for that matter, I don't think it's even their oldest beta - the former froogle, now renamed google products), predates it by a year or so. I believe froogle entered beta around Christmas 2002 or 2003. Some google labs stuff (non-beta testing and ideas area) is even older.
Carmack is not saying the industry won't go to ray tracing, but rather that the industry won't abandon rasterization because each has strengths and weaknesses.
He believes the same thing I do - a hybrid approach is most likely, at least in the short term. A sparse voxel octtree (a voxel is a 3d pixel and an octtree is a uniform 3d structure to hold the voxels - they are sparse because most are empty [hold air]) and would work well for ray tracing because it sounds like you'd need to cast rays to find the voxel. I'm not sure why/how it would save on overlapping edges unless the voxel itself holds color (texture) information and is fragment level in detail. Still, that seems like it would be an incredibly large data structure, so I'm sure he's doing some trick that I can't think of at the moment.
I don't think a game porting company is a viable business model. I don't actually do game development, but let me start with what I know...
Aspyr started that way and still makes a good amount of money porting to mac. In the 1980s before the crash almost every decent game was ported to other platforms profitably.
Until you have to start thinking about consoles. Don't the PS3 and the Wii both have some sort of GL implementation available, even if it's not their best API?
Yes, the PS3 and Wii are OpenGL, as are a number of handhelds (PSP, DS, etc). Those all have common hardware and APIs, however, so it is much easier to target them.
...but there are a couple of open alternatives -- some "open" as in "BSD license", so they can be included in commercial games.
I have yet to find a hardware accelerated physics engine using the BSD license. I believe ODE and Tokamak are software only. I'm not sure where Physx stands since the nVidia buyout, but it is available on all nVidia 8 series hardware through patch. Still, I think that would restrict it to nVidia hardware.
I know several developers that dislike OpenAL. I'm not exactly sure why, as I found it fine, but tbh, I haven't used a lot of sound APIs and haven't done extensive work with any. I'm more of a graphics person on a computer.
Is that across all windowed 3D, or only windowed GL? Because windowed GL would have to run with some OpenGL wrapper around Direct3D, if I understand it right -- you can't actually run both at once, and Aero itself is Direct3D
WDDM changed a bit during development. Originally, Aero and OpenGL were not going to be compatible, but MS backpedaled on that and now gives a rendering surface area that is compatible with OpenGL (the compositing itself is done in software, I believe, because GDI and DirectX are on the same layer but GDI is software only). You still get a performance hit because both are hardware and use different APIs, so it does a hardware context switch on each cycle (thus the 10-15% speed hit). Incidentally, I get my performance hit whether I'm using Aero or not and the basic interface is supposed to be software, so it makes me wonder (could be laptop hardware). It is definitely using a hardware context because my code uses Geometry Shaders.
or we're one big EMP pulse away from losing almost 2 zettabytes of data.
and technicallly, there is only one SI definition of zettabyte, which is 10^21. The binary definitions used by the IEC like the zettabyte=2^70 are being renamed to avoid ambiguity (proposed to be zebibyte for zeta binary byte).
I'm just the opposite, I find I need to be more selective with games and music and if they don't grab me immediately, they aren't worth my time. For instance I loved Gothic I & II, but III's controls were so hideously broken I just gave up after an hour of play (combat was frustrating and always 10 on 1, generally a sign of bad game design). MoO3 from a couple of years ago I despised so much I shelved it after playing it once (despite hearing about a patch that supposedly fixed things) - MoO2 was one of my favorite games of all time. I enjoy UT3 and The Witcher, games I bought late last year, though I almost gave up on The Witcher due to display driver issues (it crashes the Vista driver after playing a bit, though the driver usually recovers). I tried to play the Orange Box a bit, but like most Valve games, it makes me motion sick (UT3 is the first of that series that makes me a little sick - UT2004 I could play for days).
I still listen to and buy new music, but I never was a prolific buyer like some friends with thousands of CDs. In fact, I get sick of listening to my old collection, which to be quite honest, isn't that large at maybe 200 CDs and 50 song downloads. My wife neither plays game nor buys music.
I see it more like this: 1) Loki, the only major Linux commercial game porting company failed 2) DirectX is the API of choice on the platform with the most marketshare 3) many of the latest features are inconsistently supported in OpenGL 4) most games assembled via APIs licensed per platform
on point 3, we have to wait for OpenGL 3.1 (3.0 streamlines the API but does not include new features) for many features in the over a year old DX10 API to appear because any EXTs are not required to be compliant with the API, so manufacturers like ATI can ship a DX10 compliant card that supports geometry shaders for DX and claim the card is OpenGL 2.0 or 2.1 compliant but skip support for EXT_geometry_shader (and I'm calling them out specifically because they reportedly have done it and have a history of doing just this sort of thing, especially with nVidia designed features).
on point 4 - the problem isn't so much with OpenGL - many game engine features are built to the Windows API. Want Havok Physics? Better have a console or PC, because it won't work with Linux and has a restrictively expensive mac license (according to rumor, identical to the Windows license). Compound that with sound API, a video API (for movies and cutscenes, e.g. Bink), a terrain API like speedtree, etc. Basically, modern games are no longer built from scratch, but assembled via a number of generally commercial APIs, some of which charge extra for additional platforms, some of which are Windows only.
Note that I did not mention OpenGL and Vista being incompatible - the truth is, you only get a performance hit in Windowed mode. I don't know why compositing gives a 30% performance hit (in my testing) because it should be more like 10%, but in full screen mode where it does no compositing it runs almost as fast as XP (consistent with DX on Vista).
tell that to Jesus's mom - at least in Catholic tradition, she was always a virgin, despite having a husband and according to some records, other children. Note that many reformers like Martin Luther and John Calvin do not believe in Mary's perpetual virginity.
Technically, some parts are wrong in how they work at an atomic level in other models such as quantum mechanics (friction, for instance), but as an observed model of the world it works quite well.
In other words, you're both correct - Newton's model is wrong in some cases, because it doesn't work for, say, observing an object traveling very fast, but it is effective in other cases such as observing a dropped object at sea level traveling through air with no wind.
All GMAs before the X3000 line (note the X, because the 3000 and 3100 is NOT included) did not contain HW geometry processing (texturing and lighting, vertex shaders and in earlier processors, scheduling). The problem is that the newer generations mostly include 8 unified shaders, which puts them at the bottom of the heap and these shaders perform poorly compared to dedicated cards (partially due to shared memory). Many modern cards have hundreds of shaders, though some lower end and mobile have as little as 16. For reference, the X3500's fill rate is slightly slower than a 2 year old nVidia 7300 but is capable of running 8xxx line shaders.
The problem is most graphics are moving to heavily into shaders because they're easy to write and add significant visual improvement for minimal effort, but chipsets like the 950 do this in software. This is what Tim is talking about - basically, you need to write a shader version and a non-shader version just to work with the Intel chipset - 2 different games.
Only the most modern GMA chips can even run modern games that are heavily shader dependent and they do those poorly due to having a slow clock, few shaders, and dependence on shared system memory. If the X3500 was a baseline that would be ok, but most use the older and dirt cheap 915 or 965 lines, and the 915 isn't even Vista Aero capable due to lack of hardware scheduling (Intel complaining about this was the reason MS lowered its requirements for Vista). Intel has downplayed this probably because they're weathering the storm, so to say, believing the future is in Ray Tracing and having shared memory is a boon for that (Ray Tracing requires access to the complete scene and therefore having it in main memory is good). I do not share Intel's belief that Ray Tracing is the future and in particular, the very near future - I believe it is part of it, but not entirely. By the time you tack on certain expensive effects currently dumped into shaders (namely diffuse lighting for soft shadows, which is at least O(logn) using photon mapping and as much as O(n^3) using radiosity, if I recall correctly), performance is near rasterization.
It can do curved surfaces, but is far from perfect and really is an approximation placed into a polygon structure (the surface itself is flat, but you don't see the surface itself). The silhouettes are done using 4 of the 16 quadrics as a reasonable subset to get an estimate curvature (4 to save on storage). These are pre-calculated and therefore have some overhead (either in a file or dynamically calculated on load). It is also restricted to hard shadows because it works similarly to ray tracing - tracing lines, not patches (it's hard to describe that, but think of it like when light hits an object, it reflects a line, not a cone), however, you get these shadows for free.
Incidentally, Steep Parallax Mapping and Interval Mapping also use pseudo ray tracers, but the surface curvature is unique to Relief Mapping (the quadric technique doesn't work with them). Most other techniques can be adapted to support soft shadows, however, and most game developers I know (3 professional) think soft shadows are more important than curved surface relief maps. If surface curvature is important, they'd rather tessellate it out.
I know this isn't what you meant, at least by how you phrased the rest of your comment, but technically light bounces are handled very well by raytracers, as is proper color absorption from nearby reflections as long as you are referring to Specular (shiny) lighting. What isn't handled well is non-point source diffuse (soft), which is why many ray tracers bolt on photon mapping or radiosity.
yes the spooks are forbidden to domestically spy without a warrant and so this is illegal. Unfortunately, Bush has allowed the NSA to do unfettered domestic spying (yes, constitutionally illegal and against the NSA charter) by executive order so if the agencies work together in any way they may have a dubiously legal way of getting the information, so proving all of the information was obtained illegally may be harder than it sounds.
In a sane world, "the decider" wouldn't be allowed to executive order himself absolute power. Not that I blame the Republicans for this trend in using executive orders to work around congress and the law - for instance, Clinton boosted FEMA's power so that it probably can declare martial law and create internment camps (as a national security directive, so it is not public). Only congress can declare martial law by law and I think internment camps are a moot point (free? ha!), so its clear to me at least that Presidents are abusing executive orders.
Ron Paul doesn't follow the Republican doctrine close enough to garner favor. He's neither highly religious nor glued to the party line. Huckabee's popularity was based around being a conservative Christian, but some of his plans just suck (FairTax, which shifts more of the tax burden from those that make over $200000 to the middle class), so I'm glad he didn't win. I've never liked people that think God is on their side, which is what bugs the hell out of me about Bush, and I had a feeling Huckabee would be the same (note that I'm not saying belief in God is a bad thing, I just don't think God picks sides).
Ron wants to move use back to the gold standard so savings can't be eroded by devaluation of the dollar (by printing more money), but moving to the gold standard would be interesting, because Presidential Executive Order 6102, dated April 5, 1933 makes it illegal to horde gold (even by note), so unless that's repealed, it gives the government the right to seize any gold savings you might have (with some exceptions, but the $100 limit is pretty nasty). Gold has stability, but also the tendency to throttle the growth of the economy. Basically, the gold standard is like investing in Government Bonds and Fiat currency is like High Yield (Junk) Bonds - slow growth and slow rewards but generally stable vs high risk, high reward unstable. Personally, I wouldn't mind seeing a dual currency, but that brings headaches of its own (like 2 price tags).
No - DMA may help in some cases, as you describe, but you can tell a Firewire drive to copy to another Firewire drive when neither has any physical memory and it will still copy much faster than USB. The lack of a centralized controller (and device registration, scheduling, etc) actually helps keep overhead down. Note that USB can't do that - Firewire is peer-to-peer, meaning each device is aware of other devices in the chain. USB is a master-slave star network and needs a host controller (e.g. a PC).
Firewire was built a hot swappable, high speed replacement for SCSI, and is really more analogous to SATA than USB, but people compare them because they're both used as external buses for peripherals. USB was designed explicitly as a low speed, low power, low cost small peripheral handler (e.g. mice and keyboards) to replace a variety of miscellaneous specialized plugs such as game ports, parallel port, serial port, etc, and thus cost was most important and speed least. Firewire put speed first and cost last. As far as Firewire goes, I think a battle may be coming, with SATA's external plug eSATA, as I expect it to make some gains in the peripheral market, especially in storage. eSATA actually has an advantage over Firewire, because the actual device used for storage is often IDE and therefore Firewire has some conversion to do (ATA is the protocol, IDE the device - often they're used interchangeably).
The problem here is gullibility. Think of it like social engineering - someone calls and asks "We are verifying your bank account pin, can you give it to us?" and you saying sure - it's 1234! That's a lot like what this program is doing. In this case, the device at one end is saying can I have access to your memory? And the device on the other end is saying sure, despite the fact that that giving write access to memory is a lot like giving away your bank account pin (which is why it's really an OS issue, not a firewire issue). Some OS's like Linux only give read access, which means you can see what is in the account, but not take anything out, but Linux (and Windows) allow this to be set by the foreign controller, which is a bug.
DMA access should be limited to non-system memory, if allowed. Unfortunately, that isn't very controllable by current computer designs. I believe the solution proposed and implemented (I've heard about this for Windows 8, I believe) is encrypted floating addresses, so even if you have direct access to memory you don't know where to write it.
not to mention they list it as IEEE1394 port rather than firewire to avoid Apple's tariff on the name. My 9 month old laptop has one, my 2 year old homebrew desktop 2 (and I don't use either of them at the moment).
I think you overestimate the power of the tool - TomTom lists restaurants without ratings, so it's doubtful it will add tons of patrons. There are plenty of seats available, especially at dinner time (I think most people get take-out), and competition is heavy in the area. Incidentally, I noticed it's now in Google Maps, which is a fairly recent add. They still don't have another nearby Chinese restaurant that opened a year ago, however, and have one that went out of business 3 months ago - I should probably update it (the new location of TGI Fridays is correct in Google maps).
Jobs is a better example of vendor lockin - he wants everything as a disposable appliance.
It's almost impossible to find a large software company with multiple products that doesn't have some open source offerings, however, even if their main products are primarily closed source. Some examples are Apple, Microsoft [also see Codeplex], Adobe and Oracle.
Probably the best example I can think of for closed source is game companies like EA, Vivendi (Blizzard), etc. Carmack and Id are the exception, not the rule in that industry.
If you wait for a manufacturer to make all the corrections, you will wait forever because they can't check all places at all times and certainly wouldn't know all the best PoI and restaurants even if they're full time residents. For instance, both TomTom and Garmin GPS list a TGI Fridays that was a few blocks from my home as still in business when, in fact, it moved 2 miles away over 6 months ago and is being replaced by a new restaurant. There is also a fantastic Thai restaurant (it has won awards for best Thai) tucked behind a strip mall that isn't listed and I'd love to add it.
Personally, I like features like this on TomTom, but yes, an open source database would rock. Even something that pulled from google maps would be cool, IMO, as long as google maps stays free.
You'd be surprised at how some people you don't think need a lot of power actually need all of it. For instance, my mom can peg the CPU for 2-3 minutes with her large genealogy database. I do CAD for work, which requires boatloads more CPU than GPU, since the bulk of the work is tessellation and CSG, neither of which GPUs do particularly well. Yes, geometry shaders on the latest generation of hardware can do tessellation, but in practice they do it poorly due to hardware restrictions (or so I've heard - I don't actively develop in that area).
Once upon a time, high end machines came with a separate floating point processor. As cost came down, CPUs started integrating them and the FPU is now an integral part of all CPUs. GPUs are gaining the same sort of value, essentially adding sets of high speed parallel processors that perform tasks that can be offloaded from CPU. You might say that exists now with SSE/3DNow!, etc, but that has one huge flaw - separate instruction sets for each CPU. GPUs also have a much broader set of functionality (graphical and general purpose).
There's also the convergence with real-time ray tracing coming. The problem with ray tracing on GPU is that the rays need to be aware of the entire scene to be shaded properly (reflections, color bleed, etc). This is most easily done with high speed access to main memory rather than putting it all on a limited subset of memory on the GPU (GPU memory could still be used for texture lookups and other scene related things). As CPUs come out with more and more cores and higher speeds, real time ray tracing becomes closer to reality since it is highly parallel-izable and the breakpoint for when it is faster to ray trace vs shade polygons is rapidly converging. I personally don't think ray tracing will unseat polygons soon because the cost of adding in decent diffuse lighting (photon mapping is about the fastest) makes up for the speed gain, but perhaps programmers will "cheat" and do it in a shader, just as they do reflections in shaders for polygons today (usually cube maps).
LaptopVideo2Go is NOT for fixing driver problems
on
Is AMD Dead Yet?
·
· Score: 1
LaptopVideo2Go is NOT dedicated to fixing driver problems - it's dedicated to installing newer reference drivers onto your laptop without the hardware manufacturer customizing and testing them, which is required by their OEM contracts with nVidia. nVidia is NOT required to provide a driver for your laptop at all.
A bit of history - not so long ago, laptops required custom drivers and these were provided by the laptop manufacturer, not the GPU manufacturer due to specialized display hardware. Newer laptops tend to work with reference drivers, however, and no longer need specialized drivers due to standardization in the video display industry. Some people like you require the latest-and-greatest drivers to fix bugs, so they resort to the reference drivers at a possible expense of functionality. All LaptopVideo2Go does is add mobile GPU values to a list of supported machines so you can install a much newer reference driver (at the possible cost of bricking your machine) because your laptop manufacturer is too slow at releasing an official driver that suits your needs.
ATI releases a reference driver, as well, the Catalyst Mobility driver, but they allow it to be installed by anyone, as I recall (I've never had an ATI laptop, only desktops). I'm not sure if that's because they have more control over the platform, or if they have no OEM requirement for support and provide it themselves, or if they are taking a risk and letting anyone install a reference driver.
It really depends on what you're looking for - on the mid-high end, AMD's Windsor line (X2 5200+ series) compared favorably to Conroes on a price-performance ratios. The slightly better performing 6750 was about $60 more expensive than an X2 6000+ (now it's about $30) late last year when I was looking (I still have my parts and price list for about 5 systems in the $800-900 range, 3 were Intel and 2 AMD). The X2 6000+ outperformed the Conroe 6600 and 6650 in many benchmarks, as well. On the high end, however, the Core 2 Extreme wins over the Phenoms (the 9500-9700, which are pretty scarce, if available at all) on both price and performance. Intel processors also completely own in overclocking, though AMD's 9600 Black is supposed to be pretty good (still not in Intel's league, but a start).
I purchased a laptop with a dual core Conroe last year and that works great usually - when it isn't, I blame Vista (I'm doing some certification testing on Vista, so no, Ubuntu is not an option).
I was seriously considering building a new desktop with one up until my car decided it wanted $1400 in repairs and maintenance. Too bad I don't have as much talent working on cars as I do computers.
ugh - while it was incredible compared to the first movie, Ernest Goes To Camp was incredible compared to the first movie. On the one hand, you had a couple of Z-list actors and actresses that actually are better at acting than the wooden, terrible performances of the A and B list talent of the first movie, but on the other hand, many of them are completely miscast and it shows (skinny pretty girl with no muscles just doesn't cut it as a fighter, sorry).
My favorite fantasy movie characters are probably: Matthew Broderick in Ladyhawke, Arnold Schwarzenegger in Conan, almost everyone in LotR and pretty much everyone in the Princess Bride, as well. I think Brigitte Nielson was a good cast for Red Sonja, but the plot and acting were so awful it loses a vote.
Slashdot Mirror
for that matter, I don't think it's even their oldest beta -
the former froogle, now renamed google products), predates it by a year or so. I believe froogle entered beta around Christmas 2002 or 2003. Some google labs stuff (non-beta testing and ideas area) is even older.
Carmack is not saying the industry won't go to ray tracing, but rather that the industry won't abandon rasterization because each has strengths and weaknesses.
He believes the same thing I do - a hybrid approach is most likely, at least in the short term. A sparse voxel octtree (a voxel is a 3d pixel and an octtree is a uniform 3d structure to hold the voxels - they are sparse because most are empty [hold air]) and would work well for ray tracing because it sounds like you'd need to cast rays to find the voxel. I'm not sure why/how it would save on overlapping edges unless the voxel itself holds color (texture) information and is fragment level in detail. Still, that seems like it would be an incredibly large data structure, so I'm sure he's doing some trick that I can't think of at the moment.
Aspyr started that way and still makes a good amount of money porting to mac. In the 1980s before the crash almost every decent game was ported to other platforms profitably.
Yes, the PS3 and Wii are OpenGL, as are a number of handhelds (PSP, DS, etc). Those all have common hardware and APIs, however, so it is much easier to target them.
I have yet to find a hardware accelerated physics engine using the BSD license. I believe ODE and Tokamak are software only. I'm not sure where Physx stands since the nVidia buyout, but it is available on all nVidia 8 series hardware through patch. Still, I think that would restrict it to nVidia hardware.
I know several developers that dislike OpenAL. I'm not exactly sure why, as I found it fine, but tbh, I haven't used a lot of sound APIs and haven't done extensive work with any. I'm more of a graphics person on a computer.
WDDM changed a bit during development. Originally, Aero and OpenGL were not going to be compatible, but MS backpedaled on that and now gives a rendering surface area that is compatible with OpenGL (the compositing itself is done in software, I believe, because GDI and DirectX are on the same layer but GDI is software only). You still get a performance hit because both are hardware and use different APIs, so it does a hardware context switch on each cycle (thus the 10-15% speed hit). Incidentally, I get my performance hit whether I'm using Aero or not and the basic interface is supposed to be software, so it makes me wonder (could be laptop hardware). It is definitely using a hardware context because my code uses Geometry Shaders.
or we're one big EMP pulse away from losing almost 2 zettabytes of data.
and technicallly, there is only one SI definition of zettabyte, which is 10^21. The binary definitions used by the IEC like the zettabyte=2^70 are being renamed to avoid ambiguity (proposed to be zebibyte for zeta binary byte).
I'm just the opposite, I find I need to be more selective with games and music and if they don't grab me immediately, they aren't worth my time. For instance I loved Gothic I & II, but III's controls were so hideously broken I just gave up after an hour of play (combat was frustrating and always 10 on 1, generally a sign of bad game design). MoO3 from a couple of years ago I despised so much I shelved it after playing it once (despite hearing about a patch that supposedly fixed things) - MoO2 was one of my favorite games of all time. I enjoy UT3 and The Witcher, games I bought late last year, though I almost gave up on The Witcher due to display driver issues (it crashes the Vista driver after playing a bit, though the driver usually recovers). I tried to play the Orange Box a bit, but like most Valve games, it makes me motion sick (UT3 is the first of that series that makes me a little sick - UT2004 I could play for days).
I still listen to and buy new music, but I never was a prolific buyer like some friends with thousands of CDs. In fact, I get sick of listening to my old collection, which to be quite honest, isn't that large at maybe 200 CDs and 50 song downloads. My wife neither plays game nor buys music.
I see it more like this:
1) Loki, the only major Linux commercial game porting company failed
2) DirectX is the API of choice on the platform with the most marketshare
3) many of the latest features are inconsistently supported in OpenGL
4) most games assembled via APIs licensed per platform
on point 3, we have to wait for OpenGL 3.1 (3.0 streamlines the API but does not include new features) for many features in the over a year old DX10 API to appear because any EXTs are not required to be compliant with the API, so manufacturers like ATI can ship a DX10 compliant card that supports geometry shaders for DX and claim the card is OpenGL 2.0 or 2.1 compliant but skip support for EXT_geometry_shader (and I'm calling them out specifically because they reportedly have done it and have a history of doing just this sort of thing, especially with nVidia designed features).
on point 4 - the problem isn't so much with OpenGL - many game engine features are built to the Windows API. Want Havok Physics? Better have a console or PC, because it won't work with Linux and has a restrictively expensive mac license (according to rumor, identical to the Windows license). Compound that with sound API, a video API (for movies and cutscenes, e.g. Bink), a terrain API like speedtree, etc. Basically, modern games are no longer built from scratch, but assembled via a number of generally commercial APIs, some of which charge extra for additional platforms, some of which are Windows only.
Note that I did not mention OpenGL and Vista being incompatible - the truth is, you only get a performance hit in Windowed mode. I don't know why compositing gives a 30% performance hit (in my testing) because it should be more like 10%, but in full screen mode where it does no compositing it runs almost as fast as XP (consistent with DX on Vista).
tell that to Jesus's mom - at least in Catholic tradition, she was always a virgin, despite having a husband and according to some records, other children. Note that many reformers like Martin Luther and John Calvin do not believe in Mary's perpetual virginity.
Technically, some parts are wrong in how they work at an atomic level in other models such as quantum mechanics (friction, for instance), but as an observed model of the world it works quite well.
In other words, you're both correct - Newton's model is wrong in some cases, because it doesn't work for, say, observing an object traveling very fast, but it is effective in other cases such as observing a dropped object at sea level traveling through air with no wind.
All GMAs before the X3000 line (note the X, because the 3000 and 3100 is NOT included) did not contain HW geometry processing (texturing and lighting, vertex shaders and in earlier processors, scheduling). The problem is that the newer generations mostly include 8 unified shaders, which puts them at the bottom of the heap and these shaders perform poorly compared to dedicated cards (partially due to shared memory). Many modern cards have hundreds of shaders, though some lower end and mobile have as little as 16. For reference, the X3500's fill rate is slightly slower than a 2 year old nVidia 7300 but is capable of running 8xxx line shaders.
The problem is most graphics are moving to heavily into shaders because they're easy to write and add significant visual improvement for minimal effort, but chipsets like the 950 do this in software. This is what Tim is talking about - basically, you need to write a shader version and a non-shader version just to work with the Intel chipset - 2 different games.
Only the most modern GMA chips can even run modern games that are heavily shader dependent and they do those poorly due to having a slow clock, few shaders, and dependence on shared system memory. If the X3500 was a baseline that would be ok, but most use the older and dirt cheap 915 or 965 lines, and the 915 isn't even Vista Aero capable due to lack of hardware scheduling (Intel complaining about this was the reason MS lowered its requirements for Vista). Intel has downplayed this probably because they're weathering the storm, so to say, believing the future is in Ray Tracing and having shared memory is a boon for that (Ray Tracing requires access to the complete scene and therefore having it in main memory is good). I do not share Intel's belief that Ray Tracing is the future and in particular, the very near future - I believe it is part of it, but not entirely. By the time you tack on certain expensive effects currently dumped into shaders (namely diffuse lighting for soft shadows, which is at least O(logn) using photon mapping and as much as O(n^3) using radiosity, if I recall correctly), performance is near rasterization.
It can do curved surfaces, but is far from perfect and really is an approximation placed into a polygon structure (the surface itself is flat, but you don't see the surface itself). The silhouettes are done using 4 of the 16 quadrics as a reasonable subset to get an estimate curvature (4 to save on storage). These are pre-calculated and therefore have some overhead (either in a file or dynamically calculated on load). It is also restricted to hard shadows because it works similarly to ray tracing - tracing lines, not patches (it's hard to describe that, but think of it like when light hits an object, it reflects a line, not a cone), however, you get these shadows for free.
Incidentally, Steep Parallax Mapping and Interval Mapping also use pseudo ray tracers, but the surface curvature is unique to Relief Mapping (the quadric technique doesn't work with them). Most other techniques can be adapted to support soft shadows, however, and most game developers I know (3 professional) think soft shadows are more important than curved surface relief maps. If surface curvature is important, they'd rather tessellate it out.
I know this isn't what you meant, at least by how you phrased the rest of your comment, but technically light bounces are handled very well by raytracers, as is proper color absorption from nearby reflections as long as you are referring to Specular (shiny) lighting. What isn't handled well is non-point source diffuse (soft), which is why many ray tracers bolt on photon mapping or radiosity.
yes the spooks are forbidden to domestically spy without a warrant and so this is illegal. Unfortunately, Bush has allowed the NSA to do unfettered domestic spying (yes, constitutionally illegal and against the NSA charter) by executive order so if the agencies work together in any way they may have a dubiously legal way of getting the information, so proving all of the information was obtained illegally may be harder than it sounds.
In a sane world, "the decider" wouldn't be allowed to executive order himself absolute power. Not that I blame the Republicans for this trend in using executive orders to work around congress and the law - for instance, Clinton boosted FEMA's power so that it probably can declare martial law and create internment camps (as a national security directive, so it is not public). Only congress can declare martial law by law and I think internment camps are a moot point (free? ha!), so its clear to me at least that Presidents are abusing executive orders.
heh - that quote and the current candidates made me think of this comic from super Tuesday week.
Ron Paul doesn't follow the Republican doctrine close enough to garner favor. He's neither highly religious nor glued to the party line. Huckabee's popularity was based around being a conservative Christian, but some of his plans just suck (FairTax, which shifts more of the tax burden from those that make over $200000 to the middle class), so I'm glad he didn't win. I've never liked people that think God is on their side, which is what bugs the hell out of me about Bush, and I had a feeling Huckabee would be the same (note that I'm not saying belief in God is a bad thing, I just don't think God picks sides).
Ron wants to move use back to the gold standard so savings can't be eroded by devaluation of the dollar (by printing more money), but moving to the gold standard would be interesting, because Presidential Executive Order 6102, dated April 5, 1933 makes it illegal to horde gold (even by note), so unless that's repealed, it gives the government the right to seize any gold savings you might have (with some exceptions, but the $100 limit is pretty nasty). Gold has stability, but also the tendency to throttle the growth of the economy. Basically, the gold standard is like investing in Government Bonds and Fiat currency is like High Yield (Junk) Bonds - slow growth and slow rewards but generally stable vs high risk, high reward unstable. Personally, I wouldn't mind seeing a dual currency, but that brings headaches of its own (like 2 price tags).
shoot - I swapped that out for Reincarnation.
He's now a... uh, some kind of fruit bat. Best of luck in your new life GG!
No - DMA may help in some cases, as you describe, but you can tell a Firewire drive to copy to another Firewire drive when neither has any physical memory and it will still copy much faster than USB. The lack of a centralized controller (and device registration, scheduling, etc) actually helps keep overhead down. Note that USB can't do that - Firewire is peer-to-peer, meaning each device is aware of other devices in the chain. USB is a master-slave star network and needs a host controller (e.g. a PC).
Firewire was built a hot swappable, high speed replacement for SCSI, and is really more analogous to SATA than USB, but people compare them because they're both used as external buses for peripherals. USB was designed explicitly as a low speed, low power, low cost small peripheral handler (e.g. mice and keyboards) to replace a variety of miscellaneous specialized plugs such as game ports, parallel port, serial port, etc, and thus cost was most important and speed least. Firewire put speed first and cost last. As far as Firewire goes, I think a battle may be coming, with SATA's external plug eSATA, as I expect it to make some gains in the peripheral market, especially in storage. eSATA actually has an advantage over Firewire, because the actual device used for storage is often IDE and therefore Firewire has some conversion to do (ATA is the protocol, IDE the device - often they're used interchangeably).
The problem here is gullibility. Think of it like social engineering - someone calls and asks "We are verifying your bank account pin, can you give it to us?" and you saying sure - it's 1234! That's a lot like what this program is doing. In this case, the device at one end is saying can I have access to your memory? And the device on the other end is saying sure, despite the fact that that giving write access to memory is a lot like giving away your bank account pin (which is why it's really an OS issue, not a firewire issue). Some OS's like Linux only give read access, which means you can see what is in the account, but not take anything out, but Linux (and Windows) allow this to be set by the foreign controller, which is a bug.
DMA access should be limited to non-system memory, if allowed. Unfortunately, that isn't very controllable by current computer designs. I believe the solution proposed and implemented (I've heard about this for Windows 8, I believe) is encrypted floating addresses, so even if you have direct access to memory you don't know where to write it.
not to mention they list it as IEEE1394 port rather than firewire to avoid Apple's tariff on the name. My 9 month old laptop has one, my 2 year old homebrew desktop 2 (and I don't use either of them at the moment).
I think you overestimate the power of the tool - TomTom lists restaurants without ratings, so it's doubtful it will add tons of patrons. There are plenty of seats available, especially at dinner time (I think most people get take-out), and competition is heavy in the area. Incidentally, I noticed it's now in Google Maps, which is a fairly recent add. They still don't have another nearby Chinese restaurant that opened a year ago, however, and have one that went out of business 3 months ago - I should probably update it (the new location of TGI Fridays is correct in Google maps).
Jobs is a better example of vendor lockin - he wants everything as a disposable appliance.
It's almost impossible to find a large software company with multiple products that doesn't have some open source offerings, however, even if their main products are primarily closed source. Some examples are Apple, Microsoft [also see Codeplex], Adobe and Oracle.
Probably the best example I can think of for closed source is game companies like EA, Vivendi (Blizzard), etc. Carmack and Id are the exception, not the rule in that industry.
If you wait for a manufacturer to make all the corrections, you will wait forever because they can't check all places at all times and certainly wouldn't know all the best PoI and restaurants even if they're full time residents. For instance, both TomTom and Garmin GPS list a TGI Fridays that was a few blocks from my home as still in business when, in fact, it moved 2 miles away over 6 months ago and is being replaced by a new restaurant. There is also a fantastic Thai restaurant (it has won awards for best Thai) tucked behind a strip mall that isn't listed and I'd love to add it.
Personally, I like features like this on TomTom, but yes, an open source database would rock. Even something that pulled from google maps would be cool, IMO, as long as google maps stays free.
You'd be surprised at how some people you don't think need a lot of power actually need all of it. For instance, my mom can peg the CPU for 2-3 minutes with her large genealogy database. I do CAD for work, which requires boatloads more CPU than GPU, since the bulk of the work is tessellation and CSG, neither of which GPUs do particularly well. Yes, geometry shaders on the latest generation of hardware can do tessellation, but in practice they do it poorly due to hardware restrictions (or so I've heard - I don't actively develop in that area).
Convergence.
Once upon a time, high end machines came with a separate floating point processor. As cost came down, CPUs started integrating them and the FPU is now an integral part of all CPUs. GPUs are gaining the same sort of value, essentially adding sets of high speed parallel processors that perform tasks that can be offloaded from CPU. You might say that exists now with SSE/3DNow!, etc, but that has one huge flaw - separate instruction sets for each CPU. GPUs also have a much broader set of functionality (graphical and general purpose).
There's also the convergence with real-time ray tracing coming. The problem with ray tracing on GPU is that the rays need to be aware of the entire scene to be shaded properly (reflections, color bleed, etc). This is most easily done with high speed access to main memory rather than putting it all on a limited subset of memory on the GPU (GPU memory could still be used for texture lookups and other scene related things). As CPUs come out with more and more cores and higher speeds, real time ray tracing becomes closer to reality since it is highly parallel-izable and the breakpoint for when it is faster to ray trace vs shade polygons is rapidly converging. I personally don't think ray tracing will unseat polygons soon because the cost of adding in decent diffuse lighting (photon mapping is about the fastest) makes up for the speed gain, but perhaps programmers will "cheat" and do it in a shader, just as they do reflections in shaders for polygons today (usually cube maps).
LaptopVideo2Go is NOT dedicated to fixing driver problems - it's dedicated to installing newer reference drivers onto your laptop without the hardware manufacturer customizing and testing them, which is required by their OEM contracts with nVidia. nVidia is NOT required to provide a driver for your laptop at all.
A bit of history - not so long ago, laptops required custom drivers and these were provided by the laptop manufacturer, not the GPU manufacturer due to specialized display hardware. Newer laptops tend to work with reference drivers, however, and no longer need specialized drivers due to standardization in the video display industry. Some people like you require the latest-and-greatest drivers to fix bugs, so they resort to the reference drivers at a possible expense of functionality. All LaptopVideo2Go does is add mobile GPU values to a list of supported machines so you can install a much newer reference driver (at the possible cost of bricking your machine) because your laptop manufacturer is too slow at releasing an official driver that suits your needs.
ATI releases a reference driver, as well, the Catalyst Mobility driver, but they allow it to be installed by anyone, as I recall (I've never had an ATI laptop, only desktops). I'm not sure if that's because they have more control over the platform, or if they have no OEM requirement for support and provide it themselves, or if they are taking a risk and letting anyone install a reference driver.
It really depends on what you're looking for - on the mid-high end, AMD's Windsor line (X2 5200+ series) compared favorably to Conroes on a price-performance ratios. The slightly better performing 6750 was about $60 more expensive than an X2 6000+ (now it's about $30) late last year when I was looking (I still have my parts and price list for about 5 systems in the $800-900 range, 3 were Intel and 2 AMD). The X2 6000+ outperformed the Conroe 6600 and 6650 in many benchmarks, as well. On the high end, however, the Core 2 Extreme wins over the Phenoms (the 9500-9700, which are pretty scarce, if available at all) on both price and performance. Intel processors also completely own in overclocking, though AMD's 9600 Black is supposed to be pretty good (still not in Intel's league, but a start).
I purchased a laptop with a dual core Conroe last year and that works great usually - when it isn't, I blame Vista (I'm doing some certification testing on Vista, so no, Ubuntu is not an option).
I was seriously considering building a new desktop with one up until my car decided it wanted $1400 in repairs and maintenance. Too bad I don't have as much talent working on cars as I do computers.
ugh - while it was incredible compared to the first movie, Ernest Goes To Camp was incredible compared to the first movie. On the one hand, you had a couple of Z-list actors and actresses that actually are better at acting than the wooden, terrible performances of the A and B list talent of the first movie, but on the other hand, many of them are completely miscast and it shows (skinny pretty girl with no muscles just doesn't cut it as a fighter, sorry).
My favorite fantasy movie characters are probably: Matthew Broderick in Ladyhawke, Arnold Schwarzenegger in Conan, almost everyone in LotR and pretty much everyone in the Princess Bride, as well. I think Brigitte Nielson was a good cast for Red Sonja, but the plot and acting were so awful it loses a vote.