One of my cousins bought an AthlonXP 2000 with a half gig of RAM, and put Windows XP on it. My P3/650 runs faster with Windows 2000, I'm really being honest.
Let's say that you have a 70-watt chip, you throw a 160-watt Peltier on it to keep it chilly.
First, you have to power the peltier. They usually require more than 12 volts, so your PSU doesn't do a whole lot for you, you need to go out and buy a higher-voltage power supply that can pump out at least 10 amps. Then you need a relay to tie that into your PC's PSU.
Now, instead of having to remove 70 watts of heat from your PSU, you need to remove 230 watts of heat (70 CPU + 160 peltier). Water cooling is the only way to do that - and if you want to move 230 watts of heat, you're not talking about some tiny radiator and a silent fan, you need something much more substantial.
When chips drew 25 or 35 watts, using a Peltier was a lot more practical than it is now. With the hottest chips putting out up to 100 watts, it's much more simple to scale compressive cooling to those levels than peltiers.
I did, however, just read an article yesterday about a fellow using dual 160-watt peltiers to cool his chip, with another couple to cool the memory. By the time he was done, it looked like an EXTREMELY expensive investment, and with the 6 or 7 radiators, it was just a bit ugly to boot!
I used to work for a very large health care company who had been using RS6000's for quite a long time. Despite the fact that much better (and more economical) options were available, they were still pumping tens of millions of dollars into newer, faster RS6000's, and at 0.5 to 1 million dollars annually into paying people to take care of them.
When you look at the real advantages of Sun's high-end machines, IBM has some offerings, but unless I'm forgetting something, there aren't many other companies where you can turn for a machine with 100+ CPU's, the capability to partition it off into different "virtual servers", and have entire motherboards burn up without causing downtime.
Seeing that a lot of the availability features are more than you can put into just a single motherboard, it's pretty doubtful - entire mainboards can burn up inside the high-end, and it just redistributes resources as necessary. That's cool.
And, you're right on the memory bandwidth - as an example, I was thinking about a 4-way Sun a couple of years ago, but a dual Xeon had (a) more CPU power, and (b) more memory bandwidth. Their integrated memory controllers aren't exactly modern. However, when you're talking about putting over 100 (!) of the chips in the machine, in aggregate, you get some pretty decent numbers.
I don't think that Corel's problem was indecision, it was mediocrity.
Their software usually worked, and did what was advertised. It was just a pain to actually MAKE it do it. They didn't have anything truly crappy, and nothing actually good - everything they created was just incredibly mediocre. And even acquisitions, like WordPerfect, turned mediocre at their hands.
Now, I'm one of the first to say that Sun is losing a very large share of its market - the lower end. They just can't offer price-competitive counterings to Xeons and Opterons.
However, they've still got the most lucrative part of their market, the ultra-high end. With their big models starting out at about a million bucks (and that's FAR from fully equipped), they've still got plenty to keep them going.
There are still lots of apps that don't cluster well, so a room full of PC's just doesn't cut it. and there are still companies willing to shell out for the hardware they need. Sun will have to scale back on the low end, there's no doubt, but that's not a problem for them. They've always preferred to make a large profit margin on smaller volume.
There's a huge difference between tech support and rogramming/sysadmin/networking/whatever. Very few people working tech support are well-qualified to do much else, and unfortunately, that stigma will follow you as well.
Your largest hurdle is going to be to actually break out of tech support into the field you want to be in. Once you've done that, and prove that you *can* do more than just tech support, then advancements will come much more easily.
Now, hear me out. In the game, all you did was look at her boobs while you wandered around the boring game. At the movie, you just look at her boobs while she wanders around the boring movie.
That's exactly the point, if someone DOES tap into the line, it's immediately detectable. In other words, as soon as someone eavesdrops even one single bit, you can terminate the process. That's pretty cool.
I generally call it "commodity" if you can go out, buy the parts from different places, and put it together yourself. Although in the quad-cpu market you *usually* have to get the motherboard and case together, it still qualifies.
I have a quad-Xeon (P3) at the office, I bought the mobo/case from SuperMicro, and the other parts from various distributors. Sure, the price was around $15K by the time I got all of the doo-dads (10-disk RAID array...), but it was still all "commodity" hardware to me.
Now, you *might* be able to buy a Sun mobo/case by itself and do the same thing, but I had various choices as to manufacturers for the quad-Xeon. With Sun, well, you have Sun. : )
It's the same thing that's been happening for the last decade. As x86 slowly creeps in on Sun/IBM/Whatever's market, they have to come up with something "bigger".
Right now, the Opteron, with embedded memory controller and gobs of I/O, has really entered what was previously a niche market that Sun made very nice profits from.
So, now that particular cash-cow has fallen to the ravages of commodity parts, they're moving their sites even higher. Sun's never been the company to make $5 profit on each of 50 million computers, they'd much rather make $300,000 each on 1,000 computers.
I've seen various motherboards with ATI Rage chips and accompanying video memory soldered on the board. That sort of solution, though, is quite a bit more expensive than just putting the extra circuitry in the chipset. Oh well. : )
However, AFAIK, you could never use the integrated video with anything in the AGP
Ssshhhh! Don't tell my machines! ; )
Several of the NForce2 motherboards I have at the office are setup with a video card in the AGP slot, and two monitors. On the Asus boards, there were a few hoops to jump through, but on the Abits, it was a lot less hastle.
If you're looking for that kind of testing and validation, don't hold your breath.
That kind of validation doesn't generally come on the inexpensive boards, and it's been my experience that even the higher-end boards (with correspondingly higher-end price tags) still aren't too much better than the low-ends.
No difference between an Asus and an Abit motherboard?
From the numerous Abit NF7-M and Asus A7N8X-VM motherboards I've used to build all of the office machines for some time, I can tell you that there's a BIG difference between an Abit and an Asus motherboard:
The Abit works.
Now, I know, that sounds a bit cynical. And I can't say that none of the Asus boards have worked. But I *can* say that the Asus boards have been quirky, odd, and just plain wankery. The Abit boards have been solid, reliable, and terrific.
As an example, I've had to add a PCI NIC to most of the Asus boards. The onboard LAN is just too flaky. I've watched as users rebooted, only to have their onboard NIC disappear, even though still enabled in the BIOS.
I'm by no means anti-Asus. In fact, the Asus boards have some tweaks in the BIOS that I really like. But my time is valuable, and the Abit boards take a lot less of my time.
... is that they took out the integrated video. The integrated video on the NForce2 chipsets was the only integrated video that was even semi-decent. True, it doesn't keep up with a GeForceFX card, but for an integrated solution, it's awfully darn nice.
And the really nice part? When/if you DO stick an even better card in the AGP slot, you can still use the onboard for a second monitor.
For quite some time now, all of the machines I've built for our office have used NForce2 IGP chipsets, for precisely those reasons. A board that costs $100 (or less), is rock-solid, has terrific driver support, stellar performance, sound, network, etc. makes my life very, very easy. In fact, $450 will put together a VERY nice system (sans monitor) based on them.
Plus, the fact that they'll play quite a few games (Q3, WarCraft III, Counter-Strike) incredibly well makes staying late very enjoyable....
As a matter of fact, I'm going to upgrade my machine at home in the next month or two, and chances are that I'll keep using the same boards!
POWER3 Released in 1998: 15 million transistors per chip The first 64-bit symmetric multiprocessor (SMP)
Didn't several companies have 64-bit multiprocessor machines out back then? Unless I'm mistaken, Sun's Starfire was before then, having up to 64 UltraSparc II's - which, as I recall, were 64-bit chips. And that's just Sun, ignoring the other players.
So, it is just that they used "SMP", as opposed to other forms of multiprocessing, or is my memory completely skewed?
Really. Here's my example: As the systems admin, I've convinced the IT manager to let me migrate the entire company over to Linux on the desktop. Terrific, right? Well, there's one itsy-bitsy hurdle....
The expensive phone system in which they've invested a very large amount of money and time requires a client app on each workstation. And you guessed it... it only runs on windows.
I've even spoken to the company that produced the software, and offered to create and *give* them a Linux version. Nope, they can't be bothered - they're just too busy.
If you're talking about Windows 9x, then early and excessive swapping had another cause.
The memory manager for Windows 9x tries to take the same philosophy as Unix - use all available memory for disk cache, and release it as necessary. The only problem was that it would not release it willingly enough.
If you watch the memory, swap, and cache usage, you'd see that it would use so much for disk cache that swapping would occur early. Theoretically, that swapped out data could also be in the disk cache, but behavioral observation indicated that it probably wasn't.
On every win9x machine I've seen, used, fixed, or tweaked, limitting the disk cache size to a more reasonable figure GREATLY improved the performance through reducing (or eliminating) swapping.
Luckily, NT/2K/XP aren't quite as bad in that regard.
Real world experience - Rally Championship 2000 - swap file on the same drive as the game...
The last time I had a game use swap was Warcraft III on a 256-meg machine, and saw slowdowns like the one you mentioned. Rather than put swap on another drive, I stuck another DIMM in the machine, and the problem was solved. : )
If you're running Windows NT/2000/XP, make the partition in question NTFS too.
Why? NTFS is slower than FAT, offsetting the gains you might have made by your other tweaks. If you're worried about someone being able to read what's on the partition easily, well, they still can if it's NTFS.
How reliable is that 2GB drive though? I guess it doesn't really matter, cause if the drive fails, the OS should move the swapfile back to the default location.
If the drive fails, you've just experienced roughly the same thing as if you'd pulled a stick of memory out of your computer while it was running.
Slashdot Mirror
It takes a "more recent" Linux distro to support SMP? Wow. I didn't realize that 6+ years was "more recent".
steve
Why does XP do half of the stuff it does?
One of my cousins bought an AthlonXP 2000 with a half gig of RAM, and put Windows XP on it. My P3/650 runs faster with Windows 2000, I'm really being honest.
steve
Power usage is the main reason.
Let's say that you have a 70-watt chip, you throw a 160-watt Peltier on it to keep it chilly.
First, you have to power the peltier. They usually require more than 12 volts, so your PSU doesn't do a whole lot for you, you need to go out and buy a higher-voltage power supply that can pump out at least 10 amps. Then you need a relay to tie that into your PC's PSU.
Now, instead of having to remove 70 watts of heat from your PSU, you need to remove 230 watts of heat (70 CPU + 160 peltier). Water cooling is the only way to do that - and if you want to move 230 watts of heat, you're not talking about some tiny radiator and a silent fan, you need something much more substantial.
When chips drew 25 or 35 watts, using a Peltier was a lot more practical than it is now. With the hottest chips putting out up to 100 watts, it's much more simple to scale compressive cooling to those levels than peltiers.
I did, however, just read an article yesterday about a fellow using dual 160-watt peltiers to cool his chip, with another couple to cool the memory. By the time he was done, it looked like an EXTREMELY expensive investment, and with the 6 or 7 radiators, it was just a bit ugly to boot!
steve
Hey, it works for some other companies. : )
I used to work for a very large health care company who had been using RS6000's for quite a long time. Despite the fact that much better (and more economical) options were available, they were still pumping tens of millions of dollars into newer, faster RS6000's, and at 0.5 to 1 million dollars annually into paying people to take care of them.
When you look at the real advantages of Sun's high-end machines, IBM has some offerings, but unless I'm forgetting something, there aren't many other companies where you can turn for a machine with 100+ CPU's, the capability to partition it off into different "virtual servers", and have entire motherboards burn up without causing downtime.
steve
steve
Seeing that a lot of the availability features are more than you can put into just a single motherboard, it's pretty doubtful - entire mainboards can burn up inside the high-end, and it just redistributes resources as necessary. That's cool.
And, you're right on the memory bandwidth - as an example, I was thinking about a 4-way Sun a couple of years ago, but a dual Xeon had (a) more CPU power, and (b) more memory bandwidth. Their integrated memory controllers aren't exactly modern. However, when you're talking about putting over 100 (!) of the chips in the machine, in aggregate, you get some pretty decent numbers.
steve
I don't think that Corel's problem was indecision, it was mediocrity.
Their software usually worked, and did what was advertised. It was just a pain to actually MAKE it do it. They didn't have anything truly crappy, and nothing actually good - everything they created was just incredibly mediocre. And even acquisitions, like WordPerfect, turned mediocre at their hands.
steve
Now, I'm one of the first to say that Sun is losing a very large share of its market - the lower end. They just can't offer price-competitive counterings to Xeons and Opterons.
However, they've still got the most lucrative part of their market, the ultra-high end. With their big models starting out at about a million bucks (and that's FAR from fully equipped), they've still got plenty to keep them going.
There are still lots of apps that don't cluster well, so a room full of PC's just doesn't cut it. and there are still companies willing to shell out for the hardware they need. Sun will have to scale back on the low end, there's no doubt, but that's not a problem for them. They've always preferred to make a large profit margin on smaller volume.
steve
1. PuTTY.
2. Civilization 2
3. WarCraft III
4. Too busy playing Civ or WC3.
steve
There's a huge difference between tech support and rogramming/sysadmin/networking/whatever. Very few people working tech support are well-qualified to do much else, and unfortunately, that stigma will follow you as well.
Your largest hurdle is going to be to actually break out of tech support into the field you want to be in. Once you've done that, and prove that you *can* do more than just tech support, then advancements will come much more easily.
steve
Now, hear me out. In the game, all you did was look at her boobs while you wandered around the boring game. At the movie, you just look at her boobs while she wanders around the boring movie.
Sounds like a 100% adaptation, to me.
steve
That's exactly the point, if someone DOES tap into the line, it's immediately detectable. In other words, as soon as someone eavesdrops even one single bit, you can terminate the process. That's pretty cool.
steve
I generally call it "commodity" if you can go out, buy the parts from different places, and put it together yourself. Although in the quad-cpu market you *usually* have to get the motherboard and case together, it still qualifies.
I have a quad-Xeon (P3) at the office, I bought the mobo/case from SuperMicro, and the other parts from various distributors. Sure, the price was around $15K by the time I got all of the doo-dads (10-disk RAID array...), but it was still all "commodity" hardware to me.
Now, you *might* be able to buy a Sun mobo/case by itself and do the same thing, but I had various choices as to manufacturers for the quad-Xeon. With Sun, well, you have Sun. : )
steve
It's the same thing that's been happening for the last decade. As x86 slowly creeps in on Sun/IBM/Whatever's market, they have to come up with something "bigger".
Right now, the Opteron, with embedded memory controller and gobs of I/O, has really entered what was previously a niche market that Sun made very nice profits from.
So, now that particular cash-cow has fallen to the ravages of commodity parts, they're moving their sites even higher. Sun's never been the company to make $5 profit on each of 50 million computers, they'd much rather make $300,000 each on 1,000 computers.
steve
You'd need an insane amount of water to store enough energy to do anything useful with it.
steve
I've seen various motherboards with ATI Rage chips and accompanying video memory soldered on the board. That sort of solution, though, is quite a bit more expensive than just putting the extra circuitry in the chipset. Oh well. : )
steve
It's about time, seeing as how a PC inside of your Mac came out ages ago...
steve
However, AFAIK, you could never use the integrated video with anything in the AGP
Ssshhhh! Don't tell my machines! ; )
Several of the NForce2 motherboards I have at the office are setup with a video card in the AGP slot, and two monitors. On the Asus boards, there were a few hoops to jump through, but on the Abits, it was a lot less hastle.
steve
If you're looking for that kind of testing and validation, don't hold your breath.
That kind of validation doesn't generally come on the inexpensive boards, and it's been my experience that even the higher-end boards (with correspondingly higher-end price tags) still aren't too much better than the low-ends.
steve
No difference between an Asus and an Abit motherboard?
From the numerous Abit NF7-M and Asus A7N8X-VM motherboards I've used to build all of the office machines for some time, I can tell you that there's a BIG difference between an Abit and an Asus motherboard:
The Abit works.
Now, I know, that sounds a bit cynical. And I can't say that none of the Asus boards have worked. But I *can* say that the Asus boards have been quirky, odd, and just plain wankery. The Abit boards have been solid, reliable, and terrific.
As an example, I've had to add a PCI NIC to most of the Asus boards. The onboard LAN is just too flaky. I've watched as users rebooted, only to have their onboard NIC disappear, even though still enabled in the BIOS.
I'm by no means anti-Asus. In fact, the Asus boards have some tweaks in the BIOS that I really like. But my time is valuable, and the Abit boards take a lot less of my time.
steve
And the really nice part? When/if you DO stick an even better card in the AGP slot, you can still use the onboard for a second monitor.
For quite some time now, all of the machines I've built for our office have used NForce2 IGP chipsets, for precisely those reasons. A board that costs $100 (or less), is rock-solid, has terrific driver support, stellar performance, sound, network, etc. makes my life very, very easy. In fact, $450 will put together a VERY nice system (sans monitor) based on them.
Plus, the fact that they'll play quite a few games (Q3, WarCraft III, Counter-Strike) incredibly well makes staying late very enjoyable....
As a matter of fact, I'm going to upgrade my machine at home in the next month or two, and chances are that I'll keep using the same boards!
steve
Help me out here. A quote from the article is:
POWER3
Released in 1998: 15 million transistors per chip
The first 64-bit symmetric multiprocessor (SMP)
Didn't several companies have 64-bit multiprocessor machines out back then? Unless I'm mistaken, Sun's Starfire was before then, having up to 64 UltraSparc II's - which, as I recall, were 64-bit chips. And that's just Sun, ignoring the other players.
So, it is just that they used "SMP", as opposed to other forms of multiprocessing, or is my memory completely skewed?
steve
"This app only runs on windows."
Really. Here's my example: As the systems admin, I've convinced the IT manager to let me migrate the entire company over to Linux on the desktop. Terrific, right? Well, there's one itsy-bitsy hurdle....
The expensive phone system in which they've invested a very large amount of money and time requires a client app on each workstation. And you guessed it... it only runs on windows.
I've even spoken to the company that produced the software, and offered to create and *give* them a Linux version. Nope, they can't be bothered - they're just too busy.
steve
If you're talking about Windows 9x, then early and excessive swapping had another cause.
The memory manager for Windows 9x tries to take the same philosophy as Unix - use all available memory for disk cache, and release it as necessary. The only problem was that it would not release it willingly enough.
If you watch the memory, swap, and cache usage, you'd see that it would use so much for disk cache that swapping would occur early. Theoretically, that swapped out data could also be in the disk cache, but behavioral observation indicated that it probably wasn't.
On every win9x machine I've seen, used, fixed, or tweaked, limitting the disk cache size to a more reasonable figure GREATLY improved the performance through reducing (or eliminating) swapping.
Luckily, NT/2K/XP aren't quite as bad in that regard.
steve
Real world experience - Rally Championship 2000 - swap file on the same drive as the game...
The last time I had a game use swap was Warcraft III on a 256-meg machine, and saw slowdowns like the one you mentioned. Rather than put swap on another drive, I stuck another DIMM in the machine, and the problem was solved. : )
steve
If you're running Windows NT/2000/XP, make the partition in question NTFS too.
Why? NTFS is slower than FAT, offsetting the gains you might have made by your other tweaks. If you're worried about someone being able to read what's on the partition easily, well, they still can if it's NTFS.
How reliable is that 2GB drive though? I guess it doesn't really matter, cause if the drive fails, the OS should move the swapfile back to the default location.
If the drive fails, you've just experienced roughly the same thing as if you'd pulled a stick of memory out of your computer while it was running.
steve