Yes, RAM is a problem, I wasn't debating that fact.
But, RenderMan has no defined implementation, just an interface, and it can just as easily be pipelined as any other 3D graphics task. In fact, a subset of the RenderMan API could be implemented in realtime on PS2 hardware, using multipass rendering techniques (there is no requirement that RenderMan be raytraced - Pixar's implementation of the RenderMan spec, Photorealistic RenderMan, doesn't perform any raytracing).
More advanced applications, such as procedural textures, 3D texturing, etc. would bog down performance considerably, but it is still possible to develop a pipeline that would perform better on the PS2 hardware than on a single 800MHz Athlon.
The Playstation 2's "T&L" hardware is actually part of the main CPU, so the Emotion Engine would certainly outperform an 800MHz Athlon (or even a 1.2GHz Athlon) at rendering tasks.
The Graphics Synthesizer would go unused, but it's the PS2's weak link, anyway.
I am sorry to rag on someone but this statement is complete and utter bullshit. There are a _litany_ of ways that companies can make money without resorting to intellectual property, and indeed the software industry shows so many such examples that this message should be treated as a troll.
There are two fallacies to this argument:
1) This of course assumes that people will be interested in paying for something that they can get for free. Take a look at Napster as an example of how this system really works in practice.
2) The software industry is not representative of other industries. When software is open sourced, companies make profits by selling services or by selling other products for profit. Many times, things that would normally seem to be open sourced to the user (e.g., search engines) are prime examples of IP - the technology is licensed to others, so the free use may continue unhindered (and thus, intellectual property is still necessary). However, how do you sell service on a movie? Is Fox going to pay a representative to wipe my ass for me?
Half the world may not give a rat's ass about US law, copyrights, or patents; however, the other 3 billion people are enough to ensure that food gets placed on the table.
Re:Katz writes about things without having 2 clues
on
Selfish Society
·
· Score: 1
The reason "geeks" get paid so much is because they have brains and know how to use them.
Ha. Just how different is web programming, installing an operating system, or fixing a network hiccup from the things automobile mechanics, plumbers, or electricians do? You have deluded yourself into believing that day-to-day technical myopia is somehow a complicated, mentally taxing field, and for that reason you deserve to be among the highest-paid, best-treated employees in the world. But here's a clue: it's not. The average geek isn't a rocket scientist, either - in fact, based on utility alone, the average automobile mechanic probably deserves at least twice as high a salary as the average web programmer.
Computer science, automotive design, and applied physics are all truly mentally taxing - don't confuse IT with them (the difference between creating solutions and simply applying them).
I don't appreciate your racist comments towards whites either.
Any race which by and large supports a political party which trumpets saving a few bucks on taxes by nixing universal health care deserves to be given a good beating. And yes, I'm white.
Normally in battles involving the RIAA or MPAA, standing up for the little guy who is protecting his first-amendment rights against the evil corporation is not only the attractive option, but the correct one. After all, the MPAA and RIAA are both strongly corrupt organizations, and both were primary motivators behind the worst bill passed since the CDA: the DMCA.
However, in this case, things are considerably different. Not only is standing up for the RIAA the correct thing, it's absolutely necessary.
Our legal system operates quite heavily on the concept of precedents: basically, when Congress doesn't dictate exactly what constitutes an infringement and what doesn't, the Courts are responsible for applying the (admittedly vague) laws on a case-by-case basis. When a Judge reaches a decision, that decision will be used as the basis in future decisions of similar nature.
Now, let's take a look at what happens if Napster wins: a service which is almost entirely used for illegal purposes is given the green light to continue business as it sees fit, without bothering to request permission before facilitating the mass-distribution of copyrighted works. If Napster wins, every court decision involving piracy from now on will probably reference the Napster decision. It wouldn't be too long before a service similar to Napster is created which facilitates the transactions of video games and other software (akin to a centralized Hotline), and based on the Napster decision, such software is protected. Soon, you would be able to download any image, movie, software, or song for free. Sounds almost utopian, right?
Wrong. Despite the fact that most members of/. (and probably most people in the world) won't ever create something of their own, chances are they will work for a company that does. Now, with everyone in the world downloading your company's software, music, movie, etc. for free over their cable modems, how is that company going to make money? The typical/. comment is that it doesn't affect the offender. However, when your employer suddenly tells you that you are going to be laid off because they can't afford to pay you due to piracy over Napster2002, you'd probably see things a little differently. Suddenly, you can't put food on the table because some pimple-faced 13 year-old is too busy saying that Napster2002 is a form of expression, and deserves to be protected, while he rationalizes away his daily 2GB of downloads.
Anyone who is really backing Napster at this point needs to give up their myopic view of the world. Yes, the RIAA is an evil corporation. But this time, they aren't attacking free speech. They are defending copyrights. You may think it's all fine and dandy to steal other's work; however, if Napster wins it would be the biggest case of chopping off a nose to spite a face in the history of the world - eventually it will come back and bite you in the ass... hard.
Given a legal system based on precedent, allowing Napster to win is a very, very dangerous one indeed.
What is actually bad about what Napster has done (aside from the debate over morality of mp3 trading)?
Well, some would claim that profiting from enabling the distribution of other people's copyrighted work is similar to the profiteering that took place in the 18th century.
Now, if Napster charged $20 per month and gave half or two-thirds of that service charge (or just the banner advertisements) to the RIAA, then things would be a little different. Or, perhaps providing the RIAA with general demographic information (e.g., users from the West Coast tend to prefer...). There are plenty of alternatives that would allow Napster to make money, the RIAA to receive what it feels is its share, and users to continue trading music.
The fact that Napster wants to keep all its income for itself, despite the fact that no one is interested in the Napster software itself (users use Napster because of the music), is a bit unscrupulous.
Onyx2 Infinite Reality3 Multi-Rack system 256MB to 256GB RAM up to 128 R12K processors and 16 visualization units Up to 200M poly/sec up to 7GPixel/sec fill rate 320MB frame buffer, 1GB texture RAM 8-tap AA
Onyx2 Infinite Reality3 Single-Rack system 256MB to 16GB RAM up to 8 R12K processors and two visualization units 13.1M poly/sec per visualization unit up to 480MB frame buffer 8-tap AA
Onyx2 Infinite Reality3 Deskside system 256MB to 8GB RAM up to 4R12K processors 80 or 160MB frame buffer 64MB texture RAM 13.1M poly/sec 8-tap AA
In other words, in a machine that takes up about 4 cubic feet of space and requires a single outlet, Sony can outperform a vastly more expensive (and much larger), top-of-the-line SGI. However, SGI still has the GSCube whipped for RAM. In Sony's favor, though, the GSCube is designed to be a workstation, something which the Onyx2 IR3 can not claim.
For some reason, I don't think the GSCube was designed with audio in mind. It may have something to do with the fact that it's a graphics workstation, but that's just my guess.
If the Microsoft Authorization Data Specification v. 1.0 for Microsoft Windows 2000 Operating System is, as Microsoft claims, a trade secret, then it ISN'T registered with the US Copyright Office, which requires a copy of the document to be submitted when you apply for copy protection. Once something has been submitted to the Copyright Office, anyone can view it, thus removing "Trade Secret" status from it.
What Microsoft has done is attach a "Copyright (C) 2000,..." message to their specification, regardless of whether it is legally copyrightable or not. The questions Andover's lawyers ask why Microsoft feels it qualifies for intellectual property protection since it has essentially copyrighted an already-copyrighted protocol. While the text of the specification probably can be protected, whether the extensions to Kerberos are protected or not is a completely different issue, and one which is on considerably less-defined legal ground. The lawyers didn't spend a week coming up with something completely irrelevant.
The point is that they did copyright it. Slashdot is in the wrong.
However, if the material isn't legally copyrightable, then any claim Microsoft has to intellectual property protection is lost. Many of the questions in this letter directly question how Microsoft can claim copyright protection on an already copyrighted standard.
AFAIK, there are no legal precedents for this type of business model (the embrace and extend model that Microsoft uses so frequently), and taking this issue to court may bring an end to Microsoft E&E strategy altogether, provided that Andover wins.
But I have a question for the legally inclined. How binding are all of these thinly-veiled hostilities? For example, what would have happened if Andover hadn't replied to Microsoft's letter? Were they obligated to under law? And similarly, is Microsoft required to respond in kind?
IANAL, but I am doing some research on the DMCA (a manifestation of pure evil), and this is how I think the case would unfold if Andover didn't go to its legal team: Prior to the DMCA, if a copyrighted document were found on a common carrier, the company whose rights were infringed could sue the poster, but would need to go to a federal court to get an injunction (Cease and Desist order) demanding that the carrier (in this case, Slashdot) remove the infringing material.
Under the DMCA, a company need only notify the carrier in writing (an e-mail is sufficient) that a copyrighted document is on their site. If the carrier does not remove it, they can be held as liable as the original poster. If Andover hadn't gone for legal counseling and [i]refused[/i] to comply with Microsoft's demands, then there is the very realistic chance that criminal charges could be filed, in addition to Microsoft's inevitable civil suit.
Needless to say, this is only one of many articles in the DMCA that should be stricken from the US Code as quickly as possible.
Compared to it's bitrate. I don't know if it scales up to as high bitrate as DVD, but at least I think a mpeg4 movie scaled that have 5 times the resolution of a normal dvd movie, would be better. Although, I gueess you would need a powerfull computer...
This is actually incorrect, and anyone with an eye accustomed to watching DVDs on properly calibrated setups will argue with this.
The problem with DivX isn't it's resolution, but the way that it compresses things. The logo on their web page already exhibits motion artifacts on the rack focus, which suggests that there would be some serious artifacts in the backgrounds of Geri's game if they actually showed the trailer.
Additionally, the image is extremely soft. Detail is lost around the edges (as if they convolve the source signal with a 5x5 gaussian kernel or something similar during encoding to decrease the amount of space video will require after transforming it).
Also, color banding is present along flat surfaces. This isn't as noticable in 100 kbps DivX as in 12kbps Sorensen video, but it's visible enough to be unwatchable for movie enthusiasts.
And finally, watching The Matrix in anything but 5.1 sound loses so much due to the transition it almost isn't worthwhile.
At 1/10 the compression, DivX definitely provides impressive quality, but don't fool yourself into believing that the codec provides DVD-quality video. It is highly optimized for medium-bitrate situations and streaming (e.g., MPEG4), not for high-bitrate environments (e.g., MPEG2). This of course means that DivX scaled to DVD bitrates wouldn't really improve it all that much from its current position.
Yes, absolutely! GPL or multi-million-dollar proprietary patentware, doesn't matter. I don't believe in product liability. If you want a warranty, buy one. Consider it insurance. If someone wants to offer you a warranty, that's fine, but nobody should have to. There's no hypocrisy involved here. I agree with the court's decision.
I'm going to have to disagree with you here - product liability is a very, very good thing when used properly. In this case, M.A Mortensen purchased the program Softworks for Bid Analysis, which was designed to create construction bids. Mortensen submitted a bid generated by this program, which had erroneously quoted a final price $1.95 million lower than it should have. In another case (NCR back in the mid-80s), an inventory management program was sold for $180,000 to a $4 million-a-year company, with some known severe bugs (e.g., if two users tried to access the database at once the program would crash and corrupt data, items out of stock would be listed as in stock, items in stock would be listed as out of stock, etc.), and because of these faults the annual revenue plummeted to under $1 million.
Now, I can understand if you try to use Windows 3.1 as a nuclear reactor control system or download some free program to keep track of your $3 billion inventory, but using commercial construction bid software to generate construction bids, or extremely expensive inventory management software to manage inventory is very reasonable use for the software. Yes, some bugs should be expected, but the program should do what it was designed to do. Allowing software manufacturers to run the industry in a caveat emptor (buyer beware) fashion will not be a good thing, for software or consumers. This court ruling is practically a green light to high-tech con artists around the country - hopefully some higher courts will disagree with this decision.
As another example of software liability, we need only look at Microsoft Outlook. Microsoft included a 'feature' (VBScript) that allows e-mails to perform many system functions undetected. Several times this feature has been exploited for viral purposes, with each subsequent exploitation becoming increasingly malicious (Melissa, ILOVEYOU, Mother's Day). How many times must corporate e-mail be shut down (each time costing BILLIONS of dollars) before Microsoft will be forced to accept some of the responsiblity for the havoc their programs have enabled - even if the acceptance is simply removing the offending feature?
Product liability is absolutely necessary, so companies like Microsoft don't add FORMAT_HARDDISK() 'innovations' to their e-mail 'features,' and so egregious abuses of the term "software engineering," such as the M.A. Mortensen and the NCR cases listed above, aren't the modus operandi for the industry. Small businesses need to be able to buy point-of-sale and networking software confident that they won't lose 80% of their revenue as a direct (or even indirect) result of the software, or else a free-market economy in today's high-tech world will not work.
AOL can claim (and rightfully so) that their intent was to provide an internet service. The fact that MP3s reside on their service is an unexpected byproduct.
Napster, OTOH, was created with the intent to trade MP3s. Claiming that you are creating MP3 trading software incognizant of the rather obvious possibilities for unscrupulous people to abuse the network to trade legal MP3s is impossible. Napster created their software fully aware that it could aide and incite people to pirate music. There is a very large legal difference between the two, and if Napster didn't do absolutely everything within its power to safeguard against illegal transactions, they could be in some serious trouble.
Re:Don't worry about it, Napster's a different iss
on
MP3.com Loses In Court
·
· Score: 1
Napster is an entirely different issue. Napster is a network, which allows users of the network to connect to each other and download stuff. Napster itself doesn't host songs on its own "site" like MP3.com's MP3 Anywhere or whatever it was called, was doing. So, it's very likely that Napster will be ruled some form of "common carrier" and therefore not liable for what its users do amongst themselves, since its network has a legitimate use in allowing users to distribute non-copyrighted sound files. To make Napster police its network to decide what is and is not copyright infringement would put an undue burden on Napster and similar networks, and is not what a judge would order. In fact, if a network does police itself, then it would become liable--but if it just provides an open service, it can be considered a common carrier. Just my 2 pence...
First, IANAL, but I have read quite a few laws and court decisions...
It would be sooo nice if the case were that simple; however, Napster is not just a common carrier, since they also provided the software Napster Music Community. Although it would seem like the First Amendment would guarantee Napster the right to make and distribute whatever software they like, a Supreme Court decision in the mid-60s (Brandenberg v. Ohio) has set a precedent that could help the RIAA in a rather major way...
Basically, the decision stated that a work that could aide or incite illegal activity, with the author's knowledge and intent to cause such activity, is not protected by the First Amendment. Napster will be unable to argue that they were unaware that their software could be used for trading illegal MP3s. Since this is a civil case it will probably be up to Napster to prove that they didn't intend to help people trade pirate music, rather than the RIAA needing to prove the opposite.
Basically, the big 3 'trading' programs (Napster, Gnutella, and Hotline) are all on extremely shaky legal ground, and any lawsuits brought against them have considerable merit, especially if the authors don't build in "safeguards" to protect themselves. Napster is the most defendable, but that really doesn't say very much (Gnutella and Hotline are almost guaranteed to crumble if a lawsuit is ever brought against them).
Realistically, there is probably nothing that NVIDIA developed that they wouldn't be willing to open source - their 3D pipeline is an almost identical copy of SGI's (remember that lawsuit 2 years ago?).
However, welcome to the depths of cross-licensing hell. NVIDIA is free to disclose its own patents and inventions, open source its proprietary stuff, whatever. The problem is that modern drivers include quite a bit of code licensed from some other company, which WOULDN'T appreciate an open source driver. I believe John Carmack has commented several times that the DMA and AGP techniques NVIDIA uses to get peak performance from its cards is licensed from some other place. Without the licensed code, the driver speed goes to hell. With it, the chance of an open-sourced driver goes to hell. And now that NVIDIA has added a few more cross-licenses to its belt (S3TC, etc.) don't expect any open-source miracles to happen.
If you are really desperate to learn 3D programming, buying a good theory book like Computer Graphics: Principles and Practice would teach you infinitely more than looking at code for a driver would, anyway. Granted, this doesn't help you when Linux changes kernel builds from 2.39.049.beta.help-me-please-I-can't-help-myself-f rom-rebuilding-my-kernel-every-week.pre to 2.40.woohoo-stable-for-two-more-weeks.final, but the average (future) Linux user probably isn't interested in changing kernels dozens of times, they are more of the Microsoft mentality - install it, use it, and don't touch it. NVIDIA has provided drivers that will encourage people to port 3D applications to Linux, which will result in more 3D cards being supported under Linux, etc. Even though the drivers are closed source, don't confuse yourself into believing that this is anything but a tremendous boon to the future of 3D support under Linux.
Re:Let's play a little game...
on
ATI Radeon 256
·
· Score: 1
True, with 2 640x480x64bit frame buffers, there really isn't much space left for textures.
But the average PS2 resolution is 640x240x32 (16 bit color, 16 bit Z), which is as high as NTSC televisions will go. There is also a 3.2GB bi-directional bus between the GS and the EE. Technically, it would be possible to render a section of the framebuffer in the GS, and then page it out to the EE. Then the EE can do video post-processing (non-photorealistic rendering and complex antialiasing come to mind as possible applications) and output a framebuffer to the GS for display.
There are plenty of techniques to work around the 4MB VRAM on the PS2, you just need to know about parallel processing, multithreading, and paging. Unfortunately, since most games developers are still self-taught, these aren't common knowledge in the industry. When Phil Harrison guest lectured at a CS class of mine he basically went so far as to say that over half of current game developers don't have nearly what it takes to get the most of the PS2.
Re:Let's play a little game...
on
ATI Radeon 256
·
· Score: 1
Well, if every triangle is projected exactly the same and there have uniform animation, right.
Typically it will end up being 1 matrix per triangle list (to include animation and translation), which will probably drop the matrix read bandwidth requirements by a factor of 30-40.
Let's play a little game...
on
ATI Radeon 256
·
· Score: 5
... called answer the question.
I've got a graphics card by some manufacturer (it really doesn't matter who) that has a 1.6 Gigatexel fill rate. Now, given the propensity for developers to use 32-bit textures, that means that each and everyone one of the 1.6 billion texels I process every second must be accompanied by its own 4-byte read. Now, how much memory bandwidth does this require? And how much bandwidth is on the card?
Now, let's start expanding on this... 30 million triangles / second, given triangle lists equates to about 16 million vertices. At 3 floats (x,y,z) per vertex, and 4 bytes per float, that's another 192MB/sec of bandwidth we don't have. Now, in order to actually use textures, each of those vertices also needs texture coordinates, which add another 2 floats, or 128MB/sec. And then for lighting we need a vector normal to each vertex... that's another 3 floats, so 192MB/sec. Now, in order to actually project these coordinates onto the screen, every vertex needs to be multiplied by a 4x4 matrix, or 16 more floats. Whooppee! That's another GIGABYTE of bandwidth down the tubes. Then to actually display this, since I have 2 texture units per pixel pipeline, my card delivers 800MegaPixel fill rates, which at 8 bytes per pixel (24-bit RGB + 8-bit alpha, 24-bit Z, 8-bit stencil) is another 6.4GB/sec of bandwidth.
So, when all is said and done, to reach the theoretical maximum of my card, I need 14.3GB/sec memory bandwidth minimum. Add in things like texture filtering (multiple texel reads per texel write) or alpha blending and you can break 20GB/sec easily.
Multiply all this by about 10 for Microsoft's X-Box (which somehow claims to shovel 14.4 Gigatexel performance across a 6.4GB/sec unified bus), and you'll know why any and all paper specs for the X-Box are completely ridiculous.
There is only one architecture currently available or in production that actually has the bandwidth to support its theoretical maximums, and there's no way in hell it'll fit in an AGP slot. It's manufactured by Sony, can currently be bought in Japan for about $400, is slightly larger than a bread box, and provides 48GB/sec of bandwidth, albeit at a slight hit to the actual frame buffer size.
So, in the spirit of the industry, I'm announcing my new video card. It has 400Gigatexel performance, and can transform 100 billion triangles every second. Unfortunately, due to current memory and bus technologies, you'll never see more than about 500 megatexels and 2.5 million triangles, anyway.
You can keyframe a skeleton just like you can keyframe a mesh. Basically, you define a motion for the skeleton, and and time duration in which that motion occurs. Voila! A keyframe. Motion capture is often viewed as massive auto-keyframing.
In fact, including mocap, keyframing is far and away the most popular technique in modern computer graphics. Not many people use procedural animation (which is possible, although not terribly realistic yet). Basically, if the actual motion is stored as discrete samples along the motion curve which are then interpolated, you've got keyframing.
Skeletal animation is preferred because you only need to keyframe (typically) 24-50 degrees of freedom for a human object. This is much easier for artists than having to manually handle 10,000 NURBS surfaces, and also makes capturing the motion really easily. Skinning, such as included in the Radeon, comes into play because the model is only defined once. If you look at your own skin as you bend your elbow, the skin on the outside will stretch while the skin on the inside will contract. Since polygons are not soft, skinning by using matrix interpolation is used to ensure that no seams emerge at any of the joints.
Well, to be fair, the patent is not for Final Fantasy or role playing games, but the Active Time Battle System, which Square 'invented' in Final Fantasy IV (II in the US), and their method of storing hitpoints and calculating action responses.
Well, crypto isn't the only reason supercomputers are restricted - military and nuclear simulations exist, too. In fact, the latter two reasons are probably a more direct threat to national security than China trying to crack a 1024-bit encryption key. And those same applications are extremely floating-point intensive, which just happens to be the forte of the Emotion Engine.
Typical teraflop supercomputers use several thousand CPUs - 825 standard-issue 300Mhz Emotion Engines would break 5 Teraflops (recall that the French Atomic Energy Commission just purchased an Alpha supercomputer with 5 teraflop performance and it required 2,000 1.25Ghz chips). Since the Emotion Engine has relatively few transistors and a fairly low clockspeed, getting high yields shouldn't be a problem, and it is cheap to produce. Remember, Sony builds the entire Playstation 2 for about $550, and with the.18 micron process that cost will decrease considerably, to probably much less than $200 per chip. A mere 250 CPUs would be needed if the EE were running at a 1Ghz clockspeed.
Yes, in the wrong hands, the Emotion Engine could present a very real threat to governments.
It's a modified (aka, reduced) MIPS III instruction set with 128-bit registers and multimedia extensions (PADD, PMUL, etc).
The main chip has the standard-issue 1xFMAC and 1xFDIV floating point unit.
Additionally, there are two more coprocessors: VU0 and VU1. VU0 can run in independent or MIPS coprocessor mode (typically used in MIPS coprocessor mode) with 1xFDIV and 3xFMAC. VU1 can only run in independent mode, and adds an elementary function unit (1xFDIV, 1xFMAC) to its standard 1xFDIV and 4xFMAC. VU1 has its own internal instruction and data cache
Total floating point performance: 6.2GFlops at 300Mhz, or roughly equivalent to a 1.5Ghz Athlon.
The EE is currently fabricated on a.25 micron process, but Sony just finished development of.18 micron fabs for EE production. Once volume production is started, and the initial Playstation 2 lauch hysteria dies down (early 2001 in the US/Europe, probably), multimedia and scientific computers based on faster Emotion Engines will be released. The workstation model should have 4 parallel Emotion Engines running at a slightly higher clockspeed (better than 25GFlop performance).
Basically, as long as you're doing floating point operations, this chip would rock.
Okay, as an import PS2 owner, let me tell you that games DO exist, they're a helluva lot of fun, and look light-years better than everything I've ever seen on PCs, and much better than everything I've played on Dreamcast.
Playing region 1 DVDs is a bit difficult (I didn't keep any non-dual shock Playstation controllers around, unfortunately, so I have to use the DualShock 2 hacks, which are a pain in the ass - ~35% effectiveness), but the quality is just as good as, if not better than, my Panasonic A120's.
You forget that Sony is targeting the PS2 for mass-market use... 95% of the US doesn't have a DVD player, and it's frequently listed as one of the most-wanted accessories. $300 buys you an excellent DVD player, an unbelievable video game system, and (in the very near future) a full internet appliance, if you want. It works with all your old Playstation games (okay, 24 unpopular games don't work, but that's beside the point) and hardware, and can output a progressive signal if ever Sony codes it into the DVD driver. Plus, it's already a run-away success with ~2 million sold in 1 month.
Several aftermarket add-ons have been successful - the Sega CD, Jaguar CD, Light Guns, the Nintendo 64 4MB Memory Pak, Sony Multitap, and Nintendo 64 Rumble Paks. It is quite possible to create accessories that developers use, they just have to be worthwhile. The Famicom Disk and the Sattelaview for the Famicom and Super Famicom were unsuccesful since no one was interested in downloading text games from satellite or in a format that offered no advantages over the cartridge. Similarly, I would be surprised if the Nintendo Robot offered anything to anyone. Lots of failed add-ons exist (Sega 32X), but in general, if there is enough of a use for hardware add-on, developers will support it at least secondarily.
However, internet support is a massive add-on that is very worthwhile - plus, if you are developing a multi-player game, adding in support for broadband internet access is pretty easy. Since Sony has refused to allow a modem connection for the PS2, developers will NOT need to worry about optimizing network code for high-latency low-bandwidth connections, so adding internet support to Gran Turismo 2001 or Madden 2002 won't be terribly time-consuming. Net games are already in development for the PS2 (Final Fantasy XI, anyone?)... I doubt that the internet add-on will be anything but successful.
As for the hard drive... From what I've read of Sony's white papers, the HD is not designed for heavy games use. Sony envisions a broadband network where you can setup the PS2 to download a DVD movie while you play Tekken, and cache it on the hard drive, or store tons of MP3s. Its use will be primarily as multimedia storage for the internet/application side of the PS2, not for gaming. It will be a convenience for the PS2 owner, not the console developer, so it should be reasonably successful - especially for people who don't own computers and use the PS2 as their primary internet device.
Also, all current information points to an external 50GB+ Firewire hard disk released at the same time a PCMCIA Ethernet card is released - early 2001. Sony is already losing enough money on every PS2 that they sell to include $200 more hardware in the box.
Slashdot Mirror
But, RenderMan has no defined implementation, just an interface, and it can just as easily be pipelined as any other 3D graphics task. In fact, a subset of the RenderMan API could be implemented in realtime on PS2 hardware, using multipass rendering techniques (there is no requirement that RenderMan be raytraced - Pixar's implementation of the RenderMan spec, Photorealistic RenderMan, doesn't perform any raytracing).
More advanced applications, such as procedural textures, 3D texturing, etc. would bog down performance considerably, but it is still possible to develop a pipeline that would perform better on the PS2 hardware than on a single 800MHz Athlon.
The Playstation 2's "T&L" hardware is actually part of the main CPU, so the Emotion Engine would certainly outperform an 800MHz Athlon (or even a 1.2GHz Athlon) at rendering tasks. The Graphics Synthesizer would go unused, but it's the PS2's weak link, anyway.
I am sorry to rag on someone but this statement is complete and utter bullshit. There are a _litany_ of ways that companies can make money without resorting to intellectual property, and indeed the software industry shows so many such examples that this message should be treated as a troll.
There are two fallacies to this argument:
1) This of course assumes that people will be interested in paying for something that they can get for free. Take a look at Napster as an example of how this system really works in practice.
2) The software industry is not representative of other industries. When software is open sourced, companies make profits by selling services or by selling other products for profit. Many times, things that would normally seem to be open sourced to the user (e.g., search engines) are prime examples of IP - the technology is licensed to others, so the free use may continue unhindered (and thus, intellectual property is still necessary). However, how do you sell service on a movie? Is Fox going to pay a representative to wipe my ass for me?
Half the world may not give a rat's ass about US law, copyrights, or patents; however, the other 3 billion people are enough to ensure that food gets placed on the table.
The reason "geeks" get paid so much is because they have brains and know how to use them.
Ha. Just how different is web programming, installing an operating system, or fixing a network hiccup from the things automobile mechanics, plumbers, or electricians do? You have deluded yourself into believing that day-to-day technical myopia is somehow a complicated, mentally taxing field, and for that reason you deserve to be among the highest-paid, best-treated employees in the world. But here's a clue: it's not. The average geek isn't a rocket scientist, either - in fact, based on utility alone, the average automobile mechanic probably deserves at least twice as high a salary as the average web programmer.
Computer science, automotive design, and applied physics are all truly mentally taxing - don't confuse IT with them (the difference between creating solutions and simply applying them).
I don't appreciate your racist comments towards whites either.
Any race which by and large supports a political party which trumpets saving a few bucks on taxes by nixing universal health care deserves to be given a good beating. And yes, I'm white.
Normally in battles involving the RIAA or MPAA, standing up for the little guy who is protecting his first-amendment rights against the evil corporation is not only the attractive option, but the correct one. After all, the MPAA and RIAA are both strongly corrupt organizations, and both were primary motivators behind the worst bill passed since the CDA: the DMCA.
/. (and probably most people in the world) won't ever create something of their own, chances are they will work for a company that does. Now, with everyone in the world downloading your company's software, music, movie, etc. for free over their cable modems, how is that company going to make money? The typical /. comment is that it doesn't affect the offender. However, when your employer suddenly tells you that you are going to be laid off because they can't afford to pay you due to piracy over Napster2002, you'd probably see things a little differently. Suddenly, you can't put food on the table because some pimple-faced 13 year-old is too busy saying that Napster2002 is a form of expression, and deserves to be protected, while he rationalizes away his daily 2GB of downloads.
However, in this case, things are considerably different. Not only is standing up for the RIAA the correct thing, it's absolutely necessary.
Our legal system operates quite heavily on the concept of precedents: basically, when Congress doesn't dictate exactly what constitutes an infringement and what doesn't, the Courts are responsible for applying the (admittedly vague) laws on a case-by-case basis. When a Judge reaches a decision, that decision will be used as the basis in future decisions of similar nature.
Now, let's take a look at what happens if Napster wins: a service which is almost entirely used for illegal purposes is given the green light to continue business as it sees fit, without bothering to request permission before facilitating the mass-distribution of copyrighted works. If Napster wins, every court decision involving piracy from now on will probably reference the Napster decision. It wouldn't be too long before a service similar to Napster is created which facilitates the transactions of video games and other software (akin to a centralized Hotline), and based on the Napster decision, such software is protected. Soon, you would be able to download any image, movie, software, or song for free. Sounds almost utopian, right?
Wrong. Despite the fact that most members of
Anyone who is really backing Napster at this point needs to give up their myopic view of the world. Yes, the RIAA is an evil corporation. But this time, they aren't attacking free speech. They are defending copyrights. You may think it's all fine and dandy to steal other's work; however, if Napster wins it would be the biggest case of chopping off a nose to spite a face in the history of the world - eventually it will come back and bite you in the ass... hard.
Given a legal system based on precedent, allowing Napster to win is a very, very dangerous one indeed.
Well, some would claim that profiting from enabling the distribution of other people's copyrighted work is similar to the profiteering that took place in the 18th century.
Now, if Napster charged $20 per month and gave half or two-thirds of that service charge (or just the banner advertisements) to the RIAA, then things would be a little different. Or, perhaps providing the RIAA with general demographic information (e.g., users from the West Coast tend to prefer ...). There are plenty of alternatives that would allow Napster to make money, the RIAA to receive what it feels is its share, and users to continue trading music.
The fact that Napster wants to keep all its income for itself, despite the fact that no one is interested in the Napster software itself (users use Napster because of the music), is a bit unscrupulous.
Taken from SGI's site:
Onyx2 Infinite Reality3 Multi-Rack system
256MB to 256GB RAM
up to 128 R12K processors and 16 visualization units
Up to 200M poly/sec
up to 7GPixel/sec fill rate
320MB frame buffer, 1GB texture RAM
8-tap AA
Onyx2 Infinite Reality3 Single-Rack system
256MB to 16GB RAM
up to 8 R12K processors and two visualization units
13.1M poly/sec per visualization unit
up to 480MB frame buffer
8-tap AA
Onyx2 Infinite Reality3 Deskside system
256MB to 8GB RAM
up to 4R12K processors
80 or 160MB frame buffer
64MB texture RAM
13.1M poly/sec
8-tap AA
In other words, in a machine that takes up about 4 cubic feet of space and requires a single outlet, Sony can outperform a vastly more expensive (and much larger), top-of-the-line SGI. However, SGI still has the GSCube whipped for RAM. In Sony's favor, though, the GSCube is designed to be a workstation, something which the Onyx2 IR3 can not claim.
For some reason, I don't think the GSCube was designed with audio in mind. It may have something to do with the fact that it's a graphics workstation, but that's just my guess.
If the Microsoft Authorization Data Specification v. 1.0 for Microsoft Windows 2000 Operating System is, as Microsoft claims, a trade secret, then it ISN'T registered with the US Copyright Office, which requires a copy of the document to be submitted when you apply for copy protection. Once something has been submitted to the Copyright Office, anyone can view it, thus removing "Trade Secret" status from it.
..." message to their specification, regardless of whether it is legally copyrightable or not. The questions Andover's lawyers ask why Microsoft feels it qualifies for intellectual property protection since it has essentially copyrighted an already-copyrighted protocol. While the text of the specification probably can be protected, whether the extensions to Kerberos are protected or not is a completely different issue, and one which is on considerably less-defined legal ground. The lawyers didn't spend a week coming up with something completely irrelevant.
What Microsoft has done is attach a "Copyright (C) 2000,
The point is that they did copyright it. Slashdot is in the wrong.
However, if the material isn't legally copyrightable, then any claim Microsoft has to intellectual property protection is lost. Many of the questions in this letter directly question how Microsoft can claim copyright protection on an already copyrighted standard.
AFAIK, there are no legal precedents for this type of business model (the embrace and extend model that Microsoft uses so frequently), and taking this issue to court may bring an end to Microsoft E&E strategy altogether, provided that Andover wins.
But I have a question for the legally inclined. How binding are all of these thinly-veiled hostilities? For example, what would have happened if Andover hadn't replied to Microsoft's letter? Were they obligated to under law? And similarly, is Microsoft required to respond in kind?
IANAL, but I am doing some research on the DMCA (a manifestation of pure evil), and this is how I think the case would unfold if Andover didn't go to its legal team:
Prior to the DMCA, if a copyrighted document were found on a common carrier, the company whose rights were infringed could sue the poster, but would need to go to a federal court to get an injunction (Cease and Desist order) demanding that the carrier (in this case, Slashdot) remove the infringing material.
Under the DMCA, a company need only notify the carrier in writing (an e-mail is sufficient) that a copyrighted document is on their site. If the carrier does not remove it, they can be held as liable as the original poster. If Andover hadn't gone for legal counseling and [i]refused[/i] to comply with Microsoft's demands, then there is the very realistic chance that criminal charges could be filed, in addition to Microsoft's inevitable civil suit.
Needless to say, this is only one of many articles in the DMCA that should be stricken from the US Code as quickly as possible.
Compared to it's bitrate. I don't know if it scales up to as high bitrate as DVD, but at least I think a mpeg4 movie scaled that have 5 times the resolution of a normal dvd movie, would be better. Although, I gueess you would need a powerfull computer...
This is actually incorrect, and anyone with an eye accustomed to watching DVDs on properly calibrated setups will argue with this.
The problem with DivX isn't it's resolution, but the way that it compresses things. The logo on their web page already exhibits motion artifacts on the rack focus, which suggests that there would be some serious artifacts in the backgrounds of Geri's game if they actually showed the trailer.
Additionally, the image is extremely soft. Detail is lost around the edges (as if they convolve the source signal with a 5x5 gaussian kernel or something similar during encoding to decrease the amount of space video will require after transforming it).
Also, color banding is present along flat surfaces. This isn't as noticable in 100 kbps DivX as in 12kbps Sorensen video, but it's visible enough to be unwatchable for movie enthusiasts.
And finally, watching The Matrix in anything but 5.1 sound loses so much due to the transition it almost isn't worthwhile.
At 1/10 the compression, DivX definitely provides impressive quality, but don't fool yourself into believing that the codec provides DVD-quality video. It is highly optimized for medium-bitrate situations and streaming (e.g., MPEG4), not for high-bitrate environments (e.g., MPEG2). This of course means that DivX scaled to DVD bitrates wouldn't really improve it all that much from its current position.
Yes, absolutely! GPL or multi-million-dollar proprietary patentware, doesn't matter. I don't believe in product liability. If you want a warranty, buy one. Consider it insurance. If someone wants to offer you a warranty, that's fine, but nobody should have to. There's no hypocrisy involved here. I agree with the court's decision.
I'm going to have to disagree with you here - product liability is a very, very good thing when used properly. In this case, M.A Mortensen purchased the program Softworks for Bid Analysis, which was designed to create construction bids. Mortensen submitted a bid generated by this program, which had erroneously quoted a final price $1.95 million lower than it should have. In another case (NCR back in the mid-80s), an inventory management program was sold for $180,000 to a $4 million-a-year company, with some known severe bugs (e.g., if two users tried to access the database at once the program would crash and corrupt data, items out of stock would be listed as in stock, items in stock would be listed as out of stock, etc.), and because of these faults the annual revenue plummeted to under $1 million.
Now, I can understand if you try to use Windows 3.1 as a nuclear reactor control system or download some free program to keep track of your $3 billion inventory, but using commercial construction bid software to generate construction bids, or extremely expensive inventory management software to manage inventory is very reasonable use for the software. Yes, some bugs should be expected, but the program should do what it was designed to do. Allowing software manufacturers to run the industry in a caveat emptor (buyer beware) fashion will not be a good thing, for software or consumers. This court ruling is practically a green light to high-tech con artists around the country - hopefully some higher courts will disagree with this decision.
As another example of software liability, we need only look at Microsoft Outlook. Microsoft included a 'feature' (VBScript) that allows e-mails to perform many system functions undetected. Several times this feature has been exploited for viral purposes, with each subsequent exploitation becoming increasingly malicious (Melissa, ILOVEYOU, Mother's Day). How many times must corporate e-mail be shut down (each time costing BILLIONS of dollars) before Microsoft will be forced to accept some of the responsiblity for the havoc their programs have enabled - even if the acceptance is simply removing the offending feature?
Product liability is absolutely necessary, so companies like Microsoft don't add FORMAT_HARDDISK() 'innovations' to their e-mail 'features,' and so egregious abuses of the term "software engineering," such as the M.A. Mortensen and the NCR cases listed above, aren't the modus operandi for the industry. Small businesses need to be able to buy point-of-sale and networking software confident that they won't lose 80% of their revenue as a direct (or even indirect) result of the software, or else a free-market economy in today's high-tech world will not work.
AOL can claim (and rightfully so) that their intent was to provide an internet service. The fact that MP3s reside on their service is an unexpected byproduct.
Napster, OTOH, was created with the intent to trade MP3s. Claiming that you are creating MP3 trading software incognizant of the rather obvious possibilities for unscrupulous people to abuse the network to trade legal MP3s is impossible. Napster created their software fully aware that it could aide and incite people to pirate music. There is a very large legal difference between the two, and if Napster didn't do absolutely everything within its power to safeguard against illegal transactions, they could be in some serious trouble.
Napster is an entirely different issue. Napster is a network, which allows users of the network to connect to each other and download stuff. Napster itself doesn't host songs on its own "site" like MP3.com's MP3 Anywhere or whatever it was called, was doing. So, it's very likely that Napster will be ruled some form of "common carrier" and therefore not liable for what its users do amongst themselves, since its network has a legitimate use in allowing users to distribute non-copyrighted sound files. To make Napster police its network to decide what is and is not copyright infringement would put an undue burden on Napster and similar networks, and is not what a judge would order. In fact, if a network does police itself, then it would become liable--but if it just provides an open service, it can be considered a common carrier. Just my 2 pence...
First, IANAL, but I have read quite a few laws and court decisions...
It would be sooo nice if the case were that simple; however, Napster is not just a common carrier, since they also provided the software Napster Music Community. Although it would seem like the First Amendment would guarantee Napster the right to make and distribute whatever software they like, a Supreme Court decision in the mid-60s (Brandenberg v. Ohio) has set a precedent that could help the RIAA in a rather major way...
Basically, the decision stated that a work that could aide or incite illegal activity, with the author's knowledge and intent to cause such activity, is not protected by the First Amendment. Napster will be unable to argue that they were unaware that their software could be used for trading illegal MP3s. Since this is a civil case it will probably be up to Napster to prove that they didn't intend to help people trade pirate music, rather than the RIAA needing to prove the opposite.
Basically, the big 3 'trading' programs (Napster, Gnutella, and Hotline) are all on extremely shaky legal ground, and any lawsuits brought against them have considerable merit, especially if the authors don't build in "safeguards" to protect themselves. Napster is the most defendable, but that really doesn't say very much (Gnutella and Hotline are almost guaranteed to crumble if a lawsuit is ever brought against them).
Realistically, there is probably nothing that NVIDIA developed that they wouldn't be willing to open source - their 3D pipeline is an almost identical copy of SGI's (remember that lawsuit 2 years ago?).
f rom-rebuilding-my-kernel-every-week.pre to 2.40.woohoo-stable-for-two-more-weeks.final, but the average (future) Linux user probably isn't interested in changing kernels dozens of times, they are more of the Microsoft mentality - install it, use it, and don't touch it. NVIDIA has provided drivers that will encourage people to port 3D applications to Linux, which will result in more 3D cards being supported under Linux, etc. Even though the drivers are closed source, don't confuse yourself into believing that this is anything but a tremendous boon to the future of 3D support under Linux.
However, welcome to the depths of cross-licensing hell. NVIDIA is free to disclose its own patents and inventions, open source its proprietary stuff, whatever. The problem is that modern drivers include quite a bit of code licensed from some other company, which WOULDN'T appreciate an open source driver. I believe John Carmack has commented several times that the DMA and AGP techniques NVIDIA uses to get peak performance from its cards is licensed from some other place. Without the licensed code, the driver speed goes to hell. With it, the chance of an open-sourced driver goes to hell. And now that NVIDIA has added a few more cross-licenses to its belt (S3TC, etc.) don't expect any open-source miracles to happen.
If you are really desperate to learn 3D programming, buying a good theory book like Computer Graphics: Principles and Practice would teach you infinitely more than looking at code for a driver would, anyway. Granted, this doesn't help you when Linux changes kernel builds from 2.39.049.beta.help-me-please-I-can't-help-myself-
True, with 2 640x480x64bit frame buffers, there really isn't much space left for textures.
But the average PS2 resolution is 640x240x32 (16 bit color, 16 bit Z), which is as high as NTSC televisions will go. There is also a 3.2GB bi-directional bus between the GS and the EE. Technically, it would be possible to render a section of the framebuffer in the GS, and then page it out to the EE. Then the EE can do video post-processing (non-photorealistic rendering and complex antialiasing come to mind as possible applications) and output a framebuffer to the GS for display.
There are plenty of techniques to work around the 4MB VRAM on the PS2, you just need to know about parallel processing, multithreading, and paging. Unfortunately, since most games developers are still self-taught, these aren't common knowledge in the industry. When Phil Harrison guest lectured at a CS class of mine he basically went so far as to say that over half of current game developers don't have nearly what it takes to get the most of the PS2.
Well, if every triangle is projected exactly the same and there have uniform animation, right.
Typically it will end up being 1 matrix per triangle list (to include animation and translation), which will probably drop the matrix read bandwidth requirements by a factor of 30-40.
... called answer the question.
I've got a graphics card by some manufacturer (it really doesn't matter who) that has a 1.6 Gigatexel fill rate. Now, given the propensity for developers to use 32-bit textures, that means that each and everyone one of the 1.6 billion texels I process every second must be accompanied by its own 4-byte read. Now, how much memory bandwidth does this require? And how much bandwidth is on the card?
Now, let's start expanding on this... 30 million triangles / second, given triangle lists equates to about 16 million vertices. At 3 floats (x,y,z) per vertex, and 4 bytes per float, that's another 192MB/sec of bandwidth we don't have. Now, in order to actually use textures, each of those vertices also needs texture coordinates, which add another 2 floats, or 128MB/sec. And then for lighting we need a vector normal to each vertex... that's another 3 floats, so 192MB/sec. Now, in order to actually project these coordinates onto the screen, every vertex needs to be multiplied by a 4x4 matrix, or 16 more floats. Whooppee! That's another GIGABYTE of bandwidth down the tubes. Then to actually display this, since I have 2 texture units per pixel pipeline, my card delivers 800MegaPixel fill rates, which at 8 bytes per pixel (24-bit RGB + 8-bit alpha, 24-bit Z, 8-bit stencil) is another 6.4GB/sec of bandwidth.
So, when all is said and done, to reach the theoretical maximum of my card, I need 14.3GB/sec memory bandwidth minimum. Add in things like texture filtering (multiple texel reads per texel write) or alpha blending and you can break 20GB/sec easily.
Multiply all this by about 10 for Microsoft's X-Box (which somehow claims to shovel 14.4 Gigatexel performance across a 6.4GB/sec unified bus), and you'll know why any and all paper specs for the X-Box are completely ridiculous.
There is only one architecture currently available or in production that actually has the bandwidth to support its theoretical maximums, and there's no way in hell it'll fit in an AGP slot. It's manufactured by Sony, can currently be bought in Japan for about $400, is slightly larger than a bread box, and provides 48GB/sec of bandwidth, albeit at a slight hit to the actual frame buffer size.
So, in the spirit of the industry, I'm announcing my new video card. It has 400Gigatexel performance, and can transform 100 billion triangles every second. Unfortunately, due to current memory and bus technologies, you'll never see more than about 500 megatexels and 2.5 million triangles, anyway.
You can keyframe a skeleton just like you can keyframe a mesh. Basically, you define a motion for the skeleton, and and time duration in which that motion occurs. Voila! A keyframe. Motion capture is often viewed as massive auto-keyframing.
In fact, including mocap, keyframing is far and away the most popular technique in modern computer graphics. Not many people use procedural animation (which is possible, although not terribly realistic yet). Basically, if the actual motion is stored as discrete samples along the motion curve which are then interpolated, you've got keyframing.
Skeletal animation is preferred because you only need to keyframe (typically) 24-50 degrees of freedom for a human object. This is much easier for artists than having to manually handle 10,000 NURBS surfaces, and also makes capturing the motion really easily. Skinning, such as included in the Radeon, comes into play because the model is only defined once. If you look at your own skin as you bend your elbow, the skin on the outside will stretch while the skin on the inside will contract. Since polygons are not soft, skinning by using matrix interpolation is used to ensure that no seams emerge at any of the joints.
Well, to be fair, the patent is not for Final Fantasy or role playing games, but the Active Time Battle System, which Square 'invented' in Final Fantasy IV (II in the US), and their method of storing hitpoints and calculating action responses.
Well, crypto isn't the only reason supercomputers are restricted - military and nuclear simulations exist, too. In fact, the latter two reasons are probably a more direct threat to national security than China trying to crack a 1024-bit encryption key. And those same applications are extremely floating-point intensive, which just happens to be the forte of the Emotion Engine.
.18 micron process that cost will decrease considerably, to probably much less than $200 per chip. A mere 250 CPUs would be needed if the EE were running at a 1Ghz clockspeed.
Typical teraflop supercomputers use several thousand CPUs - 825 standard-issue 300Mhz Emotion Engines would break 5 Teraflops (recall that the French Atomic Energy Commission just purchased an Alpha supercomputer with 5 teraflop performance and it required 2,000 1.25Ghz chips). Since the Emotion Engine has relatively few transistors and a fairly low clockspeed, getting high yields shouldn't be a problem, and it is cheap to produce. Remember, Sony builds the entire Playstation 2 for about $550, and with the
Yes, in the wrong hands, the Emotion Engine could present a very real threat to governments.
It's a modified (aka, reduced) MIPS III instruction set with 128-bit registers and multimedia extensions (PADD, PMUL, etc).
.25 micron process, but Sony just finished development of .18 micron fabs for EE production. Once volume production is started, and the initial Playstation 2 lauch hysteria dies down (early 2001 in the US/Europe, probably), multimedia and scientific computers based on faster Emotion Engines will be released. The workstation model should have 4 parallel Emotion Engines running at a slightly higher clockspeed (better than 25GFlop performance).
The main chip has the standard-issue 1xFMAC and 1xFDIV floating point unit.
Additionally, there are two more coprocessors: VU0 and VU1. VU0 can run in independent or MIPS coprocessor mode (typically used in MIPS coprocessor mode) with 1xFDIV and 3xFMAC. VU1 can only run in independent mode, and adds an elementary function unit (1xFDIV, 1xFMAC) to its standard 1xFDIV and 4xFMAC. VU1 has its own internal instruction and data cache
Total floating point performance: 6.2GFlops at 300Mhz, or roughly equivalent to a 1.5Ghz Athlon.
The EE is currently fabricated on a
Basically, as long as you're doing floating point operations, this chip would rock.
Okay, as an import PS2 owner, let me tell you that games DO exist, they're a helluva lot of fun, and look light-years better than everything I've ever seen on PCs, and much better than everything I've played on Dreamcast.
Playing region 1 DVDs is a bit difficult (I didn't keep any non-dual shock Playstation controllers around, unfortunately, so I have to use the DualShock 2 hacks, which are a pain in the ass - ~35% effectiveness), but the quality is just as good as, if not better than, my Panasonic A120's.
You forget that Sony is targeting the PS2 for mass-market use... 95% of the US doesn't have a DVD player, and it's frequently listed as one of the most-wanted accessories. $300 buys you an excellent DVD player, an unbelievable video game system, and (in the very near future) a full internet appliance, if you want. It works with all your old Playstation games (okay, 24 unpopular games don't work, but that's beside the point) and hardware, and can output a progressive signal if ever Sony codes it into the DVD driver. Plus, it's already a run-away success with ~2 million sold in 1 month.
It will be huge.
Here is where you are mistaken -
Several aftermarket add-ons have been successful - the Sega CD, Jaguar CD, Light Guns, the Nintendo 64 4MB Memory Pak, Sony Multitap, and Nintendo 64 Rumble Paks. It is quite possible to create accessories that developers use, they just have to be worthwhile. The Famicom Disk and the Sattelaview for the Famicom and Super Famicom were unsuccesful since no one was interested in downloading text games from satellite or in a format that offered no advantages over the cartridge. Similarly, I would be surprised if the Nintendo Robot offered anything to anyone. Lots of failed add-ons exist (Sega 32X), but in general, if there is enough of a use for hardware add-on, developers will support it at least secondarily.
However, internet support is a massive add-on that is very worthwhile - plus, if you are developing a multi-player game, adding in support for broadband internet access is pretty easy. Since Sony has refused to allow a modem connection for the PS2, developers will NOT need to worry about optimizing network code for high-latency low-bandwidth connections, so adding internet support to Gran Turismo 2001 or Madden 2002 won't be terribly time-consuming. Net games are already in development for the PS2 (Final Fantasy XI, anyone?)... I doubt that the internet add-on will be anything but successful.
As for the hard drive... From what I've read of Sony's white papers, the HD is not designed for heavy games use. Sony envisions a broadband network where you can setup the PS2 to download a DVD movie while you play Tekken, and cache it on the hard drive, or store tons of MP3s. Its use will be primarily as multimedia storage for the internet/application side of the PS2, not for gaming. It will be a convenience for the PS2 owner, not the console developer, so it should be reasonably successful - especially for people who don't own computers and use the PS2 as their primary internet device.
Also, all current information points to an external 50GB+ Firewire hard disk released at the same time a PCMCIA Ethernet card is released - early 2001. Sony is already losing enough money on every PS2 that they sell to include $200 more hardware in the box.