Of ccourse you can run multiple operating systems simultaneously, though VMWare is kind of the amaturish application for it. IBM's Power systems and Sun's UltraSparc's can do this now, as can HP's Superdome. Intel is just looking to add a hardware acceleration effort to the Itanium so that systems like the Superdome (which now uses Itaniums) can do this a little bit better.
Running multiple partitions has become a rather important feature for high-end systems, and how well a system handles this is definitely a selling feature. Since Intel and HP want the Superdome to continue to compete well in the top-end of servers (and presumably Intel also wants other Itanium customers to follow suit), adding improving performance of multiple partitions is a good idea.
Of course, how this compares to what IBM and Sun have up their sleeves is another question altogether.
As for VMWare, as I mentioned, it's the kind of amature and low-end way of doing things, mainly on low-end servers and desktops using x86 processors. While Intel still hopes that IA-64 will eventually filter down to the desktop and maybe this new 'vanderpool' technology will come with it at some point in time, that's a long ways in the future.
Worse yet. The article doesn't specify this, but I'd be 99% certain that this "Vanderpool" stuff is being developped for the Itanium line of processors, not for x86.
I don't think that even Dvorak is calling for a super-secret IA-64 version of OS X!
They haven't benchmarked that probably because no one seems to have bothered anything on Macs that would make 8GB of memory worthwhile. Almost all of the tests I've seen have just been Photoshop and iMovie. If you're lucky they throw in some other media encoding tests, but that's it.
If you want to show off 8GB of memory, you need either some high-end workstation applications or some server applications. The G5 should be able to run some of this stuff, though I don't know how widely available the software is. Once Linux is somewhat more functional on the G5, we might see some more comparisons.
Also, regardless of how much of a performance hit PAE casues on the Xeon, it's just plain old UGLY and should have never been born in the first place. It's a nasty kludge that exists solely because of missing capabilities (ie the lack of true 64-bit capabilities for server tasks). Fortunately the real solution for x86 was released 5+ months ago by AMD, with their Opteron and now Athlon64 processors. Any comparison of dual G5s vs. dual Xeons should throw a dual Opteron system into the mix as well. Install Linux on all three and go to town.
Saying that the P4 was slower than the P3 is highly misleading, and in many cases just downright wrong. The PIII only ever managed to clock to 1.13GHz on a 180nm fab process, and it took Intel a LONG time and one recall before they made it that far. The P4 on a 180nm fab process clocked to 2.0GHz. A 2GHz P4 beat a 1.13GHz PIII on basically everything you could throw at it.
The PIII 'Tualatin', a PIII that was produced on a 130nm fab process, came out later than the P4, and even than it had a heck of a time competing in most tests. Sure, the bottom end P4 was a little slower than the top-end PIII, but that bottom end P4 was both cheaper and available earlier than the top-end P3.
Uhh, that could be because the Mac isn't faster, at least not if you live outside of the Steve Jobs Reality Distortion Field.
Don't get me wrong, it's a good chip, and it's plenty fast, but it's the blazing speed-demon that Apple makes it out to be. At best it's about on-par with the top-end P4 systems. A dual Mac G5 runs more or less neck and neck with a dual Xeon 3.06GHz for the most part. Some things are quicker on one system (sometimes by significant margins), some are quicker on others, but overall they're about the same.
Of course, it's slightly tough to document this since Apple hardly ever publishes any industry standard benchmarks. They put out some SPEC scores earlier, but they were pretty sad, easily beaten by low-end PC processors.
Note though: "about the same" is a VERY good thing for Apple! This is a HUGE improvement over the G4, which couldn't even keep up with the low-end Celeron processors in bargin-basement PCs.
GameCube is a 32-bit game console. It uses a custom 32-bit PowerPC chip from IBM. Sony's Playstation2 uses a MIPS32 core (obviously 32-bit) and the XBox uses an Intel Celeron-ish processor, again 32-bit.
The reason why people say that these consoles are 64 or 128-bit is because of the video subsystem. Similarly, all of the top-end video cards for PCs are 128 or 256-bit cards.
Just as a FWIW, the Athlon64 and Athlon64 FX are not first generation chips. They are currently on Revision C. Revision A (actually there were a few A versions, at least A0 and A1, possibly A2 as well) was the test silicon that was never meant for market. Revision B is what the first Opterons shipped as. The Athlon64 and Athlon64 FX are revision C of the exact same die.
Right now AMD is using the exact same silicon for all three of these processors. This is good in tha they've already had quite some time to get the bugs worked out (A0 silicon was out two years ago), though there's a slight downside in that the die is rather large. Even the regular Athlon64 has a 128-bit wide memory controller and 3 Hypertransport links on-die, it's just that half of the memory controller and two of the HT links are disabled.
All that being said though, the CPU is damn near the bottom of the list of components that cause system instability, regardless of what revision it is or who makes it. Things like RAM and the power supply are the big issues on the hardware front, though drivers are were things really cause problems. Here Intel and AMD are on equal footing except for motherboard drivers. Unfortunately no one has all that great of a reputation for motherboard drivers. Traditionally Intel has been seen as the best, but it was more of a "least bad" if you asked me. VIA had all kinds of troubles and ALi was worse. But things are a bit different now. First, Intel's current chipsets have very mature and stable drivers, and second, nVidia enter the market by making some chipsets for AMD processors, and they have totally raised the bar for everyone else, including Intel. Back in the days of previous crappy Intel drivers (ie the early i8xx drivers or the piss-poor first few releases of the PIIX4 drivers) the competition was mainly VIA and SiS, producing crappy drivers of their own. Intel's current drivers may be stable, but if they fall back on some of their old instability problems then nVidia (and by connection, AMD) could eat their lunch.
Re:Threat to Athlon64: Prescott (not Pentium 4)
on
Athlon 64 Debuts
·
· Score: 1
The Prescott's "64-bit instructions" are not AMD64 (aka x86-64), nor are they "Intel-64", nor are they likely to be even 64-bit instructions. What they are, is a rumor based on speculation based on the fact that no one is quite sure how Intel is planning on using a second ALU that is missing an AGU.
Not only is this a whole lot of unsubstatiated rumors, it is also based on not a whole heck of a lot. What's more, if this second ALU is designed to allow 64-bit operation, it's a REALLY odd-ball way to go about things. It also wouldn't offer any advantage over 32-bit systems unless there are a lot of chances to the existing ALU, and if you're making significant changes to the existing ALU to make it support 64-bit operations, why not just make it a 64-bit ALU?
Long story short, don't hold your breath for 64-bit support on the Prescott, or even on any near-future Xeons based on the Prescott core. On the flip side, this new core does fix some of the short-comings of the original Pentium4 core, and it should offer decent performance (at least on-par with the P4EE IMO).
Rogers service seems to be very up and down depending on the city you live in. I had Rogers cable internet for two years while living in Guelph (roughly halfway between Toronto and London mentioned in the parent message for those who don't know Canadian geography), and I can probably count on one hand the number of times the service went out. Speed was quite decent two, though it gradually got worse as time went on and more customers started going to broadband. I don't think this had anything to do with the cable being shared, much more to do with the upstream pipe being shared.
I also spent about 4 months with Sympatico DSL service, and it was no where near as good. Of course, it ended up being that half of my problems were caused by the DSL modem overheating (something that has occured multiple times with several different units in different cities). Hint: anyone who has an Efficient Networks SpeedStream 5360 DSL modem (default modem that Bell Sympatico uses these days), put it on it's side! The modem tends to overhead when it's sitting flat, but it also comes with little feet on it's side to stand upright, and it usually doesn't overheat like that. Since turning my DSL modem on it's side, my service has improved ten-fold.
Note that Sympatico just recently (June or July) REMOVED their service caps. I guess after a 1-year trial, they decided that it wasn't worth it.
Similarly Rogers cable was just about to put a usage cap in place, but decided against it at about the same time that Sympatico removed their cap. In short, been there, done that, and apparently the caps don't work very well.
Hotmail gets a bad reputation because it is attacked FAR more than any other mail server out there, with the possible exception of AOL. The problems with Hotmail are two-fold:
1. There are so many users of hotmail that you can easily end up with a previously used address (so even if you never give out your e-mail address, the previous owner of that address may have signed up to all sorts of crap). What's more, anytime someone puts out their hotmail address with a minor typo (either intentionally or accidentaly), it is usually a real address belonging to someone else.
2. Hotmail is CONSTANTLY being dictionary-probed by spammers. They have been subjected to this sort of dictionary-probe attack for over a year now. This is especially a problem for people with short (6 characters or less) usernames. If you have a username that is in any way related to a word or name and is fairly short, you will be probed.
Another major problem with Hotmail is that until recently it always opened all remote "images" by default. Almost all spam now comes with a "tracking image", which is just an HTML "IMG" url that points to a script to record your e-mail address. End result, if you open the message, the spammers know they have a live address even if you don't click on anything. Hotmail now has the option to disable remote image loading, though I don't know if it's turned on by default or not.
Retail cost for 1100 dual G5's (with 4GB of memory) is $5.85 million alone. Discount that by 11% and you get $5.2 million for the G5's all on their own. Then there's the inifiniband cards, which will run you easily $1000 a port, so that's another $1.1 million or more. Add in all the extras for cabling, the racks, the fans and ducting systems, etc. etc., and you'll quickly see that having a complete system cost of only $5.2 million IS heavily discounted.
As for the Deja Vu thing, I wouldn't have even mentioned it except that the last time someone pointed out this major short-comings of Apple's G5 for this application a bunch of people started screaming that this "Deja Vu" would save the day. Deja Vu is run at the software level on top of the operating system. All it can do is look at what data is there. There's absolutely NO WAY to for it to check to see if the data is correct unless they are running every single calculation twice at the same time and comparing results, and if you're doing that, you're just wasting your time and money when you could have bought a 550 node cluster with ECC memory and got the same result.
One of the really nice bonus off-shoots of the new Athlon64 and Opteron processors with their on-die memory controllers is that ALL of these systems will work with ECC memory (assuming the motherboard manufacturer's don't go way out of their way to try and screw this up).
Err, please TRY to find a 32-bit FPU in ANY processor. Trust me, it won't happen. Intel's old 8087 FPU could handle 80-bit data. This chip was used alongside an 8-bit processor.
I read the article, and it looks like it was a REAL amature decision to go with these systems.
Sure, the systems probably offered good performance for the (heavily discounted) price. However, they extremely large size of the desktop systems means that they're going to spend a lot more money on physical space to keep these systems. They also have rather high power consumption for super computer use due to all the completely useless (for a supercomputer) extras that the G5s come with (eg the high-end gaming video card). They are also likely to run into a number of problems with failing fans. Each of those G5s has 9 fans, and fans are about the least reliable part of any modern computer.
What's more, they now have 4.4 terabytes of non-ECC memory. Given the Soft Error Rate of today's memory, it's likely that they'll be getting at least one soft error every few days in this cluster. Now, maybe they'll get lucky and the error won't affect things, but they are leaving a LOT to chance here! This looks like a MAJOR oversight if you ask me. Ohh, and before someone starts talking about that "Deja Vu" technology, it does absolutely nothing to detect or correct memory soft errors.
In a lot of countries people have switched over to metric in a very half-assed manner, using metric (err, SI) units for some things and using imperial units for other things. I'm in Canada, and there are definitely a few definitely a few odd imperial measurements being thrown around.
I think the pint is pretty much a standard for beer the world over. Here you can occasionally get quart bottles of beer as well (primarily in dive bars, such as the one I frequented last night, which had fine quarts of Molson Ex. and Labatt 50:> ). Most places also tend to talk about people's height in terms of feet and inches, while weight is usually thought of in pounds here. We also have a rather odd tendency to think of water temperature (for pools, lakes, etc.) in farenheit while air temperature in Celcius.
I was recently living in Ireland, and they are similarly half-assed converted to metric. They tend to talk about people's weight in stone, and many older Irish people I met seemed to have a very tough time with temps in Celcius, still sticking to farenheit. The one thing that really cracked me up there though was that all their speed limit signs and all the speed gauges in cars were always in miles/hour, but almost all of the distance signs on highways were in kilometers!
UltraSparc is "higher percision"?! I think you've been hitting the crack-pipe a bit too much! Both chips offer 64-bit floating point registers, and with the P4 you can even use 80-bit fp values (though to the best of my knowledge, no one does). They have the exact same percision, as does every other major processor design on the plannet, just like they have for quite some time!
Now you think Intel bought the Alpha to "silence" it? Intel bought the corpse of the Alpha post-mortem. It died through YEARS of mismanagement and poor marketing at Digital. Intel gave them a helping hand by first buying the fab for the Alpha, since Digital was doing a piss-poor job of actually manufacturing the chips. They didn't get the rest of the chip design until after Digital was bought out by Compaq and than being prepped to be merged with HP. This was years after all real talent behind the Alpha had left.
Yeah, Alpha was a great processor in it's day, but it's day is LONG since gone. Ohh, and the Itanium2 beats the snot out of EV6 in every benchmark that compares the two! Look at Spec, look at TPC, look at any numbers you want, the EV6 loses them all. Even the "new" EV7 (a chip that should have arrived long ago, but just came out at the start of this year) is struggling pretty badly, mainly because after all the mismangement of Alpha lead no one to care about it anymore.
As for most old Alpha users, my guess is that most are switching to x86, and possibly thinking about x86-64. Heck, even Cray is switching to Opterons for their next big machine. When the Alpha first came out it might have been able to beat x86 solidly, but now, even with boatloads of bandwidth, the EV7 has a heck of a time beating x86.
I should probably point out that while throwing all of these bit-widths around, that's only half of the picture. The other half is, of course, the number of data transfers per second (ie the clock speed x the number of transfers per clock cycle).
A good example of this is to compare the memory bandwidth of the USIII and the P4. The USIII uses a 256-bit bus to memory, while the P4 uses only a 64-bit wide data bus (shared for memory and I/O). However, the USIII's bus runs at only 75MT/s (75MHz SDR). The P4's bus runs at 800MT/s (200MHz QDR). End result is that even though the P4's data bus is only 64-bits wide to the chipset, it offers considerably more bandwidth than the UltraSparc bus.
Similarly AMD is able to get lots of bandwidth out of their Hypertransport links, even though they max out at only 16-bits in each direction. A 32-bit HT link can manage 6.4GB/s of aggregate bandwidth (3.2GB/s in each direction) due to it's 1600MT/s (800MHz DDR) speed.
For data buses, narrow but fast point-to-point connections are taking over from wide bus slow buses. Even hard drives are moving in this direction with Serial ATA and Serial Attached SCSI.
Err, no. The internal data bus of the P4 is 256-bits wide, at least if you're talking about it's L2 cache bus. L1 cache doesn't really have a "bus", especially not the P4's trace cache (it's replacement for an L1 i-cache), but if my memory serves me correctly, the L1 d-cache of the P4 can read or write a pair of 64-byte values in 2 clock cycles. I guess that makes it's "bus" 128 bytes (not bits) wide. I don't know the bus width of this new L3 cache on this P4 "Extreme", aka a XeonMP, but I would guess it's 64-bits wide.
I haven't got a clue as to the internal data bus of the USIII, but I would guess that it's either 128-bit or 256-bit wide. Side note: the Power4 uses a MASSIVE 1024-bit wide internal bus, one of the reasons for it's impressive performance.
The only situation where the USIII has 64-bits and the P4 has 32-bits is if you are talking about integer registers or memory pointer width, neither of which are going to play a role in Spec CFP scores.
Err, Intel has got to be one of the most consistantly profitable companies of all time! There is hardly ever a QUARTER (that's three months) were they don't post $2 billion in profit!
Ohh, and almost all that profit comes from their x86 chip sales. All other divisions of the company tend to mostly just break even. Chip industry dying indeed! I think ANY company would like to die just such a death, sounds to me like they're already in corporate heaven.
I actually kind of wonder about this chip. I wouldn't really expect the extra cache to help DVD->Div-X encoding all that much, and even gaming isn't likely to see that huge of an improvement. Large cache like this tends to really benefit server applications, while desktop applications benefit most from low-latency and high bandwidth memory connections, as well as raw processing power.
It will be interesting to see if this chip really does significantly improve performance for desktop systems or not. I'm not holding my breath for any earth-shattering performance improvements on the desktop end of things, though it would probably make a hell of a chip for a single processor server.
In particular, it outperforms the UltraSPARC III even though the latter has a 2-to-1 advantage in the width of its databus -- 64 bits versus 32 bits.
Err.. The P4 has a 64-bit data bus. The UltraSparcIII has quite a different databus (due to it's integrated memory controller), but when you look at memory bandwidth, the USIII has 2.4GB/s of memory bandwidth while the P4 has 6.4GB/s.
What changed the x86 chips from also-ran losers in FP performance to the kings of the hill? SSE.
Less than 5% of SpecFP scores make use of SSE. The performance comes mainly from the P4 having a lot of memory bandwidth. The only chips with more memory bandwidth are the Alpha 21364, the Power4, the Itanium2 and the Opteron. Ohh, take a guess as to which chips get higher SpecFP scores than the Pentium4 does.
The chips were designed from the ground-up to run x86. That's all that they were ever intended to run, the whole code-morphing thing was just a publicity stunt. Sure, it might be technically possible to run other architectures, but the performance would be even worse then the (already very poor) x86 performance and besides, other than/. readers, who would care? So what if the chip can run ARM code when you can buy dirt-cheap ARM processors that are faster and consume less power. Same for MIPS or PowerPC embedded chips.
Slashdot Mirror
Of ccourse you can run multiple operating systems simultaneously, though VMWare is kind of the amaturish application for it. IBM's Power systems and Sun's UltraSparc's can do this now, as can HP's Superdome. Intel is just looking to add a hardware acceleration effort to the Itanium so that systems like the Superdome (which now uses Itaniums) can do this a little bit better.
Running multiple partitions has become a rather important feature for high-end systems, and how well a system handles this is definitely a selling feature. Since Intel and HP want the Superdome to continue to compete well in the top-end of servers (and presumably Intel also wants other Itanium customers to follow suit), adding improving performance of multiple partitions is a good idea.
Of course, how this compares to what IBM and Sun have up their sleeves is another question altogether.
As for VMWare, as I mentioned, it's the kind of amature and low-end way of doing things, mainly on low-end servers and desktops using x86 processors. While Intel still hopes that IA-64 will eventually filter down to the desktop and maybe this new 'vanderpool' technology will come with it at some point in time, that's a long ways in the future.
Worse yet. The article doesn't specify this, but I'd be 99% certain that this "Vanderpool" stuff is being developped for the Itanium line of processors, not for x86.
I don't think that even Dvorak is calling for a super-secret IA-64 version of OS X!
They haven't benchmarked that probably because no one seems to have bothered anything on Macs that would make 8GB of memory worthwhile. Almost all of the tests I've seen have just been Photoshop and iMovie. If you're lucky they throw in some other media encoding tests, but that's it.
If you want to show off 8GB of memory, you need either some high-end workstation applications or some server applications. The G5 should be able to run some of this stuff, though I don't know how widely available the software is. Once Linux is somewhat more functional on the G5, we might see some more comparisons.
Also, regardless of how much of a performance hit PAE casues on the Xeon, it's just plain old UGLY and should have never been born in the first place. It's a nasty kludge that exists solely because of missing capabilities (ie the lack of true 64-bit capabilities for server tasks). Fortunately the real solution for x86 was released 5+ months ago by AMD, with their Opteron and now Athlon64 processors. Any comparison of dual G5s vs. dual Xeons should throw a dual Opteron system into the mix as well. Install Linux on all three and go to town.
Saying that the P4 was slower than the P3 is highly misleading, and in many cases just downright wrong. The PIII only ever managed to clock to 1.13GHz on a 180nm fab process, and it took Intel a LONG time and one recall before they made it that far. The P4 on a 180nm fab process clocked to 2.0GHz. A 2GHz P4 beat a 1.13GHz PIII on basically everything you could throw at it.
The PIII 'Tualatin', a PIII that was produced on a 130nm fab process, came out later than the P4, and even than it had a heck of a time competing in most tests. Sure, the bottom end P4 was a little slower than the top-end PIII, but that bottom end P4 was both cheaper and available earlier than the top-end P3.
Uhh, that could be because the Mac isn't faster, at least not if you live outside of the Steve Jobs Reality Distortion Field.
Don't get me wrong, it's a good chip, and it's plenty fast, but it's the blazing speed-demon that Apple makes it out to be. At best it's about on-par with the top-end P4 systems. A dual Mac G5 runs more or less neck and neck with a dual Xeon 3.06GHz for the most part. Some things are quicker on one system (sometimes by significant margins), some are quicker on others, but overall they're about the same.
Of course, it's slightly tough to document this since Apple hardly ever publishes any industry standard benchmarks. They put out some SPEC scores earlier, but they were pretty sad, easily beaten by low-end PC processors.
Note though: "about the same" is a VERY good thing for Apple! This is a HUGE improvement over the G4, which couldn't even keep up with the low-end Celeron processors in bargin-basement PCs.
GameCube is a 32-bit game console. It uses a custom 32-bit PowerPC chip from IBM. Sony's Playstation2 uses a MIPS32 core (obviously 32-bit) and the XBox uses an Intel Celeron-ish processor, again 32-bit.
The reason why people say that these consoles are 64 or 128-bit is because of the video subsystem. Similarly, all of the top-end video cards for PCs are 128 or 256-bit cards.
Ahh, sweet, sweet torrent.. My DSL line is happy now that it's been maxed out at ~170KB/s :>
Just as a FWIW, the Athlon64 and Athlon64 FX are not first generation chips. They are currently on Revision C. Revision A (actually there were a few A versions, at least A0 and A1, possibly A2 as well) was the test silicon that was never meant for market. Revision B is what the first Opterons shipped as. The Athlon64 and Athlon64 FX are revision C of the exact same die.
Right now AMD is using the exact same silicon for all three of these processors. This is good in tha they've already had quite some time to get the bugs worked out (A0 silicon was out two years ago), though there's a slight downside in that the die is rather large. Even the regular Athlon64 has a 128-bit wide memory controller and 3 Hypertransport links on-die, it's just that half of the memory controller and two of the HT links are disabled.
All that being said though, the CPU is damn near the bottom of the list of components that cause system instability, regardless of what revision it is or who makes it. Things like RAM and the power supply are the big issues on the hardware front, though drivers are were things really cause problems. Here Intel and AMD are on equal footing except for motherboard drivers. Unfortunately no one has all that great of a reputation for motherboard drivers. Traditionally Intel has been seen as the best, but it was more of a "least bad" if you asked me. VIA had all kinds of troubles and ALi was worse. But things are a bit different now. First, Intel's current chipsets have very mature and stable drivers, and second, nVidia enter the market by making some chipsets for AMD processors, and they have totally raised the bar for everyone else, including Intel. Back in the days of previous crappy Intel drivers (ie the early i8xx drivers or the piss-poor first few releases of the PIIX4 drivers) the competition was mainly VIA and SiS, producing crappy drivers of their own. Intel's current drivers may be stable, but if they fall back on some of their old instability problems then nVidia (and by connection, AMD) could eat their lunch.
The Prescott's "64-bit instructions" are not AMD64 (aka x86-64), nor are they "Intel-64", nor are they likely to be even 64-bit instructions. What they are, is a rumor based on speculation based on the fact that no one is quite sure how Intel is planning on using a second ALU that is missing an AGU.
Not only is this a whole lot of unsubstatiated rumors, it is also based on not a whole heck of a lot. What's more, if this second ALU is designed to allow 64-bit operation, it's a REALLY odd-ball way to go about things. It also wouldn't offer any advantage over 32-bit systems unless there are a lot of chances to the existing ALU, and if you're making significant changes to the existing ALU to make it support 64-bit operations, why not just make it a 64-bit ALU?
Long story short, don't hold your breath for 64-bit support on the Prescott, or even on any near-future Xeons based on the Prescott core. On the flip side, this new core does fix some of the short-comings of the original Pentium4 core, and it should offer decent performance (at least on-par with the P4EE IMO).
Rogers service seems to be very up and down depending on the city you live in. I had Rogers cable internet for two years while living in Guelph (roughly halfway between Toronto and London mentioned in the parent message for those who don't know Canadian geography), and I can probably count on one hand the number of times the service went out. Speed was quite decent two, though it gradually got worse as time went on and more customers started going to broadband. I don't think this had anything to do with the cable being shared, much more to do with the upstream pipe being shared.
I also spent about 4 months with Sympatico DSL service, and it was no where near as good. Of course, it ended up being that half of my problems were caused by the DSL modem overheating (something that has occured multiple times with several different units in different cities). Hint: anyone who has an Efficient Networks SpeedStream 5360 DSL modem (default modem that Bell Sympatico uses these days), put it on it's side! The modem tends to overhead when it's sitting flat, but it also comes with little feet on it's side to stand upright, and it usually doesn't overheat like that. Since turning my DSL modem on it's side, my service has improved ten-fold.
Note that Sympatico just recently (June or July) REMOVED their service caps. I guess after a 1-year trial, they decided that it wasn't worth it.
Similarly Rogers cable was just about to put a usage cap in place, but decided against it at about the same time that Sympatico removed their cap. In short, been there, done that, and apparently the caps don't work very well.
Hotmail gets a bad reputation because it is attacked FAR more than any other mail server out there, with the possible exception of AOL. The problems with Hotmail are two-fold:
1. There are so many users of hotmail that you can easily end up with a previously used address (so even if you never give out your e-mail address, the previous owner of that address may have signed up to all sorts of crap). What's more, anytime someone puts out their hotmail address with a minor typo (either intentionally or accidentaly), it is usually a real address belonging to someone else.
2. Hotmail is CONSTANTLY being dictionary-probed by spammers. They have been subjected to this sort of dictionary-probe attack for over a year now. This is especially a problem for people with short (6 characters or less) usernames. If you have a username that is in any way related to a word or name and is fairly short, you will be probed.
Another major problem with Hotmail is that until recently it always opened all remote "images" by default. Almost all spam now comes with a "tracking image", which is just an HTML "IMG" url that points to a script to record your e-mail address. End result, if you open the message, the spammers know they have a live address even if you don't click on anything. Hotmail now has the option to disable remote image loading, though I don't know if it's turned on by default or not.
Retail cost for 1100 dual G5's (with 4GB of memory) is $5.85 million alone. Discount that by 11% and you get $5.2 million for the G5's all on their own. Then there's the inifiniband cards, which will run you easily $1000 a port, so that's another $1.1 million or more. Add in all the extras for cabling, the racks, the fans and ducting systems, etc. etc., and you'll quickly see that having a complete system cost of only $5.2 million IS heavily discounted.
As for the Deja Vu thing, I wouldn't have even mentioned it except that the last time someone pointed out this major short-comings of Apple's G5 for this application a bunch of people started screaming that this "Deja Vu" would save the day. Deja Vu is run at the software level on top of the operating system. All it can do is look at what data is there. There's absolutely NO WAY to for it to check to see if the data is correct unless they are running every single calculation twice at the same time and comparing results, and if you're doing that, you're just wasting your time and money when you could have bought a 550 node cluster with ECC memory and got the same result.
One of the really nice bonus off-shoots of the new Athlon64 and Opteron processors with their on-die memory controllers is that ALL of these systems will work with ECC memory (assuming the motherboard manufacturer's don't go way out of their way to try and screw this up).
Err, please TRY to find a 32-bit FPU in ANY processor. Trust me, it won't happen. Intel's old 8087 FPU could handle 80-bit data. This chip was used alongside an 8-bit processor.
I read the article, and it looks like it was a REAL amature decision to go with these systems.
Sure, the systems probably offered good performance for the (heavily discounted) price. However, they extremely large size of the desktop systems means that they're going to spend a lot more money on physical space to keep these systems. They also have rather high power consumption for super computer use due to all the completely useless (for a supercomputer) extras that the G5s come with (eg the high-end gaming video card). They are also likely to run into a number of problems with failing fans. Each of those G5s has 9 fans, and fans are about the least reliable part of any modern computer.
What's more, they now have 4.4 terabytes of non-ECC memory. Given the Soft Error Rate of today's memory, it's likely that they'll be getting at least one soft error every few days in this cluster. Now, maybe they'll get lucky and the error won't affect things, but they are leaving a LOT to chance here! This looks like a MAJOR oversight if you ask me. Ohh, and before someone starts talking about that "Deja Vu" technology, it does absolutely nothing to detect or correct memory soft errors.
In a lot of countries people have switched over to metric in a very half-assed manner, using metric (err, SI) units for some things and using imperial units for other things. I'm in Canada, and there are definitely a few definitely a few odd imperial measurements being thrown around.
:> ). Most places also tend to talk about people's height in terms of feet and inches, while weight is usually thought of in pounds here. We also have a rather odd tendency to think of water temperature (for pools, lakes, etc.) in farenheit while air temperature in Celcius.
I think the pint is pretty much a standard for beer the world over. Here you can occasionally get quart bottles of beer as well (primarily in dive bars, such as the one I frequented last night, which had fine quarts of Molson Ex. and Labatt 50
I was recently living in Ireland, and they are similarly half-assed converted to metric. They tend to talk about people's weight in stone, and many older Irish people I met seemed to have a very tough time with temps in Celcius, still sticking to farenheit. The one thing that really cracked me up there though was that all their speed limit signs and all the speed gauges in cars were always in miles/hour, but almost all of the distance signs on highways were in kilometers!
Please allow me to be the first to say:
What the hell are you talking about?!?!
UltraSparc is "higher percision"?! I think you've been hitting the crack-pipe a bit too much! Both chips offer 64-bit floating point registers, and with the P4 you can even use 80-bit fp values (though to the best of my knowledge, no one does). They have the exact same percision, as does every other major processor design on the plannet, just like they have for quite some time!
Now you think Intel bought the Alpha to "silence" it? Intel bought the corpse of the Alpha post-mortem. It died through YEARS of mismanagement and poor marketing at Digital. Intel gave them a helping hand by first buying the fab for the Alpha, since Digital was doing a piss-poor job of actually manufacturing the chips. They didn't get the rest of the chip design until after Digital was bought out by Compaq and than being prepped to be merged with HP. This was years after all real talent behind the Alpha had left.
Yeah, Alpha was a great processor in it's day, but it's day is LONG since gone. Ohh, and the Itanium2 beats the snot out of EV6 in every benchmark that compares the two! Look at Spec, look at TPC, look at any numbers you want, the EV6 loses them all. Even the "new" EV7 (a chip that should have arrived long ago, but just came out at the start of this year) is struggling pretty badly, mainly because after all the mismangement of Alpha lead no one to care about it anymore.
As for most old Alpha users, my guess is that most are switching to x86, and possibly thinking about x86-64. Heck, even Cray is switching to Opterons for their next big machine. When the Alpha first came out it might have been able to beat x86 solidly, but now, even with boatloads of bandwidth, the EV7 has a heck of a time beating x86.
I should probably point out that while throwing all of these bit-widths around, that's only half of the picture. The other half is, of course, the number of data transfers per second (ie the clock speed x the number of transfers per clock cycle).
A good example of this is to compare the memory bandwidth of the USIII and the P4. The USIII uses a 256-bit bus to memory, while the P4 uses only a 64-bit wide data bus (shared for memory and I/O). However, the USIII's bus runs at only 75MT/s (75MHz SDR). The P4's bus runs at 800MT/s (200MHz QDR). End result is that even though the P4's data bus is only 64-bits wide to the chipset, it offers considerably more bandwidth than the UltraSparc bus.
Similarly AMD is able to get lots of bandwidth out of their Hypertransport links, even though they max out at only 16-bits in each direction. A 32-bit HT link can manage 6.4GB/s of aggregate bandwidth (3.2GB/s in each direction) due to it's 1600MT/s (800MHz DDR) speed.
For data buses, narrow but fast point-to-point connections are taking over from wide bus slow buses. Even hard drives are moving in this direction with Serial ATA and Serial Attached SCSI.
Err, no. The internal data bus of the P4 is 256-bits wide, at least if you're talking about it's L2 cache bus. L1 cache doesn't really have a "bus", especially not the P4's trace cache (it's replacement for an L1 i-cache), but if my memory serves me correctly, the L1 d-cache of the P4 can read or write a pair of 64-byte values in 2 clock cycles. I guess that makes it's "bus" 128 bytes (not bits) wide. I don't know the bus width of this new L3 cache on this P4 "Extreme", aka a XeonMP, but I would guess it's 64-bits wide.
I haven't got a clue as to the internal data bus of the USIII, but I would guess that it's either 128-bit or 256-bit wide. Side note: the Power4 uses a MASSIVE 1024-bit wide internal bus, one of the reasons for it's impressive performance.
The only situation where the USIII has 64-bits and the P4 has 32-bits is if you are talking about integer registers or memory pointer width, neither of which are going to play a role in Spec CFP scores.
Err, Intel has got to be one of the most consistantly profitable companies of all time! There is hardly ever a QUARTER (that's three months) were they don't post $2 billion in profit!
Ohh, and almost all that profit comes from their x86 chip sales. All other divisions of the company tend to mostly just break even. Chip industry dying indeed! I think ANY company would like to die just such a death, sounds to me like they're already in corporate heaven.
Forget all of those, I'm still waiting for my VTEC P4 Type-R!
I actually kind of wonder about this chip. I wouldn't really expect the extra cache to help DVD->Div-X encoding all that much, and even gaming isn't likely to see that huge of an improvement. Large cache like this tends to really benefit server applications, while desktop applications benefit most from low-latency and high bandwidth memory connections, as well as raw processing power.
It will be interesting to see if this chip really does significantly improve performance for desktop systems or not. I'm not holding my breath for any earth-shattering performance improvements on the desktop end of things, though it would probably make a hell of a chip for a single processor server.
In particular, it outperforms the UltraSPARC III even though the latter has a 2-to-1 advantage in the width of its databus -- 64 bits versus 32 bits.
Err.. The P4 has a 64-bit data bus. The UltraSparcIII has quite a different databus (due to it's integrated memory controller), but when you look at memory bandwidth, the USIII has 2.4GB/s of memory bandwidth while the P4 has 6.4GB/s.
What changed the x86 chips from also-ran losers in FP performance to the kings of the hill? SSE.
Less than 5% of SpecFP scores make use of SSE. The performance comes mainly from the P4 having a lot of memory bandwidth. The only chips with more memory bandwidth are the Alpha 21364, the Power4, the Itanium2 and the Opteron. Ohh, take a guess as to which chips get higher SpecFP scores than the Pentium4 does.
In a word, never.
/. readers, who would care? So what if the chip can run ARM code when you can buy dirt-cheap ARM processors that are faster and consume less power. Same for MIPS or PowerPC embedded chips.
The chips were designed from the ground-up to run x86. That's all that they were ever intended to run, the whole code-morphing thing was just a publicity stunt. Sure, it might be technically possible to run other architectures, but the performance would be even worse then the (already very poor) x86 performance and besides, other than