Remember the good old days of 3DFX, the Voodoos 1-5, how big they grown, how suddenly they fell? Every heard of the rare Voodoo 5 6000, the last card 3DFX ever produced? If yes, you might be interested in this post, spotted by Derek2k. This post shows how the card was developed, the specifications and what bench marks it achived in Quake 3. Voodoo5 6000 video card will never be available in the retail channel. No websites and magazines have ever received a Voodoo5 6000 card from 3dfx to preview and publish benchmarking results. In this page, we would like to shed some sidelights on this somewhat undissected card. Press Read More for further info!! Thanks Derek2k for this one one ;-)
Due to a problem with Geocities, I have copied the posted here to view. At the end of this post, you can view the original post which is still credited The development of Voodoo5 6000 had gone through several observable changes on the card as a results of problems encountered concerning heat dissipation, external power supply, stability and mainboard compatibility. On the first demo sample (Fig. 1) of Voodoo5 6000, collocations of the four video processors (VSA-100) were found entirely different comparing to that of the original design (Fig. 2) that 3dfx revealed 5 months before. In November 2000, 3dfx displayed the final version of Voodoo5 6000 3dfx at Comdex 2000. On this Voodoo5 6000 card, the Intel PCI-to-PCI bridge chip that had been on there before was replaced by a HINT PCI-to-PCI bridge chip (Fig. 1). At the Comdex, 3dfx announced that it wasn't going to bring the Voodoo5 6000 card to the retail channel and it has licensed Voodoo5 6000 technology to a Visual Simulation system manufacturer Quantum3D. On 15 December 2000, NVIDIA acquired "core business" of 3dfx and the dissolution of 3dfx company started.
Fig. 1 The picture of Voodoo5 6000
(click for large image)
Fig. 2 Original design of Voodoo5 6000
(click for large image) We didn't known exactly how fast Voodoo5 6000 run until FiringSquad posted a few Quake3 test results. To this day, one and only Voodoo5 6000 review published was the one at Voodoo Extreme. Here we would like to introduce a speculative method that will help us know more about the performances of Voodoo5 6000. First off, let see some facts. (1) Each pair of video chip and memory on Voodoo5 6000 works essentially independently: Voodoo5 6000 card has 4 video chips (VSA-100). Each chip has its own pathway to its 32MB of essentially independent SDRAM. Voodoo5 6000 runs with a mode that the textures in any scene must be duplicated in each set of 32MB SDRAM. (2) N sample FSAA essentially render a scene N-number of times and blends those scenes into one to remove jaggies. This is the multi-sample mechanism applied by Voodoo5 6000 to achieve FSAA. Now, according to the table of effective frame output listed below: Table 1. Effective Frame Output of Different Cards Effective Frame Output Voodoo5 4500
(1 x VSA100) Voodoo5 5500
(2 x VSA100) Voodoo5 6000
(4 x VSA100) 1 x (no FSAA ) 1 x Frame / 1 chip 1 x Frame / 2 chips 1 x Frame / 4 chips 1 x (2 sample FSAA) / 2 x Frame / 2 chips 2 x Frame / 4 chips 1 x (4 sample FSAA) / 4 x Frame / 2 chips 4 x Frame / 4 chips 1 x (8 sample FSAA) / / 8 x Frame / 4 chips We can speculate the performances of Vodoo5 6000 as:
Voodoo5 6000, 2 x FSAA = Voodoo 5 5500, no FSAA
Voodoo5 6000, 4 x FSAA = Voodoo 5 5500, 2 x FSAA
Voodoo5 6000, 8 x FSAA = Voodoo 5 5500, 4 x FSAA
This speculation method has been confirmed to be practicable by Voodoo5 5500/Voodoo5 4500. Table 2. Quake3 Performances Voodoo5 5500, 2xFSAA* Voodoo5 4500, no FSAA** 640x480x32 84 86 1024x768x32 37 37 * http://www.anandtech.com/showdoc.html?i=1382&p=7
** http://www.anandtech.com/showdoc.html?i=1382&p=8 Fig. 3 shows the Quake3 performances of Voodoo5 6000 and GeForce2 Ultra. The data were got from the benchmarking results of Voodoo5 5500 and GeForce2 Ultra on a Pentium III 1GHz system. Our results agree quite well with those published at Voodoo Extreme. It should be noticed that this speculation method could only get practicable results under 32bit color mode, because of the somewhat deviant behaviors of Voodoo5 cards on 16bit color mode.
Figure 3. Quake3 Performances of Voodoo5 6000 and GeForce2 Ultra
"demo001.dm3" test, "Normal" settings, color depth = 32bit, Screen_refresh_rate = 75Hz,
V-Sync disabled, Pentium III 1.0GHz, 128MB RDRAM, Windows 98 SE
**sources of benchmarking results for speculation: http://www3.anandtech.com/printarticle.html?i=1276,
http://www.anandtech.com/printarticle.html?i=1298 Someone speculates the performances of Voodoo5 6000 simply by the formula:
Voodoo5 6000, (no/ 2/ 4) x FSAA = 2 x [Voodoo 5 5500, (no/ 2/ 4) x FSAA]
The deficiency of this speculation is that it ignores the fact that a video card performances are limited severely by CPU at the low resolutions display, thus there are generally significant warp on the data got under those circumstances. From the benchmarking and speculation results, it can be know that Voodoo5 6000 is no over +5% faster than GeForce2 Ultra, while it has been show that there will be at least a +25% performance lead of NV20 over GeForce2 Ultra in the NV20 preview. So it should be smarter to wait for a just coming NV20 than rush out to bid that Quantum 3D Voodoo5 6000 PC (if it was there) at eBay News Source: Warp 2 Search
Further Info: Sidelights of Voodoo5 6000
Due to a problem with Geocities, I have copied the posted here to view. At the end of this post, you can view the original post which is still credited The development of Voodoo5 6000 had gone through several observable changes on the card as a results of problems encountered concerning heat dissipation, external power supply, stability and mainboard compatibility. On the first demo sample (Fig. 1) of Voodoo5 6000, collocations of the four video processors (VSA-100) were found entirely different comparing to that of the original design (Fig. 2) that 3dfx revealed 5 months before. In November 2000, 3dfx displayed the final version of Voodoo5 6000 3dfx at Comdex 2000. On this Voodoo5 6000 card, the Intel PCI-to-PCI bridge chip that had been on there before was replaced by a HINT PCI-to-PCI bridge chip (Fig. 1). At the Comdex, 3dfx announced that it wasn't going to bring the Voodoo5 6000 card to the retail channel and it has licensed Voodoo5 6000 technology to a Visual Simulation system manufacturer Quantum3D. On 15 December 2000, NVIDIA acquired "core business" of 3dfx and the dissolution of 3dfx company started.
Fig. 1 The picture of Voodoo5 6000
(click for large image)
Fig. 2 Original design of Voodoo5 6000
(click for large image) We didn't known exactly how fast Voodoo5 6000 run until FiringSquad posted a few Quake3 test results. To this day, one and only Voodoo5 6000 review published was the one at Voodoo Extreme. Here we would like to introduce a speculative method that will help us know more about the performances of Voodoo5 6000. First off, let see some facts. (1) Each pair of video chip and memory on Voodoo5 6000 works essentially independently: Voodoo5 6000 card has 4 video chips (VSA-100). Each chip has its own pathway to its 32MB of essentially independent SDRAM. Voodoo5 6000 runs with a mode that the textures in any scene must be duplicated in each set of 32MB SDRAM. (2) N sample FSAA essentially render a scene N-number of times and blends those scenes into one to remove jaggies. This is the multi-sample mechanism applied by Voodoo5 6000 to achieve FSAA. Now, according to the table of effective frame output listed below: Table 1. Effective Frame Output of Different Cards Effective Frame Output Voodoo5 4500
(1 x VSA100) Voodoo5 5500
(2 x VSA100) Voodoo5 6000
(4 x VSA100) 1 x (no FSAA ) 1 x Frame / 1 chip 1 x Frame / 2 chips 1 x Frame / 4 chips 1 x (2 sample FSAA) / 2 x Frame / 2 chips 2 x Frame / 4 chips 1 x (4 sample FSAA) / 4 x Frame / 2 chips 4 x Frame / 4 chips 1 x (8 sample FSAA) / / 8 x Frame / 4 chips We can speculate the performances of Vodoo5 6000 as:
Voodoo5 6000, 2 x FSAA = Voodoo 5 5500, no FSAA
Voodoo5 6000, 4 x FSAA = Voodoo 5 5500, 2 x FSAA
Voodoo5 6000, 8 x FSAA = Voodoo 5 5500, 4 x FSAA
This speculation method has been confirmed to be practicable by Voodoo5 5500/Voodoo5 4500. Table 2. Quake3 Performances Voodoo5 5500, 2xFSAA* Voodoo5 4500, no FSAA** 640x480x32 84 86 1024x768x32 37 37 * http://www.anandtech.com/showdoc.html?i=1382&p=7
** http://www.anandtech.com/showdoc.html?i=1382&p=8 Fig. 3 shows the Quake3 performances of Voodoo5 6000 and GeForce2 Ultra. The data were got from the benchmarking results of Voodoo5 5500 and GeForce2 Ultra on a Pentium III 1GHz system. Our results agree quite well with those published at Voodoo Extreme. It should be noticed that this speculation method could only get practicable results under 32bit color mode, because of the somewhat deviant behaviors of Voodoo5 cards on 16bit color mode.
Figure 3. Quake3 Performances of Voodoo5 6000 and GeForce2 Ultra
"demo001.dm3" test, "Normal" settings, color depth = 32bit, Screen_refresh_rate = 75Hz,
V-Sync disabled, Pentium III 1.0GHz, 128MB RDRAM, Windows 98 SE
**sources of benchmarking results for speculation: http://www3.anandtech.com/printarticle.html?i=1276,
http://www.anandtech.com/printarticle.html?i=1298 Someone speculates the performances of Voodoo5 6000 simply by the formula:
Voodoo5 6000, (no/ 2/ 4) x FSAA = 2 x [Voodoo 5 5500, (no/ 2/ 4) x FSAA]
The deficiency of this speculation is that it ignores the fact that a video card performances are limited severely by CPU at the low resolutions display, thus there are generally significant warp on the data got under those circumstances. From the benchmarking and speculation results, it can be know that Voodoo5 6000 is no over +5% faster than GeForce2 Ultra, while it has been show that there will be at least a +25% performance lead of NV20 over GeForce2 Ultra in the NV20 preview. So it should be smarter to wait for a just coming NV20 than rush out to bid that Quantum 3D Voodoo5 6000 PC (if it was there) at eBay News Source: Warp 2 Search
Further Info: Sidelights of Voodoo5 6000
