Wow a sensible post that actually got moderated up. At last!
I'm sick of the blatant bias in the/. moderation on this issue, and this is one of the few exceptions. The moderators don't appear to have clue 1 about OpenGL graphics issues and have let Precision Insight confuse the DRI & OS issues and now are moderating up an unjustified onslaught against nVidia, and why? They have the temerity to develop great drivers for Linux. NVidia are not leaning on the community and throwing half baked rubbish over the wall but have put in the time and dollars to write a real quality OpenGL implementation.
These drivers are FAST, I mean REALLY fast! They are also extremely functional. This is probably the fastest most functional OpenGL implementation available for Linux, so set the OS issue aside and thank nVidia for seeing Linux as important enough to put in the development effort.
Ask them nicely for source code, begin that next step now that we have great hardware accelerated OpenGL on Linux. Don't flame off at them like rabid animals, appreciate what's been done but ask for more.
There is a very sound technical reason for the difference at various resolutions. The GeForce has hardware T&L which outperforms the CPU T&L of the voodoo. The voodoo on the other hand has more raw textured fill performance. This means that the voodoo can beat the GeForce on an application which is fill limited, but a GeForce can beat the voodoo on an application which is geometry (T&L) limited. In very simple terms the higher resolution applications tend to be fill limited and reducing the resolution places a greater emphasis on T&L performance. So, the results you see are not likely to change greatly with newer drivers because 3Dfx have highly optimized software SSE & 3DNow! T&L code after years of development.
The GeForce wins on geometry (T&L-transform & lighting), the Voodoo wins on textured fill. Bear in mind that this was an SLI version of the card with two VSA-100 parts.
If you want high resolution go for the 2 part 3Dfx card if you want all round performance go for the GeForce. A single part Voodoo card is going to be a poor performer.
One thing the article didn't touch on is the CPU speed dependency for the voodoo, this system had an 800 MHz processor, if you have a slower processor or one without SSE instructions you can expect the voodoo to be worse at some of the intermediate resolutions because it will be more T&L bound. The GeForce has much less dependency on the CPU because it offloads the T&L to the CPU, in addition the CPU is able to do other stuff while the card is busy in a well written application. The other point to note is that with a FASTER PIII the voodoo will begin to catch up to the GeForce, even at the lower resolutions, so a 1GHz PIII would work more to the voodoo's advantage at least in the benchmarks.
So, if you're upgrading your PIII 500 or any early Celeron system (the latest Celerons have SSE older ones don't) you should really go for the GeForce, if you are building the latest 1GHz power system then the voodoo looks like a good bet especially if you are running at high resolution. If you're CPU somewhere in between then decide what's more important to you, geometry or fill.
It's just the gif compression which is at issue. You can still use gif files but you can't use compressed files unless you get a license. This is tragic since it's a relatively obvious optimal binary level dictionary based compression, and any number of alternatives compression schemes would have done quite nicely instead. Even if gifs had given you some kind of choice of compression we'd all be in better shape.
What's really odious is the way Unisys is trying to enforce it's patent rights here. Not only do they want fees from major application providers but they are now chasing end users too, basically anyone who wants to use the gif format with compression. This is a rather dubious approach since web designers often use packages which have ALREADY licensed the LZW algorithm. Unisys have a pretty screwed up notion of what their I.P. is worth and are darned fortunate that gif files happened to use their compression algorithm, they basically struck gold by pure accident. The rest of us are fortunate that gif compression is optional but it still isn't ideal.
In some respects you have to admire their guts. They're in it for the money, plain and simple. They know that and also know how long the gravy train will run before it derails and they're making the most of it.... simply because they can. There are no relationships to be managed and no follow-on business, they just gotta milk this cow until it dies. That is the nature of corporations in a situation like this, they're in it for the money, it's no use wagging your finger at them, they almost have a duty to maximize the benefit from this little accident.
What is surprising is the momentum behind the LZW in gif files, PNG took a while but a simple modification of the gif format to allow another compression algorithm would have been a viable alternative for gifs.
PNG is long overdue and very welcome for many reasons, it's just MUCH better on so many levels, gif won't be missed. If this patent nonsense helps displace gifs with PNG files then Unisys will actually be doing everyone a favor, and that is wonderfully ironic.
It is clear even from the slashdot article that the complaint was about a refusal to kill the posting once alerted to it. I don't see how that could be interpreted to mean UK ISP's must now monitor all posts. It means they'd better kill defamatory posts if the victim of the post complains, that seems perfectly reasonable.
Just put yourself in the position of being defamed like that and the ISP simply telling you they don't give a damn and are going to keep sending out the misleading info. Demon are HUGE in the UK so this would have gone out to a LOT of people.
Serves Demon right. Maybe next time they'll listen to the victim.
The ongoing cost of ground station equipment and staffing makes this impractical.
They shut the service down not just because of the debt they had to service but also because of operational costs. It's amazing but after all the upfront costs they couldn't even pay the rent.
Nobody is just going to let you 'tinker' with satellites. You might take out another bird or deorbit the thing into downtown L.A., and who's to say you won't do it deliberately or because someone kidnapped your grandmother. Got clearance? Got secure facilities? Got a clue about safely operating a satellite network?
It is actually worth something to have less space junk and light pollution up there, and that's more valuable than anything tinkering could offer, and a heck of a lot safer for the rest of us.
I agree that it undermines PI, but not OpenGL on XFree86, it has a great infrastructure already which nVidia are using. There's no splint here. One OpenGL driver is as good as another if it can play with XFree86 and export the right functions. The issue here for PI is it doesn't use their DRI, but they don't use the SGI SI, hey big deal, it doesn't matter so long as the functionality is provided. LaMonice is trying to make this look like a rift when it isn't. The real danger is forcing companied like nVidia to cow tow to PI because of the disinformation being spread.
Nothing Open Source is being undermined here. We have an infrastructure already, we don't need PI's infrastructure. It's usefull to some, but it's insane to force people with perfectly working driver infrastructures to use the PI code base.
The bottom line is nVidia won't be using PI's DRI and PI are upset at that. Don't let Precision Insight lead you by the nose here.
For everyone else it doesn't matter. The nVidia drivers are based on the SGI SI. Recently Open Sourced for free OS'. The drivers will supply quality OpenGL transparently through GLX. You won't be able to tell the difference.
The only danger here is Precision Insight managing to dominate the OpenGL driver business through their own DRI interface. OK it's Open Source, but why the heck is LaMonica trying to ram it down everyones throat?
The whole purpose of XFree86 release 4.0 is that you can plug in any driver. So, now PI are insisting you plug in their infrastructure?
This makes no sense, why develop a great framework like XFree86 4.0 then hobble it through Precision Insight politics? Besides, the DRI may be Open Source but is all the driver code it requires? In other words this is a bigger red herring than it first appears, because ultimately some lower level code remains proprietary. It's just a flagrant attempt by Precision Insight to carve themselves a niche in the Linux graphics world.
This is very dirty politics. Precision Insight should earn their place, not try to use dirty politics to strongarm nVidia into using their driver infrastructure. NVidia obviously think they have something technically superior and the apps will work transparently, Precision Insight KNOW this but choose to paint a sinister picture. Get off your soap box LaMonica.
The whole point about the 4.0 release is you can plug any driver into it's infrastructure, nVidia plan to DO this and now they get attacked. This is absolutely crazy!
It seems to me that Precision Insight want's their own Mesa based driver infrastructure to dominate and their real complaint has nothing to do with Open Source. The nVidia drivers due out will be high quality and will export the same functions everyone else does, it is proprietary code which they aren't obliged to give away it isn't based on Mesa, it's based on the SGI SI. So it will 'just work', and it will work with XFree86 release 4.0. So what's the problem here? The problem that I see is Precision Insight playing a political game to force the hand of an IHV to do their bidding. I smell a rat. Linux should be able to tollerate companies supporting their platform without this kind of political nonsense. Look at what's going on here, nVidia make great cards and decide to support Linux with real completed drivers which pass the conformance tests. Precision Insight are attacking them because they intend to deliver a driver which uses an infrastructure which isn't based on Mesa and isn't Open Source. That's nuts.
The key point here is that the OpenGL implementation will export quality OpenGL functions and with the finalized ABI and the XFree86 4.0 infrastructure it doesn't matter a hoot what the underlying implementation is based on. In fact it may pass more of the conformance tests than Mesa currently does.
In other words this is a smoke screen to cover a self serving adgenda by Precision Insight. nVidia are supporting Linux and supporting it well, let's not discourace them by letting Precision Insight further their own adgenda through turning public opinion against nVidia.
It seems to me that the reason for filtering these sites is that they contain data or links to data which help circumvent the filtering.
Given that the objective of the filter might be undermined through access to these sites it might seem reasonable to filter them in all categories, because any category of filtering could be disabled using the information on the sites.
I don't advocate or condone any of this, filtering is inevitably a ham fisted tool. I just think it's important to give a ballanced view, and this article is not ballanced.
What would you suggest Mattel do apart from follow your adgenda and close up shop? They sell filtering software, that software has been circumvented, and they have filtered the sites which publish how to circumvent the filtering. This seems like something people who purchase the software would reasonably expect them to do.
The article claims that a cheap TV is better than a computer monitor in it's opening paragraph, that claim is absolute rubbish. The very opposite is true, even the highest quality TV is nowhere near the resolution of even a modest monitor. TV's are designed to resolve 480 lines or less and the ntsc signal is interlaced. Even the latest HDTV's are designed for 1020 lines interlaced, and it's not clear that many of them are actually up to the job of fully resolving those images.
D3D is a part of DirectX and most DirectX games use D3D now, obviously 2D games like Starcraft would be an exception as would a game which uses OpenGL for 3D and DirectX for sound & peripheral input for example Quake2. This means that, in many cases, the claim that DirectX works flawlessly in not compatible with D3D not working well.
The only way you are likely to get D3D working reasonably well with the required hardware acceleration on a broad range of graphics cards is to leverage the OpenGL driver work being done now. What is required is a D3D emulation layer on OpenGL, (some preliminary work exists out there for OpenGL on D3D and vice versa). Some areas would be forever problematic but at some point in the future when XFree86 version 4 is out with many drivers available based on Mesa or the SGI SI or homegrown code, you'd have a reasonable chance of getting functional workalike D3D via emulation code on OpenGL.
You could still implement a native D3D implementation for some cards instead of using the emulation layer for OpenGL, but the emulation layer would be the right approach to getting broad and reasonable hardware acceleration quickly by leveraging existing driver work.
The SGI systems are cache coherent so YES you can. When a write happens there is an overhead updating the the other CPU's but the 'locking' is handled automatically and transparently by the architecture, unlike more traditional very large architectures.
An application may choose to implement additional locks, but at least on a single image system the latency is measured in microseconds or less, not milliseconds like a cluster.
Some systems are suitable for a cluster, but only a fool would claim clusters can solve all scalability problems.
No you won't but there are some customers who do stuff besides alnalyzing SETI data with their machines.
The reason the SGI is so fast at processing SETI results is because of all the fast Fourier Transforms the analysis requires.
That should give you a clue as to why the SGI systems are much better suited to a broad range of scientific codes and outperform competing systems on real applications. Then if you have large data sets and a big problem the SGI systems will really help you because of their high bandwidth ccNUMA architecture. You won't see that with the relatively small problem sets the SETI software processes.
Here SGI is giving a hugely valuable piece of code away and doing a great service to the OS cause. So I come on slashdot to check the response, I sort by highest first to see the intelligent comments and what do I see?
Some diatribe by a bitter ex SGI employee the first thing most readers will see. Moderated to a five! I wouldn't have minded if it wasn't all completely baseless opinion. I've worked for SGI for 5 years, I see guys like this come and go. They've were saying MIPS was dead years ago, and the processors still kicks ass. In fact since SGI spun off MIPS last year it's been doing MUCH better than SGI on the stock market.
Please check the relevance or validity of a post to the subject and check the facts in that post before moderating it 5 and makig it the first thing everyone sees when seeking more information. 2 years at SGI doesn't give everyone insight into it's future and the glaring mismatch between the value of MIPS and the claims made it the post should have at least alerted the moderator to the posters obvious bias.
Earlier you said MIPS and IRIX was dead but the revenue from these systems is HUGE. The desktop sales of O2 and OCTANE are still growing. For some businesses SGI is essential. MIPS & IRIX will never gain wintel share but that's not the intention. To say they are dead is a very emotional and maybe you like a black and white world because the complexity of reality is daunting but it is completely inaccurate to use the word 'dead' here.
You have also said that the 320 & 540 were axed prematurely. If you had a clue about business you'd realize that an expensive late product which only has niche applications which can exploit it's unique architecture is going to have trouble. But that wasn't the real problem with these systems. The real problem was building the NEXT one, to do that you have to engineer new chips, rewrite parts of NT again and other additional engineering costs, not to mention cut special deals with Intel and MS. All everyone else has to do is go buy the latest ACME mother board. To sustain that kind of R&D you need enough people willing to pay a premium for the enhanced capability the R&D brings. The sad truth is that those customers did not exist in sufficient numbers. Halting the 320 & 540 was the best decision SGI made in relation to the whole NT thing.
Without knowing much about the intended use for this machine it's difficult to say what problems lie in store when you switch from the Octane, however it looks capable. The only serious and obvious problem I see is the graphics card. If you've been using an Octane it is more than likely that you have been deploying some kind of OpenGL graphics application on that system. I'd hate to see you get egg on your face even though you are switching from an SGI system so; make sure that you can get hardware accelerated OpenGL on the graphics card you use. You have some choices there, with commercially available X servers which support GLX or Mesa based hardware acceleration in Xfree86, things are progressing rapidly on the GLX/OpenGL acceleration front, just make sure you have a plan in place to get acceleration which is compatible with your hardware, unfortunately this doesn't work out of the box yet, so you'll have to play with X servers, drivers and Mesa, depending on your card & approach. You can always switch the card so it may make sense to get your initial development systems and remain flexible about the graphics cards. See what works best for your applications during the port then finalize on that card and driver setup. By the time you deploy you can be sure the driver situation will have improved immensely so a graphics card final selection before fielding the solution would probably be appropriate, and you only have to swap out the card in your development systems which will be a relatively small portion of the system cost on a small number of systems.
You (Anonymous Coward-LiNuX MaN) wrote: "SGI IRIX, solaris, hpux, and all those are just simply not scalable, reliable, and arent secure at all compared to a red hat distro."
You are clearly wrong here, just take for example the issue which is least subjective and easiest to demonstrate; scalability, the SGI systems running IRIX scale to hundreds of processors on a single image system with realizable performance and they scale to thousands on clustered solutions. Support for 64 bit addressing and file systems allow massive memory capacity and huge high performance file systems.
With RH Linux you can just about benefit from two CPU's if you patch with the latest kernel. Forget big memory addressing, forget massive high performance file systems and massive single files. If you have a specific class of problem you might be able to cobble a beowulf cluster together with relatively low bandwidth, high latency interconnect and allow custom application software to benefit. This is hardly great scalability and doesn't even compare well with Solaris or HP-UX.
I don't doubt things will improve but for goodness sake, be realistic about the work involved to get there. Even against NT there's a lot of work to be done on Linux scalability issues. Linux has a long way to go before it matches the performance of an O.S. like IRIX.
Overselling Linux like this does not help your credibility or the Linux cause and your fantastic claims of problems with other more mature operating systems are counterproductive.
You're thinking of the OTHER Cray. Sure Seymour founded Cray but when he passed away it was after leaving and starting up ANOTHER Cray company, not the one SGI bought.
I disagree that the best staff are contract employees who know their worth. I know some contractors and they are a mixed bunch on the whole. A variety of circumstances lead them to contracting. Good employees are hard to find and then recruit. You need a fairly exhaustive interview process and you need good interviewers, so building that critical mass is the first obstacle. I know full time employees who are world renowned, and have head hunters calling all the time, they stay doing what they do because of the nature of the work, and their passion for a vision of the future, not only for the money. One observation relevant these days is that nobody ever got rich contracting, you can get rich on a startup stock options so who is the more ambitious financially? Who cares more about the long term success of the company?
If you want the best individuals you need a leading project with inspired vision and you need to find the best people and do what it takes to hire them. Offering a long term incentive like stock options seems to be one popular way of generating focus on executing well and staying around.
I am an immigrant worker in the high tech industry, I'm not on an H1 visa but I am on an L1 (company transfer). The cost of employing someone like myself to a U.S. company can be considerable. There are recruitment costs, legal fees applying for a temporary Visa status, relocation costs and perhaps ongoing legal fees associated with permanent residency applications. All of this adds up to a heck of a bill for my employer, and the initial visa is typically term limited to 3 years or maybe 5 with extension.
I am not looking for another position but I have no doubt I could obtain one within a few days and I am frequently contacted by head hunters like everyone else around here. The only real guarantee my employer has against my leaving is my continued satisfaction and loyalty after a significant initial outlay.
The bottom line is that there are checks up the wazoo before immigrant workers can work here and then significant restrictions are applied. On top of that there significant barriers to this type of recruitment because of the additional costs involved. It is NOT less expensive for a company to hire from overseas, and to do so they must demonstrate that the position cannot be filled by a resident who has the skills.
Slashdot Mirror
Wow a sensible post that actually got moderated up. At last!
/. moderation on this issue, and this is one of the few exceptions. The moderators don't appear to have clue 1 about OpenGL graphics issues and have let Precision Insight confuse the DRI & OS issues and now are moderating up an unjustified onslaught against nVidia, and why? They have the temerity to develop great drivers for Linux. NVidia are not leaning on the community and throwing half baked rubbish over the wall but have put in the time and dollars to write a real quality OpenGL implementation.
I'm sick of the blatant bias in the
These drivers are FAST, I mean REALLY fast! They are also extremely functional. This is probably the fastest most functional OpenGL implementation available for Linux, so set the OS issue aside and thank nVidia for seeing Linux as important enough to put in the development effort.
Ask them nicely for source code, begin that next step now that we have great hardware accelerated OpenGL on Linux. Don't flame off at them like rabid animals, appreciate what's been done but ask for more.
Anandtech & your conclusion is wrong.
There is a very sound technical reason for the difference at various resolutions. The GeForce has hardware T&L which outperforms the CPU T&L of the voodoo. The voodoo on the other hand has more raw textured fill performance. This means that the voodoo can beat the GeForce on an application which is fill limited, but a GeForce can beat the voodoo on an application which is geometry (T&L) limited. In very simple terms the higher resolution applications tend to be fill limited and reducing the resolution places a greater emphasis on T&L performance. So, the results you see are not likely to change greatly with newer drivers because 3Dfx have highly optimized software SSE & 3DNow! T&L code after years of development.
There is no surprise here.
The GeForce wins on geometry (T&L-transform & lighting), the Voodoo wins on textured fill. Bear in mind that this was an SLI version of the card with two VSA-100 parts.
If you want high resolution go for the 2 part 3Dfx card if you want all round performance go for the GeForce. A single part Voodoo card is going to be a poor performer.
One thing the article didn't touch on is the CPU speed dependency for the voodoo, this system had an 800 MHz processor, if you have a slower processor or one without SSE instructions you can expect the voodoo to be worse at some of the intermediate resolutions because it will be more T&L bound. The GeForce has much less dependency on the CPU because it offloads the T&L to the CPU, in addition the CPU is able to do other stuff while the card is busy in a well written application. The other point to note is that with a FASTER PIII the voodoo will begin to catch up to the GeForce, even at the lower resolutions, so a 1GHz PIII would work more to the voodoo's advantage at least in the benchmarks.
So, if you're upgrading your PIII 500 or any early Celeron system (the latest Celerons have SSE older ones don't) you should really go for the GeForce, if you are building the latest 1GHz power system then the voodoo looks like a good bet especially if you are running at high resolution. If you're CPU somewhere in between then decide what's more important to you, geometry or fill.
Big deal, they aren't compressed are they?
So who cares?
It's just the gif compression which is at issue. You can still use gif files but you can't use compressed files unless you get a license. This is tragic since it's a relatively obvious optimal binary level dictionary based compression, and any number of alternatives compression schemes would have done quite nicely instead. Even if gifs had given you some kind of choice of compression we'd all be in better shape.
What's really odious is the way Unisys is trying to enforce it's patent rights here. Not only do they want fees from major application providers but they are now chasing end users too, basically anyone who wants to use the gif format with compression. This is a rather dubious approach since web designers often use packages which have ALREADY licensed the LZW algorithm. Unisys have a pretty screwed up notion of what their I.P. is worth and are darned fortunate that gif files happened to use their compression algorithm, they basically struck gold by pure accident. The rest of us are fortunate that gif compression is optional but it still isn't ideal.
In some respects you have to admire their guts. They're in it for the money, plain and simple. They know that and also know how long the gravy train will run before it derails and they're making the most of it.... simply because they can. There are no relationships to be managed and no follow-on business, they just gotta milk this cow until it dies. That is the nature of corporations in a situation like this, they're in it for the money, it's no use wagging your finger at them, they almost have a duty to maximize the benefit from this little accident.
What is surprising is the momentum behind the LZW in gif files, PNG took a while but a simple modification of the gif format to allow another compression algorithm would have been a viable alternative for gifs.
PNG is long overdue and very welcome for many reasons, it's just MUCH better on so many levels, gif won't be missed. If this patent nonsense helps displace gifs with PNG files then Unisys will actually be doing everyone a favor, and that is wonderfully ironic.
It is clear even from the slashdot article that the complaint was about a refusal to kill the posting once alerted to it. I don't see how that could be interpreted to mean UK ISP's must now monitor all posts. It means they'd better kill defamatory posts if the victim of the post complains, that seems perfectly reasonable.
Just put yourself in the position of being defamed like that and the ISP simply telling you they don't give a damn and are going to keep sending out the misleading info. Demon are HUGE in the UK so this would have gone out to a LOT of people.
Serves Demon right. Maybe next time they'll listen to the victim.
The ongoing cost of ground station equipment and staffing makes this impractical.
They shut the service down not just because of the debt they had to service but also because of operational costs. It's amazing but after all the upfront costs they couldn't even pay the rent.
Nobody is just going to let you 'tinker' with satellites. You might take out another bird or deorbit the thing into downtown L.A., and who's to say you won't do it deliberately or because someone kidnapped your grandmother. Got clearance? Got secure facilities? Got a clue about safely operating a satellite network?
It is actually worth something to have less space junk and light pollution up there, and that's more valuable than anything tinkering could offer, and a heck of a lot safer for the rest of us.
I agree that it undermines PI, but not OpenGL on XFree86, it has a great infrastructure already which nVidia are using. There's no splint here. One OpenGL driver is as good as another if it can play with XFree86 and export the right functions. The issue here for PI is it doesn't use their DRI, but they don't use the SGI SI, hey big deal, it doesn't matter so long as the functionality is provided. LaMonice is trying to make this look like a rift when it isn't. The real danger is forcing companied like nVidia to cow tow to PI because of the disinformation being spread.
Nothing Open Source is being undermined here. We have an infrastructure already, we don't need PI's infrastructure. It's usefull to some, but it's insane to force people with perfectly working driver infrastructures to use the PI code base.
This is about PI control.
This is a self serving article.
The bottom line is nVidia won't be using PI's DRI and PI are upset at that. Don't let Precision Insight lead you by the nose here.
For everyone else it doesn't matter. The nVidia drivers are based on the SGI SI. Recently Open Sourced for free OS'. The drivers will supply quality OpenGL transparently through GLX. You won't be able to tell the difference.
The only danger here is Precision Insight managing to dominate the OpenGL driver business through their own DRI interface. OK it's Open Source, but why the heck is LaMonica trying to ram it down everyones throat?
The whole purpose of XFree86 release 4.0 is that you can plug in any driver. So, now PI are insisting you plug in their infrastructure?
This makes no sense, why develop a great framework like XFree86 4.0 then hobble it through Precision Insight politics? Besides, the DRI may be Open Source but is all the driver code it requires? In other words this is a bigger red herring than it first appears, because ultimately some lower level code remains proprietary. It's just a flagrant attempt by Precision Insight to carve themselves a niche in the Linux graphics world.
This is very dirty politics. Precision Insight should earn their place, not try to use dirty politics to strongarm nVidia into using their driver infrastructure. NVidia obviously think they have something technically superior and the apps will work transparently, Precision Insight KNOW this but choose to paint a sinister picture. Get off your soap box LaMonica.
The whole point about the 4.0 release is you can plug any driver into it's infrastructure, nVidia plan to DO this and now they get attacked. This is absolutely crazy!
It seems to me that Precision Insight want's their own Mesa based driver infrastructure to dominate and their real complaint has nothing to do with Open Source. The nVidia drivers due out will be high quality and will export the same functions everyone else does, it is proprietary code which they aren't obliged to give away it isn't based on Mesa, it's based on the SGI SI. So it will 'just work', and it will work with XFree86 release 4.0. So what's the problem here? The problem that I see is Precision Insight playing a political game to force the hand of an IHV to do their bidding. I smell a rat.
Linux should be able to tollerate companies supporting their platform without this kind of political nonsense. Look at what's going on here, nVidia make great cards and decide to support Linux with real completed drivers which pass the conformance tests. Precision Insight are attacking them because they intend to deliver a driver which uses an infrastructure which isn't based on Mesa and isn't Open Source. That's nuts.
The key point here is that the OpenGL implementation will export quality OpenGL functions and with the finalized ABI and the XFree86 4.0 infrastructure it doesn't matter a hoot what the underlying implementation is based on. In fact it may pass more of the conformance tests than Mesa currently does.
In other words this is a smoke screen to cover a self serving adgenda by Precision Insight. nVidia are supporting Linux and supporting it well, let's not discourace them by letting Precision Insight further their own adgenda through turning public opinion against nVidia.
It seems to me that the reason for filtering these sites is that they contain data or links to data which help circumvent the filtering.
Given that the objective of the filter might be undermined through access to these sites it might seem reasonable to filter them in all categories, because any category of filtering could be disabled using the information on the sites.
I don't advocate or condone any of this, filtering is inevitably a ham fisted tool. I just think it's important to give a ballanced view, and this article is not ballanced.
What would you suggest Mattel do apart from follow your adgenda and close up shop? They sell filtering software, that software has been circumvented, and they have filtered the sites which publish how to circumvent the filtering. This seems like something people who purchase the software would reasonably expect them to do.
The article claims that a cheap TV is better than a computer monitor in it's opening paragraph, that claim is absolute rubbish. The very opposite is true, even the highest quality TV is nowhere near the resolution of even a modest monitor. TV's are designed to resolve 480 lines or less and the ntsc signal is interlaced. Even the latest HDTV's are designed for 1020 lines interlaced, and it's not clear that many of them are actually up to the job of fully resolving those images.
D3D is a part of DirectX and most DirectX games use D3D now, obviously 2D games like Starcraft would be an exception as would a game which uses OpenGL for 3D and DirectX for sound & peripheral input for example Quake2. This means that, in many cases, the claim that DirectX works flawlessly in not compatible with D3D not working well.
The only way you are likely to get D3D working reasonably well with the required hardware acceleration on a broad range of graphics cards is to leverage the OpenGL driver work being done now. What is required is a D3D emulation layer on OpenGL, (some preliminary work exists out there for OpenGL on D3D and vice versa). Some areas would be forever problematic but at some point in the future when XFree86 version 4 is out with many drivers available based on Mesa or the SGI SI or homegrown code, you'd have a reasonable chance of getting functional workalike D3D via emulation code on OpenGL.
You could still implement a native D3D implementation for some cards instead of using the emulation layer for OpenGL, but the emulation layer would be the right approach to getting broad and reasonable hardware acceleration quickly by leveraging existing driver work.
You don't even know what image processing is! That's largely performed on 2D workstations these days. It's already very low end business.
You seem to have forgotten that SGI owns CRAY, it makes the fastest highest bandwidth systems in the world.
The computational markets and the 3D graphics markets are largely treated as separate businesses inside SGI.
No, you're fired!
The SGI systems are cache coherent so YES you can. When a write happens there is an overhead updating the the other CPU's but the 'locking' is handled automatically and transparently by the architecture, unlike more traditional very large architectures.
An application may choose to implement additional locks, but at least on a single image system the latency is measured in microseconds or less, not milliseconds like a cluster.
Some systems are suitable for a cluster, but only a fool would claim clusters can solve all scalability problems.
No you won't but there are some customers who do stuff besides alnalyzing SETI data with their machines.
The reason the SGI is so fast at processing SETI results is because of all the fast Fourier Transforms the analysis requires.
That should give you a clue as to why the SGI systems are much better suited to a broad range of scientific codes and outperform competing systems on real applications. Then if you have large data sets and a big problem the SGI systems will really help you because of their high bandwidth ccNUMA architecture. You won't see that with the relatively small problem sets the SETI software processes.
Well I'm gobsmacked.
Here SGI is giving a hugely valuable piece of code away and doing a great service to the OS cause. So I come on slashdot to check the response, I sort by highest first to see the intelligent comments and what do I see?
Some diatribe by a bitter ex SGI employee the first thing most readers will see. Moderated to a five! I wouldn't have minded if it wasn't all completely baseless opinion. I've worked for SGI for 5 years, I see guys like this come and go. They've were saying MIPS was dead years ago, and the processors still kicks ass. In fact since SGI spun off MIPS last year it's been doing MUCH better than SGI on the stock market.
Please check the relevance or validity of a post to the subject and check the facts in that post before moderating it 5 and makig it the first thing everyone sees when seeking more information. 2 years at SGI doesn't give everyone insight into it's future and the glaring mismatch between the value of MIPS and the claims made it the post should have at least alerted the moderator to the posters obvious bias.
Shame on you slashdot!
Earlier you said MIPS and IRIX was dead but the revenue from these systems is HUGE. The desktop sales of O2 and OCTANE are still growing. For some businesses SGI is essential. MIPS & IRIX will never gain wintel share but that's not the intention. To say they are dead is a very emotional and maybe you like a black and white world because the complexity of reality is daunting but it is completely inaccurate to use the word 'dead' here.
You have also said that the 320 & 540 were axed prematurely. If you had a clue about business you'd realize that an expensive late product which only has niche applications which can exploit it's unique architecture is going to have trouble. But that wasn't the real problem with these systems. The real problem was building the NEXT one, to do that you have to engineer new chips, rewrite parts of NT again and other additional engineering costs, not to mention cut special deals with Intel and MS. All everyone else has to do is go buy the latest ACME mother board. To sustain that kind of R&D you need enough people willing to pay a premium for the enhanced capability the R&D brings. The sad truth is that those customers did not exist in sufficient numbers. Halting the 320 & 540 was the best decision SGI made in relation to the whole NT thing.
Not the first?
You seem to be overlooking the funding SGI gave to Precision Insight for their DRI work on XFree86 4.0. SGI was there with the earliest.
Without knowing much about the intended use for this machine it's difficult to say what problems lie in store when you switch from the Octane, however it looks capable. The only serious and obvious problem I see is the graphics card.
If you've been using an Octane it is more than likely that you have been deploying some kind of OpenGL graphics application on that system. I'd hate to see you get egg on your face even though you are switching from an SGI system so; make sure that you can get hardware accelerated OpenGL on the graphics card you use. You have some choices there, with commercially available X servers which support GLX or Mesa based hardware acceleration in Xfree86, things are progressing rapidly on the GLX/OpenGL acceleration front, just make sure you have a plan in place to get acceleration which is compatible with your hardware, unfortunately this doesn't work out of the box yet, so you'll have to play with X servers, drivers and Mesa, depending on your card & approach.
You can always switch the card so it may make sense to get your initial development systems and remain flexible about the graphics cards. See what works best for your applications during the port then finalize on that card and driver setup. By the time you deploy you can be sure the driver situation will have improved immensely so a graphics card final selection before fielding the solution would probably be appropriate, and you only have to swap out the card in your development systems which will be a relatively small portion of the system cost on a small number of systems.
You (Anonymous Coward-LiNuX MaN) wrote:
"SGI IRIX, solaris, hpux, and all those are just simply not scalable, reliable, and arent secure at all compared to a red hat distro."
You are clearly wrong here, just take for example the issue which is least subjective and easiest to demonstrate; scalability, the SGI systems running IRIX scale to hundreds of processors on a single image system with realizable performance and they scale to thousands on clustered solutions. Support for 64 bit addressing and file systems allow massive memory capacity and huge high performance file systems.
With RH Linux you can just about benefit from two CPU's if you patch with the latest kernel. Forget big memory addressing, forget massive high performance file systems and massive single files. If you have a specific class of problem you might be able to cobble a beowulf cluster together with relatively low bandwidth, high latency interconnect and allow custom application software to benefit. This is hardly great scalability and doesn't even compare well with Solaris or HP-UX.
I don't doubt things will improve but for goodness sake, be realistic about the work involved to get there. Even against NT there's a lot of work to be done on Linux scalability issues. Linux has a long way to go before it matches the performance of an O.S. like IRIX.
Overselling Linux like this does not help your credibility or the Linux cause and your fantastic claims of problems with other more mature operating systems are counterproductive.
I'll elaborate; it's spelled "silicon" unless you've attended the Dan Quale school of spelling.
You're thinking of the OTHER Cray. Sure Seymour founded Cray but when he passed away it was after leaving and starting up ANOTHER Cray company, not the one SGI bought.
I disagree that the best staff are contract employees who know their worth. I know some contractors and they are a mixed bunch on the whole. A variety of circumstances lead them to contracting. Good employees are hard to find and then recruit. You need a fairly exhaustive interview process and you need good interviewers, so building that critical mass is the first obstacle. I know full time employees who are world renowned, and have head hunters calling all the time, they stay doing what they do because of the nature of the work, and their passion for a vision of the future, not only for the money. One observation relevant these days is that nobody ever got rich contracting, you can get rich on a startup stock options so who is the more ambitious financially? Who cares more about the long term success of the company?
If you want the best individuals you need a leading project with inspired vision and you need to find the best people and do what it takes to hire them. Offering a long term incentive like stock options seems to be one popular way of generating focus on executing well and staying around.
I am an immigrant worker in the high tech industry, I'm not on an H1 visa but I am on an L1 (company transfer). The cost of employing someone like myself to a U.S. company can be considerable. There are recruitment costs, legal fees applying for a temporary Visa status, relocation costs and perhaps ongoing legal fees associated with permanent residency applications. All of this adds up to a heck of a bill for my employer, and the initial visa is typically term limited to 3 years or maybe 5 with extension.
I am not looking for another position but I have no doubt I could obtain one within a few days and I am frequently contacted by head hunters like everyone else around here. The only real guarantee my employer has against my leaving is my continued satisfaction and loyalty after a significant initial outlay.
The bottom line is that there are checks up the wazoo before immigrant workers can work here and then significant restrictions are applied. On top of that there significant barriers to this type of recruitment because of the additional costs involved. It is NOT less expensive for a company to hire from overseas, and to do so they must demonstrate that the position cannot be filled by a resident who has the skills.