $2 per pound.
The thing is, if pass on the steak and buy the ground chuck instead, they give you 10% off everything else in your cart.
Which is pretty unfair to the company that's selling the sirloin steak for 1/2 the price of ground chuck.
Ultimately, if the courts and justice department allow this sort of behavior to continue, our economy will be ruined the way the Soviet Union's was under communism,. For a market economy to flourish fair and open markets are required. If Intel is allowed to act as a sort of commisariat of chips, getting whatever price it wants for whatever products it decides it wants to limit the public to, the US will ultimately lose.
PS - in servers, of course, the difference is actually that large. A Xeon is flat out crap compared to a dual core Opteron. In other areas, Intel doesn't lag as far behind and in notebooks they are competitive (although Intel sells dead end products vs. AMD's "ready for the future" products they are still competitive).
Re: ULi Electronics anticipates its sales in the third quarter will climb more than 20% sequentially, driven by rising shipments of K8-based chipsets, as well as southbridges for use with ATI northbridges.
Good to see that the chipset co's have been showing the common sense to fill orders for chipsets for AMD (that has been partnering with them) first, and then filling orders for Intel chipsets (which has the stated goal of taking all of their chipset business from them) as capacity permits.
Re: Tenchusatsu wrote - It wouldn't be too hard to disconnect 3/4 of the register file, 1/2 of the reordering buffer file, and 1/2 of each execution unit
That would take 12 to 18 months (or longer) between design, prototyping, and validation.
And those resources are stretched pretty tight at Intel these days.
Their alternatives were to delay Xeon a year or do what they did - put an extra $1.50 worth of heatsink and fan in each boxed prescott processor box and let their customers suffer with noisy, power eating PCs.
Re: Intel's dual-core Pentium 4 a rush job
I don't think any of the info in that article was news to anyone here.
I hope the engineer doesn't get in trouble for being frank.
Re: I don't think the Inq has all that great of sources
Yeah? The Inq was telling the world about Intel's clone of AMD64 when Intel's ceo was telling investors that no such project existed at Intel.
Apparently the Inq knows more about what goes on inside Intel than Intel's corporate officers.
Re: The Inq may not like the simplicity of the MCM approach
By the end of 2006, AMD will very likely have quad core chips. There's nothing inherently wrong with MCM's but they aren't "simple." The simple, elegant, and inexpensive solution is put it all on one chip of silicon.
The ugly, expensive, kludge is to solder together a module.
Intel's costs will be roughly triple those of AMD if AMD is selling a single chip against two Intel chips, a module, and the labor to assemble and test the module.
AMD had a very tough time trying to make money when they had to ship modules (early Athlons with off die cache) against coppermine with its simple (and cheap to produce) on-die cache. AMD was shipping a CPU, 2 cache chips, and a board that had to be produced, assembled and tested against Intel's single chip.
Now it may be Intel's turn to try to compete the hard way.
Re: makes you pretty old, like geezer material,
That's me, all right.
But still fighting the good fight.
It started with warring against the closed mind types who didn't want to let those newfangled "PCs" in when we had perfectly good terminals that worked just great with the VAXs and the Mainframe. PC SAS was a real breakthrough in getting the serious work to move on "down."
Now it's dual core Opterons vs. Xeons. 80% more performance with 33% less power use - but you have to talk for hours before you can get people to go with it. "I didn't see it advertised on TV with Burger King and cheap home loans! How do I know it will work? Those guys don't take me out to lunch and send me free golf balls like Intel does, why should I buy from them?"
Why are so many "technical" people so terrified of new technology?
Re: called Cloverton. It is a four core, two Woodcrest die MCM, and it is slated for Q4 2006
What does sticking two chips on a circuit board have to do with a tri-core or quad core competitor from AMD?
If Intel has been forced to go "off silicon" to get four cores into a socket as late as Q4 2006, they must be getting some disappointing news on future power consumption.
Re: <i.My first car, was a 1937 Ford coup. Gas was #0.16 a gallon. Had to work hard to afford the gas. I remember when it went up to .18 cents, peopple were shocked
Man, you sure got me beat! I was worked part time at a gas station when I was in high school, and remember a good price for gas was 29.9 cents per gallon (it sort of rolled off the tounge), but we'd go below that during price wars with neighboring stations.
We gave away green stamps and free glasses with fill ups, on top of that. That was a full service price that included pumping the gas, cleaning the windows, and checking oil, water, and tires.
Re: IPF has a little over 10% share. In certain sub-segments like HPTC IPF share is reprotedly at 20-30% of the RISC plus IPF market.
Thanks for the link. It still makes it somewhat hard to see how IPF is doing. Intel gives away the chips for the HPTC market, which buys them sales. As far as I know, they actually sell to the commercial market (at least, I've not been offered free Itaniums by anyone but the HPTC guys). So in trying to track how IPF is doing, the question becomes - how much of 10% of the total RISC market is 20% to 30% of the HPTC market?
At best, it looks like IPF is treading water, at great expense to Intel (still, that's better than I was expecting). At worst, it's fading fairly fast.
VLIW always looked interesting, but it comes from the perspective that the software is the easy part and the hardware is the hard part. History seems to show that it's the other way around.
Itanium's IPC will be slipping in forthcoming versions, according to Intel's roadmaps, which probably indicates that they've decided that while Prescott's pipeline was way too long, McKinley's may have been about 50% too short. So an otherwise unqualified note that the clock is going up may not indicate as much as one would hope.
Re: With Intel running smack into the heat wall, there is no reason for them to run "high power transistors" dedicated to 64-bit, unless there are no such things except in your own mind.
Surprise !
Intel copied AMD's spec for AMD64 and named their reverse engineered copycat chip EM64T.
You should check it out - Intel has been adding a cloned version of AMD64 to all of its chips, from Celeron to Xeon, except Pentium M, where they couldn't fit all the high speed transistors needed by a 64-bit chip into the power budget.
Re: That must really be a poor guy
$1.23 in 1995 and $1.33 in 1997 isn't all that far back.
But you're right, I should have used a base of $1.50 (2002), rather than $1.25. And I used nominal prices, which is wrong, but it's still the way most people think/view the world.
Regards,
Dan
Re: today's leak of a threat from oil tankers
That should cheer up Monday morning's market.
Re: The question HP is asking itself is when will x86-64 be good enough? They are hoping soon, as they are not replacing PA-RISC or Alpha sales on a 1 for 1 basis with IPF which is frustrating for them. Time will tell.
Are you sure and do you have a link? That would be the strongest evidence one could find that IPF is or isn't doing well, but I've not seen any numbers either way.
One would expect that the burst of pent up demand for ANY hardware from HP to update their HPUX and True64 user bases will be largely exhausted around now. So we're going to be starting to see how IPF is really being viewed.
I posted the above with the notion that HPUX and TRUE64 users have for some time been reluctant to buy dead end hardware and also unceration about the Itanium "new direction" but that a lot of them were finally squeezed into Itanium due to short term requirements (needing either to expand capacity or replace failing hardware that can support their OS). Most of the "squeezing" has now taken place, so going forward we should start seeing "voluntary" purchases of Itanium systems.
So how's the HPUX / True64 unit market share been doing? That should be a good indication on how Itanium is doing.
Re: So no chance for any other processor architecture to succeed except some form of x86?
Think "Esperanto."
What makes Esperanto easier to learn than other languages?
Esperanto has a number of features that make it relatively easy to learn:
A regular and phonetic spelling system. Where the Chinese school child must spend years learning the relationship between the spoken and written language, and the American school child must spend an almost equally long period learning to spell, the Esperanto system (one letter = one sound) can be learned in about half an hour. This also includes a regular system of accentuation.
A regular and exception-free formal grammar. Doubters correctly insist that the grammar of any real language cannot be completely described, as Esperanto speakers sometimes claim for themselves, with a mere sixteen grammatical rules that can be written on the back of a postcard, and they are entirely correct; but what they fail to mention is that, in fact, most of the grammar, and certainly all of the most important grammar, is available in these rules. Learn eleven invariable grammatical endings and how they are used, and (with a vocabulary) you will immediately be able to invent grammatically correct, usable and useful sentences in Esperanto.
A regular, one-to-one and easily learned system of forming new words from words you already know. This is particularly useful because it allows you to take a fairly small basic vocabulary (the usual figures is about 500 items, including word-roots, particles, and affixes) and carry on long and fairly complex discussions about a wide range of topics, including technical ones. While modern Esperanto has a considerably larger overall vocabulary of unique roots (officially, about 9000 at last count), many of these are simply synonymous with words that can be formed from the most basic roots, and it is always considered acceptable (and usually elegant) to create your own words rather than borrowing somebody else's.
http://www.webcom.com/~donh/efaq.html
Re I found that only around 2% of electricity is generated from oil in the US
Those numbers can be a little misleading. There are a lot of dual fuel and tri-fuel plants in the US, and they switch fuel sources depending on the relative cost of oil, coal, and natural gas. Some plants, particularly in the Northeast, remain fuel oil only, so regionally, even with present prices, some areas get a significant amount of their electric power from ol fired plants.
To a degree, it's a moot point, since the price of Oil has kept the multi fuel plants from burning any oil since the late '70's. Almost all of the new capacity added in the past couple of decades has been natural gas plants that aren't dual fuel, more for reasons of capital cost and emissions than fuel cost.
The problem with natural gas is that there's rarely any on site storage, so as soon as the gas stops, so does the electricity. Oil, Gas, and (of course) nuclear plants can all run for a while even if there's a supply disruption since they all store days to years of fuel on site.
Re: Even more remarkable considering the other huge shadow looming over I2 - the Grim Reaper.
LOL !!
Re: I think the equation changes a bit at $10/gallon.
The figure that seems to pop up all the time to get people to change their habits is $4 per gallon (and it's already at $3.50 in some California markets.
Think of the poor guy that bought an Explorer when gas was a little over a buck a gallon, and uses it for a California (70 miles each way) commute. His monthly gasoline bill has already gone from $250 per month to $650 per month.
That's enough to get noticed, especially since it's after tax dollars.
For someone with a more typical 30 mile (each way) commute, in an area with more typical fuel prices, it's not so bad. Gas is up by about $2 per gallon so that's an extra $160 per month.
At $4 per gallon nationally, the typical driver would be paying an extra $225 per month. If the family has 2 or 3 cars the other cars are probably driven fewer mile per day, but the family fuel bill is still up by $300 to $400 per month. Pre-tax the increment for a fair number of families would be pushing $6,000 per year.
That "deal" on the new Chevy Tahoe or Ford Explorer may not be looking so great as the gas bills keep showing up month after month.
Re: When Intel deposed the FTC expert
LOL !!!
Nice fantasy - next you'll be telling us that Joss Stone is pestering you to let her have a child with you, but you're too busy.
Re: No this isn't discovery... This is solicitation
AMD's discovery is normal and their requests for document protection is normal but their open appeal to others is unusual.
You're right about that but most cases don't potentially affect so many persons and organizations all over the world.
Re: all things will be revealed in the fullness of time.
Oh yeah? Tell it to Jimmy Hoffa's body....
Signed,
Mark Felt
Re: When Xeon moved to 64-bit, the power envelope did not change all that much. Same thing when P4 moved to 64-bit.
Those chips have a 64-bit core with the 64-bit instruction api turned off in microcode. The same thing AMD did with its early socket 754 Semprons - they were 64-bit core (and used the power of a 64-bit chip) but had access to the 64-bit instructions turned off in the bios or through a fusible link.
The high power transistors are still there and get used every time you do an ADD, DEC, XOR, SUB, etc.
Re: If you had the proof, you would have presented it. Stop telling fibs,
Apparently you're incapble of clicking on the "who wrote" button that I pointed you too. Let me help you out.
I posted: You want to compare power consumption of a chip with half the pipeline width
http://www.investorshub.com/boards/read_msg.asp?message_id=7313332
You replied: Care to explain how you consider that to be half as long?
http://www.investorshub.com/boards/read_msg.asp?message_id=7314620
Looks like wbmw "tells fibs."
Re: You tried to pass off the 64-bit feature of K8 as being a significant contributer to power,
It is.
It nearly doubles the size of the core, and nearly doubles the number of high power transistors in the chip.
It only increases the total size of the chip by about 5%, but it nearly doubles the number of high power transistors used in the chip.
Your link is entirely content free -
It says Intel will propose its penalties within 30 days - there's nothing about the actual agreement in there.
All it says is that Intel will come up with a plan to stop their criminal behavior and that the plan will be posted - so where's the plan? What's the agreement? What did they agree to do? That was either never done, or has remained secret.
The fact that the commission suddenly decided to have Intel pick its own punishment is telling - what political or economic pressures were brought to bear to pervert the FTC into coming up with such a lame "order" after years of investigation?
Why not have the Fox draw up a plan to protect the chicken house - and then keep that plan secret from the farmer.
Your link sas only Respondent shall submit, within thirty (30) days of the date this agreement is signed by Respondent, an initial report, pursuant to § 2.33 of the Commission's Rules, signed by the Respondent, setting forth in detail the manner in which the Respondent will comply with the Order when and if entered.
The design of these upcoming processors is largely derived from the Pentium M notebook line of chips.
Intel has continues to have a problem here. A large part of the Pentium M's power savings come from using an obsolete 32-bit core with 1/4 the register space of AMD64 and EM64T. That means it can perform well in some areas (especially if your test is graphic card or cache size limited) but (besides being unable to run current and future versions of Windows and Linux) it has very weak floating point.
These are easily fixed problems, in terms of perfomrance, but fixing them will significantly add to power use.
=
=
=
=
=
=
=
=
=
=
=
=
=
=
=
Re: good one, Dan
Thanks.
Didn't mean to nag you too much but your perspective (cheering results that the rest of the world saw as very bad news) does show you're using something of a filter to view the world.
Re: If you have a case, do you need to "solicit" further information?
If the attorneys didn't proceed with discovery, it would be so negligent they'd be guilty of malpractice.
Discovery is always part of the case, as is continuing to look for additional cases and evidence for and against your position right through the morning of the trial.
My wife's an attorney - this is the way it works.
Re: I would not state something that silly.
Sure you would. And you did.
Go back through the links.
Re: 64-bit only takes up 5%
When companes like AMD and Intel make chips, they use several types of transistors, that use very different levels of power.
A 4mb cache Xeon uses only a little more power than a 1mb cache Xeon (on the same process) while having almost 4 times as many transistors. Bottom line is that most of the transistors on a modern CPU use almost no power.
The I/O transistors use some power, and the high speed core transistors use a lot of power. That's why doubling the number of cores, even though it may only add 5% to the total count, can increase power use by 50% to 75%.
Re: Same old same old. Make up the facts
You guys make it up, then you whine that you're asked to back it up.
Re: FYI, pipeline length has nothing to do with the processing width of instructions
LOL !! you're the one who suggested that it did.
I posted about pipeline width.
Time for a little remedial reading training for wbmw?
Re: Unless you can find a link
You posted that it wasn't secret. Either present something other than that's the way it is in your personal fantasy world or stop pretending Intel didn't insist on a secret plea before it admitted to abusing its monopoly.
Re: Dell did pretty well.
Talk about Intel colored glasses.
Re: extending x86 to 64-bits deserves neither applause for innovation nor masterful planning.
Bull - just about everyone, especially Intel, was shouting that the X86 architecture was inefficient and doomed.
AMD did a brilliant job of interleaving new features, particularly registers and memory architecture, into a structure that maintained full performance compatibility with Intel's obsolete design.
How many times did Intel rework the X86 core without figuring out how to increase the number of registers?
AMD, in particular, Dirk Meyer's team at AMD, did what Intel swore was impossible.
RE: Everyone should drop him a note and ask why. I did.
Better yet, AMD should subpoena him and his boss and ask them why under oath..