Search<%BANNER[cool_130_l]%>
<%BANNER[left_130x300]%>
<%BANNER[left_130x130_2]%>
InformationX-bit Labs for mobile users! Do not forget that we are running a special version of X-bit Labs web-site for users of mobile and handheld devices: http://pda.xbitlabs.com. Check out our news and articles from smartphones and PDAs to be always updated on the latest computer and technology news. <%BANNER[right_130x600]%>
|
<%BANNER[top_768x90]%>
|
||||||||||||||||||||||
|
<%BANNER[banner_468x60]%>
Articles: VideoS3 Savage2000 PreviewCategory: Video by FastSite [ 09/08/1999 | 12:00 AM ] There is one single day left till the announcement of a new nVidia GPU. However, suddenly S3 offered the world a nice surprise: a new revolutionary product. Here is a bit on this newcomer. Table of contents:
A few weeks ago there was hardly any info about the new product from S3 except the unprovedrumors. Remember GeForce 256 (codename NV10) from nVidia and Napalm from 3dfx: in fact, therewas and there still is so much credible info and interesting suppositions. However, the newchip from S3 couldn't boast being so known. For example, there was an opinion that the newchip will be called GX4 and that it will supposedly have an integrated geometric acceleratorresponsible for transformation and lighting. And a week before the official press release oneof S3 managers accidentally said that we could expect the products such as GX4 very soon. The30th of August finally came and the Web started swarming with the info about the new chip fromS3 - Savage2000. Why did S3 choose particularly this date - 30 August, and not 30 September,for instance? Everything is very simple. The matter is that not so long ago S3 was announcedto be buying a very well known company - Diamond Multimedia. Of course, the announcement of anew product with really promising features accompanied by the grown attention of the mass media,analysts and managers of the company's partners (as well as competitors) has every chance to exertpositive influence not only over the general image of S3 but also over the company's shares. So,the launching of a new chip short before closing a very serious deal seems to be a really cleverstrategic move. Especially bearing in mind the rumors that not all Diamond Multimediashareholders approve of this deal. However, now we have every reason to predict that S3will succeed in buying Diamond Multimedia, which will also mean getting a perfect sales andmarketing network with all the necessary support. Besides, S3 will also acquire progressivetechnologies in the field of telecommunications and multimedia as well as a number of wellpromoted trade marks. Moreover, the logo of Savage2000 includes a stylized diamond, i.e.Diamond :-). Besides, there was another reason for S3 to announce Savage2000 exactly on 30 August.The matter is that the next day, 31 August, was planned by nVidia to become the turning pointfor the whole world: it was going to announce its new super product. And this event was infact very well prepared. The tension started growing long before this day and up to the lastminute the curtain of mystery was down, which was thoroughly controlled by the company. As aresult, on 31 August, 1999 nVidia officially announced their new offspring GeForce 256. Ofcourse, all the info about the GPU, which is the official name of nVidia's new integratedchipset, was prepared in advance. However, S3 with the launching of its Savage2000 managedto slightly spoil nVidia's triumph. Remember that nVidia web-site proudly states that GPUGeForce 256 is the first integrated solution for the mass market with the hardwarerealization of T&L block (Transformation and Lighting). In other words, GeForce 256 canboast an integrated geometric accelerator. But the previous day S3 already announces Savage2000,their version of a GPU, which also has an integrated geometric accelerator. That is why thepriority belongs to S3 despite all the matters. The present article is devoted to Savage2000chip (or GPU, if you like it more), because it was announced before GeForce 256. Thereforewe will turn to GeForce just in case we really need it for some purpose or better explanation.And a bit later we will undoubtedly dedicate a separate article solely to this chip fromnVidia. And now back to our today's hero - Savage2000. Let's take a closer look at the productoffered by S3 this time. You will hardly find a guy, who may be greatly surprised at the all-in-one integrated chipas well as at the particular chip modifications intended for different market sectors. Here S3decided to follow a beaten track, which is not in the least reproachful or shameful. TheSavage2000 chips family is an integrated 2D/3D/Video solution. For now S3 has announced onlytwo chip versions: Savage2000 and Savage2000+. However, we would like to stress that it isonly "for now", because the history gave us to understand: the need or sudden technologicalachievements may force new modifications to appear despite all the plans or predictions.Savage2000 chip is intended for OEM market and Savage2000+ - for the consumer market. Thechips differ only in working frequencies of the graphics core and memory bus. The mini tablebelow may help you to compare them:
And these differences seem to be the only ones, all the other features of both chipversions are absolutely identical. That is why later on in this article we will speaksimply about Savage2000 implying the whole chip family. So, let's study the specs of Savage2000: General Characteristics
 2D Acceleration Features
3D Acceleration Features
Flat Panel Desktop Monitor Support
DVD/DTV Support
Full Software Support
And now we would like to dwell on the most interesting details. We decided not todiscuss the 2D part since it is hard to imagine anything new there. Of course, you canspoil something if you do your best, but as for S3 Company, they don't have any problemswith 2D for a considerable while already. So, we suggest passing over to the most excitingpart right away. First a few words about Savage2000 architecture. In fact, Savage2000 appears to be the first, which can boast two key architectural solutionsalready realized: single-pass Quad-Texture engine and S3TL. The first one is a graphics engine,which allows superposing four textures: two over one displayed pixel per time step. Moreover,there is also a regime when all the four textures are superposed over one single pixelsimultaneously. Note that this option (four textures superposition) isn't yet supportedby the applications but it is a very good step forward, which will be very useful in thefuture. It is worth mentioning here that the possible appearance of a new technologycan't be completely excluded. Namely we mean the technology when the rendering iseffectuated by two pipelines simultaneously in two different frame buffers and then theresults obtained are combined and displayed on the monitor. It may also turn out a technologysimilar to T-Buffer from 3dfx. Well, time will show. All this means that Savage2000 is equippedwith two rendering pipelines with two texturing blocks in each and at the same time the superposingof up to four textures over one pixel is also supported. Just for you to compare. The today's leaders in the 3D graphics market don't have suchbrilliant opportunities. For example, nVidia Riva TNT has two rendering pipelines with onetexture block. The double pipeline architecture of Savage2000 together with the Single-passQuad-Texture engine is expected to provide high scene fillrate in both types of applications:those using multitexturing and those, which do not resort to this technology. The higher is thefillrate the higher is the fps, which will undoubtedly increase the gaming performance at higherresolutions. In other words, there is every reason to believe that Savage2000 based graphicscards will be very fast at least compared to the present day leaders in the market. We would like to remind you that the graphics cards based on Savage2000 chip can have up to64MB of local graphics memory. The size of this local memory together with the high performanceof the chip itself can provide the user with comfortable gaming conditions at 1600x1200 in 32bitcolor depth with 32bit Z-bufferization, triple bufferization and with the 32MB free localgraphics memory. However, this memory can be used not just for texture storing but mainly fortexture caching. In this case AGP texturing will also remain demanded. The giant textures willbe loaded from the AGP memory for static geometric scenes. And from the texture cache of thelocal memory will also load the textures for dynamic geometry. Note that Savage2000 of coursesupports the texture compression technology S3TC licensed by Microsoft and included into D3D,and due to texture compression you can save 192MB (!) of textures in only 32MB local memoryprovided for texture caching. In this context it is very appropriate that Savage2000 can boasta fully fledged 128-bit local graphics memory interface, instead of that specific 64-bit one aswe saw by Savage4. However, despite all the innovations Savage2000 chip can't boast hardware support ofEnvironmental Mapped Bump Mapping or Dot Produce Bump Mapping. This is actually quite strange.In fact, the words "Hardware Bump Mapping" in the specs imply one and the same Embossing BumpMapping, which can't boast being extremely interesting. And it's a pity really. S3 engineerswere probably too busy trying to finish the work as soon as possible, or simply didn't payenough attention to the realization of progressive bump mapping techniques when the projectof a new chip was accepted. And now we are slowly going to the sweetest part. But today we will begin with the qualityof the image created. If we take a look back at Savage4, we will hardly deny that despite allthe drawbacks (especially concerning the performance) Savage4 based graphics cards provide veryhigh quality 3D image. We shall even say, practically the best 3D image. There are no reasons forus to think that Savage2000 based graphics cards will fail to cope up with this task as good asits predecessors did. Besides, Savage2000 is equipped with an integrated geometric coprocessorS3TL (S3 Transformation & Lighting). Note that the support of a special geometric coprocessor bythe graphics cards is a long awaited and violently discussed point in the Internet. And now wehave a beautiful opportunity to discuss the first officially announced 3D graphics chip with theintegrated geometric accelerator. Since S3TL is one of the most interesting peculiarities ofSavage2000 we will pay particular attention to this issue. Especially keeping in mind that weexpect this geometric coprocessor to provide not only significant performance increase, but alsohigher realism and better image quality. Here we suggest a short theoretical tour. What is Hardware Transformation and Lighting (T&L)?The process of generating a 3D scene from the supplied data (a 3D "world") is carried out by twofunctional units. The first functional unit is a "geometry engine", which transforms the 3Dcoordinates of the world into 2D coordinates for the monitor. This has to be done because themonitor on which the 3D world is to be viewed is a 2D device, which cannot directly accept 3Dcoordinates. The second functional unit is a "rasterization engine" which performs the task ofdrawing now the 2D world onto the screen. In addition to transforming 3D coordinates into 2Dcoordinates, the geometry engine can be used to generate lighting information for the world -storing information about light sources and calculating the amount of light each source castson any given point in the world. While rasterization engine (which includes the 3D renderingpipeline) is a part of a standard 3D video card ubiquitous on consumer PCs over the last fewyears, it has only been the high-end workstation markets that have had access to hardwaregeometry calculations. Now, with the announcement of its Savage2000 accelerator featuring acompletely integrated geometry engine, S3 is bringing hardware transformation and lightingtechnology to mainstream PC market for the very first time. S3TLA tightly integrated design, S3's highly efficient geometry engine is optimized to workdirectly with S3's advanced rasterizing engine. This integration is expected to provide thehighest possible performance during the 3D scene rendering. The chart below represents thescheme of Savage2000 chip:  First, the 3D data making up the scene to be rendered is accepted, which occurs directly fromthe CPU or, more efficiently, from AGP/local video memory, so that the CPU is never involved andall the calculations are carried out by the geometric coprocessor integrated into the graphicschip. And the free CPU resource can be used for some other purposes, for instance for thedevelopment of a more progressive artificial intelligence (AI), the simulation of the world'sphysics (gravitational force, air density, etc.) in games. This data then passes through atransformation stage, in which it is converted from the original 3D scene data (known as"object space") into 3D data as seen in the view to be rendered ("view space"). In the viewspace each point can be lit from the stored lighting data, before transforming the data intothe position of the eye of the user - the correct position for viewing (known, obviously, as"eye space"). From here it is converted to 2D screen coordinates and corrected for distance(an operation known as perspective transformation). Finally, the resulting data is clippedto the screen - eliminating all data that is off screen to avoid wasting effort in later rendering.The transformation and lighting calculations require very powerful computing capacities. They can becarried out either by the CPU or by the geometric coprocessor. These several stages of coordinatestransformation and lighting calculations are made as a pipeline and the calculations are directlyeffected by the FPU block. As a result we get the hardware T&L realization called S3TL. The FPUperformance is the main factor that influences the amount of polygons used for the tessellation,i.e. the presentation of the uneven surfaces with a series of triangles. The more powerful is theFPU of the geometric coprocessor, the more polygons can be used to present the object and as a resultthe more realistic image you will get in the end.  Unfortunately, S3 never announced the computing capacity of its S3TL. But the notorious GeForcefrom nVidia allows processing up to 15 million triangles per second. We can probably supposethat S3TL has quite a comparative potential. The OpenGL Lighting ModelS3's hardware lighting model is based on the OpenGL lighting model (which with a very smallnumber of changes - also supported in hardware - is equivalent to the Direct3D lighting model).There are quite a bewildering number of parameters that can be used to control each individuallight (S3 supports the OpenGL standard of 8 simultaneously), but OpenGL lights can be dividedinto three basic types: Directional lights, Point lights, and Spot lights. Let's dwell on thelighting types especially taking into account the growing popularity of this topic, which islikely to be very acute during the next couple of years. Directional LightsA directional light is the simplest type of light that can be used. Here, the light sourceis treated as a point an infinite distance from the object being viewed. The light rays thatarrive at the object are therefore parallel. An example of a directional light is the effectof the sun at the surface of the planets in its orbit. Although it is not a point source,and it is not an infinite distance away, the rays that arrive at the surface are sufficientlyclose to parallel that the effect is the same. (The simplification here is pretty rough,because we all know that the light rays do not spread linearly, however, it is quite possiblefor the current 3D graphics). All that's required to specify a directional light is itsdirection, color and intensity (brightness). Point LightsWhile directional lights are excellent for modeling light sources completely outside thescene to be rendered, light originating from an object inside the rendered scene usually needsto be modeled by a point light, which adds position and attenuation to the parameters requiredto specify the light. A typical point light source would be a candle, which casts light in auniform sphere in all directions. The addition of attenuation allows the intensity of the lightto decrease with the distance from the object on which the light is cast. In software lighting,i.e. with the CPU, point lights are usually significantly slower than directional lights -requiring approximately twice as much time to calculate. This is not the case with hardwarelighting, i.e. when we involve the power of the geometric accelerator (S3TL), where the twotypes of lighting have very similar performance. Spot LightsSpot lights are the most complex type of lights as they add the ability to define a coneoutside, of which the light source cannot shine. This is useful for modeling searchlights,flashlights, headlights or any other application where the area of light fall must be carefullycontrolled. Spot lights require the most computational effort to generate the resulting colorvalues and, as a result, are rarely used in traditional software methods (with CPU). Withhardware, there is still a performance penalty, but it is relatively small and certainlyless than a factor of 2. With spot lights, two separate colors are calculated for each vertex: the non-reflective or"diffuse" component, and the reflected or "specular" component. The diffuse component considersonly the material of the surface and the colors of the light while the specular componentsimulates the light reflecting directly from the object such as the reflection of room lightson a polished table. To find the final color value for lighting a vertex, first a total diffuse and a totalspecular color are generated by summing the respective components created by every light inthe scene. An ambient color contribution, simulating other lighting in the scene, is thenadded to the diffuse color and then these final diffuse and specular values are applied tothe surface. Direct3D and standard OpenGL differ slightly in the way that they do this.Direct3D applies the diffuse color before texture and the specular color after texture,while OpenGL applies both beforehand. An enhanced OpenGL model exists that works the sameway as Direct3D and this is generally thought to provide a more realistic effect. For more details, including formulae that describe the lighting process, see the current GLspecification. At the time of writing this could be found at:www.opengl.org/Documentation/OpenGL12.html(Section 2.13.1) While it is possible to support more than 8 light sources simultaneously, OpenGL onlyguarantees support for 8 lights and applying more than 8 lights to an object tends (as thecontributions from all the lights are summed) to result in one thing - white! Games, as filmsand TV, use more dramatic lighting, which comes from using one or two powerful, tinted lightswithin a scene for global illumination, while using other highly localized light sources toprovide additional special effects. Applications can easily determine the most importantlight sources to apply to each individual object and will use many light sources within ascene without needing more than 8 for any single object. What Gives the Use of S3TL?With the advent of hardware geometry, 3D cards are now capable of rendering entire sceneswith little or no CPU intervention, which will revolutionize the 3D applications industry.Instead of spending 50-95% of its time moving and transforming data, static geometry can beprepositioned in AGP or video memory, and fetched directly by the 3D engine with massive bandwidth.As well as saving CPU cycles, this technique significantly reduces system memory bus traffic, whichreduces competition for the same memory being accessed through AGP. Even if the static geometry(such as walls, floor, labyrinth ceiling in the game levels) is calculated in AGP memory,system memory bandwidth requirements are reduced by a minimum of two times, and may even reachfour times or more. Finally, dynamic geometry (for instance the monsters or heroes moving) stillgets the transformation benefits, without the bandwidth improvements, for a 40-50% speedenhancement. This is very important for game developers in particular, who are urgently seeking more CPUtime for the powerful, CPU-hungry artificial intelligence (AI) and physics engines currently indevelopment. Why Use S3TL When Texture-Based Lighting is Available?Rapidly emerging as an extremely important technology in recent years, texture-based lighting(lighting maps) has shown an amazing ability to render realistic looking scenes. A powerful toolthat we expect to continue to be very important, light-mapping is a technique that is complemented,not eliminated, by hardware lighting support S3TL. Lightmaps are excellent for generating static lighting solutions as offline-renderingengines can use slow radiosity-based lighting engines that produce very realistic results,but take excessive time. Radiosity is a graphics technology of creating natural lighting(including indirect lighting) basing on a physical model of light spreading in a complexarchitectural environment. This technology allows calculating exact ways the light goesand effectuating the colorful scene visualization. However, lighting maps don't produce particularly good dynamic lighting solutions. Theeffects produced are CPU-intensive, tend to be restricted to casting unrealistic spheres oflight around objects and the "grid" of the lightmap texture is often visible. Dynamicallyrelighting an entire world - for example, changing the lighting when a light source is destroyed -are very difficult problems to solve. The picture below shows the example of aliasing (visualization error) of Quake2 lightmap.  Note the jagged edges around the light sphere cast on the ground by the firing of thegun. Hardware lighting allows dynamic lights to be easily and quickly calculated by the 3D enginein the process of rasterization, and their effect can be added to that of the lightmap used bythe standard rendering pipelines with multitexturing support. The equation for generatingaccurate static lighting with the help of the lightmaps (in multitexturing) with the addeddynamic lighting calculated by the geometric accelerator looks as follows: ({diffuse + lightmap} * base_texture) + specular, which sounds in the following way. The diffuse color component is added to the lightmaptexel and the resulting sum is superposed over the base-texture, and after that the reflectedspecular component color is added. This can be easily programmed into any Direct3D compliant multitexturing hardware using onlytwo textures. The illustration below shows the example of the lightmaps used together with dynamicdiffuse lighting.  The wall is a lightmapped surface. The bright lighting spots are the vertex dynamic lighting. Hardware vertex lighting is also the first step towards consumer 3D accelerators thatsupport per-pixel (often mistakenly called "Phong") lighting in hardware - an approach thatshould be able to provide much better results than multitextured solutions. Introducinghardware vertex lighting to developers now will create a solid base of applications thatsupport hardware lighting and are ready to take the next step into photorealistic 3D. What about Faster CPU's with 3D-Optimized Instruction Sets?While the introduction of technology designed to accelerate 3D transform and lightingoperations from Intel(SSE) and AMD(3DNow!) has significantly increased the ability of theCPU (to be more precise the FPU block) to handle geometry tasks, a 3D hardware geometryengine still has a very valuable part to play. In a typical case, the number of CPU cyclestaken to generate a vertex on these 3D-enhanced processors has been reduced by 3-4 times.Yes, as the CPU has to use the same bus to connect with system memory and the 3D chipset,feeding the data to the 3D chip was already a significant performance bottleneck before theintroduction of transform enhancements. The ideal situation is for the chip to feed itself by picking up its own geometry data fromvideo memory or AGP memory, without ever being touched by the CPU. New instruction sets makeminor enhancements by better prefetching for reads, but are fundamentally restricted by theavailable data bandwidth across memory buses. As a result these enhanced mathematics engineson the CPU can be used for the more general tasks for which they are better suited: 3D worldphysics (mass, inertia, etc.), AI, dynamic modeling, procedural texture generation. And the main advantage provided by the special geometric coprocessor of the graphics cardis a greater range of opportunities for the software developers to put into practice theirdreams and to give way to imagination. Applications Support of S3TLTaking the best advantage of hardware geometry features isn't just a hardware issue. Supportis also needed at the application level - both in implementation and design philosophy (in otherwords, the support should be introduced in the drivers, interfaces and the applications).Implementation issues are the easiest to solve. Some existing applications will benefit immediately. Any game making use of OpenGL willgain an immediate performance benefit. Among games like that are Quake2 and Quake3 Arena from iDSoftware, Inc. This benefit gets larger if the application uses OpenGL display lists to compilegeometry into a format that can be natively understood by the 3D geometry on the hardware.Microsoft DirectX 7 also natively supports hardware geometry as applications that previouslyused Direct3D transform pipeline can be ported to hardware geometry now with just trivialchanges. And, re-implementing existing software-transform engines to use hardware transformationis usually not a difficult task. All in all, about 30% of the games launching for the 1999 holidayseason will support hardware transformation, i.e. by the geometric accelerator. Of course, this doesn't really take advantage of what can be done with hardware geometry(read S3TL). In particular, most Direct3D and OpenGL applications are still doing lightinginternally (utilizing the system CPU, of course) instead of passing the information to API toallow accelerated hardware lighting. Applications can quickly be enhanced for hardware coordinates transformation. The vastincrease in polygon counts available (doubled at a minimum; while ten times or more is mostlikely) will encourage developers to do so. Some changes are very simple. Model dynamiclevel-of-detail changes can be made at farther distances from the user, improving distantobject fidelity. Models can also be re-tiled to show smoother detail and better edges. Thisis likely to be an easy process, since the ultra high-detail models used to generate FMV (FullMotion Video) sequences for games can instead be used in the games themselves. Finally, hardwarelighting can be used to replace applications' internal lighting pipelines, and additional lightsand effects are usually an easy enhancement both in implementation effort and visible effect! In the near term, application developers - particularly game programmers - will use a similarparadigm to the one employed in the early days of hardware rasterization (when the first 3Daccelerators came out into the mass market), where the application will be able to use bothhardware (with the geometric accelerator) and software (utilizing the CPU) to do transformationsat the choice of the user. In practice, while running the game you will be able to define if youwant to use the geometric accelerator or not. It will be quite natural that as high performancemass-market hardware geometry engines - with a demonstrable increase in performance oversoftware geometry processing - become available, S3 expects the crossover period, whenthe applications are made for two single methods of geometry calculations, to be short. Let's Sum UpWell, we can state a number of facts without the slightest hesitation or doubt: the worldhas definitely changed, but it was S3 who really did it just a day before nVidia. Yes, S3 Inc.was the first to officially announce the chipset with the integrated geometric accelerator S3TL.Of course, we don't want to deny the merits of nVidia in terms of the huge advertising campaignof GeForce 256, which was so remarkably arranged. But the priority now belongs to S3. You knowthat GeForce 256 from nVidia, as Savage2000, has an integrated geometric coprocessor, which isthe main connecting element between these two new products. We feel sure enough to predict thatthe term "GPU" (GraphicsProcessor Unit) will become very popular in the near future, althoughthe name can hardly influence the performance. ;-) We believe there is every reason to say that Savage2000 will undoubtedly be faster than anyof the today's 3D graphics chips. Even if we disregard the S3TL. As we have already mentioned,Savage2000 has two rendering pipelines with two texturing blocks in each. This peculiarityallows Savage2000 to operate two pixels and to superpose two textures over each pixel pertime step. There has never been anything of the kind in the mass market before. The appearanceof the geometric accelerator S3TL will allow increasing the visualization quality significantlydue to the hardware lighting and rising the overall performance due to the hardware calculationsof the polygon vertex coordinates. Unfortunately, there are no exact numbers, which could showthe performance of S3TL in coordinates transformation. And to tell the truth we don't know howmany polygons the geometric accelerator integrated into Savage2000 can process. For the timebeing we can only try to guess, but we believe that this value lies somewhere between 12 and18 mln triangles per second. This value is indirectly proven by S3 saying that S3TL will allowincreasing the number of polygons used in the scene rendering by 4-10 times compared to thesoftware geometry. By the way, GeForce 256 also has a few peculiarities, which appear for the first time in theproduct aimed at the mass market. Here we would like to point out the four-pipeline architecture,which lets GeForce 256 render four pixels per time step. Now it is pretty hard to foresee whichgraphics card and on which GPU will work faster and provide better 3D image. We would refrain fromexpressing our opinion on the matter since we believe that there is not so much time left beforethe first graphics accelerators based on these chipsets appear. As far as we know, the first cardsbased on Savage2000 are expected to come out in November this year. Of course, if nothingextraordinary and fatal happens in September, the first graphics card based on Savage2000 willbe from Diamond. The retail version on Savage2000+ GPU with 64MB local graphics memory, AGP 4x,TV-out and maybe also with a special port for digital monitors is supposed to cost about $250. As for the graphics cards based on GeForce 256, they are expected to step into the market abit earlier, probably in October already. And the price will be almost the same since thememory is also almost of the same size. The new chips from S3 will cost impressively little money: $29 for Savage2000 and $35 forSavage2000+ if the supply is at least 10,000 pieces. Note that Savage2000 will be manufactured with a more progressive technology - 0.18 micron,than GeForce 256, which will be made with 0.22 micron process. It means that GeForce 256 chipswill be more expensive than the one from S3. However, the cost and the profit are not the onlynotions, and thee is also such thing as competition, which seems to cause no problems for bothsides. By the way, we think that it is exactly the 0.22 micron technology, which forced nVidiato limit the clock frequency of its chip at 120MHz, otherwise, the chip could simply getoverheated. This supposition is proven by Leadtek's statement that their cards based onGeForce 256 will be equipped with temperature monitoring devices. nVidia will supposedlyshift to 0.18 micron technology only next year and GeForce's clock frequency may get higherthen. As for S3, the shift to 0.18 micron technology is not a novelty for this company. It hasbeen using this process to manufacture the chipsets for mobile solutions for half a yearalready. Frankly speaking, we don't feel like telling you a lot about GeForce 256 now because we areplanning to make it a separate topic for discussion in our next articles. That is why don't getlost till then :-). And for a few more minutes back to S3. In our article we supposed that Savage2000 would be ableto process from 12 to 18 million triangles per second. What made us think so? The speed of thetriangle vertex coordinates calculation depends on the computing capacity of the FPU block,doesn't it? Well, our opinion is based on the fact that S3 is known to have everythingnecessary to create such a powerful floating point unit. And where did S3 manage to getall these opportunities? Again the things are very simple. In 1997 at the auction was sold apackage of patents developed by Exponential Technology Company, which was working on x86 PowerPCcompatible processor ordered and financed by IBM. So, the company suffered great technical andfinancial difficulties, which resulted into the urgent sale of its all property and in the firstplace of its patented technologies. And at the auction some company, which wished to remainanonymous, bought the whole package for 10,000,000 USD. This event gave way to numerous rumorssuspecting Intel to be this unknown buyer, which seemed quite logical. However, it suddenlyturned out to be not Intel but S3. This meant that from then on S3 got the real chance to developits own x86 processor because the technologies, which were at the company's disposal, allowedcarrying out such work. Some analysts even predicted that Intel was about to acquire a new powerfulcompetitor, however, S3 chose another way. In the beginning of 1999 S3 and Intel signed across-license agreement and established long-term partnership, which made S3 Intel's AGP 4xvalidation partner. Besides, Intel bought about 25% of S3 shares. In other words, the companiesmaintained close friendly relations. Moreover, there appeared some rumors about another secretagreement. Anyway, each of them got what they wanted. That is why we have almost no doubts thatthe FPU block by Savage2000 is rather powerful, and probably even cooler than that of GeForce256. The announcement of the new graphics chips generation Savage2000 by S3 has once again proventhat it deserves being called an innovator and technological leader. If you remember it was S3that developed the texture compression technology S3TC, which was licensed by Microsoft and thenbecame a standard and was included in DirectX. It was S3 that designed the first graphicsaccelerator with single-pass tri-linear filtering support. And finally, it was S3 that appearedthe first to announce a new graphics accelerators generation with the integrated geometriccoprocessor (S3TL). That is why all it has to do now is just to develop the achieved successand to strengthen its position. Everything will solely depend not that greatly on the timelylaunching of the graphics cards based on Savage2000, but mostly on the high quality drivers tothem. Especially ICD OpenGL. Anyway, the first prize will get the company, which manages to winthe heart of the end user. But don't forget that 3dfx also doesn't keep hands in pockets andvery soon, maybe already in October, it will also please our sight with a new chipset. As for us, all we can do is to have patience and to wait for the first graphics cards onSavage2000. Latest materials in category: Video
<%BANNER[banner_468x30]%>
Article Rating |
Category NewsCategory: Video Friday, July 25, 200812:31 pm Channel Vendors Demand Graphics Cards Suppliers to Recall Potentially Faulty Nvidia GeForce Graphics Cards. Resellers Want to Return Potentially Faulty Nvidia GeForce Graphics Boards to Makers Thursday, July 17, 20085:48 am Microsoft Preps to Unveil DirectX 11 Features in Several Days. ATI, Nvidia, Microsoft to Discuss DirectX 11 Techniques at XNA, Siggraph Wednesday, July 16, 200812:30 pm New Generation ATI Radeon for Mainstream, Mobile Markets are Ready. PCI-SIG Approves ATI RV730, M98-L, M96 Graphics Chips 7:22 am EVGA and XFX Reimburse Price Difference on GeForce GTX 200 after Price Collapse. EVGA and XFX to Return Money to GeForce GTX 200 Purchasers Tuesday, July 15, 20084:23 pm Startup Promises to Revolutionize Multi-GPU Technology Early Next Year. LucidLogix Unveils Hydra Distributed Processing Engine All Latest News <%BANNER[right_130x130_1]%>
|
||||||||||||||||||||||
<%BANNER[foot_728x90]%> | |||||||||||||||||||||||
Print article
Send link to friend
Digg this
Post to del.icio.us
Post to Slashdot
All of us are impatiently awaiting the announcements of new nVidia and 3dfx graphicschipsets. And it is quite understandable, because they promise us much more realistic 3Dgraphics. Sometimes they even speak about photorealistic graphics. But on the other handthe 3D graphics market can boast another powerful figures and "dark horses" as well. Wehave already touched upon the possible future of the graphics accelerators market and wehave also expressed the opinion that nVidia and 3dfx don't need to hurry up with theannouncements of the new products and technologies. They are most likely to decide on awaiting game. And in this respect all our expectations came true. S3 Inc. appeared thefirst company, which had officially announced its integrated chipset from the new chipsetgeneration. Besides, S3 has also prepared a name for its coming offspring: Savage2000. Whyso? Well, first of all Savage trade mark is pretty famous in the market, and secondly, 2000is a very popular number, which symbolizes the beginning of the new millennium. Franklyspeaking, you can argue if the choice of such a name is really justified. Take, for instance,Savage3D and Savage4: they are mostly intended for low-cost solutions market. While Savage2000,and we will undoubtedly return to it once again a bit later, is aimed at least at customerswith average income but higher demand for 3D accelerators quality and performance. As forthe number 2000, it is quite an ordinary thing and shouldn't be regarded as something soimpressive. All in all, S3 marketing department didn't spend too much time imagining thebest move to make, or maybe they just didn't have enough time. Anyway, seems that everythingis clear with the name. And now let's speak a little bit about the time S3 chose for thechipset launching: why now? Why did they decide to announce Savage2000 at this particulartime and not a month later for instance?
