Talk:Scan-Line Interleave

My sources suggest that 3dfx was not, in fact, the first to combine the power of two video cards. Around 1990, Silicon Graphics produced the SkyWriter, which contained twin graphics subsystems that could be coupled to interleave frame generation. Citations follow.

SkyWriter Technical Report - Hyperpipeline

Citation identifying publishing date of above report

Siggimoo 21:05, 31 August 2006 (UTC)

Cables

It's not true that a pair of Voodoo cards communicated by ribbon cable AND Vga-to-vga cable. Two voodoo cards are connected esclusively by the ribbon cable, and only one of them is connected to the 2D graphics board via VGA to VGA cable (the monitor is connected to this card's output), —The preceding unsigned comment was added by 82.193.5.35 (talk) 11:47, 16 March 2007 (UTC).

Accuracy

I think that this page is somewhat inaccurate. I owned a Voodoo II SLI setup in 1998, and my experience was completely different. The setup was easy, it significantly increased performance over a single-card solution, and the game incompatibilities were minimal. This article definitely needs sources.

Aerst2 (talk) 20:27, 20 March 2008 (UTC)

The hardware configurations the Voodoo2's were used in varied wildly, you can't tack it on one person's experience whether it's easy to get the cards to work well or not. I personally have four Voodoo2 cards (2 x 8MB, 2 x 12MB) and they're all a nightmareish experience to get working properly.

Windows 98 didn't have any virtualized hardware resources, so many devices in the system had to share IRQs and other resources. Voodoo2's took a big linear chunk of I/O, which usually couldn't be changed and often conflicted with just about every other device in the system. You more or less had to strategically install devices one after another in specific order so they found the resources they liked and nothing tried to edge in on them. 207.119.2.32 (talk) 06:26, 21 December 2010 (UTC)

I want to remove the Criticism section

The Criticism section is full of inaccurate information and original research. I plan on removing the entire section unless somebody can defend what is written there.

The primary flaw in the Criticism section is that it concentrates the discussion on SLI's effect on frame rates (FPS). Voodoo 2 SLI was never marketed as increasing FPS. Therefor I don't see how you can criticize its lack of effect in that area. The primary benefit to using Voodoo 2 SLI is enabling the 1024x768 resolution on Voodoo 2 (a single Voodoo 2 card could only produce a maximum resolution of 800x600). This information seems to be missing from the article.The Goat (talk) 14:57, 30 September 2008 (UTC)

Criticism updated with references

I've added two references that support the arguments made in the article, that is to say, little to no performance increase in frame rates. In regard to the above section here in discussion, it is true that the SLI enabled higher resolutions, but that's another matter. 3dfx did in fact market the arrangement as a performance increase across the board. —Preceding unsigned comment added by 97.118.240.6 (talk) 21:52, 13 December 2008 (UTC)

For example, from quantum 3d http://www.quantum3d.com/press/htmlarchive/3-31-98a.html "The Quantum3D SBSLI approach combines two Voodoo2 graphics subsystems that operate in parallel to provide twice the fill rate performance of single Voodoo2 based graphics accelerators"

and

"“The release of Quantum3D's 16MB and 24MB, dual-processor Obsidian2 X-series products will bring a new standard of performance to the market. We're proud to include the Obsidian2 in our fastest systems, giving our customers the highest level of graphics performance ever offered outside of military simulators."

But the benchmarks referenced with this arrangement do not reflect this.

Aside from that, increased resolutions are based largely upon RAM, not fill rates. Citing increased fill rates points towards the assertion of an increase in frames per second, rather than resolution. In other words, the arrangement, when citing fill rates, is being marketed as an increase in performance speed.

It's very difficult to find 3dfx commercials on SLI or images from magazines, as the company was bought out years ago and all of their public references yanked. Third parties were at the time however, given the impression by 3dfx that 2 cards meant double the frame rates, as is obvious by their websites claiming this, and linking those claims to now defunct 3dfx articles. For example, "The Voodoo2 introduced Scan-Line Interleave (SLI) to the gaming market. In SLI mode, two Voodoo2 boards were connected together, each drawing half the scan lines of the screen. For the price of a second Voodoo2 board, users could essentially double their 3D throughput. " http://www.nationmaster.com/encyclopedia/3Dfx#SLI —Preceding unsigned comment added by 97.118.240.6 (talk) 22:12, 13 December 2008 (UTC)

I would recommend adding to one of the sections on the article outlining the proven benefits of the SLI arrangement, such as the increase in resolution as you've pointed out and is easily verified, as opposed to an article deletion that now has references. —Preceding unsigned comment added by 97.118.240.6 (talk) 22:01, 13 December 2008 (UTC)

I still support removing the section entirely. 3dfx never marketed SLI as increasing framerates. The guru3d.com article you linked to clearly states that SLI increases 3d performance by enabling 1024x768 resolution. The other criticisms are pure hogwash. Not compatible with every computer because it requires two PCI slots? So every article on wikipedia concerning computer expansion boards should include criticisms complaining that an expansion slot is required. No PC hardware is 100% compatible with every PC ever made. That is not a legitimate criticism.The Goat (talk) 15:30, 15 January 2009 (UTC)

From the article that you say, "The guru3d.com article you linked to clearly state SLI increases 3d performance by enabling 1024x768 resolution".

Taken from the article; "combines two Voodoo2 graphics subsystems that operate in parallel to provide twice the fill rate performance of single Voodoo2 based graphics accelerators—without requiring dual graphics boards (and associated PCI slots) and inter-board SLI interconnection cabling assemblies. On an Intel Pentium II 333MHz system running id Software's Quake-II benchmark, “Demo 1”, the X-16 and X-24 achieve 84.4 frames per second at 1024x768 resolution—the highest of any product benchmarked to date. " —Preceding unsigned comment added by 97.124.207.215 (talk) 00:36, 3 March 2009 (UTC)

The issue of hardware compatibility is a real one and not a matter of opinion. You seem to be bothered that this some how criticizes the company or technology as a byproduct, but it's obviously not a criticism of either of those issues, rather an explanation as to the one of the difficulties 3dfx had in proliferating the concept. Furthermore, if you bothered to check any of the references (which it seems you haven't) you'll see that there are sites linking back to the original 3dfx articles (which are now gone), and quoting 3dfx articles that you can cross-reference with other linking sites to see that many third parties were reporting the same claims held by the manufacturer; that the fill rate would be doubled with two cards. This is a direct claim of fps increase with the SLI arrangement. Fill rate was the biggest factor in determining fps. A doubling of fill rate would mean an increase in fps, so that is a claim to performance increase outside of resolution gains.

The criticism section is entirely relevant to the topic as SLI was a commercial failure and ultimately led to the demise of the leading company in that sector. It's also important to understand why a similar technology, "Scalable Link Interface" which uses multiple graphics cards albeit with different implementation is so popular, while this endeavor was not. —Preceding unsigned comment added by 97.124.207.215 (talk) 00:27, 3 March 2009 (UTC)

I still think you should add the improved resolution. List criticism but no benefits? sounds like someones being preferential to me, not to mention how wiki articles are suppose to be NEUTRAL. complete failure? 3dfx used it again with the Voodoo 5 and 6, these cards were good performers if the game didn't use hardware T&L(which you can turn off or emulate it with a patch), and then the agp corruption bug with the 6(which even matched performence with later cards like the geforce3 if you could get your hands one of them). Also this leading to the demise of the company? come on, the voodoo 2 was good, LOTS of things lead to 3dfx's demise. even without sli they were still good then. like SLI is now, its just something for the hardcore enthusiast. —Preceding unsigned comment added by 76.123.100.84 (talk) 12:28, 6 March 2009 (UTC)

Get your facts straight. Voodoo 6? No such thing. The voodoo 5 was an even bigger failure than the SLI concept to begin with. The company was already irreversibly ruined by the time the v5 went to market. V5 matched performance of Geforce 3? No, sorry, it was on par with the Geforce 2, and some flavors of the 2 (the Geforce 2 GTS even) smoked the v5. That's without considering that the v5 was stuck at dx7 at second generation, and introduced as dx6 in a dx7 mainstream world. —Preceding unsigned comment added by 97.118.241.44 (talk) 01:34, 18 March 2009 (UTC)

I said the Voodoo 6 matched performance with it, not the Voodoo 5. Get your facts straight, Voodoo 6000 exists, but theyre prototypes and really rare: http://www.voodoo3dfx.com/Evap/SS_Foro_Recopilacio/2_v6k_Foro.jpg —Preceding unsigned comment added by 76.123.100.84 (talk) 17:08, 26 March 2009 (UTC)

This http://www.tomshardware.com/reviews/voodoo,85-19.html refutes what has been claimed in this article. I have removed the performance criticism for 2 reasons. 1) the guru3d reference did not note what resolution the benchmarks were run and the thg article only tested SLI at 640x480. 2) From the link I gave, SLI provided obvious benefits at 800x600, but less obvious at 640x480.

So in effect what you found was a source that claimed something different, and changed the article to support your source over other verified sources. I would think that with conflicted accounts from different sources that it would be discussed from both angles, as opposed to just subjectively removing the one you happen to disagree with. I notice that you also happened to remove my sources which support everything which was stated in the article. —Preceding unsigned comment added by 174.16.82.207 (talk) 22:19, 18 April 2009 (UTC)

"I said the Voodoo 6 matched performance with it, not the Voodoo 5. Get your facts straight, Voodoo 6000 exists, but theyre prototypes and really rare" The voodoo 6000 is a voodoo 5, it's not a voodoo 6. Besides as you stated, it's a prototype, so that hardly relevant here unless you wish to discuss and compare all experimental technology from all sources with a consumer product. Ridiculous. —Preceding unsigned comment added by 174.16.82.207 (talk) 22:16, 18 April 2009 (UTC)

I gave you my reason for removing your article. Learn to read better. —Preceding unsigned comment added by 72.229.118.39 (talk) 04:16, 25 April 2009 (UTC)

The problem is not so much with not understanding what you're saying (i.e. "learn to read better"), rather that your reasons lack merit and are inappropriate justifications for article deletions. —Preceding unsigned comment added by 97.118.201.253 (talk) 18:11, 16 October 2009 (UTC)

Removing unsupported reference

In the criticism section it states, "(It can now be seen that the original 3DFX SLI was very effective, more so than even current implementations by Nvidia & ATI, capable of achieving near 100% performance increase over a single Voodoo2 card. The reason this could not be observed before was that the CPU was the limiting factor.)" But, there is no indication on the reference supporting this statement. As a matter of fact, benchmarks with SLI cards show almost no difference. On page 3 of the thread the images show little difference. Images are located here; http://i40.tinypic.com/2r559ns.gif http://i43.tinypic.com/f595l5.gif

Since the statement is not supported by the reference (3), it is being removed. — Preceding unsigned comment added by 63.225.125.86 (talk) 03:51, 29 August 2012 (UTC)

Remove San Francisco Rush from the article

@Voodood as far as I am aware, arcade SF rush does not use SLI. The quote you cited reads:

Atari is as professional as they come with its impressive new San Francisco Rush , a racer built around the Voodoo Graphics dual chips

I don't think this is referring to using SLI; I think this is just referring to the fact that the 3DFX Voodoo has two chips, one for the pixel pipeline and one as a separate texturing unit (the FBI and the TMU)

https://vintage3d.org/3dfx1.php

Voodoo Graphics is made of two chips- usual pixel pipeline and separate texturing unit. This allowed 3dfx to scale solution with versatile chip counts for arcade machine[s, a]nd very high end PC boards from Obsidian.

https://thandor.net/list_object/list_chipset/1

The Voodoo Graphics chipset consists of two chips. A TMU and FBI chip. The TMU is the Texture Mapping Unit and places the textures onto 'things'. The FBI (Frame Buffer Interface) combines colors, adds fog and is able to do alpha blending, dithering and stuff like that. These two chips together are the actual Voodoo Graphics chipset.

This is not the same as SLI, which involves using two 3DFX Voodoo chipsets in unison. The TMU and the FBI together are only part of one Voodoo chipset.

Additionally, checking the MAME documentation at https://github.com/mamedev/mame/blob/1fed8407104656580e8454e884510989704324a5/src/mame/midway/seattle.cpp#L50, the game indeed just uses a singular Voodoo 1 chipset of just the TMU and the FBI.

I think San Francisco Rush should be removed from the article, but I wanted to ask you first, just in case I may have missed something in my research.

Additionally, while "Such hardware was cheaper to develop for than proprietary systems, and Atari used the savings to sell the game at a lower price to arcade operators." is correct, it's not really relevant to the article, and is better suited for an article focusing more broadly on the Voodoo 1 itself, or the Voodoo line as a whole, like https://en.wikipedia.org/wiki/3dfx (as far as I'm aware, the Voodoo 1 does not have its own article yet). Heroponrikibestest (talk) 05:06, 1 September 2024 (UTC)

Well, frankly, I'll remove it. Coming to think of it, it's jarring to put it in the article. Voodood 06:29, 1 September 2024 (UTC)

Thank you! Sorry about that. It's nice to see the VSA-100/Voodoo5 citations, too; this article definitely needed more sources, so I'm glad to see two more. Heroponrikibestest (talk) 06:37, 1 September 2024 (UTC)

You're welcome! ;P Ooh Yeeah! Voodood 06:39, 1 September 2024 (UTC)

Wanted to issue a minor self correction, for anyone viewing this later- slightly misread the mame documentation at the time. San fransico rush actually uses two TMU chips and one FBI chip, which is also where I assume the original confusion over it being SLI came from. That said, no reverting needs to be done, as, at least as far as I've seen, this still isn't an example of SLI, which specifically refers to linking two seperate cards/chipsets together. Arcade SF rush, while certainly more advanced than a "normal" voodoo 1 chipset, still doesn't seem to be doing anything that would be considered SLI. As a point of comparison, the Voodoo 2 also has 2 TMUs and one FBI (as already noted on its respective article) and it would obviously be incorrect to claim all Voodoo 2 cards are already doing SLI simply because they have 2 TMUs. Heroponrikibestest (talk) 16:08, 1 November 2024 (UTC)