Why bother buying a computer motherboard, cpu and case? Maybe the case to store it in, but you could make a full fledged computer with just a graphics card they are so powerful.
GPU's vs CPU's is a bit like having 5000 highly trained monkeys vs 5 highly trained people. If your task is easy enough for the GPU, it'll do it blazingly fast. On the other hand, for some tasks the CPU is still the better option.
No, it shows that people (and sadly many scientists) don't understand their statistics. It showed a central tendency when looking at the geometric averages of faces. Of course higher sample sizes would have less variance, too bad the analysis didn't take this into account. Psychologists in particular need to start looking a little deeper at their SPSS results.
Interestingly, in the latest number of American Scientist there's an article which highlights this issue, using the study by Kanazawa (from TFA: "This builds on previous work by Satoshi Kanazawa"). In the article they calculate the 95 percent confidence interval based on the data used by Kanazawa. Kanazawa reported that attractive parents were 4.7% more likely to have girls, however the confidence interval reveals that estimates as low as -3.9% and as high as 13.3% are consistent with the data.
Even simpler mathematical analysis would include such techniques as seeing which one takes up more disk space. Last I checked, that was very highly correlated with compression level.
The problem is that there are many choices left to the compression program which affect the quality/size trade-off. A high quality compression program might generate optimized quantization tables for that specific image, resulting in a superior image at lower bitrate compared to say the standard libjpeg implementation.
Excuse me. Java pre-dates VB by a considerable amount. You haven't been around long either have you?
The initial project that became Java started in 1991, the same year Visual Basic 1.0 was released. Java was released to the public in 1995, the same year Visual Basic 4.0 was released. How do you figure that Java predates VB by a considerable amount?
OK, maybe what I did was simple enough, but I just don't see what's so inherently hard about parellel programming. Surely I am missing something.
For me, the two things that are hardest are designing an efficient parallel algorithm for the target platform and ensuring fast but proper synchronization.
For instance, if your target is a GPU, then you have a bunch of execution units in parallel, but communication between them is limited. You have to take this into consideration when designing the algorithm.
If your target is regular CPU's, then you might have a handful for execution units and communication can be fast. However you need to ensure proper synchronization. Locks can be very expensive on some platforms, and so you want to reduce the use of them, especially in highly congested parts of your application. However this can be very difficult to do correctly, so that there aren't any race conditions or data corruption.
In general, using threads can be very easy however for all but the most trivial issues it can be a bit tricky to do it efficiently, so that it's actually worthwhile to use threads. Imho.
Interesting turnaround. The original motivation for the GPU was to allow the CPU to offload expensive graphics computation to a dedicated processor. Now it appears that that newer GPUs are allowed to offload their computation back to the CPU again.
No, that is not what's going on. The GPUs have become so fast at doing their jobs that the CPU can't feed them fast enough. That's where the new features in DX11 will help. It will make it possible to efficiently use multiple threads to feed the GPUs. This has been an issue in DX10 and earlier.
I read some while ago that the LHC was the first particle accelerator powerful enough to basically destroy itself if the beam was dumped directly into the walls.
The energy stored in the entire beam will be around 350MJ, which, if I did the conversion correctly, is equivalent to about 83kg TNT. Of course it won't be able to dump all of it in an instant (at least not in the same location), but I imagine it could still be quite destructive if it fails.
[...]restart[...]restart[...]restart[...]restart[...]restart[...]restart[...]restart. That's what takes hours when building a PC.
You know that in the vast majority of cases, there's no problem just saying "no" when the installer asks you to restart? I install most of my drivers in one session, THEN restart. Most programs in one session etc.
But in any case, it does take time to set up the PC from scratch after you've built it.
Except that power companies charge by the unit. So do water companies.
Here in Norway, if your household has main fuses above a certain rating (around 100A afaik) or the yearly use exceeds some limit, you'll have to pay per kW as well as per kWh.
IBM modified Deep Blue after Kasparov won his first match. This meant that, after finally learning how Deep Blue played (on the fly!), Kasparov basically had to play a brand new opponent.
What's unfair with radically changing the playing style between two matches? If the opponent can't reliably predict your moves, doesn't that work to your advantage? I can't see how there's anything preventing Kasparov from doing the same.
For example your teammates in GRAW1 are a bunch of idiots. Occasionally handy but mostly dead because they wandered into the line of fire.
You're talking about the XBox version or something? Cause on the PC, GRAW2 team AI is significantly worse. At least in GRAW1 they would correctly identify incoming fire and return it. In GRAW2 there's countless times where I've seen them sitting still in the open, facing the wrong direction, yelling "taking fire" until they die.
Story from today. My dad calls me because his anti-virus internet protection gizmo thingy on his computer (F-Secure, required by his employer) asked if "tunnel service" could be allowed to connect to the internet.
I was on my way home, so couldn't look it up. I asked where it was located, and it was in something like \Windows\System32\Downloads. Now the "Downloads" directory made me suspicious.
Long story short, IT guys at his work later confirmed that it was safe (part of their VPN/Citrix solution or something).
Now, I consider myself at least an average power user, having programmed for windows since the win3.11 days. Even I had a hard time deciding if this was a "good" or "bad" application.
Now put yourself in the shoes of someone who uses a computer to get things done. Like most people use their cars. How are they to determine if "tunnel service" is good or bad? And how to determine if it's the "good" "tunnel service" and not a trojan with the same name?
Immediate mode is bad because it limits the driver from performing certain optimizations. The driver can't know that you always send it the same static vertices, so it can't, for instance, compile the shebang into an optimized triangle list which is stored in video RAM, eliminating the upload of the vertex data every frame. Instead you should use display lists or Vertex Buffer Objects (VBO).
Fixed function is bad because almost none (all?) of the current hardware works like that any more. They will instead convert the fixed function stuff to shaders and use that. Afaik that will also limit the stuff the driver can optimize, and may result in a lot of unnecessary computations. Instead, use vertex and pixel shaders.
Though of the two, immediate mode is by far the worst.
I'm working on a physically-based renderer, and I absolutely understand what you mean about realistic vs real. We find we have to insert additional light sources etc to make a good image, just as you would in a professional photo shoot.
My experience is that if you set a programs ( notepad for example ) affinity to the second core and then set it to 'realtime', Windows slows to an agonising crawl with the first core usage at 0% and the second at 100%.
You should almost never use REALTIME_PRIORITY_CLASS, because this interrupts system threads that manage mouse input, keyboard input, and background disk flushing. This class can be appropriate for applications that "talk" directly to hardware or that perform brief tasks that should have limited interruptions.
Looks like a recipe for failure to me. What's the secret?
I know that for high performance RC cars and such, you can buy "matched packs", where they've matched the cells in the pack based on their behavior (discharge rate, peak voltage or something like that). For such a huge setup as this, I wouldn't be surprised if they've done something similar.
So how exactly does that contradict what I said. It seems like WARP10 is just the DX10 version of the reference rasterizer.
You said that "by far the more useful thing is that it's probably better for development because the driver developers will have a reference point of how the graphics are supposed to render".
I see WARP as more useful for developing actual applications than for debugging drivers.
Also, larger game companies will be able to point out these differences to get bug fixes out of the graphics card companies. "Your graphics card renders this incorrectly with regards to the reference, fix it" is much more forceful than "your graphics card behaves differently than your competitor".
DirectX already contains a reference rasterizer, which is better suited for that. This thing seems instead to be meant for applications that doesn't necessarily need more than "interactive" frame rates, but do need to run on a broad class of machines. Or for easing development of applications which could benefit from hardware acceleration when available (image processing f.i.).
We don't see WARP10 as a replacement for graphics hardware, particularly as reasonably performing low end Direct3D 10 discrete hardware is now available for under $25. The goal of WARP10 was to allow applications to target Direct3D 10 level hardware without having significantly different code paths or testing requirements when running on hardware or when running in software.
And yes, I could try to do CSS surgery, but it would be a fair amount of work to figure out what to override.
Depends on your browser I guess... In Opera you could simply go "View -> Style -> High Contrast (B/W)" and everything is turned into readable black on white.
I figured they were just an updated binary format, just like every other Office upgrade
At least now know why it takes ages to save a document compared to the old format (this can be especially noticable when working on image heavy presentations).
...to finally share proper doc of the old standards. This just means they feel confident that MS Office 2007 will take firm enough root to ensure that the old game of catch up for FOSS projects will stay the same.
TFA says 5n-5, but I don't get it because if n=1, then zero stars would be seen. The relevant part comes after that:
... all they could show was that such a function couldn't have more than 5n-5 solutions, if n was bigger than one. In other words, their formula simply does not apply for the case when n = 1.
Slashdot Mirror
Small correction: I should have said fossil fuel instead of just oil.
how do you know the CO2 rise is man made in the first place, and not oh say from the oceans which are the largest stores of CO2?
IIRC it's from the ratio of carbon isotopes. The ratio found in oil being different from the ratio in CO2 produced by living things.
Why bother buying a computer motherboard, cpu and case? Maybe the case to store it in, but you could make a full fledged computer with just a graphics card they are so powerful.
GPU's vs CPU's is a bit like having 5000 highly trained monkeys vs 5 highly trained people. If your task is easy enough for the GPU, it'll do it blazingly fast. On the other hand, for some tasks the CPU is still the better option.
No, it shows that people (and sadly many scientists) don't understand their statistics. It showed a central tendency when looking at the geometric averages of faces. Of course higher sample sizes would have less variance, too bad the analysis didn't take this into account. Psychologists in particular need to start looking a little deeper at their SPSS results.
Interestingly, in the latest number of American Scientist there's an article which highlights this issue, using the study by Kanazawa (from TFA: "This builds on previous work by Satoshi Kanazawa"). In the article they calculate the 95 percent confidence interval based on the data used by Kanazawa. Kanazawa reported that attractive parents were 4.7% more likely to have girls, however the confidence interval reveals that estimates as low as -3.9% and as high as 13.3% are consistent with the data.
Even simpler mathematical analysis would include such techniques as seeing which one takes up more disk space. Last I checked, that was very highly correlated with compression level.
The problem is that there are many choices left to the compression program which affect the quality/size trade-off. A high quality compression program might generate optimized quantization tables for that specific image, resulting in a superior image at lower bitrate compared to say the standard libjpeg implementation.
Excuse me. Java pre-dates VB by a considerable amount. You haven't been around long either have you?
The initial project that became Java started in 1991, the same year Visual Basic 1.0 was released. Java was released to the public in 1995, the same year Visual Basic 4.0 was released. How do you figure that Java predates VB by a considerable amount?
OK, maybe what I did was simple enough, but I just don't see what's so inherently hard about parellel programming. Surely I am missing something.
For me, the two things that are hardest are designing an efficient parallel algorithm for the target platform and ensuring fast but proper synchronization.
For instance, if your target is a GPU, then you have a bunch of execution units in parallel, but communication between them is limited. You have to take this into consideration when designing the algorithm.
If your target is regular CPU's, then you might have a handful for execution units and communication can be fast. However you need to ensure proper synchronization. Locks can be very expensive on some platforms, and so you want to reduce the use of them, especially in highly congested parts of your application. However this can be very difficult to do correctly, so that there aren't any race conditions or data corruption.
In general, using threads can be very easy however for all but the most trivial issues it can be a bit tricky to do it efficiently, so that it's actually worthwhile to use threads. Imho.
Interesting turnaround. The original motivation for the GPU was to allow the CPU to offload expensive graphics computation to a dedicated processor. Now it appears that that newer GPUs are allowed to offload their computation back to the CPU again.
No, that is not what's going on. The GPUs have become so fast at doing their jobs that the CPU can't feed them fast enough. That's where the new features in DX11 will help. It will make it possible to efficiently use multiple threads to feed the GPUs. This has been an issue in DX10 and earlier.
I read some while ago that the LHC was the first particle accelerator powerful enough to basically destroy itself if the beam was dumped directly into the walls.
The energy stored in the entire beam will be around 350MJ, which, if I did the conversion correctly, is equivalent to about 83kg TNT. Of course it won't be able to dump all of it in an instant (at least not in the same location), but I imagine it could still be quite destructive if it fails.
[...]restart[...]restart[...]restart[...]restart[...]restart[...]restart[...]restart. That's what takes hours when building a PC.
You know that in the vast majority of cases, there's no problem just saying "no" when the installer asks you to restart? I install most of my drivers in one session, THEN restart. Most programs in one session etc.
But in any case, it does take time to set up the PC from scratch after you've built it.
Except that power companies charge by the unit. So do water companies.
Here in Norway, if your household has main fuses above a certain rating (around 100A afaik) or the yearly use exceeds some limit, you'll have to pay per kW as well as per kWh.
IBM modified Deep Blue after Kasparov won his first match. This meant that, after finally learning how Deep Blue played (on the fly!), Kasparov basically had to play a brand new opponent.
What's unfair with radically changing the playing style between two matches? If the opponent can't reliably predict your moves, doesn't that work to your advantage? I can't see how there's anything preventing Kasparov from doing the same.
For example your teammates in GRAW1 are a bunch of idiots. Occasionally handy but mostly dead because they wandered into the line of fire.
You're talking about the XBox version or something? Cause on the PC, GRAW2 team AI is significantly worse. At least in GRAW1 they would correctly identify incoming fire and return it. In GRAW2 there's countless times where I've seen them sitting still in the open, facing the wrong direction, yelling "taking fire" until they die.
Have a brain when using the PC.
It works for all operating systems.
Story from today. My dad calls me because his anti-virus internet protection gizmo thingy on his computer (F-Secure, required by his employer) asked if "tunnel service" could be allowed to connect to the internet.
I was on my way home, so couldn't look it up. I asked where it was located, and it was in something like \Windows\System32\Downloads. Now the "Downloads" directory made me suspicious.
Long story short, IT guys at his work later confirmed that it was safe (part of their VPN/Citrix solution or something).
Now, I consider myself at least an average power user, having programmed for windows since the win3.11 days. Even I had a hard time deciding if this was a "good" or "bad" application.
Now put yourself in the shoes of someone who uses a computer to get things done. Like most people use their cars. How are they to determine if "tunnel service" is good or bad? And how to determine if it's the "good" "tunnel service" and not a trojan with the same name?
Immediate mode is bad because it limits the driver from performing certain optimizations. The driver can't know that you always send it the same static vertices, so it can't, for instance, compile the shebang into an optimized triangle list which is stored in video RAM, eliminating the upload of the vertex data every frame. Instead you should use display lists or Vertex Buffer Objects (VBO).
Fixed function is bad because almost none (all?) of the current hardware works like that any more. They will instead convert the fixed function stuff to shaders and use that. Afaik that will also limit the stuff the driver can optimize, and may result in a lot of unnecessary computations. Instead, use vertex and pixel shaders.
Though of the two, immediate mode is by far the worst.
Interesting, thank's for the reply.
I'm working on a physically-based renderer, and I absolutely understand what you mean about realistic vs real. We find we have to insert additional light sources etc to make a good image, just as you would in a professional photo shoot.
Of the rendering jobs we submit, approximately 0.001% use raytracing exclusively, about 0.5% make use of raytracing extensions.
Are you using radiosity for GI, or are you not using any GI for the remaining 99.499%?
My experience is that if you set a programs ( notepad for example ) affinity to the second core and then set it to 'realtime', Windows slows to an agonising crawl with the first core usage at 0% and the second at 100%.
From MSDN:
You should almost never use REALTIME_PRIORITY_CLASS, because this interrupts system threads that manage mouse input, keyboard input, and background disk flushing. This class can be appropriate for applications that "talk" directly to hardware or that perform brief tasks that should have limited interruptions.
Looks like a recipe for failure to me. What's the secret?
I know that for high performance RC cars and such, you can buy "matched packs", where they've matched the cells in the pack based on their behavior (discharge rate, peak voltage or something like that). For such a huge setup as this, I wouldn't be surprised if they've done something similar.
So how exactly does that contradict what I said. It seems like WARP10 is just the DX10 version of the reference rasterizer.
You said that "by far the more useful thing is that it's probably better for development because the driver developers will have a reference point of how the graphics are supposed to render".
I see WARP as more useful for developing actual applications than for debugging drivers.
Also, larger game companies will be able to point out these differences to get bug fixes out of the graphics card companies. "Your graphics card renders this incorrectly with regards to the reference, fix it" is much more forceful than "your graphics card behaves differently than your competitor".
DirectX already contains a reference rasterizer, which is better suited for that. This thing seems instead to be meant for applications that doesn't necessarily need more than "interactive" frame rates, but do need to run on a broad class of machines. Or for easing development of applications which could benefit from hardware acceleration when available (image processing f.i.).
From the MSDN page on WARP:
We don't see WARP10 as a replacement for graphics hardware, particularly as reasonably performing low end Direct3D 10 discrete hardware is now available for under $25. The goal of WARP10 was to allow applications to target Direct3D 10 level hardware without having significantly different code paths or testing requirements when running on hardware or when running in software.
And yes, I could try to do CSS surgery, but it would be a fair amount of work to figure out what to override.
Depends on your browser I guess... In Opera you could simply go "View -> Style -> High Contrast (B/W)" and everything is turned into readable black on white.
I figured they were just an updated binary format, just like every other Office upgrade
At least now know why it takes ages to save a document compared to the old format (this can be especially noticable when working on image heavy presentations).
...to finally share proper doc of the old standards. This just means they feel confident that MS Office 2007 will take firm enough root to ensure that the old game of catch up for FOSS projects will stay the same.
I guess that whole ISO voting stuff on OOXML just passed you by?
... all they could show was that such a function couldn't have more than 5n-5 solutions, if n was bigger than one. In other words, their formula simply does not apply for the case when n = 1.