Silicon Investor (SI) -- The First Internet Community

STOCKTALK

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor. We ask that you disable ad blocking while on Silicon Investor in the best interests of our community. If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.

Politics : Formerly About Advanced Micro Devices -- Ignore unavailable to you. Want to Upgrade?

To: Paul Engel who wrote (82693)	12/11/1999 8:43:00 PM
From: Dan3	Read Replies (3) \| Respond to of 1571546

Re: when SLUDGEHummer appears, AMD will have a NEGATIVE 18 month "lead" Paul, Many of the possible benefits of larger word and address sizes were achieved in the move from 8 to 16 bits (essentially unlimited number of instructions and modes). Moving to 32 bits allowed for the use of flat memory spaces for all but a handful of systems. Moving to 64 bits allows for very large virtual memory sizes and more data to be moved in a single clock by a single instruction. Very large virtual memory sizes are needed primarily by terabyte sized databases - even 5 years from now, this will matter to fewer than 1% of systems. The other benefit, making extended (64bit) move, load, etc. instructions available to the compiler should not be too hard to implement. Many of the postulated benefits of the Itanium are due not to any inherent advantage of going from 32 to 64 bits, (unlike the moves from 8 to 16 to 32 bits) but from having the CPU execute several 8 to 32 bit operations in parrallel which is the VLIW paradigm or a sort of generalized 3DNow/SSE instruction set. It is providing a compiler that can parrallelize instructions in this way that has occupied Intel for many years as a compiler was developed to take advantage of the planned architecture of the Itanium. What I find intriguing is that Sledgehammer may come out from the start with two cores per chip. 64 bits should handle any address requirements for a long time, and if instructions are provided to synchronize the cores to allow for 128 moves, loads, etc. AMD could be shipping a 128bit processor a few months after Intel ships its 64 bit processor. The dual cores would let the chip take advantage of years of development of multithreaded operating systems and compilers without so much as a re-compile. The synch instructions would let it move data at twice the per/clock rate of Itanium. My understanding from early published reports is that Itanium is pretty large and uses a lot of power - it may be several more years before it will be practical to put multiple Itanium cores on a single die. So AMD could be the sole provider of 128 bit chips for quite a while. Maybe 4 willamettes on a core? Oops, only 32 bit addressing, too bad. Two Mckinleys on a die? How long till that can ship? Cheers, Dan

To: Paul Engel who wrote (82693)	12/11/1999 11:26:00 PM
From: kash johal	Read Replies (3) \| Respond to of 1571546

Paul, Re:"The problem for AMD is that all the major OS's have already been ported to IA/64 bit architecture, and Intel is shipping IA/64 bit systems to hardware AND software developers. Thus, by the time SLUDGEHummer comes out, all the major players will have had 18 months of IA/64 development under their belts, and will be shipping their applications running under EPIC/IA/64 versions of Windows 2000, Solaris, Linux, Monterey, etc. That puts AMD well behind the 64 bit 8-ball - late again to the new market ! Even when AMD was EARLY to market with 3DNOT!, they barely attracted anybody but gamers to use that software, giving up a nine month lead to Intel's SSE. Now, when SLUDGEHummer appears, AMD will have a NEGATIVE 18 month "lead" - I doubt that any major developers will shift gears from IA64 - unless the IA64 is a total bust. So far, the ITaniun sounds like it will be what Intel expected it to be - far from a bust." You make some excellent comments. If windoze and Linux are not ported to Sledgehammer then I agree it will be a dead duck. As far as Itanium, I agree it will be big, very big for Intel in the 2002-2005 arena and allow penetration into "big-iron" apps. However my understanding is that Itanium systems will run in the $100K plus range. I don't expect sledgehammer to ever play in that arena. I can see it play in the $2-25K server market ie where Intels Xeons/cascades play. I can also see it play on high-end desktops. Lots of folks may buy it even though they only need 32 bits - just so they can be 64 bit compatible. If the die size adder of 5% is true - I can see AMD moving its high end CPU's to sledgehammer. Lots of folks hate to buy an expensive CPU if they think it will be out of date and not be able to run some future s/w. An example is folks who are buying HDTV's even though hardly anything comes in HDTV. And even when transmissions start common delivery methods such as cable/DirecTV won't work for many years due to bandwidth issues. But I do agree that timing of sledgehammer will be key. AMD must deliver it in early 2001 or mid 2001 at the latest. And they need to give some indication of major software support - such as msft in 2000. regards, Kash

To: Paul Engel who wrote (82693)	12/13/1999 1:24:00 AM
From: Charles R	Read Replies (2) \| Respond to of 1571546

<Now, when SLUDGEHummer appears, AMD will have a NEGATIVE 18 month "lead" - I doubt that any major developers will shift gears from IA64 - unless the IA64 is a total bust.> If moving people to new instruction sets were so easy why didn't MIPS, Alpha, Intel's own 860 succeed? The fraction of applications that get ported to IA-64 will probably be less than 10% of x86 applications 5 years out. If Intel doesn't counter with x86-64, AMD will become the owner of x86 extensions. And Intel might get a chance to see what it was like to be SGI or DEC a few years back wondering why a brand new instruction set does not take off. <So far, the ITaniun sounds like it will be what Intel expected it to be - far from a bust.> I have hardly met a guy from Intel who is not pleased by how well the Merced platform seems to be coming up so far. Everything other than MHz seem to be going its way. And, from an external view point, Sun's bumbling with UltraSparc may be the best thing that has happened to Merced. Score one for Intel.