The PS2 was maybe oversimplified as it is a complexe thing with a lot of different execution units, call it the main bus being hoged by texture transfers and the GS being stalled because it is waiting for texture upload to finish.
Or are you talking about game designers should just design their games around the limitations of the platform and just use a smaller amount of texture for better fillrate etc.? That would maybe help on the paper but very likely it will look crappier than with better texture and some lost fillrate.
If it was something else: explain! otherwise it is just a worthless flaimbait.
A better development kit isn't going to fix that. PS2 developers can code in C++ and a higher level graphic sdk, but that isn't going to get them anywhere near good performance.
The problem is the very ambitious architecture of the PS2. The GS (graphic synthesizer) got just 4mb of very fast ram. While that enabled Sony to have extremely high theoretical fill rate by embeding the RAM into the GS and connecting it with a 2560-Bit bus, it is also not nearly enough to store all the textures and the framebuffer. That results in the PS2 having to spend a lot of time transfering textures between GS and regular RAM. Because changing out the texture takes a lot of time, you need to order your triangles in a way that minimizes the texture changes, which is a lot of trouble and hurts performance for sure. The PS2 EE (the main CPU) also got just 16kb cache, which is clearly not enough. Memory access to stuff not in the cache is extremely expensive and the Rambus RAM with its high-bandwidth but also high-latency access profile isn't going to help.
Because of that a PS2 coder needs to spend a lot of time on optimizing algorithms for ordered local data access and rewriting stuff in assembler to be able to fit the whole routine into the cache.
A interessting document from Sony about PS2 performance is here: (PDF only sorry) http://www.scee.sony.co.uk/sceesite/files/presenta tions/PSP/HowFarHaveWeGot.pdf
While marketing said 66 million polys/second, even after all these years the fastest real world Sony seems to know about is 125k polys @ 60 Hz, which translates into 7.5 million polys/second while the average recent game seems to do just 3 million polys/s
Better SDKs aren't able to help here. The problem are hardware limitations. And while the hard-to-optimize-for design will sure enable programmers to squeeze out quite a bit of additional performance, but it will never be able to reach the real-world performance of XBox and Gamecube.
And Sony even has better DevKits now, but as you can see their feature isn't C++ or something similiar to DirectX but instead tools to analyse how the cpus is stalled by cache misses etc.
Imho the PS2 is similiar in design to the first Pentium 4, ambitious, marketing-driven design with very high theoretical peak performance but low real world performance.
BTW: Gamecubes marketing is exactly the opposite, Nintendo claimed 7-12 million polys/second while one of their launch games 'Rogue Leader' was pushing 15 million polys/second in some scenes.
Sure, they can still relicense it. They just need permission from the authors of all contributed code parts or need to remove the parts where they couldn't get permission. I think there are some company driven GPL'ed projects where they will only accept contributions to the main code base where the authors grants them permission to relicense the contributed code.
Of course if really important and substantial parts were contributed in a early development phase, that could be nearly impossible to do without a nearly complete rewrite because a lot of code is based on the contributed code.
If it were different, it would cause a lot of problems. If you now contribute to a GPL'ed project, you know your work will always stay free. If the original author could just change the license at will, he could change that at will too. He could try to earn money with your code and you couldn't do anything legally to stop him. If you are fine with that, you could just use the BSD license and even grant that right to everyone. But if you contributed to a GPL'ed project because you like the ideas behind the GPL, you would likely only contribute to a small selected list of projects where you are sure the project leader won't do anything stupid like that. (e.g.: linux kernel)
"Socialism produces worse living conditions than a free market."
If you are talking about socialism/communism like in the soviet union, yes. But if you look at western european social market economies, the answer is no. While a completely free market may produces better living conditions in the average, it also causes a bigger variance than social market economies. IMHO preventing extreme poverty is worth slightly worser living conditions for everyone else.
TCO depends a lot on wages vs. hardware/software costs. If the wages are high, it can make sense to spend more money on hardware/software if that causes less downtime and less work for the adminstration. But if the wages are low, work-intensive solutions with slightly more downtime can become acceptable and have a lower TCO in the end. In countries like brazil I wouldn't be so sure if the TCO of a LTSP is really lower.
Not unlikely hardware is more expensive than in the US and the wages are a lot lower for sure. And students are working on these PCs, so downtime is almost free. I believe such a four head solution also provides better response than a LTSP installation. Video playback and similiar stuff should be possible on a four head installation.
It puts 10 MP (specifically a cannon, which may be important, as lense and CCD design do have profound effects on the digital camera) as being the superior to 35mm film in every possible catagory, hence a 20 MP camera like this one fujifilm camera> would outperform a 70mm film in every possible catagory
You obviously can't calculate. 70mm film got four times the area of 35mm film so you would need a 40 MP camera to beat 70mm film by the same relative margin.
Swapping can easily kill the framerate more than a slow videocard. Especially when a game is playable most of the time, but the framerate goes down to single digit numbers from time to time not enough ram is often the problem.
Trying to fix swapping with a faster HDD is obviously stupid, more ram is needed.
Asymmetrical burdens on encoder and decoder are not related to the transform used. DCT and inverse DCT need the same amount of calculations and most wavelet transform implementation have symmetrical performance, too.
Asymmetric performance is mostly caused by motion estamination. The encoder needs to find motion vectors for every macroblock while the decoder just reads them from the bitstream. Finding good motion vectors is quite hard and is causing most of the asymmetry between encoder and decoder. But motion compensation is used for nearly all good video codecs no matter if DCT, Wavelet or another transform is used.
DCT is only really good if you want to code periodic signals, but most picture signals aren't periodic, so DCT isn't really optimal. But because of the periodic nature of the DCT a lot of calculation time can be saved by exploiting symmetries in the transform matrix. A brute force 8x8 DCT would require 1024 multiplications, while a optimized algorithms needs less than the half number. On 8x8 blocks DCT is much faster than a good wavelet transform. (There is the haar wavelet transform, which is faster, but also so bad that it is useless for compression)
So a decent wavelet transform needs more calculation power. But wavelet transform can be done in O(n) (or O(n^2) for 2D transforms) while DCT scales with O(n*log(n)). For that reason DCT always needs to operate on small blocks for decent performance while wavelet doesn't take a performance hit from operating on the whole image and is able to use that for better compression.
Wavelet transforms need more processing power than DCT, and this processing power is needed both in encoder and decoder. The performance would still be asymmetrical but the relative difference between encoder and decoder performance would shrink while the absolute performance difference between them would stay about the same.
I have often seen the same behavoir but I think it just shows how we need to optimize GUIs to be more accessible to the general public.
Appearantly many people think in tasks and want one programm for each task. We need a gui where the users are easily able to find which programm is able to do what they want and guide them how to use it.
Good post, but for the record you can get legitimate 26-episode series for under $100, just not really new and mainstream stuff like GitS:SAC. See ADV's Princess Nine Boxset for an example of really cheap legimate anime, it is sold for about $40 at several online retailers for the full 26 episode series, otoh it is older and not mainstream as it is a series about Japanese high school girls trying to play in the boys baseball league.
Official and legimate release are also sometimes regionfree, for example ADV's release of Neon Genesis Evangelion or CPM's release of "Now and Then, Here and There".
4 Disc for a 26 episode series is also sometimes done, e.g.: Fruits Basket by Funimation or Blue Seed by ADV.
Deciding if something is bootleged isn't that easy. Regionfree, low disc count and bad english all together is a almost 100% secure check for bootlegs, but some legitimate release are mistoken as a bootleg if you think one fulfilled criteria is enough. If you want to check for bootlegs check for chinese subs and regionfree, that combination is a 99% secure criteria or try if you can find the DVD in a known to be legimate shop. (Amazon isn't if you include external sellers)
Sony and Pioneer Players can be made codefree too
on
The Borg MegaCube
·
· Score: 4, Informative
Sony and Pioneer Player can made codefree with a new firmware. The codefree Sony firmware also disables User Prohibitions, so you can skip to the next chapter whenever you want and you can always change subtitels, audio etc.
Many recent Pioneer Players also can be made codefree with a code entered with the remote control. Sony and Pioneer are both manufacturing really nice dvd players and many retailers here(Germany) are offering them preloaded with a codefree firmware, so there is really no reason to avoid these players. You can get almost every player in a codefree version here, no matter which brand.
In the the very nice anime series "Serial Experiments Lain" there is a designer drug like nano-machine based "Accela" substance. Accela causes a change of consciousness and seems to connect people to the "Wired", a huge omnipresent network, without additional tech.
You could build a flexible limiter. For most situations you only need to drive over the limit for a very short time. The limiter could allow you to drive over the limit for 30 seconds and start limiting if you drive too fast for a longer time.
If you need to drive an injured person to the hospital or something like that, there could be an emergency switch that disables the limiter completely. After that emergency you would need to get the limiter reenabled by the local police and you would get a ticket if your emergency was something like oversleept and didn't want to be late.
PS(2) games are also region coded and you need a modchip to play back US or Japan games on a australian PS(2). Many games are never released in PAL regions and many gamers there got modchips to play US games because they are released much sooner and are usually better. (bad ntsc->pal conversation, borders, wrong speed etc.)
Using regioncoding they can still sell shitty outdated, overpriced games in Europe and Australia.
I would recommend pyDance or Stepmania for exercise in a fun, hasslefree and open source flavor. You just need a dance mat and a PSX2PC adapter to start. You can do it at home, you can start on a slow and easy level and get better while seeing the success in your scores and a half hour can easily get your shirt soaked with sweat.
Most (it used to be all) fansubbers sub stuff that hasn't been licensed in the state where the fansubbers is living.(usually usa) Harry Potter is licensed and even a publishing date is availiable.
Linux has had Kerberos and IPSec for long time, too. This is not a Linux vs. Windows thing, it is about the best way to use Kerberos authentification for SSH.
Why you are so sure it is a bluff ? You are right, it is very unlikely that MS releases an official bootloader. So if they want even a small chance of success, they need a real threat. Just bluffing around won't work, but if they can demonstrate their exploit to MS it could might work. If their exploit is working, MS wouldn't have much to loose, cause if the exploit is released everyone can run Linux and pirated games on their Xbox. If they release a Linux bootloader they might can keep the exploit secert. They could also sell a Linux Kit like Sony is doing and keep things under control. (For example register consoles with linux bootloader and blacklist them in future games and xbox live, cause it isn't unlikely that someone finds a way to use a linux bootloader to boot pirated games)
While many people tried to develop Go software using neural networks, the strongest go playing software are still NOT (mainly) neural network based.
Neural networks are often able to choose very good moves in situations where the other expert system based Go softwares are only able to generate mediocre to bad moves, but neural network go playing software often completely fails to see the right move where it is obvious. (for humans and regular go software) Click this link for a computer go software ladder.
And here is the important tidbit:
Interestingly, for the first time ever, a neural network program has had some success, with Neuro Go coming a creditable 6th.
They had already hit the 2 billion by now if AMD hadn't appeared in the scene as it has.
Why ? AMD hasn't shipped a billion x86 chips yet, so even combined they wouldn't have shipped two billion chips yet. And without competition from AMD their CPUs would still be a lot more expensive and a bit slower, not exactly something that lets you sell more CPUs.
If you're into the whole anime thing, like I am, Bit Torrent is a godsend. BitTorrent is the biggest thing to happen to digital fansubs since DivX.
It is a big thing to happen to digital fansubs, but it isn't a godsend. Fansubs were supposed to be only available to fanboys/girls, that have watched most commercially available anime and want to watch other anime that is unlikely to get commercially available. Now most anime are licensed (but not announce) before they air in Japan. So most fansubs now are just an excuse for piracy.
Bittorrent is pushing into same direction. Bittorrent works really well for really popular stuff, that is very likely to get a US DVD release in the next year. But it works only mediocre for stuff that isn't that popular like very old anime or stuff like Mermaid Melody. Old shows or shows for childern are where fansubbing still makes some sense.
So Bittorrent only helps fansubs where they aren't needed anymore and are contraproductive.
Cg is only for SHADERS not for the real program
on
The Cg Tutorial
·
· Score: 1
Cg is a high level shader language. You only use it for shaders not for your real program. You write your stuff in C, C++ or every other language where you got OpenGL or Direct3D bindings. Cg is only used for shaders, tiny little specialiced code fragments that run directly on the graphic card and manipulate vertexes (vertex shader) or shade the inside of the triangles in new ways. (pixel shaders)
Before Nvidia created Cg, the only way creating new shaders was using low-level assembly like language. With Cg you can write shaders in a high-level c like language. Using real C(++) for shaders wouldn't work cause the video card isn't a full featured cpu.
There isn't any "raw" opengl interface under Linux, so you will always need to run two x servers. A quite ugly solution in my opionion.
This work should be done at QT and GTK+ level. Overhead would be much smaller that way. I can only see some advantages of the double X server solution if you want to run remote x clients.
Slashdot Mirror
The PS2 was maybe oversimplified as it is a complexe thing with a lot of different execution units, call it the main bus being hoged by texture transfers and the GS being stalled because it is waiting for texture upload to finish.
Or are you talking about game designers should just design their games around the limitations of the platform and just use a smaller amount of texture for better fillrate etc.? That would maybe help on the paper but very likely it will look crappier than with better texture and some lost fillrate.
If it was something else: explain! otherwise it is just a worthless flaimbait.
A better development kit isn't going to fix that. PS2 developers can code in C++ and a higher level graphic sdk, but that isn't going to get them anywhere near good performance.
a tions/PSP/HowFarHaveWeGot.pdf
The problem is the very ambitious architecture of the PS2. The GS (graphic synthesizer) got just 4mb of very fast ram. While that enabled Sony to have extremely high theoretical fill rate by embeding the RAM into the GS and connecting it with a 2560-Bit bus, it is also not nearly enough to store all the textures and the framebuffer. That results in the PS2 having to spend a lot of time transfering textures between GS and regular RAM. Because changing out the texture takes a lot of time, you need to order your triangles in a way that minimizes the texture changes, which is a lot of trouble and hurts performance for sure. The PS2 EE (the main CPU) also got just 16kb cache, which is clearly not enough. Memory access to stuff not in the cache is extremely expensive and the Rambus RAM with its high-bandwidth but also high-latency access profile isn't going to help. Because of that a PS2 coder needs to spend a lot of time on optimizing algorithms for ordered local data access and rewriting stuff in assembler to be able to fit the whole routine into the cache.
A interessting document from Sony about PS2 performance is here: (PDF only sorry)
http://www.scee.sony.co.uk/sceesite/files/present
While marketing said 66 million polys/second, even after all these years the fastest real world Sony seems to know about is 125k polys @ 60 Hz, which translates into 7.5 million polys/second while the average recent game seems to do just 3 million polys/s
Better SDKs aren't able to help here. The problem are hardware limitations. And while the hard-to-optimize-for design will sure enable programmers to squeeze out quite a bit of additional performance, but it will never be able to reach the real-world performance of XBox and Gamecube.
And Sony even has better DevKits now, but as you can see their feature isn't C++ or something similiar to DirectX but instead tools to analyse how the cpus is stalled by cache misses etc.
Imho the PS2 is similiar in design to the first Pentium 4, ambitious, marketing-driven design with very high theoretical peak performance but low real world performance.
BTW: Gamecubes marketing is exactly the opposite, Nintendo claimed 7-12 million polys/second while one of their launch games 'Rogue Leader' was pushing 15 million polys/second in some scenes.
Sure, they can still relicense it. They just need permission from the authors of all contributed code parts or need to remove the parts where they couldn't get permission. I think there are some company driven GPL'ed projects where they will only accept contributions to the main code base where the authors grants them permission to relicense the contributed code.
Of course if really important and substantial parts were contributed in a early development phase, that could be nearly impossible to do without a nearly complete rewrite because a lot of code is based on the contributed code.
If it were different, it would cause a lot of problems. If you now contribute to a GPL'ed project, you know your work will always stay free. If the original author could just change the license at will, he could change that at will too. He could try to earn money with your code and you couldn't do anything legally to stop him. If you are fine with that, you could just use the BSD license and even grant that right to everyone. But if you contributed to a GPL'ed project because you like the ideas behind the GPL, you would likely only contribute to a small selected list of projects where you are sure the project leader won't do anything stupid like that. (e.g.: linux kernel)
It was a parody of the mastercard commercials and the mastercard commercials got unrealistic prices, too.
While a completely free market may produces better living conditions in the average, it also causes a bigger variance than social market economies. IMHO preventing extreme poverty is worth slightly worser living conditions for everyone else.
And the Geforce4ti is also faster than the low end gf fx cards. There is really no good reason to get a low end fx.
TCO depends a lot on wages vs. hardware/software costs. If the wages are high, it can make sense to spend more money on hardware/software if that causes less downtime and less work for the adminstration. But if the wages are low, work-intensive solutions with slightly more downtime can become acceptable and have a lower TCO in the end. In countries like brazil I wouldn't be so sure if the TCO of a LTSP is really lower.
Not unlikely hardware is more expensive than in the US and the wages are a lot lower for sure. And students are working on these PCs, so downtime is almost free. I believe such a four head solution also provides better response than a LTSP installation. Video playback and similiar stuff should be possible on a four head installation.
You obviously can't calculate. 70mm film got four times the area of 35mm film so you would need a 40 MP camera to beat 70mm film by the same relative margin.
Swapping can easily kill the framerate more than a slow videocard. Especially when a game is playable most of the time, but the framerate goes down to single digit numbers from time to time not enough ram is often the problem.
Trying to fix swapping with a faster HDD is obviously stupid, more ram is needed.
Asymmetrical burdens on encoder and decoder are not related to the transform used. DCT and inverse DCT need the same amount of calculations and most wavelet transform implementation have symmetrical performance, too.
Asymmetric performance is mostly caused by motion estamination. The encoder needs to find motion vectors for every macroblock while the decoder just reads them from the bitstream. Finding good motion vectors is quite hard and is causing most of the asymmetry between encoder and decoder. But motion compensation is used for nearly all good video codecs no matter if DCT, Wavelet or another transform is used.
DCT is only really good if you want to code periodic signals, but most picture signals aren't periodic, so DCT isn't really optimal. But because of the periodic nature of the DCT a lot of calculation time can be saved by exploiting symmetries in the transform matrix. A brute force 8x8 DCT would require 1024 multiplications, while a optimized algorithms needs less than the half number. On 8x8 blocks DCT is much faster than a good wavelet transform. (There is the haar wavelet transform, which is faster, but also so bad that it is useless for compression)
So a decent wavelet transform needs more calculation power. But wavelet transform can be done in O(n) (or O(n^2) for 2D transforms) while DCT scales with O(n*log(n)). For that reason DCT always needs to operate on small blocks for decent performance while wavelet doesn't take a performance hit from operating on the whole image and is able to use that for better compression.
Wavelet transforms need more processing power than DCT, and this processing power is needed both in encoder and decoder. The performance would still be asymmetrical but the relative difference between encoder and decoder performance would shrink while the absolute performance difference between them would stay about the same.
I have often seen the same behavoir but I think it just shows how we need to optimize GUIs to be more accessible to the general public.
Appearantly many people think in tasks and want one programm for each task. We need a gui where the users are easily able to find which programm is able to do what they want and guide them how to use it.
Good post, but for the record you can get legitimate 26-episode series for under $100, just not really new and mainstream stuff like GitS:SAC. See ADV's Princess Nine Boxset for an example of really cheap legimate anime, it is sold for about $40 at several online retailers for the full 26 episode series, otoh it is older and not mainstream as it is a series about Japanese high school girls trying to play in the boys baseball league.
Official and legimate release are also sometimes regionfree, for example ADV's release of Neon Genesis Evangelion or CPM's release of "Now and Then, Here and There".
4 Disc for a 26 episode series is also sometimes done, e.g.: Fruits Basket by Funimation or Blue Seed by ADV.
Deciding if something is bootleged isn't that easy. Regionfree, low disc count and bad english all together is a almost 100% secure check for bootlegs, but some legitimate release are mistoken as a bootleg if you think one fulfilled criteria is enough. If you want to check for bootlegs check for chinese subs and regionfree, that combination is a 99% secure criteria or try if you can find the DVD in a known to be legimate shop. (Amazon isn't if you include external sellers)
Sony and Pioneer Player can made codefree with a new firmware. The codefree Sony firmware also disables User Prohibitions, so you can skip to the next chapter whenever you want and you can always change subtitels, audio etc.
Many recent Pioneer Players also can be made codefree with a code entered with the remote control. Sony and Pioneer are both manufacturing really nice dvd players and many retailers here(Germany) are offering them preloaded with a codefree firmware, so there is really no reason to avoid these players. You can get almost every player in a codefree version here, no matter which brand.
In the the very nice anime series "Serial Experiments Lain" there is a designer drug like nano-machine based "Accela" substance.
Accela causes a change of consciousness and seems to connect people to the "Wired", a huge omnipresent network, without additional tech.
You could build a flexible limiter. For most situations you only need to drive over the limit for a very short time. The limiter could allow you to drive over the limit for 30 seconds and start limiting if you drive too fast for a longer time.
If you need to drive an injured person to the hospital or something like that, there could be an emergency switch that disables the limiter completely. After that emergency you would need to get the limiter reenabled by the local police and you would get a ticket if your emergency was something like oversleept and didn't want to be late.
PS(2) games are also region coded and you need a modchip to play back US or Japan games on a australian PS(2). Many games are never released in PAL regions and many gamers there got modchips to play US games because they are released much sooner and are usually better. (bad ntsc->pal conversation, borders, wrong speed etc.)
Using regioncoding they can still sell shitty outdated, overpriced games in Europe and Australia.
I would recommend pyDance or Stepmania for exercise in a fun, hasslefree and open source flavor.
You just need a dance mat and a PSX2PC adapter to start. You can do it at home, you can start on a slow and easy level and get better while seeing the success in your scores and a half hour can easily get your shirt soaked with sweat.
There is also another difference:
Most (it used to be all) fansubbers sub stuff that hasn't been licensed in the state where the fansubbers is living.(usually usa) Harry Potter is licensed and even a publishing date is availiable.
Linux has had Kerberos and IPSec for long time, too. This is not a Linux vs. Windows thing, it is about the best way to use Kerberos authentification for SSH.
Why you are so sure it is a bluff ? You are right, it is very unlikely that MS releases an official bootloader. So if they want even a small chance of success, they need a real threat.
Just bluffing around won't work, but if they can demonstrate their exploit to MS it could might work. If their exploit is working, MS wouldn't have much to loose, cause if the exploit is released everyone can run Linux and pirated games on their Xbox. If they release a Linux bootloader they might can keep the exploit secert. They could also sell a Linux Kit like Sony is doing and keep things under control. (For example register consoles with linux bootloader and blacklist them in future games and xbox live, cause it isn't unlikely that someone finds a way to use a linux bootloader to boot pirated games)
Neural networks are often able to choose very good moves in situations where the other expert system based Go softwares are only able to generate mediocre to bad moves, but neural network go playing software often completely fails to see the right move where it is obvious. (for humans and regular go software)
Click this link for a computer go software ladder. And here is the important tidbit:
They had already hit the 2 billion by now if AMD hadn't appeared in the scene as it has.
Why ? AMD hasn't shipped a billion x86 chips yet, so even combined they wouldn't have shipped two billion chips yet. And without competition from AMD their CPUs would still be a lot more expensive and a bit slower, not exactly something that lets you sell more CPUs.
If you're into the whole anime thing, like I am, Bit Torrent is a godsend. BitTorrent is the biggest thing to happen to digital fansubs since DivX.
It is a big thing to happen to digital fansubs, but it isn't a godsend. Fansubs were supposed to be only available to fanboys/girls, that have watched most commercially available anime and want to watch other anime that is unlikely to get commercially available. Now most anime are licensed (but not announce) before they air in Japan. So most fansubs now are just an excuse for piracy.
Bittorrent is pushing into same direction. Bittorrent works really well for really popular stuff, that is very likely to get a US DVD release in the next year. But it works only mediocre for stuff that isn't that popular like very old anime or stuff like Mermaid Melody. Old shows or shows for childern are where fansubbing still makes some sense. So Bittorrent only helps fansubs where they aren't needed anymore and are contraproductive.
If you don't know about fansubs and the ideas behind it, read this very nice writeup on everything2
Cg is a high level shader language. You only use it for shaders not for your real program. You write your stuff in C, C++ or every other language where you got OpenGL or Direct3D bindings.
Cg is only used for shaders, tiny little specialiced code fragments that run directly on the graphic card and manipulate vertexes (vertex shader) or shade the inside of the triangles in new ways. (pixel shaders)
Before Nvidia created Cg, the only way creating new shaders was using low-level assembly like language. With Cg you can write shaders in a high-level c like language. Using real C(++) for shaders wouldn't work cause the video card isn't a full featured cpu.
There isn't any "raw" opengl interface under Linux, so you will always need to run two x servers. A quite ugly solution in my opionion.
This work should be done at QT and GTK+ level. Overhead would be much smaller that way. I can only see some advantages of the double X server solution if you want to run remote x clients.