Talk:Scalable Link Interface
This is the talk page for discussing improvements to the Scalable Link Interface article. This is not a forum for general discussion of the article's subject. |
Article policies
|
Find sources: Google (books · news · scholar · free images · WP refs) · FENS · JSTOR · TWL |
This article is rated C-class on Wikipedia's content assessment scale. It is of interest to the following WikiProjects: | ||||||||||||||||||||||||
|
Petition to use this rewrite (or similar)
editI've taken the liberty to do a draft rewrite of this page at User:ShuffyIosys/sandbox, as I see the following problems:
- Contains way too much information
- Contains information that's outdated
- Contains pieces of information that may or may not enhance the article itself.
My goal of the rewrite is:
- Condense all of the information as much as possible, without losing so much detail an average reader will not be sufficiently informed of what SLI is. For instance, I took the configurations in "Other Implementations" and left it as a bullet point list.
- Remove non-important bits of SLI's history. I don't think anyone cares that Gigabyte made a dual GPU 6600GT card. It's old news and irrelevant to the present.
- Remove information that doesn't add value to the material. A lighter example would be SLI physics. That piece of information doesn't help add value to the article, and it's not really an SLI mode to split GPUs to do PhysX or rendering. A heavier one would be mentioning that Microsoft didn't include native GPU switching on Windows 7. It puts Microsoft in a negative tone, hurting the neutrality.
Message me if you approve, have suggestions, etc.
Notice
editPlease move the comparison of SLI/CrossFire to Comparison of CrossFire and Scalable Link Interface LaVieEntiere (talk) 21:16, 16 September 2008 (UTC)
x16 lie
edit<quote> In its early implementations, motherboards capable of SLI required a special card (colloquially known as a "paddle card") which came with the motherboard. This card would fit into a socket usually located between both of the PCI-Express x16 slots. Depending on which way the card was inserted, the motherboard would either channel all 16 lanes into the primary PCI-Express x16 slot, or split lanes equally to both PCI-Express x16 slots. This was necessary as no motherboard at that time had enough PCI-Express lanes for both to have 16 lanes each. Thanks to the advancement in available PCI-Express lanes, most modern SLI-capable motherboards allow each video card to use all 16 lanes in both PCI-Express x16 slots.
The SLI bridge is used to reduce bandwidth constraints and send data between both graphics cards directly. It is possible to run SLI without using the bridge connector on a pair of low-end to mid-range graphics cards (e.g. 7100GS or 6600GT) with Nvidia's Forceware drivers 80.XX or later. Since these graphics cards do not use as much bandwidth, data can be relayed through just the chipsets on the motherboard. However, if no SLI bridge is used on two high-end graphics cards, the performance suffers severely as the chipset does not have enough bandwidth <unquote>
Absolutly incorrect.
1. SLI is pure software technology. Any hardware use was a commercial trick. Ealier nVIDIA cards did nod need x16 interface, so "paddle card" allowed some x4 or x8 lanes to be used as board-to-board interface. Later (prior GT200) this link was arranged by the "SLI bridge" and the second card needed only x1 connection (to introduce itself to the OS) to work in SLI.
2. nVIDIA cards are working at very low PCIe clock speed - that is why x8 interface is not enough for them. But the total bandwidth even in triple SLI is less than 1 GB/s (which is bandwidth of the x4 PCIe v 1.1) - this allowed nVIDIA to use 1-1.2GB/s northbridge/processor link even up to their latest chipsets. The SLI bridge is used to allow direct GPU to GPU link avoiding the use of nVIDIA inadequate PCIe bridge. But on GT200 GPU up to x4 PCIe link is also used.Stasdm (talk) 15:12, 21 January 2009 (UTC)
Unfiled Discussion
editTwo high-end video cards seem like a very expensive proposition. If most PCIe, SLI arrangements use only eight lanes per card, the more practical method would be to put two GPUs on one graphics card. That way you would still have 16 total lanes, while one card would occupy less space on the motherboard.
True, and the model for that would be the 7950 GX2. It features two GPUs on two PCBs, connected by a PCIe switch which outputs on one PCIe connector. This allows for SLI on non-nVidia motherboards, however, there is a drawback. The motherboard BIOS has to recognize the PCIe switch connected to the PCIe slot and also recognize the two GPUs beyond that. Now, if you have a SLI motherboard, that could mean Quad SLI, though nVidia is withholding support for Quad SLI. Cnet.com has stated that Quad SLI is not noticable without a high-resolution TV or monitor of at least 30 inch. In benchmarking tests, it is noted that its closest rival is two 7900 GTs in SLI. However the onboard PCIe switch has a lower latency than the PCB used to connect the two 7900 GTs, resulting in better performance. You can find the 7950 GX2 between $550-$650. The 7900 GTs sell from $250-$350. Buying a twin 7900 GT now may be cheaper than a single 7950 GX2, but that leaves you with no more PCIe X16 slots. But keep in mind that the 7950 GX2 takes one PCIe X16 plus the slot next to it (due to the second GPU and HSF). Buying a single 7900 GT now may be a better buy, as ATI is about to release its R580+ GPU sometime this fall and nVidia is going to release its G80 sometime in spring 2007, and come with them dropped prices in current GPUs. Cnet.com stated that the gap between the 7950 GX2 and the next-generation is too close and if you're going to buy a 7950 GX2, you might as well wait for Christmas. --66.68.43.166 22:42, 17 July 2006 (UTC)
nVidia is withholding support for Quad SLI becouse their chipsets would not support the bandwidth - but Quad SLI is nicely working on AMD and INTEL chipsets!Stasdm (talk) 15:16, 21 January 2009 (UTC)
3dfx's scan line interleave
edit3dfx's scan line interleave was not the same as nvidia's scalable link interface. the two technologies functioned completely differently and were developed at different times by different companies. scan line interleave should not redirect to scalable link interface, so much as phong shading should not redirect to smooth shading (thankfully it does not). 141.211.231.115 22:24, 13 Mar 2005 (UTC)
- I agree with this comment - the technologies are quite different and basically unrelated. If they are both to be in the same article, it should be something like "Technologies allowing graphics cards to work in tandem", although more succinctly-phrased... – drw25 (talk) 17:46, 22 September 2005 (UTC)
Wikify Quad-SLI section
editWhen Quad SLI was on its own page it required wikification. I'm adding that now to the section here, 'cos it isn't quite clear. It doesn't fully elaborate on what quad sli is, and the paragraph is messy and condensed. I'll leave it to people more skilled than I... --Morbid88 07:46, 10 August 2006 (UTC)
- It did seem a bit messy, so I've tried to sort it out. The quad-SLI section was mostly about dual-GPUs, which has it's own section two paragraphs above it. Skorpus McGee 21:35, 2 April 2007 (UTC)
- I too will try to help improve this section and the other ones on this page which need cleaned. Edit: I'm new at editing, but it seems to me as if this section just keeps advertising old graphics card. John Aplessed (talk) 21:06, 21 August 2010 (UTC)
Why do you need a special mobo?
editBoth SLI and CrossFire appear to need a special motherboard. I can't find any explaination of why this is. It could be because you need to have two 16x PCIe lanes, but I don't think that's the case because I've heard of systems using two 8x instead. Even more confusing, why do you need a special mobo for the nVIDIA 7950, which only uses one slot?
Can someone explain this for me? Or better yet, put a good explaination in the article?
Maury 12:11, 15 August 2006 (UTC)
- The explaination is already there more or less. You primarily need a Nvidia chipset mobo with SLI precisely because Nvidia sells chipsets and therefore they can make more money by only officially support SLI on their mobos. There is no real technical reason AFAIK why non-SLI Nvidia based mobos and non-Nvidia based mobos can't support SLI provided they have the slots and enough lanes (most of the original mobos only had enough lanes to provide 2*8x lanes which I believe is enough for most of the current gen but not sure about quad SLI). AFAIK Crossfire is a little more complex and you do need to do something to support it Nil Einne 15:40, 20 September 2006 (UTC)
- Actually, it's pretty much the other way around. In the stone age of SLI, you were able to use an Intel chipset for it, because the nForce 4 did not exist yet. ULI had also made an SLI-supporting board by the use of 'hacked' older drivers. In both cases, the performance isn't as good as a proper nForce SLI chipset, which probably means the MCP on those boards has more to do with SLI than just doling out PCI-E lanes. On the other hand, Intels 975X runs CrossFire without issue, and is just as fast as the Xpress 3200 (or whatever it's called). -Skorpus McGee 21:39, 2 April 2007 (UTC)
The only difference in the non-nVIDIA and nVIDIA early SLI implementation is that nVIDIA chipsets supported switching off parity check, which allows higher transactions speeds at very slow (1/4 of standard speed) nVIDIA PCIe implementation (all modern chipsets support this feature). Some old nVIDIA chipdets did not support parity check at all (hence the still often met recommendation not to install RAID cards into "graphics" slots).Stasdm (talk) 20:21, 13 December 2008 (UTC)
Do you still need a special nForce motherboard to run SLI (as of 2013)? The way the article is written, it gives the impression that SLI is very difficult or impossible to do out of the box unless you have a specially manufactured nForce motherboard or use 'hacks'. Based on what I've read on numerous forums, it seems that all you need (for Intel mobos) are two PCI E slots. Or am I wrong? Genuinely confused here. confused 1:48, 26 December 2013 (UTC) — Preceding unsigned comment added by 2601:A:4180:64:9C2:D1AB:561E:8527 (talk)
Images
editThis page needs images of some of the huge graphics cards that people have these days hooked up together.... I want to see a free image for quad-SLI.... I myself only have one motherboard-sized graphics card so.... -Aknorals 18:45, 20 September 2006 (UTC)
High performance cards and SLI
editThe article mentions that often with high performance cards, the benefits of SLI are diminished and that in some cases performance might even suffer. Before I go out and buy two GeForce 8800GTX's I'd really like to know if this is true of the latest generation of cards. the article does mention that the issue has been addressed but how successfully? If it's an issue which has been addressed and eliminated, or significantly reduced, then is it now misleading to say that 'high end' cards have problems? Isn't it a difficult statement anyway in light of how the definition of 'high end' changes all the time? But the bottom line is I don't want to spend £1000 on two graphics cards if I'm only going to get £600 worth of performance. --GothMoggie 15:36, 23 November 2006 (UTC)
I believe this issue is regarding the fact that Day One drivers tend to be unoptimized and thus they won't show any real benefit for multi-GPU setups. This was most prevelent in the Quad SLI implimentation when 7900GX2 drivers were only OEM and while beta drivers for the 7950GX2 existed, they showed a lack of performance boosting until NVIDIA released the official drivers. In my opinion, if you can wait, buy two video cards and go at it. However, considering CPU bottlenecks (I believe the 8800GTX still has framerate caps on certain games because CPU bottlenecks), you should just buy one 8800GTX, unless you have a 24"-30" monitor to enjoy the high resolution-higher performance output the 8800GTX can do. -- XenoL-Type 18:35, 7 January 2007 (UTC)
Caveats section
edit- Anyone else think there should be a small note here about the power consumption requirements? Not sure if it's really a caveat or not, but might be noteworthy. 24.131.120.80 19:39, 30 March 2007 (UTC)
Pretty straightforward, if you double the number of GPUs then you will increase (not sure whether or not it doubles it) the power consumption. -RK 216.9.142.201 18:48, 4 April 2007 (UTC)
Multi-GPU article
editI was wondering whether the Scalable Link Interface, Scan-Line Interleave, and ATI Crossfire articles would not be better placed in one artcle because they are similar technologies. I realize that they work differently, but they all use two graphics cards to give better performance than would otherwise be possible with a single card. Just a suggestion. I will post in Crossfire talk page as well. -RK 216.9.142.201 18:57, 4 April 2007 (UTC)
- Well, the two technologies are severely different, each with their own histories, methods, technology, compatibility, and brands. The articles are already quite complicated on their own, and I think merging them would just result in a very large and unwieldy page. A good idea would be to make a multi-GPU page anyway, have it briefly describe SLI and CrossFire, and link to the respective pages. -Skorpus McGee 11:53, 8 April 2007 (UTC)
- An excellent Idea, mention the advantages and disadvantages of all three with an abbreviated history of each, but leave the actual technical description for the individual pages. This might help to clear up the articles a little also. More discussion? I don't have time to do this myself. -RK 216.9.142.185 18:12, 10 April 2007 (UTC)
Cleaned up - still confusing?
editI've reworded the article as best as I could to try and make it easier to understand. I'd love to just remove the "confusing or unclear" tag, but I'd like someone else to give the article a once-over first. -Skorpus McGee 11:50, 8 April 2007 (UTC)
- I have read the article and it is not confusing, so I will remove the tag. -- Vision Thing -- 11:39, 30 June 2007 (UTC)
3dfx financial collapse?
editWhen 3dfx collapsed financially, its intellectual property was purchased by NVIDIA.
Yes, 3dfx was in bad weather when it came to their finances, but they did not collapse financially. nVidia made 3dfx' board of directors an offer which was accepted before any financial collapse could/would have happened. NitroX infinity 12:10, 6 October 2007 (UTC)
I made some changes to this part myself. NitroX infinity 11:47, 13 October 2007 (UTC)
3-Way SLI and Crossfire comparison
editThe second paragraph of the triple sli section is purely speculation about the prices of the ATI multi- (more than 2)gpu solution, which has not been officially announced yet. The comparison to nvidia's triple sli is not quite fair, as it says that the cost for triple sli will be higher than the triple (or quad) ati setup, simply because there is speculation that the less inexpensive ati cards will be able to run in 2+ configurations. However, there will not be the same performance out of 3 hd2400 cards as there would be with 3 8800gtx cards, so although the minimum price will likely be lower,a similarly performing ati setup (with 3 hd2900 cards) will probably be close in price to the nvidia setup. I think this section should be removed, because it is unclear, based primarily on speculation, and does not make a fair or valid comparison. --71.235.45.130 20:55, 23 October 2007 (UTC)
I'd like to point out (due to recent edits made) that this technology is limited to the 8800 GTXs and 8800 Ultras on the 680i and 780i chipset platforms as of now. Separately, this technology is called "3-way SLI," not "Tri-SLI," "Triplet SLI," "Triple SLI" or anything else. Exitios (talk) 06:23, 22 January 2008 (UTC)
As an update to the last post 3-way sli is now compatible with either the 8800GTX, 8800 Ultra and the 9800GTX. This setup will work on a 680i,780i and the new 790i chipset motherboards —Preceding unsigned comment added by 77.96.146.63 (talk) 14:50, 21 May 2008 (UTC)
Havok being dropped
editI realise that there is no citation for hardware Havok processing being dropped, but there isn't going to be one. The reality of the situation is that Havok is never going to run on a GeForce, SLI or not; most hope vanished when nVidia bought Ageia, and what little was left disappeared when Intel bought Havok.
It deserves a mention that it won't be implemented, at best with a citation needed tag. Not because it's in dispute, but because, well... it needs a citation. nVidia don't do press releases of products that we won't see. If there is some site out there which is reliable enough for a cite, I'd like to see it. -Skorpus McGee (talk) 22:26, 21 August 2008 (UTC)
Image copyright problem with Image:Sli ready.jpg
editThe image Image:Sli ready.jpg is used in this article under a claim of fair use, but it does not have an adequate explanation for why it meets the requirements for such images when used here. In particular, for each page the image is used on, it must have an explanation linking to that page which explains why it needs to be used on that page. Please check
- That there is a non-free use rationale on the image's description page for the use in this article.
- That this article is linked to from the image description page.
This is an automated notice by FairuseBot. For assistance on the image use policy, see Wikipedia:Media copyright questions. --02:13, 31 October 2008 (UTC)
Hybrid SLI on Windows 7
editAccording to this article, Microsoft's Windows 7 will not naively support Hybrid SLI, saying that Hybrid SLI is unstable and gives bad performance. http://www.custompc.co.uk/news/605145/hybrid-sli-and-crossfire-unstable-says-microsoft.html Not sure wheather this should be in the article or not.--Crab182 (talk) 22:01, 6 November 2008 (UTC)
Consumer-centric article
editThis article is focused on the consumer graphics line. The entire Nvidia Quadro series is skipped. I'm still a wiki-newbie, but with some help we could add content regarding Nvidia Quadro Plex, Mosaic Mode, features and limitations present on their high-end products. —Preceding unsigned comment added by Qdotdot (talk • contribs) 22:20, 28 January 2009 (UTC)
White Hdmi input 16x exspress —Preceding unsigned comment added by 81.68.254.73 (talk) 10:41, 2 February 2009 (UTC)
1080p by 1080i or 720p/520p
hd ready tv or full hd intact
Unnecessary marketing
edit[QUOTE] In January 2009 Mirrors Edge on Microsoft Windows by DICE and distributed by E.A., became the first major title to add Nvidia PhysX to enhance visual effects in-game and add gameplay elements. [QUOTE] —Preceding unsigned comment added by Xowets (talk • contribs) 18:07, 7 May 2010 (UTC)
4-Way SLI
editThere is no information in this article regarding 4-Way SLI. Here are a few bits to start with:
-GeForce 400 Series only
-Requires motherboard with 2 nForce 200 MCPs and 4 PCI Express x16 slots (i.e. EVGA Classified X58 4-Way SLI)
-Requires 4-Way SLI bridge (included with compatible motherboards)
-Requires Windows Vista or 7, 32 or 64-bit (Linux?)
-Requires 1200-1500W PSU —Preceding unsigned comment added by 75.135.60.144 (talk) 01:44, 2 July 2010 (UTC)
crossfire
editThe constant references to crossfire need to be removed from the article. This is an article about SLI. not a consumer guide. It seems every time there is some thing presented about SLI there has to be a bit tacked on about how the same issue affects crossfire. That is irrelevant in this article.--Crossmr (talk) 12:12, 9 November 2010 (UTC)
Dual GPU cards, 3-way SLI, 4-way SLI, and Quad-SLI
editThe GTX 590 is a dual-GPU card. Quad-SLI means when you have two dual-GPU Cards in SLI while 4-way SLI means when you have four single-GPU cards in SLI. The GTX 460 and 560 are not capable of 3-way or 4-way because they have only a single SLI connector — Preceding unsigned comment added by 174.73.123.129 (talk) 01:44, 17 June 2011 (UTC)
AMD 900-Series Chipset Mainboards can run SLI
editI have added a note in the Caveats section for the time being mentioning that as of the AMD 990FX/X chipsets on mainboards, they can now run SLI configurations. I will revise the entire Caveat to reflect this soon (hopefully within 24 hours.) Vedalken (talk) 01:46, 27 June 2011 (UTC)
- Also current mainboards with some Intel chipsets like http://www.asus.com/Motherboards/Intel_Socket_1155/P8Z68_DELUXE/ seem to support SLI. Probably this caveat can be removed. -- 87.79.68.161 (talk) 10:08, 10 January 2012 (UTC)
Article needs to be OS-agnostic
editThis article needs a rewrite to specify which OS it is referring to on which case. This is due to SLI advertised by Nvidia in Windows, Linux, BSD, Mac, Solaris; but only usable under Windows. In other OSes, although advertised since 2006 is extremely underdeveloped. I have not tested it for BSD or Mac, but those are also need to be handled. 93.129.54.108 (talk) 19:00, 21 May 2012 (UTC)
400 Mhz SLi Link?
editIs there any proof that the SLi link runs at 400 MHz? or was this assumed since the RAMDAC's were 400MHz? NVIDIA claims the SLi link is a 1GB/s link. However, they also claim the SLi link limits the pixel clock to 400MHz so 4K displays and 120Hz 1440P screens through SLi is not possible without reducing refresh rate. 96.232.200.150 (talk) 16:03, 27 June 2012 (UTC)
Yet another mainboard which supports SLI
editAs we're not an advertising platform, I thought I'd just leave it here as a side-note. There is a special SLI Krait Edition of the MSI 970a-G4x series mobos (dunno if the Sp. Ed. is a G43 or G46 one as they're almost identical). JFYI, as the article only mentions the 990FX chipset. -andy 2.242.58.68 (talk) 09:24, 18 September 2015 (UTC)
A Commons file used on this page has been nominated for speedy deletion
editThe following Wikimedia Commons file used on this page has been nominated for speedy deletion:
You can see the reason for deletion at the file description page linked above. —Community Tech bot (talk) 14:22, 31 August 2018 (UTC)
Link in Reference #21 not working, found replacement
editThe link in reference #21 isn't working, it is giving me a 404 error. But I managed to get the link to the page by going to the website in question and going through their archive. I don't know how to properly change the link in the reference (without breaking anything) so I post it here and someone who knows can update it. https://www.tomshardware.com/reviews/vga-charts-viii,1184-20.html Thank you. 2A02:908:F765:F320:7C05:9EB6:2CC5:9D5D (talk) 13:20, 5 July 2022 (UTC)