If a single-cpu will process a few hundred models, then an extra CPU (or three) still won't get you near this chip's power. It's like rendering 3D video, dedicated hardware blows away a general-purpose CPU.
If this card lives up to the hype, then it'll be a good seller.
I think that multi-core chips have an even larger potential in the mobile market than in the desktop market.
With current processers - and even some not-so-current processers - there's not really much that an average person does on a laptop that actually uses all (or even most) of the CPU cycles. DVD playback, email, web surfing, and word processing tend to be the big apps - and none of them require much of a CPU at all.
However, once a person starts trying to do several things at once, then issues like context switches and interrupts start to make the machine feel sluggish, and having an extra CPU *does* make a difference in that area - and improvements to a machine's responsiveness (as opposed to just raw throughput) tend to make the user's "experience" much more pleasant - just look at the efforts in the Linux scheduler world to provide interruptable kernel and other patches which increase responsiveness at the expense of a little bit of throughput.
So, here's my idea: Take small, modestly-powered cores, and put two - or even more - on a chip. Scale the frequency of each core independentyly, turning the unused ones as low as possible (perhaps even OFF) when not needed. You'd be able to have a much more responsive machine and run off of much less power (on average) than with current chips.
As for the responsiveness of dual-CPU machines, I have a dual Pentium 133 that I picked up years ago, and it's not bad at all to use. It's not going to fool anyone into thinking it's an Athlon64, but when I tell people that they're using an original Pentium, they're surprised at just how nice it is to use.
I applied for two jobs, one called me back very quickly, gave me an interview, and offered me a job. I told them that I was waiting to hear on another job, and that I'd like to hold off on my acceptance until I heard back.
The person started getting pushy and belligerant. I pointed out to him that if I accepted the job, and two weeks later found out that the other offer was better, it would not be fair TO ME to pass up the other job - and that it would not be fair TO HIM if I left his company after two weeks.
At that point, he started getting REALLY pushy. Almost angry. He started going into metaphors about high school dances to get me to take the job right then because he had a lot of work to do. I even offered to work for him for a few weeks FOR FREE until I heard back on the other job. He just got more and more pushy, belligerant, and bully-ish.
At that point, I came to my senses and realized that I should turn him down cold. Even if I never heard back from the other job, I did not ever, ever, EVER want to work for someone like that. I politely but firmly told him that I no longer wanted the job, and left. I've never looked back, nor have I ever regretted it.
It'll work just fine. In fact, NewIsys' Horus architecture ties multiple 4-way Opterons together into a single, coherant system with up to 32 processers - and their docs explicitly state that with dual-core chips, the 8 machines would effectively have 64 CPUs. Not bad for commodity-level stuff.
Well... it remains to be seen whether a dual-core processer will cost any less than two single-core chips.
However, even if it does, you're still missing on one of the huge, principle benefits of the Opteron architecture: The multiple memory controllers. Two single-core chips will have two memory controllers, whereas a single dual-core chip will only have one.
Now, for compiling code, that's not going to make much difference, but for more memory-intensive stuff, it certainly can.
You'll have to forgive me, I'm not the greatest cryptographer in the world. But let's say that Joe Shmoe takes a picture with his cheap 8-megapixel camera, with a very high ISO setting for lots of noise. Now, that's roughly 192 megabits of information.
Suppose he needs to encode a 1 kilobit message. that means that there's going to be one bit of signal for every 192 kilobits of image. Now, say he does the encoding to merely appear like more noise in the already noisy image.
Given that low of a signal-to-noise ratio, I really don't see how you could detect the message unless you had prior knowledge of the algorithm or locations.
The whole point is that once you stop the processes that cause agin, you *would* act, look, and feel like you were 20 (or thereabouts).
Besides.... think of the benefits of compound interest. Instead of working for fourty years, and living off of your retirement for the next ten or twenty, you could work for eighty years, then live off of a much, much larger retirement for *hundreds* of years. Not a bad idea...
You don't need a full x16 slot. Tests have shown that you don't see any performance drop with a single card until you're in an x4 slot - two x8 slots are much more than enough for current-generation video cards.
My cheap $2,000 system beat the pants off of the other servers people put together because the hardware was actually compatible.
I've had one vendor loan me a $25,000 machine for a week to evaluate it. I ran head-to-head comparisons with a $13,000 machine that I had assembled, the $13,000 unit was fully twice as fast. Reliability? Uptimes in excess of 18 months were only broken to move the machine to a different location. : )
If it will shut down the business and cost more than a new server, then spend more bucks on a unit from big corp
Why, is Big Corp's machine guaranteed to have fewer problems? Do they guarantee that things will be fixed faster?
In my (admittedly limitted) relations with various Big Corps, I haven't seen that. Back in the day when we did business with Sun (which has, admittedly, been some time), the Sun Web Server would bring our Sun hardware to a complete, total lockup requiring a power cycle to correct. We spent over two *months* in direct work with the SWS development team at Sun, and they were never able to resolve the issue. We finally switched to another web server, and that solved the case. In the mean time, I can't tell you how many times we'd had to power cycle our server. In that case, going with Big Corp cost us many times more money, and did us absolutely no good.
There have been a very few companies where the support I've received was more timely and complete than what I could find on my own, but in the majority of cases, going the "Big Corp" route simply cost me more and got me less.
Why don't you try running linux, which will ignore the BIOS and do it's own HDD geometry homework.
Wow. Amazing how Linux can solve hardware problems in software.
As many things as it can do, even Linux won't access an entire drive if your IDE controller isn't capable of 48-bit addressing. Really. If the controller itself doesn't have the capability of addressing the entire drive, you're screwed from the start. Don't believe me? Get an old P2 motherboard, plop a 200-gig drive in, boot up Knoppix (or your favorite distro), and see.
Overall, I'm not even sure why you had to make it an OS issue, seeing as how sufficiently recent versions of Windows have 48-bit addressing capability, and can use all of a large drive as well. Maybe you just couldn't pass up the chance to flame someone, who knows.
I guess my measly 8x300gb machine just won't be as cool any more. : )
As an aside, replacing motherboards to support larger disks seems like a lot more work and expense than just buying a new controller card. The two controller cards for the machine I mentioned above totalled something like $50.
In my area, a few years ago, there was an accident on I15 involving a tanker full of something flammable, spilling the load and igniting, ruining the surface of the road. No big deal, divert traffic, fix road.
So, these guys want to put oil and gas pipelines by this magehighway designed to carry tractor-trailers 4 lanes wide, huh? They better have pretty good separation.
"Yesterday, I was contacted by the leader of the spinoff project who told me that he's quiet angry at us for doing that and that it's considered unethical and rude to copy code from the spinoff.
Just about any behavior you can think of is rude to someone on the Internet.
Don't get me wrong, I'm not one of the "Do whatever you like, and tell anyone who complains that they're an oppressive nazi" sort of people. I like it when people are civil, and have some manners.
But some people are overly anal in the other direction. I'd mention specifics, but that would just engender a pedantic flame-war over the items I've mentioned... so I'll just say that even staying within some pretty sensible, civil, friendly bounds you'll still run into people who believe that you have in some way offended or slighted them.
Back in the day when Livingston Portmaster 3's were still lots and lots of money, I opened one up and saw that they were based around a 40 MHz AMD 486-clone.
Memory manufacturers, in order to have bigger numbers, specify capacity in megabits, not megabytes. So, take the number of megabits, and divide by 8. That gives you megabytes. Now, multiply by the number of chips you want to stick on the DIMM (I've seen as low as 2, and as high as 32), and you have the total capacity.
Spending $50 on a game to wonder if you'll like it or not is kind of suicide, I bought UT2k4 for $30!!!
Spending $50 on *one* game to see if I'll like it isn't all that bad. But when I have to spend $50 on five, ten, or twenty games before I find one that I like, then it's not as easy to say "Spending $1000 on games to see if you might like one isn't that bad."
I keep trying game demos, and it seems that for every 100 games published, there might be one that isn't just crap.
There's just no way I'm going to pay $50 for a game that I can't even play unless I keep forking out more money. If they want $13 to $15 per month to play the game, then they should give the game away for free.
An electric car isn't hard to make. An electric car that goes fast isn't hard to make. An electric car with a long cruising range isn't hard to make. And an electric car that goes fast *and* has a long cruising range still isn't too hard to make.
On the other hand, making an electric car that can go reasonably fast, has a reasonably long cruising range, has a reasonably long battery life span, and is reasonably affordable does seem to be pretty tough to do. If you want to do some good for the planet in the area of electric cars, work on that problem.
Not only that, the screw-in CF lamps don't work well in the cold, which makes them useless for outdoor lamps in the winter.
I rather enjoy putting them in the porch lights during the summer, so I can leave them on all night if I want, but during the winter, I have to go back to the regular incandescents.
Yeah, I know. You're not supposed to use them outside because of moisture. Well, the fixtures keep the water out of the lights. : )
and only uses 22w to produce twice the light of a 100w bulb."
A 100-watt light bulb puts out around 1500-1600 lumens. These lamps are rated at 280 and 320 lumens. A more accurate statement would be "and uses one-fifth the energy to produce one-fifth the light"
A lot of the time, I'd love to jump in the back seat of my Element with the dogs, tell the computer "drive us to the trail head", and hop out when we get there.
On another plus note, a very great deal of the congestion and slowdown occurs from two things: people wanting to drive at greatly dissimilar speeds, and people being selfish ("I'll cut off 30 cars, making them all come to a stop, to save myself 20 seconds."). With more cars driving in a synchronized pattern, I think that the overall throughput will go up greatly.
Slashdot Mirror
If a single-cpu will process a few hundred models, then an extra CPU (or three) still won't get you near this chip's power. It's like rendering 3D video, dedicated hardware blows away a general-purpose CPU.
If this card lives up to the hype, then it'll be a good seller.
Steve
I think that multi-core chips have an even larger potential in the mobile market than in the desktop market.
With current processers - and even some not-so-current processers - there's not really much that an average person does on a laptop that actually uses all (or even most) of the CPU cycles. DVD playback, email, web surfing, and word processing tend to be the big apps - and none of them require much of a CPU at all.
However, once a person starts trying to do several things at once, then issues like context switches and interrupts start to make the machine feel sluggish, and having an extra CPU *does* make a difference in that area - and improvements to a machine's responsiveness (as opposed to just raw throughput) tend to make the user's "experience" much more pleasant - just look at the efforts in the Linux scheduler world to provide interruptable kernel and other patches which increase responsiveness at the expense of a little bit of throughput.
So, here's my idea: Take small, modestly-powered cores, and put two - or even more - on a chip. Scale the frequency of each core independentyly, turning the unused ones as low as possible (perhaps even OFF) when not needed. You'd be able to have a much more responsive machine and run off of much less power (on average) than with current chips.
As for the responsiveness of dual-CPU machines, I have a dual Pentium 133 that I picked up years ago, and it's not bad at all to use. It's not going to fool anyone into thinking it's an Athlon64, but when I tell people that they're using an original Pentium, they're surprised at just how nice it is to use.
steve
I applied for two jobs, one called me back very quickly, gave me an interview, and offered me a job. I told them that I was waiting to hear on another job, and that I'd like to hold off on my acceptance until I heard back.
The person started getting pushy and belligerant. I pointed out to him that if I accepted the job, and two weeks later found out that the other offer was better, it would not be fair TO ME to pass up the other job - and that it would not be fair TO HIM if I left his company after two weeks.
At that point, he started getting REALLY pushy. Almost angry. He started going into metaphors about high school dances to get me to take the job right then because he had a lot of work to do. I even offered to work for him for a few weeks FOR FREE until I heard back on the other job. He just got more and more pushy, belligerant, and bully-ish.
At that point, I came to my senses and realized that I should turn him down cold. Even if I never heard back from the other job, I did not ever, ever, EVER want to work for someone like that. I politely but firmly told him that I no longer wanted the job, and left. I've never looked back, nor have I ever regretted it.
steve
It'll work just fine. In fact, NewIsys' Horus architecture ties multiple 4-way Opterons together into a single, coherant system with up to 32 processers - and their docs explicitly state that with dual-core chips, the 8 machines would effectively have 64 CPUs. Not bad for commodity-level stuff.
steve
Well... it remains to be seen whether a dual-core processer will cost any less than two single-core chips.
However, even if it does, you're still missing on one of the huge, principle benefits of the Opteron architecture: The multiple memory controllers. Two single-core chips will have two memory controllers, whereas a single dual-core chip will only have one.
Now, for compiling code, that's not going to make much difference, but for more memory-intensive stuff, it certainly can.
steve
If you're truly paranoid, you can disable loadable modules, thus preventing a kernel-level rootkit module from being loaded.
steve
You'll have to forgive me, I'm not the greatest cryptographer in the world. But let's say that Joe Shmoe takes a picture with his cheap 8-megapixel camera, with a very high ISO setting for lots of noise. Now, that's roughly 192 megabits of information.
Suppose he needs to encode a 1 kilobit message. that means that there's going to be one bit of signal for every 192 kilobits of image. Now, say he does the encoding to merely appear like more noise in the already noisy image.
Given that low of a signal-to-noise ratio, I really don't see how you could detect the message unless you had prior knowledge of the algorithm or locations.
steve
The whole point is that once you stop the processes that cause agin, you *would* act, look, and feel like you were 20 (or thereabouts).
Besides.... think of the benefits of compound interest. Instead of working for fourty years, and living off of your retirement for the next ten or twenty, you could work for eighty years, then live off of a much, much larger retirement for *hundreds* of years. Not a bad idea...
steve
You don't need a full x16 slot. Tests have shown that you don't see any performance drop with a single card until you're in an x4 slot - two x8 slots are much more than enough for current-generation video cards.
steve
My cheap $2,000 system beat the pants off of the other servers people put together because the hardware was actually compatible.
I've had one vendor loan me a $25,000 machine for a week to evaluate it. I ran head-to-head comparisons with a $13,000 machine that I had assembled, the $13,000 unit was fully twice as fast. Reliability? Uptimes in excess of 18 months were only broken to move the machine to a different location. : )
steve
If it will shut down the business and cost more than a new server, then spend more bucks on a unit from big corp
Why, is Big Corp's machine guaranteed to have fewer problems? Do they guarantee that things will be fixed faster?
In my (admittedly limitted) relations with various Big Corps, I haven't seen that. Back in the day when we did business with Sun (which has, admittedly, been some time), the Sun Web Server would bring our Sun hardware to a complete, total lockup requiring a power cycle to correct. We spent over two *months* in direct work with the SWS development team at Sun, and they were never able to resolve the issue. We finally switched to another web server, and that solved the case. In the mean time, I can't tell you how many times we'd had to power cycle our server. In that case, going with Big Corp cost us many times more money, and did us absolutely no good.
There have been a very few companies where the support I've received was more timely and complete than what I could find on my own, but in the majority of cases, going the "Big Corp" route simply cost me more and got me less.
steve
Aren't they on the "no-no" list for hard-drives as of late?
Depending on whom you ask, *every* hard drive manufacturer is on the "no-no" list. Name the brand, someone will give you a horror story with them.
steve
Why don't you try running linux, which will ignore the BIOS and do it's own HDD geometry homework.
Wow. Amazing how Linux can solve hardware problems in software.
As many things as it can do, even Linux won't access an entire drive if your IDE controller isn't capable of 48-bit addressing. Really. If the controller itself doesn't have the capability of addressing the entire drive, you're screwed from the start. Don't believe me? Get an old P2 motherboard, plop a 200-gig drive in, boot up Knoppix (or your favorite distro), and see.
Overall, I'm not even sure why you had to make it an OS issue, seeing as how sufficiently recent versions of Windows have 48-bit addressing capability, and can use all of a large drive as well. Maybe you just couldn't pass up the chance to flame someone, who knows.
steve
I guess my measly 8x300gb machine just won't be as cool any more. : )
As an aside, replacing motherboards to support larger disks seems like a lot more work and expense than just buying a new controller card. The two controller cards for the machine I mentioned above totalled something like $50.
steve
In my area, a few years ago, there was an accident on I15 involving a tanker full of something flammable, spilling the load and igniting, ruining the surface of the road. No big deal, divert traffic, fix road.
So, these guys want to put oil and gas pipelines by this magehighway designed to carry tractor-trailers 4 lanes wide, huh? They better have pretty good separation.
steve
"Yesterday, I was contacted by the leader of the spinoff project who told me that he's quiet angry at us for doing that and that it's considered unethical and rude to copy code from the spinoff.
Just about any behavior you can think of is rude to someone on the Internet.
Don't get me wrong, I'm not one of the "Do whatever you like, and tell anyone who complains that they're an oppressive nazi" sort of people. I like it when people are civil, and have some manners.
But some people are overly anal in the other direction. I'd mention specifics, but that would just engender a pedantic flame-war over the items I've mentioned... so I'll just say that even staying within some pretty sensible, civil, friendly bounds you'll still run into people who believe that you have in some way offended or slighted them.
steve
Back in the day when Livingston Portmaster 3's were still lots and lots of money, I opened one up and saw that they were based around a 40 MHz AMD 486-clone.
steve
Memory manufacturers, in order to have bigger numbers, specify capacity in megabits, not megabytes. So, take the number of megabits, and divide by 8. That gives you megabytes. Now, multiply by the number of chips you want to stick on the DIMM (I've seen as low as 2, and as high as 32), and you have the total capacity.
steve
Spending $50 on a game to wonder if you'll like it or not is kind of suicide, I bought UT2k4 for $30!!!
Spending $50 on *one* game to see if I'll like it isn't all that bad. But when I have to spend $50 on five, ten, or twenty games before I find one that I like, then it's not as easy to say "Spending $1000 on games to see if you might like one isn't that bad."
I keep trying game demos, and it seems that for every 100 games published, there might be one that isn't just crap.
steve
There's just no way I'm going to pay $50 for a game that I can't even play unless I keep forking out more money. If they want $13 to $15 per month to play the game, then they should give the game away for free.
steve
(D&D, natch)
You use the term "natch", and claim to have a critical eye?
Myself, I'm skeptical of people who claim to have a double-digit IQ if they use the term "natch".
steve
An electric car isn't hard to make. An electric car that goes fast isn't hard to make. An electric car with a long cruising range isn't hard to make. And an electric car that goes fast *and* has a long cruising range still isn't too hard to make.
On the other hand, making an electric car that can go reasonably fast, has a reasonably long cruising range, has a reasonably long battery life span, and is reasonably affordable does seem to be pretty tough to do. If you want to do some good for the planet in the area of electric cars, work on that problem.
steve
Not only that, the screw-in CF lamps don't work well in the cold, which makes them useless for outdoor lamps in the winter.
I rather enjoy putting them in the porch lights during the summer, so I can leave them on all night if I want, but during the winter, I have to go back to the regular incandescents.
Yeah, I know. You're not supposed to use them outside because of moisture. Well, the fixtures keep the water out of the lights. : )
steve
and only uses 22w to produce twice the light of a 100w bulb."
A 100-watt light bulb puts out around 1500-1600 lumens. These lamps are rated at 280 and 320 lumens. A more accurate statement would be "and uses one-fifth the energy to produce one-fifth the light"
steve
A lot of the time, I'd love to jump in the back seat of my Element with the dogs, tell the computer "drive us to the trail head", and hop out when we get there.
On another plus note, a very great deal of the congestion and slowdown occurs from two things: people wanting to drive at greatly dissimilar speeds, and people being selfish ("I'll cut off 30 cars, making them all come to a stop, to save myself 20 seconds."). With more cars driving in a synchronized pattern, I think that the overall throughput will go up greatly.
steve