The result is that the new CPU compares poorly to the Athlon CPU, which processes three FPU instructions per CPU cycle compared one instruction per CPU cycle on PIII's.
P4 also only does 2 FP ops/second but outperforms K7 by about 50% on SPECfp. FP performance typically has more to do with memory bandwidth than with the speed of the FP unit.
The problem is that while dual channel DDR will supply that much bandwidth, K7's FSB is quite sluggish and won't be able to take advantage of it. You will already note that currently shipping K7 systems show very minimal performance speedup with DDR over SDR (it's i820 all over again). There's no point in throwing even more bandwidth at the chip, which cannot use the bandwidth.
The most exciting DDR product is the ServerWorks GCHE chipset which has dual channel DDR supporting quad Foster processors, which can actually take advantage of the increased bandwidth.
Hmm, in real life P4/RDR systems outperform K7/DDR systems by about 3x on memory bandwidth. Since P4/RDR offers only 50% more theoretical bandwidth than K7/DDR, it appears that it is DDR that is having more problems reaching its theoretical limit.
Besides, 60% of 3.2 GB/s is still faster than 75% of 2.4 GB/s. I don't really care who is faster relative to theoretical peak, I just care who is faster. And that's P4/RDR.
Nope, you are wrong. The bandwidth of dual channel PC800 RDRAM is 3.2 GB/s (800 MHz x 16 bit x 2 channels). The bandwidth of DDR SDRAM is only 2.1 GB/s or 2.4 GB/s. The corporate push behind DDR is attempting to confuse customers by giving a name of PCxxx where xxx is the bandwidth, not the frequency.
The industry does not at all revolve around Microsoft. Every PC hardware manufacturer (such as Intel, AMD, Dell, Compaq, Gateway, and HP) are seriously counting on Windows XP for Q4 sales. All of their forward looking statements count XP as the reason why there might possibly be an economic recovery in the second half. PC sales are expected to be flat this year, but if XP is blocked, they will be severely down.
If Windows XP is blocked (and I don't see that as at all likely), there is no question that the Nasdaq will sink below 1,000.
The fact that hardware manufacturers are counting on Microsoft to deliver XP is not Microsoft's fault, it's the hardware maker's fault for depending on Microsoft. If they could build their own killer app's, they wouldn't care if XP was blocked. But, of course, that would set the industry back 20 years when every manufacturer had their own closed and proprietary software for their computers.
For thousands of years, music existed without profit. It wasn't until technology screwed it up by introducing distribution media (radio, records, etc) that people decided they could be rich by playing music. Now technology is allowing us to put it all back.
How has recording technology screwed up music? There will be more recordings made in Japan of Beethoven's Fifth this year than there were worldwide performances of it during Beethoven's lifetime. In the last ten years, there has been, by far, more music produced (and, moreover, more music available) than there ever has been before. Before recordings, only the rich elite could attend music performances. At the advent of recording, only a few performers were able to record, and there was little choice of recordings available music. With the advent of the LP, much more music was recorded, and great performers have been forever immortalized. With the CD the selection has compeletely skyrocketed. All music ever recorded is simultaneously and instanteously available. How is this bad?
I challenge you to name an important music performer before the advent of recording, or an important composer before the advent of writing. Practically every important composer, in fact, came after the advent of the printing press, the first technology which enabled mass reproduction of music.
I think it has more to do with the fact that it's mutable and ephemeral. If you buy a paper book, then you know that it'll pretty much always be there unless you lose it or it decomposes, neither of which can be the fault of the entity that sold it to you. You have no such guarantee with online materials. It might go away with no notice whatsoever. In addition, using a book requires no preparation or additional equipment.
Electricity is much more mutable and ephemeral than any information you can get on the internet, yet people have no problem paying for it. Electricity is just as intangible as data. Same goes for gasoline, food, water, and various other consumables.
And the ! GHz P3 was launched weeks before it was available. So where can I buy the P4 1.8 GHz?
Go to any major OEM website such as Gateway or Dell. They are all shipping 1.8 GHz Pentium 4 systems today, and are advertising them on their website.
(As an aside: it's really funny how some/.-er's get really confused and don't understand why some processors are not listed on pricewatch and then accuse them as not being available).
The P4's long pipeline means that bubbles take a long time to propagate off the CPU, so a single bubble results in a lot of wasted cycles. When the G4e's shorter pipeline has a bubble, it propagates through to the other end quickly so that fewer cycles are wasted. So a single bubble in the P4's 20 stage pipeline wastes at least 20 clock cycles (more if it's a bubble in one of the longer FPU pipelines), whereas a single bubble in the G4e's 7 stage pipeline wastes at least 7 clock cycles. 20 clock cycles is a lot of wasted work, and even if the P4's clock is running twice as fast as the G4e's it still takes a larger performance hit for each pipeline bubble than the shorter machine.
What the author apparently fails to grasp is the only thing which matters is wall clock time. P4 may have a 20 cycle mispredict penalty, higher than G4e's penalty of 7, but it also at about triple the clock speed. 20 cycles @ 1.8 GHz is less than 7 cycles @ 600 MHz.
This is basically another very pedestrian hate-on-P4 article with very little substance. P4 does have some performance problems (mostly to do with shifts and multiplies) and they're documented in the optimization manual, but this article does nothing to dig any deeper than what a dozen other pedestrian articles have said.
Also...
Intel was definitely paying attention, and as the Willamette team labored away in Santa Clara they kept MHz foremost in their minds.
Willamette was designed entirely in Oregon. Santa Clara had nothing to do with it, and has had nothing to do with IA-32 design since P5 (nearly 10 years ago).
The EU decided that their antitrust laws apply to US companies, since GE and Honeywell do business in Europe, even though th companies are US based, and the US approved the merger (not to mention other companies the EU is after such as Intel and Microsoft). So if the off-shore web-sites do business in the US (i.e. if they have customers in the US who viw the pages) they will definitely be subject to US law. What probably would happen is the web site would not allow US customers.
Underclocking. If you don't need a tip-top performance PC, reducing the CPU clock speed cuts down on heat. If you underclock far enough you don't even need separate CPU cooling at all. Same goes for graphics cards and, to a lesser extent, motherboard chipsets.
Don't do this! All modern microprocessors are designed with dynamic circuits which are very timing dependent. In general, processors cannot be underclocked reliably. The Pentium 4, for example, will not run reliably under about 1.2 GHz or will have problems with the L2 cache. It is safe if you run a particular processor at a lower speed bin, but definitely don't run it at a speed where it's not sold at. Additionally, typically not all bus ratios are validated. There might be bugs a different clock ratios, even lower ones than what is sold (changing the bus ratio radically changes timing conditions possible in the processor, and some bugs are only visible at certain ratios). Also, underclocking the bus is problematic also. High speed RAM is also timing sensitive and could cause corruption if underclocked.
In that case, Intel could insert some obfuscated code to detect AMD CPUs into its compilers' output and then run delay loops on AMD CPUs to create a phony lack of benchmark performance.
You seem to be confused. AMD has the choice of any compiler in the world to use when submitting SPEC benchmarks. They choose to use Intel's because it is the best. If Intel crippled support for AMD processors in its compiler, then AMD would use a different compiler. Of course, if AMD had compiler expertise they would develop their own compilers optimized for their chips. But they don't know how to develop compilers (and that will be quite a performance limiter for x86-64 since they will have to rely on GCC, which has terribl performance).
The uniprocessor-only P4. Intel spent years perfecting SMP on their earlier processors, and for what? So that AMD could beat them to the punch, running a 1.4Ghz CPU in SMP mode. Intel also embraced the slower-but-cheaper shared memory bus architecture, which is going to kill SMP performance in comparison.
You are wrong. The DP capable P4 (known as Xeon) was launched in May, and was launched well before the DP Athlon was released. Moreover, you can buy real dual Xeon systems from Dell, IBM, Compaq, and the like, yet you cannot buy a DP Athlon system from any major vendor, since no major OEM's want it.
You are absolutely clueless if you think that performance scales linearly with clock speed. There are so many other factors which go into performance than clock speed. Not to mention the fact that all processors are not clockable at the same rate. P4 is much more clockable than any other current processor (because of shorter pipeline stages), so why assume a five stage machine can clock as fast? What exactly are you trying to measure? All that says is that Sparc does more per clock than P4? So what? The P4 is still much faster because it clocks faster!
If there are layoffs, some of the VA guys came from other places (like Linuxcare) and may have the pain (as I did) of having been downsized in more than one Linux company.
Moral of the story: walk out of the interview if you find out that the company sells products which depend on Linux, contains Linux anywhere in its business plan, or in any way contains the word Linux in its ticker symbol.
You are definitely very confused. VA Linux was nowhere even within spitting distance of the biggest IPO ever. The five biggest IPO's ever are: (1) AT&T Wireless (just a couple of months ago), (2) UPS, (3) Conoco, (4) Goldman Sachs, and (5) Agere Systems.
Hardware is a notoriously low-margin business. Only a few big companies can hope to compete on price. Anybody else who wants people to buy their iron had better add some serious value. "We'll install Linux for you" doesn't even come close.
That's interesting. Intel made over $10 billion in profit last year (which FYI is more than any other tech company has ever made in one year), and had gross margins over 50% (which is absolutely unprecedented for a company of that size). All of the profit was from selling pure hardware, and they didn't do any "value added" crap either.
There is certainly no other pure service company with either profits or margins anywhere in that ballpark. IBM makes a lot of profits (not as much as Intel), but their margins are considerably lower (they just have a lot of revenues).
An UltraSparc III @ 900MHz that you cite above has a little over half the MHz of a P4 1.7GHz. So get UltraSparc's score, double it, and you get... 876, which is more than the shoddy 573 the P4 gets.
Are you new to computers?
What makes you think speed:clock is a meaningful benchmark?
In pure CPU performance, a 1.7 GHz P4 absolutely slaughters any Sparc. Please note the comparing kernel compiles is entirely meaningless because you are targeting two different architectures; i.e. the compiler is doing completely different work in each case.
Here's the respectve SPECint numbers (of which, GCC is a component):
P4 1.7 GHz - 573
UltraSparc II 450 MMhz - 225
UltraSparc III 750 MHz - 370
UltraSparc III 900 MHz - 438
And floating point:
P4 1.7 GHz - 598
UltraSparc II 450 MHz - 274
UltraSparc III 750 MHz - 373
UltraSparc III 900 MHz - 427
As you can see, Sun products are quite sluggish in comparison to commodity Intel products.
The most appropriate comparison to Sun is DEC. DEC had extremely strong software (VMS), and weak hardware (VAX), and very expensive machines. They were killed mostly because Sun put systems which were much cheaper and much faster (although of dubious quality, which has improved somewhat in the past 10 years). Sun now has reasonably strong software (Solaris), quite weak hardware (Sparc, although it is quite strong as far as scalability goes). I think Sun has quite a good chance to lose significant ground to Intel based systems with Linux or Windows, sold by Dell and Compaq. Intel is now competitive with Sparc (always has been at the low end, and just recently with Itanium at the high end). With Compaq's recent transformation, clearly they are trying to turn into a service company, and shed as much hardware design/manufacturing as possible. With this strategy they really have an excellent chance to reduce their costs and really beat Sun. You can bet Dell will continue to chip away at the low end markets also.
What DID happen to alpha? What would have happened if Compaq hadn't bought Digital, and there was still a company around that cared if Alpha lived or died. And what happens to Alpha now? I see no mention of it in this article.
If Compaq didn't buy DEC, then DEC would have most likely gone the way of SGI, and faded into gradual irrelevance. With the acquisition, the best DEC technology (VMS, Tru64) will be able to survive, and now is a lot more sustainable than it would have been under DEC.
Compaq is, according to the article, also commiting to the release of one more generation of Alpha processors. But, I think you can assume that will be the last. They will be porting their OS technology to Itanium.
I think you are basically right. As far as I can tell from the press releases, this is what's happening: (1) Compaq is cancelling development on EV8 and successors, (2) Compaq will complete EV7, (3) Intel will get all Alpha technology (CAD tools, chip designs, etc.), (4) Intel will offer Alpha engineers positions presumable on IA-64 development, (5) Compaq will completely migrate away from Alpha to Itanium.
The last one gives it away: if Compaq is not a customer of Alpha, who is? So, yes, Alpha is officially dead (with EV7). Intel's interest in Alpha is not in Alpha itself, but in the design technology.
Slashdot Mirror
The result is that the new CPU compares poorly to the Athlon CPU, which processes three FPU instructions per CPU cycle compared one instruction per CPU cycle on PIII's.
P4 also only does 2 FP ops/second but outperforms K7 by about 50% on SPECfp. FP performance typically has more to do with memory bandwidth than with the speed of the FP unit.
The problem is that while dual channel DDR will supply that much bandwidth, K7's FSB is quite sluggish and won't be able to take advantage of it. You will already note that currently shipping K7 systems show very minimal performance speedup with DDR over SDR (it's i820 all over again). There's no point in throwing even more bandwidth at the chip, which cannot use the bandwidth.
The most exciting DDR product is the ServerWorks GCHE chipset which has dual channel DDR supporting quad Foster processors, which can actually take advantage of the increased bandwidth.
Hmm, in real life P4/RDR systems outperform K7/DDR systems by about 3x on memory bandwidth. Since P4/RDR offers only 50% more theoretical bandwidth than K7/DDR, it appears that it is DDR that is having more problems reaching its theoretical limit.
Besides, 60% of 3.2 GB/s is still faster than 75% of 2.4 GB/s. I don't really care who is faster relative to theoretical peak, I just care who is faster. And that's P4/RDR.
Nope, you are wrong. The bandwidth of dual channel PC800 RDRAM is 3.2 GB/s (800 MHz x 16 bit x 2 channels). The bandwidth of DDR SDRAM is only 2.1 GB/s or 2.4 GB/s. The corporate push behind DDR is attempting to confuse customers by giving a name of PCxxx where xxx is the bandwidth, not the frequency.
The industry does not at all revolve around Microsoft. Every PC hardware manufacturer (such as Intel, AMD, Dell, Compaq, Gateway, and HP) are seriously counting on Windows XP for Q4 sales. All of their forward looking statements count XP as the reason why there might possibly be an economic recovery in the second half. PC sales are expected to be flat this year, but if XP is blocked, they will be severely down.
If Windows XP is blocked (and I don't see that as at all likely), there is no question that the Nasdaq will sink below 1,000.
The fact that hardware manufacturers are counting on Microsoft to deliver XP is not Microsoft's fault, it's the hardware maker's fault for depending on Microsoft. If they could build their own killer app's, they wouldn't care if XP was blocked. But, of course, that would set the industry back 20 years when every manufacturer had their own closed and proprietary software for their computers.
Great, there goes every single bit of hope that the PC industry may recover in Q4.
For thousands of years, music existed without profit. It wasn't until technology screwed it up by introducing distribution media (radio, records, etc) that people decided they could be rich by playing music. Now technology is allowing us to put it all back.
How has recording technology screwed up music? There will be more recordings made in Japan of Beethoven's Fifth this year than there were worldwide performances of it during Beethoven's lifetime. In the last ten years, there has been, by far, more music produced (and, moreover, more music available) than there ever has been before. Before recordings, only the rich elite could attend music performances. At the advent of recording, only a few performers were able to record, and there was little choice of recordings available music. With the advent of the LP, much more music was recorded, and great performers have been forever immortalized. With the CD the selection has compeletely skyrocketed. All music ever recorded is simultaneously and instanteously available. How is this bad?
I challenge you to name an important music performer before the advent of recording, or an important composer before the advent of writing. Practically every important composer, in fact, came after the advent of the printing press, the first technology which enabled mass reproduction of music.
I think it has more to do with the fact that it's mutable and ephemeral. If you buy a paper book, then you know that it'll pretty much always be there unless you lose it or it decomposes, neither of which can be the fault of the entity that sold it to you. You have no such guarantee with online materials. It might go away with no notice whatsoever. In addition, using a book requires no preparation or additional equipment.
Electricity is much more mutable and ephemeral than any information you can get on the internet, yet people have no problem paying for it. Electricity is just as intangible as data. Same goes for gasoline, food, water, and various other consumables.
And the ! GHz P3 was launched weeks before it was available. So where can I buy the P4 1.8 GHz?
/.-er's get really confused and don't understand why some processors are not listed on pricewatch and then accuse them as not being available).
Go to any major OEM website such as Gateway or Dell. They are all shipping 1.8 GHz Pentium 4 systems today, and are advertising them on their website.
(As an aside: it's really funny how some
Yeah, but the P4 is not available @ 1.8 GHz yet
Yes it is. It was launched on Monday.
And even with the faster P4 (woah, 6% higher clock), 20 cycles @ 1.8 GHz is more than 7 cycles @ 733 MHz.
The point is, a 20 cycle penalty is not three times more expensive than a 7 cycle penalty, as the article implies.
The article also fails to mention that Willamette has the most advanced BPU in the world, which minimizes the number of mispredicts greatly.
The P4's long pipeline means that bubbles take a long time to propagate off the CPU, so a single bubble results in a lot of wasted cycles. When the G4e's shorter pipeline has a bubble, it propagates through to the other end quickly so that fewer cycles are wasted. So a single bubble in the P4's 20 stage pipeline wastes at least 20 clock cycles (more if it's a bubble in one of the longer FPU pipelines), whereas a single bubble in the G4e's 7 stage pipeline wastes at least 7 clock cycles. 20 clock cycles is a lot of wasted work, and even if the P4's clock is running twice as fast as the G4e's it still takes a larger performance hit for each pipeline bubble than the shorter machine.
...
What the author apparently fails to grasp is the only thing which matters is wall clock time. P4 may have a 20 cycle mispredict penalty, higher than G4e's penalty of 7, but it also at about triple the clock speed. 20 cycles @ 1.8 GHz is less than 7 cycles @ 600 MHz.
This is basically another very pedestrian hate-on-P4 article with very little substance. P4 does have some performance problems (mostly to do with shifts and multiplies) and they're documented in the optimization manual, but this article does nothing to dig any deeper than what a dozen other pedestrian articles have said.
Also
Intel was definitely paying attention, and as the Willamette team labored away in Santa Clara they kept MHz foremost in their minds.
Willamette was designed entirely in Oregon. Santa Clara had nothing to do with it, and has had nothing to do with IA-32 design since P5 (nearly 10 years ago).
The EU decided that their antitrust laws apply to US companies, since GE and Honeywell do business in Europe, even though th companies are US based, and the US approved the merger (not to mention other companies the EU is after such as Intel and Microsoft). So if the off-shore web-sites do business in the US (i.e. if they have customers in the US who viw the pages) they will definitely be subject to US law. What probably would happen is the web site would not allow US customers.
Underclocking. If you don't need a tip-top performance PC, reducing the CPU clock speed cuts down on heat. If you underclock far enough you don't even need separate CPU cooling at all. Same goes for graphics cards and, to a lesser extent, motherboard chipsets.
Don't do this! All modern microprocessors are designed with dynamic circuits which are very timing dependent. In general, processors cannot be underclocked reliably. The Pentium 4, for example, will not run reliably under about 1.2 GHz or will have problems with the L2 cache. It is safe if you run a particular processor at a lower speed bin, but definitely don't run it at a speed where it's not sold at. Additionally, typically not all bus ratios are validated. There might be bugs a different clock ratios, even lower ones than what is sold (changing the bus ratio radically changes timing conditions possible in the processor, and some bugs are only visible at certain ratios). Also, underclocking the bus is problematic also. High speed RAM is also timing sensitive and could cause corruption if underclocked.
In that case, Intel could insert some obfuscated code to detect AMD CPUs into its compilers' output and then run delay loops on AMD CPUs to create a phony lack of benchmark performance.
You seem to be confused. AMD has the choice of any compiler in the world to use when submitting SPEC benchmarks. They choose to use Intel's because it is the best. If Intel crippled support for AMD processors in its compiler, then AMD would use a different compiler. Of course, if AMD had compiler expertise they would develop their own compilers optimized for their chips. But they don't know how to develop compilers (and that will be quite a performance limiter for x86-64 since they will have to rely on GCC, which has terribl performance).
The uniprocessor-only P4. Intel spent years perfecting SMP on their earlier processors, and for what? So that AMD could beat them to the punch, running a 1.4Ghz CPU in SMP mode. Intel also embraced the slower-but-cheaper shared memory bus architecture, which is going to kill SMP performance in comparison.
You are wrong. The DP capable P4 (known as Xeon) was launched in May, and was launched well before the DP Athlon was released. Moreover, you can buy real dual Xeon systems from Dell, IBM, Compaq, and the like, yet you cannot buy a DP Athlon system from any major vendor, since no major OEM's want it.
Does AMD have an optimizing compiler for the Athlon that you can plug into VC++? If so, it should have been included in the tests this article ran.
No they do not. AMD uses Intel compilers for their SPEC scores since it is the best X86 compiler.
You are absolutely clueless if you think that performance scales linearly with clock speed. There are so many other factors which go into performance than clock speed. Not to mention the fact that all processors are not clockable at the same rate. P4 is much more clockable than any other current processor (because of shorter pipeline stages), so why assume a five stage machine can clock as fast? What exactly are you trying to measure? All that says is that Sparc does more per clock than P4? So what? The P4 is still much faster because it clocks faster!
If there are layoffs, some of the VA guys came from other places (like Linuxcare) and may have the pain (as I did) of having been downsized in more than one Linux company.
Moral of the story: walk out of the interview if you find out that the company sells products which depend on Linux, contains Linux anywhere in its business plan, or in any way contains the word Linux in its ticker symbol.
You are definitely very confused. VA Linux was nowhere even within spitting distance of the biggest IPO ever. The five biggest IPO's ever are: (1) AT&T Wireless (just a couple of months ago), (2) UPS, (3) Conoco, (4) Goldman Sachs, and (5) Agere Systems.
Hardware is a notoriously low-margin business. Only a few big companies can hope to compete on price. Anybody else who wants people to buy their iron had better add some serious value. "We'll install Linux for you" doesn't even come close.
That's interesting. Intel made over $10 billion in profit last year (which FYI is more than any other tech company has ever made in one year), and had gross margins over 50% (which is absolutely unprecedented for a company of that size). All of the profit was from selling pure hardware, and they didn't do any "value added" crap either.
There is certainly no other pure service company with either profits or margins anywhere in that ballpark. IBM makes a lot of profits (not as much as Intel), but their margins are considerably lower (they just have a lot of revenues).
An UltraSparc III @ 900MHz that you cite above has a little over half the MHz of a P4 1.7GHz. So get UltraSparc's score, double it, and you get ... 876, which is more than the shoddy 573 the P4 gets.
Are you new to computers?
What makes you think speed:clock is a meaningful benchmark?
In pure CPU performance, a 1.7 GHz P4 absolutely slaughters any Sparc. Please note the comparing kernel compiles is entirely meaningless because you are targeting two different architectures; i.e. the compiler is doing completely different work in each case.
Here's the respectve SPECint numbers (of which, GCC is a component):
P4 1.7 GHz - 573
UltraSparc II 450 MMhz - 225
UltraSparc III 750 MHz - 370
UltraSparc III 900 MHz - 438
And floating point:
P4 1.7 GHz - 598
UltraSparc II 450 MHz - 274
UltraSparc III 750 MHz - 373
UltraSparc III 900 MHz - 427
As you can see, Sun products are quite sluggish in comparison to commodity Intel products.
The most appropriate comparison to Sun is DEC. DEC had extremely strong software (VMS), and weak hardware (VAX), and very expensive machines. They were killed mostly because Sun put systems which were much cheaper and much faster (although of dubious quality, which has improved somewhat in the past 10 years). Sun now has reasonably strong software (Solaris), quite weak hardware (Sparc, although it is quite strong as far as scalability goes). I think Sun has quite a good chance to lose significant ground to Intel based systems with Linux or Windows, sold by Dell and Compaq. Intel is now competitive with Sparc (always has been at the low end, and just recently with Itanium at the high end). With Compaq's recent transformation, clearly they are trying to turn into a service company, and shed as much hardware design/manufacturing as possible. With this strategy they really have an excellent chance to reduce their costs and really beat Sun. You can bet Dell will continue to chip away at the low end markets also.
What DID happen to alpha? What would have happened if Compaq hadn't bought Digital, and there was still a company around that cared if Alpha lived or died. And what happens to Alpha now? I see no mention of it in this article.
If Compaq didn't buy DEC, then DEC would have most likely gone the way of SGI, and faded into gradual irrelevance. With the acquisition, the best DEC technology (VMS, Tru64) will be able to survive, and now is a lot more sustainable than it would have been under DEC.
Compaq is, according to the article, also commiting to the release of one more generation of Alpha processors. But, I think you can assume that will be the last. They will be porting their OS technology to Itanium.
I think you are basically right. As far as I can tell from the press releases, this is what's happening: (1) Compaq is cancelling development on EV8 and successors, (2) Compaq will complete EV7, (3) Intel will get all Alpha technology (CAD tools, chip designs, etc.), (4) Intel will offer Alpha engineers positions presumable on IA-64 development, (5) Compaq will completely migrate away from Alpha to Itanium.
The last one gives it away: if Compaq is not a customer of Alpha, who is? So, yes, Alpha is officially dead (with EV7). Intel's interest in Alpha is not in Alpha itself, but in the design technology.