An especially troublesome aspect of pairing mis-matched cards: whatever this technology is, it can't use alternate-frame rendering (AFR, where alternating frames are sent to alternating cards).
You can't make AFR work with an unbalanced pair of cards because the fastest framerate you can get is limited to 2x the speed of the slowest card. Let me explain: if you paired a new card (100fps) with one that had 1/2 the speed (50fps), if you had perfect scaling, you could technically get 150fps from the pair. But if they used AFR and took turns rendering one frame at a time, the most you could see from the pair is twice the old card's framerate, or 100fps.
AFR could only give you 150fps in this case if the game engine could tell you the future, and supply a frame for rendering one frame in-advance to the older card. This is simply not possible, and the game engine supplies frames only in real time.
So, this means that you have to use a better method to balance the workload between the two (or more) cards. But this was already attempted - Nvidia offered split-screen rendering for SLI, and for a time ATI offered tiled rendering, but both methods had major compatibility issues (and split-screen rendering was notorious for tearing artifacts). Today, both ATI and Nvidia support AFR as the default rendering mode, and very seldom do they step oiutside this circle.
The fact is, other methods besides AFR have already been considered, but in the end AFR is the method the industry has consistently fallen back on. The only real downside of AFR is you need balanced cards, or it won't give you any benefit. Unfortunately, I'm hesitant to believe that ANY company can come up with a better solution for SLI, even a third-party solutions like these.
The Atom being used for this is a horrible thing. It's designed as a low-power processor but so much so that under real use it uses considerably more power with less capability than an athlon 64.
That's not the CPU, it's the chipset.
The board in question uses the 945GC chipset, while all mobile devices use the Mobile 945GM or better. The Mobile 945GM features a much cooler 7w TDP (see section 11.2), while the 945GC is 22w TDP. You can even get as low as 5w TDP if you pay extra.
Why the difference? The 945GC is the rejects pile. If they can't get the chip to run an the lower voltages specced for the 945GM (about 1.05v), then they attempt to bin it at the much higher clock and voltage of the 945GC.
The Intel Atom board with the 945GC is intended for ultra-cheap PCs (sub $150), and is mostly marketed in third-world countries. the board itself is NOT designed for low power, although there's nothing stopping a third party from making such a beast.
Correct me if I'm wrong, but didn't the Genesis/Megadrive come out between the NES and SNES? That means for a while Sega did have the upper hand.
Yes, but they squandered their lead. While the Megadrive made System 16 arcade ports easy, most of these games had already been ported to the Master System, which made the system less appealing to Master System owners. And Sega had trouble getting third-party giants to jump ship, so the game selection as a whole was poor.
By the time Sega got off their ass and created some good original games, third-parties started to take notice of the system. But by then any potential for a lead was gone.
Not only that, but even after the SNES was released, the Genesis and SNES were very different machines with their own pros/cons.
Of course.
Megadrive was based off System 16. The idea behind system 16 was to get as much excitement on the screen for as little memory as possible, since it was extremely expensive in the early 1980s. Megadrive could move a lot of pixels very fast, with double-buffering, and this made up for the marginal color depth and FM synth.
The SNES didn't have an arcade legacy to support, and memory was much cheaper by the early 90s, so they packed it with twice the ram and better video/sound.
Why should Apple, (or anyone else) be forced to license a product to a competitor just because the competitor thinks it would make its competing product better?
Because any user can walk into a store and buy it off the shelf. Psystar is supplying these very same licenses to people who purchase their hardware.
Either you don't sell the OS at all (only bundled with your hardware, or given away for free), or you have to make sales non-discriminatory. This is the same sort of restriction levied on Microsoft with Windows sales.
That's an inaccurate description. There's nothing stopping you from re-working a half-node shrinked design. Yes, it is often used as a stop-gap where all they do is a shrink, but that's not always the case: look at ATI's Radeon 4800 series on 55nm (an all-new design built on a half-node between 65nm and 45nm). And for further examples, Nvidia released the very successful 6600GT and later the 7800 series on the 110nm half-node (between 130nm and 90nm). These are all NEW designs, not simple die shrinks.
Here is why they really call it "half-node."
There is a rhyme and reason to why we jump from process to process..50 ->.35 ->.25 ->.18 ->.13 ->.09 ->.065 ->.045 ->.032 micron. These are referred to as full-node process jumps.
The reason we make such jumps is because each jump is rougly double the density of the previous; that is to say, you can fit twice as many transistors into the same area (example: 90nm^2 / 65nm^2 = 1.91). They choose nice round numbers, so it's never exactly twice as dense. This large jump in density justifies the incredible investment into new fab tech, especially if you are a company with few fabs.
But companies like TSMC have lots of fabs, and they can afford what are called half-node processes (the density is increased by only 1.5, or %50). Examples of half-node processes include.22 micron (GeForce 256),.15 micron (Radeon 9700 Pro),.11 micron (6600GT, 7800GTX),.08 micron (Radeon HD 2900XT), and now 55nm (Radeon HD 4800 series). You'll note that EVERY one of the above examples is new technology, not a simple die-shrink of an older core.
Customers like half-node processes because it gives them more choices, and lets them reduce power without waiting for the full-node process jump.
I saw the same thing. If you get no search results, try re-submitting your request. It never happened to me twice-in-a-row.
I think it may be that they're overloaded from all the traffic. You do know that the news of their existence has been spewed on just about every website, including all the major news outlets (MUCH more traffic than Slashdot). I'm amazed they can keep up, considering they really don't have the money for the kind of hardware Google has!
Now it doesn't search your entire browsing history or your bookmarks, just URLs you have actually typed into the bar. You can visit any site you want, and so long as you don't type the address in, it won't show up in the awful bar results.
Not that my girlfriend has a problem with porn. I did this mostly because I have some less open-minded friends who use my computer, and who might be uncomfortable seeing my entire browsing history. I actually did the same thing for my girlfriend's computer, for exactly the same reason.
But that is an incredibly poor assumption. The chipset bears some responsibility for that power increase at load.
Let me try to illustrate why:
Have you ever watched the temperatures of your system go up while under load? Sure, the CPU temperature climbs, but so does the chipset temperature - SIGNIFICANTLY. On a desktop board with a desktop CPU, the climb in chipset power at load is small compared to the CPU, and is often lost in the noise. But when paired with a low-power CPU like the Atom, the chipset power rise at load becomes significant.
Really, do you think that the sudden load of memory accesses and routing of data has no effect on the chipset's power consumption?
I firmly believe that the 945GC is responsible for half of that 4w increase, which puts the Atom right around the promised 2w TDP mark.
Yes, Intel is selling the Atom with a craptastic 945GC chipset, which doesn't help things. Intel is still manufacturing the 945GC chipset on an old 130nm fab, and is practically dumping these chips for the sub-$200 computer market. This makes the D945GCLF dirt-cheap, but also much less efficient than the CPU itself would imply.
As I have posted earlier, you can pair-up Atom with many other lower-power chipsets. Options available include the mobile GM945, which features a TDP about 10w lower than the 945GC. Or if you want absolute low power, there's always the Poulsbo chipset (part of Centrino Atom), which sips power at 2.3w TDP.
I have to agree with the author on one thing: if Intel is strong-arming the industry and locking-down Atom boards, this is bad for all of us. I've been really looking forward to the low-power performance Atom could bring to the MiniITX market, but so far that hasn't materialized.
You are of course referring to the D945GCLF with the desktop Atom and 945GC.
Intel is practically dumping the 945GC because it's built in an old 130nm fab, a fab Intel finds no monetary reason to upgrade. The desktop atom chip is much more frugal, with only a 4w TDP, so it is just the chipset holding it back.
The intent of the D945GCLF is not to be an ultra-low power board, but to be a CHEAP board to feed the $100-150 PC market, and find a use for old fab tech. There are much more efficient bridge chips available from Intel that can be used with the Atom: you can use standard mobile GM945 chips (10w less than 945GC), or if you want ultra-low power there is always the Poulsbo chipset featured in the Atom Centrino platform (2.3w total).
Hey, if manufacturers aren't making a low-enough power Atom board you crave and want, bug them to hell and let them know there's a market! Intel isn't the only MiniITX board maker in town, you know.
The Atom delivers MORE performance/watt than Via's solutions.
Atom destroys the C7 in performance/watt. The C7 has a relatively poor memory architecture and has piss-poor SSE performance in comparison to the Atom. The C7 also cannot match the low consumption of the Atom at a similar clock speed.
If you really think the C7 has good performance/watt, just see how a DESKTOP Intel Celeron keeps pace with the C7 in terms of power consumption. The power consumption is within a few watts, and the Intel chip delivers more than double the performance in some benchmarks. Never mind that the Intel board costs less than anything Via has on-offer.
The Via Nano is closer: it delivers better per-clock integer performance (2 integer units vs one on the Atom), but the two chips deliver the same per-clock SSE performance. When you consider that the Nano uses 3-10x the power of the Atom at various clock speeds, you begin to see how Via can't compete on performance/watt.
Of course, the only car I really consider ugly are those box-things. Dunno the name of them, I'm not a car guy, but they look just like large metal rectangles with a spot cut out for the hood/windshield.
Really, it's a better Camry: same engine, similar curb weight, more passenger space, less cost, all without the grandpa-inspired handling of the Camry.
And yes, you will get poor gas mileage if all you do is drive city, but so does the Camry. If you drive mixed like I do, you'll get good mileage (despite the boxy frame). My combined mileage is 27 mpg (and I drive FAST), which is about average for the performance level of the xB. The older model gets even better mileage, but you don't want to try driving a box around on a 1.6l engine.
Personally, I find it a bit hard to believe that a civilization is smart enough to travel interstellar distances but too stupid to use basic camouflage.
Camouflage from what?
Aliens visiting us could see in completely different wavelengths. Honestly, the only reasion we see in the visible range is because it's the most intense light that passes through our atmosphere. If we had a different atmosphere or a different sun, you bet we'd see in a different range. It wouldn't be the first time that a major assumption was made about alien species.
And speaking of major assumptions, what if the aliens can't see? Would they even conceive of the need for visual camouflage?
The fact is, most Bluetooth headsets are Class 2 devices, which have a maximum power of 2.5 mW. This is orders of magnitude less than the emissions from a cell phone, which can peak at 500 mW.
If the emissions from a cell phone are simply "questionable" in terms of cancer, there's no way a signal with 100x less power is. But on the flip side, the power difference between the two is so large that you COULD see them claiming cell emissions are "bad" while not seeing any problem with the much lower power emitted by Bluetooth Class 2 devices.
Yeah, I did the same thing. I made programs to solve all sorts of problems from class on my TI-82, usually while sitting there in class.
Further, some of my games made use of stuff from said classes. The Scorched Earth clone I made used randomized trig curves to build maps, and of course I used the ballstic parabola equation to plot the trajectory of the shots. I did this all with TIBASIC, and although it was slow, it was playable.
If I were to make a recommendation, I'd say get him a TI-89. It has much more horsepower than my TI-82, so you can do more in real-time than my old TI-82. It will also last him through college if he takes a technical road. BUT: you will need to help him get started, show him some of the more interesting aspects, and help him write a few of his first programs.
Does it really matter when the existing solution (water with ions) also conducts electricity quite well?
The fact is, MOST of the best conductors of heat also conduct electricity well.
Pure water is actually a poor electrical and heat conductor; what makes it popular for cooling is that it is liquid at room temperature, cheap and easy to acquire, and is only marginally corrosive (compared to other liquid coolants).
Liquid is the key: even though water is a poor thermal conductor, if you keep it constantly moving through a system, it will move more heat than just passive conduction using a solid metal. This has to do with increased rate of heat transfer when you have a larget temperature delta.
Why not both? A vertical screen for viewing text, images, data, etc. And a horizontal one for playing with widgets, data entry, and the like.
We already have this, and it's called a laptop. The horizontal plane contains manipulation tools, and the vertical plane contains feedback.
I actually don't understand all the excitement over touchscreens for large devices myself. Touchpads were actually developed to leverage the technology of touchscreens without all the drawbacks of said large screens.
Benefits of touchpads over touchscreens:
* You have much reduced arm travel, which is one of the biggest annoyances with large touchscreens. This gets worse as screens get larger. * Your arm doesn't block the things you are looking at on the screen.
This whole love-fest over touchscreen technology isn't anything new, and the only reason why it's recently resurfaced is because touchpad recognition technology is now excellent.
Basically what I'm saying is that the iPhone et al are just touchpads with a screen attached. If you try to expand said screens, you'll run into the same problems previous touchscreens did and never solved. I don't have much hope for those problems getting resolved, so touchscreens will probably remain restricted to hand-held devices.
You know, most programmers already know how to construct state machines, and alreay create entire programs using this concept. Your ideas are not revolutionary, they only highlight the need to use asynchronous state machines over synchronous threads. Where your ideas fail is this: you want to get rid of threads completely.
Do you want to know why the programming community likes threads? There's a simple reason: state machines DO NOT SCALE. As you add more capabilities to a state machine, the number of states you have to add for each new functionality can increase geometrically. Petty soon, the number of independent state variables defining the machine gets unmanageable.
What do you do when that happens? Well, in the real world, you break up that state machine into multiple threads. You basically draw an artificial line between units that should not be separated, simply because leaving them attached is not feasable or maintainable. Yes, operations within each thread happen asynchnously, but the multiple threads are now dependent on each other, because you were forced to break-up the complex system.
In the end, you're left with the same problem we've always had: balancing threads with performance. Yes, you can build threads that are asynchronous; this has always been possible using state machines. What you cannot do is completely toss threads; threads exist to make progamming problems more accessible to humans. Unless you can take humans out of the loop, threads are here to stay.
You think your client, one of the richest men in the world, is a vigilante who likes to dress up as a bat and beat criminals to a bloody pulp with this bare hands.
Oh, and the movie was awesome. I never considered Batman to be a truely dark character before this film; previous directors have tried to make Batman dark, and they have failed.
(*) Fact: CPUs for quite some time are capable of decoding HD video in real time - it's the transfer (1) to/from RAM and (2) video hardware which takes most of the time.
Interesting fact, and I most certainly believe it. In fact, the marginal system memory bandwidth of most PCs is the very reason we have video cards (and now even sound cards) with local cache. We make the system memory slow because we want lots of it. We can afford to make local memories faster because there is less of it.
I have to agree that there's really no need for this kind of co-processor - dual-core CPUs can decode almost anything worthwhile in real-time these days, and even embedded GPUs can render assistance decoding video. This device has a fast local memory, but unless the results are staying local (batch processing), you're going to run into the limits of system memory bandwidth streaming the results to something that can actually use it.
I'f like to make one further note: quad cores on the desktop are already offering more horsepower than this thing anyway, and mobile dual-cores are within spitting distance:
Assumptions:
Each SPU has 2 128-bit vector units. Four SPUs means eight vector units. Runs at around 1.5 GHz.
Conroe: two 128-bit vector units per-core. Dual-core platforms have about %85 of the horsepower running at 2.6 GHz. Four cores would blow Toshiba's SPU device out of the water, assuming you aren't bandwidth-limited.
I'm sure he meant from *this generation* but it's what I instantly thought of as well.
That's exactly what I meant. You can attach all the memory you want to a video card, true, but there is a limit to how much you can conceivably use.
When you use more memory, you use more memory bandwidth; this is an undisputable fact. If you double the resolution of textures in a scene, the texture memory bandwidth you need to render that scene doubles. If you double your resolution, the amount of framebuffer writes doubles.
Since memory bandwidth is finite, you can only add so much before it becomes worthless. This is why the 2GB would be FAR MORE useful on a DDR5 card like the 4870, or else the GTX 280 with a 512-bit memory bus.
Don't get me wrong; the 4850 has just enough memory bandwidth to perform efficiently in it's current configuration, and I like it so much I bought one myself. But anything above 1GB on the 4850 will be wasted, because you cannot magically increase the bandwidth.
Heh, you don't have to tell me that we'll always need MORE MEMORY. My first 3D accelerator featured an amazing 4MB of memory! But along the same line as above, I wouldn't have put 16MB of ram on my first card, because it would have been worthless with the limited 64-bit bus.
Actually, it's pointless for FPS style games. They'll never use even a GB of that memory effectively because the games are designed around people with 512MB at the high end.
They're only doing this because DDR3 is much cheaper than the DDR5 on the 4870. A 2GB 4850 with DDR3 is cheaper than a 1GB 4870 with DDR5. Me, I can't see the value of getting a card with more than 1GB, even for future games.
The only reason I see to buy this card is maybe there are drivers optimized for professional work where the memory requirements are much higher (3D modelers and the like).
There won't be. This card is marketed as a 4850, not a FireGL, which means it won't be all that useful or professionals. Without the drivers to accelerate professional applications, the extra memory is largely useless.
Slashdot Mirror
An especially troublesome aspect of pairing mis-matched cards: whatever this technology is, it can't use alternate-frame rendering (AFR, where alternating frames are sent to alternating cards).
You can't make AFR work with an unbalanced pair of cards because the fastest framerate you can get is limited to 2x the speed of the slowest card. Let me explain: if you paired a new card (100fps) with one that had 1/2 the speed (50fps), if you had perfect scaling, you could technically get 150fps from the pair. But if they used AFR and took turns rendering one frame at a time, the most you could see from the pair is twice the old card's framerate, or 100fps.
AFR could only give you 150fps in this case if the game engine could tell you the future, and supply a frame for rendering one frame in-advance to the older card. This is simply not possible, and the game engine supplies frames only in real time.
So, this means that you have to use a better method to balance the workload between the two (or more) cards. But this was already attempted - Nvidia offered split-screen rendering for SLI, and for a time ATI offered tiled rendering, but both methods had major compatibility issues (and split-screen rendering was notorious for tearing artifacts). Today, both ATI and Nvidia support AFR as the default rendering mode, and very seldom do they step oiutside this circle.
The fact is, other methods besides AFR have already been considered, but in the end AFR is the method the industry has consistently fallen back on. The only real downside of AFR is you need balanced cards, or it won't give you any benefit. Unfortunately, I'm hesitant to believe that ANY company can come up with a better solution for SLI, even a third-party solutions like these.
The Atom being used for this is a horrible thing. It's designed as a low-power processor but so much so that under real use it uses considerably more power with less capability than an athlon 64.
That's not the CPU, it's the chipset.
The board in question uses the 945GC chipset, while all mobile devices use the Mobile 945GM or better. The Mobile 945GM features a much cooler 7w TDP (see section 11.2), while the 945GC is 22w TDP. You can even get as low as 5w TDP if you pay extra.
Why the difference? The 945GC is the rejects pile. If they can't get the chip to run an the lower voltages specced for the 945GM (about 1.05v), then they attempt to bin it at the much higher clock and voltage of the 945GC.
The Intel Atom board with the 945GC is intended for ultra-cheap PCs (sub $150), and is mostly marketed in third-world countries. the board itself is NOT designed for low power, although there's nothing stopping a third party from making such a beast.
Correct me if I'm wrong, but didn't the Genesis/Megadrive come out between the NES and SNES? That means for a while Sega did have the upper hand.
Yes, but they squandered their lead. While the Megadrive made System 16 arcade ports easy, most of these games had already been ported to the Master System, which made the system less appealing to Master System owners. And Sega had trouble getting third-party giants to jump ship, so the game selection as a whole was poor.
By the time Sega got off their ass and created some good original games, third-parties started to take notice of the system. But by then any potential for a lead was gone.
Not only that, but even after the SNES was released, the Genesis and SNES were very different machines with their own pros/cons.
Of course.
Megadrive was based off System 16. The idea behind system 16 was to get as much excitement on the screen for as little memory as possible, since it was extremely expensive in the early 1980s. Megadrive could move a lot of pixels very fast, with double-buffering, and this made up for the marginal color depth and FM synth.
The SNES didn't have an arcade legacy to support, and memory was much cheaper by the early 90s, so they packed it with twice the ram and better video/sound.
Why should Apple, (or anyone else) be forced to license a product to a competitor just because the competitor thinks it would make its competing product better?
Because any user can walk into a store and buy it off the shelf. Psystar is supplying these very same licenses to people who purchase their hardware.
Either you don't sell the OS at all (only bundled with your hardware, or given away for free), or you have to make sales non-discriminatory. This is the same sort of restriction levied on Microsoft with Windows sales.
Sure we wang!
That's an inaccurate description. There's nothing stopping you from re-working a half-node shrinked design. Yes, it is often used as a stop-gap where all they do is a shrink, but that's not always the case: look at ATI's Radeon 4800 series on 55nm (an all-new design built on a half-node between 65nm and 45nm). And for further examples, Nvidia released the very successful 6600GT and later the 7800 series on the 110nm half-node (between 130nm and 90nm). These are all NEW designs, not simple die shrinks.
Here is why they really call it "half-node."
There is a rhyme and reason to why we jump from process to process. .50 -> .35 -> .25 -> .18 -> .13 -> .09 -> .065 -> .045 -> .032 micron. These are referred to as full-node process jumps.
The reason we make such jumps is because each jump is rougly double the density of the previous; that is to say, you can fit twice as many transistors into the same area (example: 90nm^2 / 65nm^2 = 1.91). They choose nice round numbers, so it's never exactly twice as dense. This large jump in density justifies the incredible investment into new fab tech, especially if you are a company with few fabs.
But companies like TSMC have lots of fabs, and they can afford what are called half-node processes (the density is increased by only 1.5, or %50). Examples of half-node processes include .22 micron (GeForce 256), .15 micron (Radeon 9700 Pro), .11 micron (6600GT, 7800GTX), .08 micron (Radeon HD 2900XT), and now 55nm (Radeon HD 4800 series). You'll note that EVERY one of the above examples is new technology, not a simple die-shrink of an older core.
Customers like half-node processes because it gives them more choices, and lets them reduce power without waiting for the full-node process jump.
I saw the same thing. If you get no search results, try re-submitting your request. It never happened to me twice-in-a-row.
I think it may be that they're overloaded from all the traffic. You do know that the news of their existence has been spewed on just about every website, including all the major news outlets (MUCH more traffic than Slashdot). I'm amazed they can keep up, considering they really don't have the money for the kind of hardware Google has!
Set the following field to true:
browser.urlbar.matchOnlyTyped
Now it doesn't search your entire browsing history or your bookmarks, just URLs you have actually typed into the bar. You can visit any site you want, and so long as you don't type the address in, it won't show up in the awful bar results.
Not that my girlfriend has a problem with porn. I did this mostly because I have some less open-minded friends who use my computer, and who might be uncomfortable seeing my entire browsing history. I actually did the same thing for my girlfriend's computer, for exactly the same reason.
But that is an incredibly poor assumption. The chipset bears some responsibility for that power increase at load.
Let me try to illustrate why:
Have you ever watched the temperatures of your system go up while under load? Sure, the CPU temperature climbs, but so does the chipset temperature - SIGNIFICANTLY. On a desktop board with a desktop CPU, the climb in chipset power at load is small compared to the CPU, and is often lost in the noise. But when paired with a low-power CPU like the Atom, the chipset power rise at load becomes significant.
Really, do you think that the sudden load of memory accesses and routing of data has no effect on the chipset's power consumption?
I firmly believe that the 945GC is responsible for half of that 4w increase, which puts the Atom right around the promised 2w TDP mark.
Yes, Intel is selling the Atom with a craptastic 945GC chipset, which doesn't help things. Intel is still manufacturing the 945GC chipset on an old 130nm fab, and is practically dumping these chips for the sub-$200 computer market. This makes the D945GCLF dirt-cheap, but also much less efficient than the CPU itself would imply.
As I have posted earlier, you can pair-up Atom with many other lower-power chipsets. Options available include the mobile GM945, which features a TDP about 10w lower than the 945GC. Or if you want absolute low power, there's always the Poulsbo chipset (part of Centrino Atom), which sips power at 2.3w TDP.
I have to agree with the author on one thing: if Intel is strong-arming the industry and locking-down Atom boards, this is bad for all of us. I've been really looking forward to the low-power performance Atom could bring to the MiniITX market, but so far that hasn't materialized.
I think it's more common today than you think. My 2008 Scion xB has one. So do lots of other cars, especially hybrids.
Honestly, if I didn't have the gauge, I'd never believe I was getting 27 mpg average driving a box.
You are of course referring to the D945GCLF with the desktop Atom and 945GC.
Intel is practically dumping the 945GC because it's built in an old 130nm fab, a fab Intel finds no monetary reason to upgrade. The desktop atom chip is much more frugal, with only a 4w TDP, so it is just the chipset holding it back.
The intent of the D945GCLF is not to be an ultra-low power board, but to be a CHEAP board to feed the $100-150 PC market, and find a use for old fab tech. There are much more efficient bridge chips available from Intel that can be used with the Atom: you can use standard mobile GM945 chips (10w less than 945GC), or if you want ultra-low power there is always the Poulsbo chipset featured in the Atom Centrino platform (2.3w total).
Hey, if manufacturers aren't making a low-enough power Atom board you crave and want, bug them to hell and let them know there's a market! Intel isn't the only MiniITX board maker in town, you know.
The Atom delivers MORE performance/watt than Via's solutions.
Atom destroys the C7 in performance/watt. The C7 has a relatively poor memory architecture and has piss-poor SSE performance in comparison to the Atom. The C7 also cannot match the low consumption of the Atom at a similar clock speed.
If you really think the C7 has good performance/watt, just see how a DESKTOP Intel Celeron keeps pace with the C7 in terms of power consumption. The power consumption is within a few watts, and the Intel chip delivers more than double the performance in some benchmarks. Never mind that the Intel board costs less than anything Via has on-offer.
The Via Nano is closer: it delivers better per-clock integer performance (2 integer units vs one on the Atom), but the two chips deliver the same per-clock SSE performance. When you consider that the Nano uses 3-10x the power of the Atom at various clock speeds, you begin to see how Via can't compete on performance/watt.
Of course, the only car I really consider ugly are those box-things. Dunno the name of them, I'm not a car guy, but they look just like large metal rectangles with a spot cut out for the hood/windshield.
Hey, don't hate the xB until you tried driving one. I admit that I don't really like the original, but I actually bought a 2008 model because of the redesign.
Really, it's a better Camry: same engine, similar curb weight, more passenger space, less cost, all without the grandpa-inspired handling of the Camry.
And yes, you will get poor gas mileage if all you do is drive city, but so does the Camry. If you drive mixed like I do, you'll get good mileage (despite the boxy frame). My combined mileage is 27 mpg (and I drive FAST), which is about average for the performance level of the xB. The older model gets even better mileage, but you don't want to try driving a box around on a 1.6l engine.
Personally, I find it a bit hard to believe that a civilization is smart enough to travel interstellar distances but too stupid to use basic camouflage.
Camouflage from what?
Aliens visiting us could see in completely different wavelengths. Honestly, the only reasion we see in the visible range is because it's the most intense light that passes through our atmosphere. If we had a different atmosphere or a different sun, you bet we'd see in a different range. It wouldn't be the first time that a major assumption was made about alien species.
And speaking of major assumptions, what if the aliens can't see? Would they even conceive of the need for visual camouflage?
The fact is, most Bluetooth headsets are Class 2 devices, which have a maximum power of 2.5 mW. This is orders of magnitude less than the emissions from a cell phone, which can peak at 500 mW.
If the emissions from a cell phone are simply "questionable" in terms of cancer, there's no way a signal with 100x less power is. But on the flip side, the power difference between the two is so large that you COULD see them claiming cell emissions are "bad" while not seeing any problem with the much lower power emitted by Bluetooth Class 2 devices.
Yeah, I did the same thing. I made programs to solve all sorts of problems from class on my TI-82, usually while sitting there in class.
Further, some of my games made use of stuff from said classes. The Scorched Earth clone I made used randomized trig curves to build maps, and of course I used the ballstic parabola equation to plot the trajectory of the shots. I did this all with TIBASIC, and although it was slow, it was playable.
If I were to make a recommendation, I'd say get him a TI-89. It has much more horsepower than my TI-82, so you can do more in real-time than my old TI-82. It will also last him through college if he takes a technical road. BUT: you will need to help him get started, show him some of the more interesting aspects, and help him write a few of his first programs.
Does it really matter when the existing solution (water with ions) also conducts electricity quite well?
The fact is, MOST of the best conductors of heat also conduct electricity well.
Pure water is actually a poor electrical and heat conductor; what makes it popular for cooling is that it is liquid at room temperature, cheap and easy to acquire, and is only marginally corrosive (compared to other liquid coolants).
Liquid is the key: even though water is a poor thermal conductor, if you keep it constantly moving through a system, it will move more heat than just passive conduction using a solid metal. This has to do with increased rate of heat transfer when you have a larget temperature delta.
Why not both? A vertical screen for viewing text, images, data, etc. And a horizontal one for playing with widgets, data entry, and the like.
We already have this, and it's called a laptop. The horizontal plane contains manipulation tools, and the vertical plane contains feedback.
I actually don't understand all the excitement over touchscreens for large devices myself. Touchpads were actually developed to leverage the technology of touchscreens without all the drawbacks of said large screens.
Benefits of touchpads over touchscreens:
* You have much reduced arm travel, which is one of the biggest annoyances with large touchscreens. This gets worse as screens get larger.
* Your arm doesn't block the things you are looking at on the screen.
This whole love-fest over touchscreen technology isn't anything new, and the only reason why it's recently resurfaced is because touchpad recognition technology is now excellent.
Basically what I'm saying is that the iPhone et al are just touchpads with a screen attached. If you try to expand said screens, you'll run into the same problems previous touchscreens did and never solved. I don't have much hope for those problems getting resolved, so touchscreens will probably remain restricted to hand-held devices.
You know, most programmers already know how to construct state machines, and alreay create entire programs using this concept. Your ideas are not revolutionary, they only highlight the need to use asynchronous state machines over synchronous threads. Where your ideas fail is this: you want to get rid of threads completely.
Do you want to know why the programming community likes threads? There's a simple reason: state machines DO NOT SCALE. As you add more capabilities to a state machine, the number of states you have to add for each new functionality can increase geometrically. Petty soon, the number of independent state variables defining the machine gets unmanageable.
What do you do when that happens? Well, in the real world, you break up that state machine into multiple threads. You basically draw an artificial line between units that should not be separated, simply because leaving them attached is not feasable or maintainable. Yes, operations within each thread happen asynchnously, but the multiple threads are now dependent on each other, because you were forced to break-up the complex system.
In the end, you're left with the same problem we've always had: balancing threads with performance. Yes, you can build threads that are asynchronous; this has always been possible using state machines. What you cannot do is completely toss threads; threads exist to make progamming problems more accessible to humans. Unless you can take humans out of the loop, threads are here to stay.
You think your client, one of the richest men in the world, is a vigilante who likes to dress up as a bat and beat criminals to a bloody pulp with this bare hands.
Well, he is a crazy nut, but in his downtime, he hangs out with his friend Piderman.
I believe this is your sandwich!
Oh, and the movie was awesome. I never considered Batman to be a truely dark character before this film; previous directors have tried to make Batman dark, and they have failed.
(*) Fact: CPUs for quite some time are capable of decoding HD video in real time - it's the transfer (1) to/from RAM and (2) video hardware which takes most of the time.
Interesting fact, and I most certainly believe it. In fact, the marginal system memory bandwidth of most PCs is the very reason we have video cards (and now even sound cards) with local cache. We make the system memory slow because we want lots of it. We can afford to make local memories faster because there is less of it.
I have to agree that there's really no need for this kind of co-processor - dual-core CPUs can decode almost anything worthwhile in real-time these days, and even embedded GPUs can render assistance decoding video. This device has a fast local memory, but unless the results are staying local (batch processing), you're going to run into the limits of system memory bandwidth streaming the results to something that can actually use it.
I'f like to make one further note: quad cores on the desktop are already offering more horsepower than this thing anyway, and mobile dual-cores are within spitting distance:
Assumptions:
Each SPU has 2 128-bit vector units. Four SPUs means eight vector units. Runs at around 1.5 GHz.
Conroe: two 128-bit vector units per-core. Dual-core platforms have about %85 of the horsepower running at 2.6 GHz. Four cores would blow Toshiba's SPU device out of the water, assuming you aren't bandwidth-limited.
Yup, you got my vote.
Go Team Shake!
I'm sure he meant from *this generation* but it's what I instantly thought of as well.
That's exactly what I meant. You can attach all the memory you want to a video card, true, but there is a limit to how much you can conceivably use.
When you use more memory, you use more memory bandwidth; this is an undisputable fact. If you double the resolution of textures in a scene, the texture memory bandwidth you need to render that scene doubles. If you double your resolution, the amount of framebuffer writes doubles.
Since memory bandwidth is finite, you can only add so much before it becomes worthless. This is why the 2GB would be FAR MORE useful on a DDR5 card like the 4870, or else the GTX 280 with a 512-bit memory bus.
Don't get me wrong; the 4850 has just enough memory bandwidth to perform efficiently in it's current configuration, and I like it so much I bought one myself. But anything above 1GB on the 4850 will be wasted, because you cannot magically increase the bandwidth.
Heh, you don't have to tell me that we'll always need MORE MEMORY. My first 3D accelerator featured an amazing 4MB of memory! But along the same line as above, I wouldn't have put 16MB of ram on my first card, because it would have been worthless with the limited 64-bit bus.
Actually, it's pointless for FPS style games. They'll never use even a GB of that memory effectively because the games are designed around people with 512MB at the high end.
They're only doing this because DDR3 is much cheaper than the DDR5 on the 4870. A 2GB 4850 with DDR3 is cheaper than a 1GB 4870 with DDR5. Me, I can't see the value of getting a card with more than 1GB, even for future games.
The only reason I see to buy this card is maybe there are drivers optimized for professional work where the memory requirements are much higher (3D modelers and the like).
There won't be. This card is marketed as a 4850, not a FireGL, which means it won't be all that useful or professionals. Without the drivers to accelerate professional applications, the extra memory is largely useless.