Both AMD and Intel were formed out of Fairchild. Jerry Sanders was a salesman at Fairchild, and left Fairchild to start AMD. Robert Noyce and Gordon Moore were the top engineers at Fairchild (Noyce built the first silicon integrated circuit there, and Moore was one of the most prominent engineers there), and they went on to form Intel.
The tradition continues today: AMD lives mostly off of sleazy marketing (PR ratings, anyone?) which Intel has made, and continues to make, very significant innovations (first semiconductor RAM, first microprocessor, first OOO processor, first MT processor, AGP, PCI, USB) while I can't think of a significant 'first' or industry standard which AMD invented.
Re:Hate to say, sounds like a dot-bomb strategy...
on
HP Buys Compaq
·
· Score: 5, Interesting
Compaq ate Digital, sold the StrongARM to Intel who buried it b/c it was an order of magnitude faster than Intel's low-power chips.
Huh? The guys upstairs in WCCG doing StrongARM and XScale (StrongARM, renamed) would be very interested in knowing that Intel buried their product. The fact is, StrongARM is generally acknowledged as one of Intel's key acquisitions in the last few years, and has a highly bright future ahead of it (at some point, it is likely to replace DragonBall in the Palm). It's a heck of a lot more successful than when DEC owned it, that's for sure.
FYI, the entire original StrongARM team walked out as soon as they were acquired by Intel. That's their fault, not Intel's. The Alpha team seems to have been a lot more cooperative (a whole bunch of them were just named Intel Fellows last week).
Re:Hate to say, sounds like a dot-bomb strategy...
on
HP Buys Compaq
·
· Score: 1
Wow. How much did AMD pay you for that post? Or are you just trying to boost the price of your stock portfolio?
I think the only real interesting 'battle' is Unix. Both PA-RISC and Alpha are already dead (in favor of Itanium), and the others are just boring PC brands.
The other interesting thing is VMS and Tandem. These are going to be ported to Itanium, as was announced a couple of months ago, but how exactly will they fit into HP's overall strategy?
What does HP want from Compaq? In the past few weeks they were talking about exiting the PC business. Will the consolidated company be a PC powerhouse or a server powerhouse, or both?
Re:Very few mergers succeed. Combine two weaklings
on
HP Buys Compaq
·
· Score: 3, Interesting
I agree. It seems like HP is going to become the Computer Associates on the hardware side, and just buy up all of these failing companies with proprietary projects, and milk them to the death. Merging two companies of this tremendous size seems a recipe for disaster.
I've never been impressed with HP's products (other than printers, which are the best), particularly their servers or workstations.
That's too bad. HP's PA-RISC line has always been absolutely top of the line and has always held its own against Alpha. For whatever reason, HP has never been viewed as the 'hip' comany that DEC was, and seemed more stodgy and conservative (almost IBMish).
Re:Hate to say, sounds like a dot-bomb strategy...
on
HP Buys Compaq
·
· Score: 2
I agree that this deal seems really shady for HP. Compaq has a huge product line (VMS, Tru64, Tandem, and PC stuff) and HP has its own huge product line (they still have their proprietary HP9000 stuff, right? plus HPUX and PC stuff).
In an age where the common strategy is to streamline business lines, obviously HP is taking a different approach. I really have to wonder exactly what they're thinking here. It seems that Compaq's strategy with DEC was to transition customers on the proprietary platforms to Windows NT but that didn't work. HP must have something better up its sleeve.
Re:Implications for alpha?
on
HP Buys Compaq
·
· Score: 2
Actually, this transaction completely explains the fate of Alpha: HP did not want to buy a company which had a directly competing product (Alpha vs. Itanium), and which did not even use its (HP's) parts in their products. So, HP convinces Compaq to give up Alpha to Intel, and switch to using HP processors.
Although there was a lot of confusion about Compaq's decision to drop Alpha initially, it is now crystal clear.
Re:How will they market that?
on
Itanium Update
·
· Score: 2
Remember that Itanium is marketed solely towards IT people, who know that gigahertz does not equal performance, and who do real performance studies before deploying a system. Look at the success of HP and Compaq whose chips are reasonably fast, yet have slow megahertz ratings (or Sun and IBM, whose chips are slower, and have low megahertz ratings, but sell very well).
It is only the consumer market which looks at gigahertz. Which means that Intel will have to make a high megahertz version if it expects Itanium to enter the consumer market.
Re:Pentium 4 Multithreading?
on
Itanium Update
·
· Score: 2
Not really. On the 486 the FP unit was a discrete part of the chip, which could potentially have a defect, and thus be disabled. It was done only to increase yield. MT on P4 is spread completely through the chip, and it is unlikely that a defect would prevent MT from working but let the chip run in single-thread mode (since almost all parts of the chip are shared in MT). The reason MT is not enabled is not a manufacturing issue (like on 486SX) but mostly for paranoia about pioneering a totally new feature.
Re:328 registers???
on
Itanium Update
·
· Score: 3, Interesting
328 *physical* registers, not logical (ISA accessible). with 128 context switches will hurt big time ia64. yet another bad design decision of the itanic.
A context switch happens one in a blue moon. Fast context switches are not going to make up for sluggish performance for the real work the machine is doing between context switchs. Registers are considerably faster than cache; the absolutely fastest cache in the world is P4's L1 cache which has a load latency of 2 cycles, and on most architectures it is 3 cycles. Putting 128 qwords into registers is an absolutely dramatic speedup for programs which have a working set more than 8 dwords (all that IA-32 gives you).
Intel's upcoming Tulloch chipset support quad channel RDRAM (droool....). RDRAM is much more scalable than DDR due to lower pin count. Dual channel RDRAM has fewer pins than single channel DDR, which makes motherboard design much cheaper.
There's a fatal flaw in your assessment; Intel has licensed the P4 bus to both ALi and SiS and both are releasing DDR chipsets within the next month. The problem is not that Intel doesn't want others producing chipsets, it's that it doesn't want people producing chipsets without a bus license. There's really nothing more to this than what the headline says.
Mainly for cost. DDR is cheaper than RDRAM, and Intel wants to push the P4 into sub-$800 systems. It can achieve this with SDR (which is much lower performance than DDR) but not with RDRAM (at least not with today's prices). Hopefully, Rambus will go out of business so people can build RDRAM without having to pay royalty. RDRAM is much more scalable than DDR, and should theoretically be heaper (due to muh lower pin count). Note that Intel will continue to develop RDRAM chipsets for customers who need optimal performance.
Uh, have you actually looked at the Anandtech article about the Via P4/DDR chipset? Intel's i850 (RDRAM) chipset outpferoms Via's DDR chipset on every benchmark.
Intel uses it to sell it's Xeon chips to businesses at much lower clock rates and higher prices than the P4; Intel uses it to explain why Itanium runs at 800MHz; AMD's new chip runs at 1.5GHz, but they say it outperforms a 2GHz P4; Alphas run at 1GHz but are acknowledged to be much faster than a P4; Sparcs run at 900MHz, yet are also acknowledged to be better performers than a P4.
There seems to be some confusion. SPARC, Athlon, Alpha, and Itanium are not faster performers than P4 (except the Itanium which beats P4 at FP).
The best selling 32-bit microprocessor family, at least as of 1999, was ARM. Paul DeMone has a solid article with plenty of historical and technical detail.
Of course, since only a tiny percentage of ARM processors power personal computers, it is irrelevant to a discussion on personal computers.
How many chips does the average computer have now, about four? They are all documented but it's less interesting than the original PC since fewer chips means you won't be able to see how it works and how the different functionality connects together. Plus, publishing schematics would add cost to building the system, and frankly very few people want it.
The PC only barely achieved parity with contemporary systems, and PC architecture systems lagged behind contemporaries for over a decade. It wasn't until he mid-ninties, with PCI and 32 bit processors that PC hardware caught up with where the like of the Macintosh and Amiga, never mind real workstations, had been years before.
First of all, when the IBM PC debuted it was the technologically the best personal computer available, if only because it was 16 bit and could address 640k, while the others (Apple II, TRS-80, Kaypro, Commodore 64, etc.) were only 8 bit and could only address 64k. This enabled much serious business applications to be written for the IBM PC than the other computers.
Workstations started to become available by around then, but were much more expensive. It wasn't until several years after it debuted, that it was surpassed in some aspects by Macintosh and Amiga.
Although it's true that PC's took a while to catch up (though I'd place the parity date closer to the early-90's than the mid-90's), for the past several years they have even outpaced the high-end workstations in terms of features and performance.
The first personal computer was probably DEC's PDP-8/m (started shipping in 1972) which pre-dated the Altair and Apple by several years.
That said, 'PC' as understood today means 'IBM PC compatible' (as opposed to Apples or workstations), and today's PC's are direct descendants of the original IBM PC 5150. The PC is by far the most widely used and most important architecture in use today. The 5150 was not the first personal computer, but was the first PC.
Are you new to computers? The memory you get from malloc isn't actually allocated until you access it. There is no translation for the pages at first, so the first access gets a page fault at which point the OS maps a PTE to it. And, no, you could not achieve that with your first 5GB hard drive because most likely you were running IA-32 which supports only 32 bit linear addresses. On Linux a process has a maximum of 3GB of memory so that's the maximum you would be able to allocate (subtracting the tiny overhead from malloc, as well as the other code and data the process has).
Clearly this is false since both VIA and SiS are coming out with P4/DDR chipsets this month (and Intel with a P4/SDR one). The only thing delaying the DDR version of Brookdale is validating it (this is Intel's first DDR chipset while VIA and SiS have more experience with it).
The high end always has a lower price/performance ratio than the low end. Workstations cost 5x-10x more than PC's, but only perform marginally better. But it's often worth it.
Case in point: I run simulations of microprocessors for a living. For this application, we had about a 3x speedup moving from P3/SDRAM to P4/RDRAM, since the application's bottleneck is memory bandwith. A typical job moved from 10 hours to 3 hours, and the cost of the speedup ($52 for a 256MB system) is absolutely peanuts compared to the productivity increase.
Patterson & Hennesey's Computer Architecture: A Quantitative Approach is absolutely indispendable.
Intel's three volume X86 manual set is completely required for anybody doing development on X86 systems.
The Alpha ARM is good reading. So is the VAX manual.
Something like the DECsystem-10 System Description makes good midnight reading for a totally different perspective (my favorite current computer book), if not something remotely applicable to today's systems.
Slashdot Mirror
Both AMD and Intel were formed out of Fairchild. Jerry Sanders was a salesman at Fairchild, and left Fairchild to start AMD. Robert Noyce and Gordon Moore were the top engineers at Fairchild (Noyce built the first silicon integrated circuit there, and Moore was one of the most prominent engineers there), and they went on to form Intel.
The tradition continues today: AMD lives mostly off of sleazy marketing (PR ratings, anyone?) which Intel has made, and continues to make, very significant innovations (first semiconductor RAM, first microprocessor, first OOO processor, first MT processor, AGP, PCI, USB) while I can't think of a significant 'first' or industry standard which AMD invented.
Compaq ate Digital, sold the StrongARM to Intel who buried it b/c it was an order of magnitude faster than Intel's low-power chips.
Huh? The guys upstairs in WCCG doing StrongARM and XScale (StrongARM, renamed) would be very interested in knowing that Intel buried their product. The fact is, StrongARM is generally acknowledged as one of Intel's key acquisitions in the last few years, and has a highly bright future ahead of it (at some point, it is likely to replace DragonBall in the Palm). It's a heck of a lot more successful than when DEC owned it, that's for sure.
FYI, the entire original StrongARM team walked out as soon as they were acquired by Intel. That's their fault, not Intel's. The Alpha team seems to have been a lot more cooperative (a whole bunch of them were just named Intel Fellows last week).
Wow. How much did AMD pay you for that post? Or are you just trying to boost the price of your stock portfolio?
I think the only real interesting 'battle' is Unix. Both PA-RISC and Alpha are already dead (in favor of Itanium), and the others are just boring PC brands.
The other interesting thing is VMS and Tandem. These are going to be ported to Itanium, as was announced a couple of months ago, but how exactly will they fit into HP's overall strategy?
What does HP want from Compaq? In the past few weeks they were talking about exiting the PC business. Will the consolidated company be a PC powerhouse or a server powerhouse, or both?
I agree. It seems like HP is going to become the Computer Associates on the hardware side, and just buy up all of these failing companies with proprietary projects, and milk them to the death. Merging two companies of this tremendous size seems a recipe for disaster.
I've never been impressed with HP's products (other than printers, which are the best), particularly their servers or workstations.
That's too bad. HP's PA-RISC line has always been absolutely top of the line and has always held its own against Alpha. For whatever reason, HP has never been viewed as the 'hip' comany that DEC was, and seemed more stodgy and conservative (almost IBMish).
I agree that this deal seems really shady for HP. Compaq has a huge product line (VMS, Tru64, Tandem, and PC stuff) and HP has its own huge product line (they still have their proprietary HP9000 stuff, right? plus HPUX and PC stuff).
In an age where the common strategy is to streamline business lines, obviously HP is taking a different approach. I really have to wonder exactly what they're thinking here. It seems that Compaq's strategy with DEC was to transition customers on the proprietary platforms to Windows NT but that didn't work. HP must have something better up its sleeve.
Actually, this transaction completely explains the fate of Alpha: HP did not want to buy a company which had a directly competing product (Alpha vs. Itanium), and which did not even use its (HP's) parts in their products. So, HP convinces Compaq to give up Alpha to Intel, and switch to using HP processors.
Although there was a lot of confusion about Compaq's decision to drop Alpha initially, it is now crystal clear.
Remember that Itanium is marketed solely towards IT people, who know that gigahertz does not equal performance, and who do real performance studies before deploying a system. Look at the success of HP and Compaq whose chips are reasonably fast, yet have slow megahertz ratings (or Sun and IBM, whose chips are slower, and have low megahertz ratings, but sell very well).
It is only the consumer market which looks at gigahertz. Which means that Intel will have to make a high megahertz version if it expects Itanium to enter the consumer market.
Not really. On the 486 the FP unit was a discrete part of the chip, which could potentially have a defect, and thus be disabled. It was done only to increase yield. MT on P4 is spread completely through the chip, and it is unlikely that a defect would prevent MT from working but let the chip run in single-thread mode (since almost all parts of the chip are shared in MT). The reason MT is not enabled is not a manufacturing issue (like on 486SX) but mostly for paranoia about pioneering a totally new feature.
328 *physical* registers, not logical (ISA accessible). with 128 context switches will hurt big time ia64. yet another bad design decision of the itanic.
A context switch happens one in a blue moon. Fast context switches are not going to make up for sluggish performance for the real work the machine is doing between context switchs. Registers are considerably faster than cache; the absolutely fastest cache in the world is P4's L1 cache which has a load latency of 2 cycles, and on most architectures it is 3 cycles. Putting 128 qwords into registers is an absolutely dramatic speedup for programs which have a working set more than 8 dwords (all that IA-32 gives you).
One of the article's points, one that I strongly agree with, is how overpowered the machines are that most people buy.
Maybe if LinuxWorld got some decent powered machines, they wouldn't be Slashdotted already.
Intel's upcoming Tulloch chipset support quad channel RDRAM (droool....). RDRAM is much more scalable than DDR due to lower pin count. Dual channel RDRAM has fewer pins than single channel DDR, which makes motherboard design much cheaper.
There's a fatal flaw in your assessment; Intel has licensed the P4 bus to both ALi and SiS and both are releasing DDR chipsets within the next month. The problem is not that Intel doesn't want others producing chipsets, it's that it doesn't want people producing chipsets without a bus license. There's really nothing more to this than what the headline says.
Mainly for cost. DDR is cheaper than RDRAM, and Intel wants to push the P4 into sub-$800 systems. It can achieve this with SDR (which is much lower performance than DDR) but not with RDRAM (at least not with today's prices). Hopefully, Rambus will go out of business so people can build RDRAM without having to pay royalty. RDRAM is much more scalable than DDR, and should theoretically be heaper (due to muh lower pin count). Note that Intel will continue to develop RDRAM chipsets for customers who need optimal performance.
Because they can't outperform Via
Uh, have you actually looked at the Anandtech article about the Via P4/DDR chipset? Intel's i850 (RDRAM) chipset outpferoms Via's DDR chipset on every benchmark.
Intel uses it to sell it's Xeon chips to businesses at much lower clock rates and higher prices than the P4; Intel uses it to explain why Itanium runs at 800MHz; AMD's new chip runs at 1.5GHz, but they say it outperforms a 2GHz P4; Alphas run at 1GHz but are acknowledged to be much faster than a P4; Sparcs run at 900MHz, yet are also acknowledged to be better performers than a P4.
There seems to be some confusion. SPARC, Athlon, Alpha, and Itanium are not faster performers than P4 (except the Itanium which beats P4 at FP).
Let's have a look:
P4/1.8GHz: SPECint - 574, SPECfp - 618
Athlon/1.4GHz: SPECint - 495, SPECfp - 426
Alpha/1001MHz: SPECint - 561, SPECfp - 585
SPARC/900MHz: SPECint - 439, SPECfp - 439
Itanium/800MH: SPECint - 314, SPECfp - 655
The best selling 32-bit microprocessor family, at least as of 1999, was ARM. Paul DeMone has a solid article with plenty of historical and technical detail.
Of course, since only a tiny percentage of ARM processors power personal computers, it is irrelevant to a discussion on personal computers.
How many chips does the average computer have now, about four? They are all documented but it's less interesting than the original PC since fewer chips means you won't be able to see how it works and how the different functionality connects together. Plus, publishing schematics would add cost to building the system, and frankly very few people want it.
The PC only barely achieved parity with contemporary systems, and PC architecture systems lagged behind contemporaries for over a decade. It wasn't until he mid-ninties, with PCI and 32 bit processors that PC hardware caught up with where the like of the Macintosh and Amiga, never mind real workstations, had been years before. First of all, when the IBM PC debuted it was the technologically the best personal computer available, if only because it was 16 bit and could address 640k, while the others (Apple II, TRS-80, Kaypro, Commodore 64, etc.) were only 8 bit and could only address 64k. This enabled much serious business applications to be written for the IBM PC than the other computers. Workstations started to become available by around then, but were much more expensive. It wasn't until several years after it debuted, that it was surpassed in some aspects by Macintosh and Amiga. Although it's true that PC's took a while to catch up (though I'd place the parity date closer to the early-90's than the mid-90's), for the past several years they have even outpaced the high-end workstations in terms of features and performance.
The first personal computer was probably DEC's PDP-8/m (started shipping in 1972) which pre-dated the Altair and Apple by several years.
That said, 'PC' as understood today means 'IBM PC compatible' (as opposed to Apples or workstations), and today's PC's are direct descendants of the original IBM PC 5150. The PC is by far the most widely used and most important architecture in use today. The 5150 was not the first personal computer, but was the first PC.
Are you new to computers? The memory you get from malloc isn't actually allocated until you access it. There is no translation for the pages at first, so the first access gets a page fault at which point the OS maps a PTE to it. And, no, you could not achieve that with your first 5GB hard drive because most likely you were running IA-32 which supports only 32 bit linear addresses. On Linux a process has a maximum of 3GB of memory so that's the maximum you would be able to allocate (subtracting the tiny overhead from malloc, as well as the other code and data the process has).
Clearly this is false since both VIA and SiS are coming out with P4/DDR chipsets this month (and Intel with a P4/SDR one). The only thing delaying the DDR version of Brookdale is validating it (this is Intel's first DDR chipset while VIA and SiS have more experience with it).
The high end always has a lower price/performance ratio than the low end. Workstations cost 5x-10x more than PC's, but only perform marginally better. But it's often worth it.
Case in point: I run simulations of microprocessors for a living. For this application, we had about a 3x speedup moving from P3/SDRAM to P4/RDRAM, since the application's bottleneck is memory bandwith. A typical job moved from 10 hours to 3 hours, and the cost of the speedup ($52 for a 256MB system) is absolutely peanuts compared to the productivity increase.
Hmm, seems a little software centric to me.
Patterson & Hennesey's Computer Architecture: A Quantitative Approach is absolutely indispendable.
Intel's three volume X86 manual set is completely required for anybody doing development on X86 systems.
The Alpha ARM is good reading. So is the VAX manual.
Something like the DECsystem-10 System Description makes good midnight reading for a totally different perspective (my favorite current computer book), if not something remotely applicable to today's systems.