Until recently, if you wanted additional video card power, you had to buy a new card that was both more powerful and more expensive. Instead of investing all that money on a newer and faster card, why not just buy a duplicate of the one you now have?
There is a workaround, though, if you have an Nvidia card and use their multi-GPU technology “SLI.”
By utilizing a SLI bridge, Nvidia has developed a proprietary parallel processing technology that allows multiple GPUs to connect and work within the same PC.
Scalable Link Interface (SLI) is the acronym for the multi-GPU technology’s key concept of scalability.
Using multiple GPUs to render frames simultaneously for a single display output, a machine’s graphics processing power can theoretically be considerably increased in this way.
The SLI technology was not always owned and developed by Nvidia, as it is now. When 3Dfx was at its peak in the 1990s, it competed with Nvidia and ATI in the field of 3D graphics processing and graphics card technology. SLI was the invention of 3Dfx (now AMD).
Three graphics processing units (GPUs) were combined on a single card for the first time with 3Dfx’s introduction of Scan-Line Interface (SLI) in 1998. The Voodoo2 was the first product to utilize SLI. As the first consumer-oriented video card solely focused on gaming, the Voodoo2 paved the path for today’s thriving market for high-end gaming gear.
A ribbon cable was used to connect two Voodoo2 cards so that they could share data and produce the screen by alternately generating horizontal pixel lines. Compared to single-card competition, the technique provided 1024768 resolutions, a significant improvement.
After an intensive marketing campaign, SLI technology claimed to double the processing capabilities of a personal computer (PC), In reality, there were numerous flaws with the technology, most of which were caused by the necessity of sacrificing PCI slots in order to accommodate two Voodoo2 units at the expense of other components.
To add insult to injury, there were also display imperfections such as incomplete frames and tearing in the actual rendering process itself. The prohibitive cost of acquiring two cards (around $300 each at the time) proved a barrier to the widespread adoption of the tech, and it remained the preserve of niche enthusiasts eager to brag about their gaming setup.
Due to a messy merger with STB Systems, sluggish sales, and the inability to compete with the exponentially stronger GeForce and Radeon line of cards, 3Dfx’s fortunes were swiftly going downhill. Nvidia purchased the pioneering graphics card manufacturer in late 2000, and production and support were soon to be discontinued.
It wasn’t until 2004 that Nvidia reintroduced the SLI technology as the Scalable Link Interface, designed for PCI-e slots rather than the older PCI slots.
For many recent Nvidia GPUs, SLI is supported by the majority of motherboard chipsets, including the Intel X and Z series.
However, the core principle of using two parallel GPUs stays same despite a plethora of new and improved hardware.
Specifics and Features
When using SLI, one card serves as the “master” and all other cards serve as “slaves.” The master card is responsible for delivering the completed 3D visuals to the display. 2-Way, 3-Way, and 4-Way SLI are all examples of this technology’s ability to connect two, three, and four GPUs at simultaneously.
The master-slave relationship and data flow between two identical GPUs are ensured by a SLI bridge. Allowing the units to communicate directly, rather than through the motherboard’s chipset, avoids the problem of vying for bandwidth. Bridges come in four varieties:
- Standard SLI bridge with a 400 MHz pixel frequency and a transmission capacity of up to 1GB per second.
- SLI LED bridge with a 540 MHz pixel clock and the capacity to emit RGB LED lights when in operation is similar to the aforementioned standard bridge.
- This high-bandwidth SLI bridge operates at 650 MHz and has transfer speeds approaching 2GB per second. There’s a good chance you’ve seen this bridge before.
- Nvidia’s RTX GPUs use the NVLink bridge, the most modern and most advanced form of interconnect. It can transfer data at speeds of up to 100 GB per second.
Using SLI is as simple as distributing the rendering effort equally across both GPUs. Most of the time, one card will work on a specific group of frames while the other does the opposite. Together, they form a complete frame that is ready to be displayed on a computer monitor.
There are now three SLI modes in use:
- It is possible to divide the rendering load into horizontal pieces based on 3D complexity and distribute it to the GPUs in this manner.
- Rendering alternate frames sequentially with each GPU is called Alternate Frame Rendering (AFR). All of the frames in the game can be rendered by one or more graphics processing units (GPUs). As a result, the frame rate is raised.
- SLI Antialiasing, in which the antialiasing capabilities are improved two-fold by distributing the processing task among the GPUs in an offset manner. Visual quality improves dramatically, but performance suffers as a result. If you’ve ever seen “SLI 8X” or “SLI 16X” in the settings of a game and wondered what it meant, now you know!
A collection of driver updates meant to get the GPU ready to work with the game and improve its visual performance are published by Nvidia as a SLI profile when they’ve established a stable base of settings that don’t break the game.
Relevance and Criticism
Since the glory days of the GTX 600, 700, and 900 series of Nvidia GPUs, SLI support has been steadily decreasing. During that time period, the majority of games attempted to implement SLI, and both Nvidia and game developers made a concerted effort to do so. SLI proven to be a cost-effective method of doubling a computer’s graphics capability for consumers..
Dual-GPU setups are becoming less and less necessary for most gamers, as more powerful single GPUs, such as the 1000 and 2000 series, are capable of handling the demands of “ultra” settings in today’s games.
Because SLI is only used by a small percentage of gamers, only a small number of new titles can benefit from profile building. Rather than coding their games to support SLI, developers choose to focus on features, bug fixes, and modifications that affect a high percentage of single-GPU gamers. For purists aiming for 5K or higher FPS and resolutions, it should at the very least not break.
There are fewer individuals buying SLI-capable machines since developers are less likely to focus on SLI support due to the lack of demand for multi GPU configurations. It’s a never-ending circle.
Games that support SLI often don’t achieve the intended 200 percent scaling, and the SLI technology is prone to micro stuttering (particularly in Alternate Frame Rendering mode) and other performance concerns that aren’t present in single GPU rendering. A lot of people are put off by this because of this.
SLI isn’t compatible with a lot of games, both new and old, therefore you’ll have to disable it to play. Price considerations are once again raised as a result of this. Many people don’t understand why it makes sense to spend twice as much money for so little return. This tendency is also influenced by the power consumption of a SLI system.
Finally, Nvidia appears to be gradually phasing away SLI support as newer models are released, which is the best indication that the technology is on the decline. Only the RTX 2080 and RTX 2080 TI of the most current RTX series enable NVLink, which is required for two SLI GPUs. Because these models are so expensive, it’s unlikely that anyone outside of techies will make use of them.
DirectX 12 does not currently support SLI, which further diminishes SLI’s prospects. The number of games that can take use of SLI is rapidly dwindling as more and more developers switch to the more flexible and up-to-date API.
SLI adherents, on the other hand, say otherwise. There is, however, a definite impression that this group is keen to justify their massive investment in a top of the line dual GPU system with a brazen and hearty dose of cognitive dissonance.
Scalable Link Interface (SLI)
So, what is the SLI acronym for? When it comes to running two graphics cards side-by-side, how does SLI benefit me? In 1998, Nvidia gave us the Scalable Link Interface (SLI), a small jewel.
After 3dfx was purchased by Nvidia, the technology was put on hold for a spell.
It wasn’t until 2004 that Nvidia re-released SLI with upgrades to make it compatible with PCIe buses, which had just become mainstream.
Multiple GPU setups raise compatibility issues (covered in more detail below), however SLI allows for the simultaneous usage of up to four cards. Using a parallel processing technique, it breaks down the information into smaller parts and allows many cards to work on various portions at the same time, resulting in a more efficient use of resources.
SLI anti-aliasing, SFR, and AFR are the main modes of operation for Nvidia graphics cards in SLI. The Nvidia Control Panel, which should have been installed on your computer when you installed your visual card drivers, allows you to pick and adjust these modes.
Split Frame Rendering (SFR)
As many cards as there are in the frame are split horizontally and the task is distributed among them. Depending on what’s in the frame, it can decide how to split the frame.
Suppose that the top half of the frame is largely vacant and static, while the bottom half is filled with moving characters. If the frame is divided 50/50, then one card would be doing more work than the other. This means that the card responsible for the upper portion of the screen may get around 60% of the screen real estate, while the card responsible for the bottom portion of the screen will only have to handle 40% of that portion. (Please keep in mind that these percentages are merely examples and not exact figures.)
Alternate Frame Rendering (AFR)
In this mode, each card renders a full frame, and the frames are queued up. As a result, if you’re using two cards, one will handle frames with an even number, and the other will handle frames with an odd number. However, it can also lead to difficulties such as micro stuttering, making it less efficient than SFR for obtaining greater frame rates.
When you witness visible stutters on your screen due to a mismatch in frame rendering timings between GPUs, this is known as micro stuttering. There will be a delay if the cards are not correctly synced or if they lack sufficient information to guess the next frame that needs to be rendered.
Rather than increasing frame rates, this setting provides improved image quality. Anti-aliasing is shared between the cards, resulting in a better overall image quality.
Allows for more advanced anti-aliasing options in games to be unlocked (such as SLI 8x, 16x, and even 32x on quad-SLI systems).
As a result of computer pixels being squares, spherical images might appear jagged at their borders. Anti-aliasing is an antidote. An anti-aliasing technique fixes this by sampling pixels surrounding the issue area, and then filling in the best possible assumption.
Compatibility and Issues
You’ll only need to buy two graphics cards to get the job done. Any two Nvidia cards will do, but what about a third? Unfortunately, things aren’t that straightforward. There’s a good chance it won’t be…
Two (or more) of a Kind
Choosing the right graphics cards is the first step; not all cards will work together. They must both have the same GPU and the same amount of video RAM in order to work together.
There is no difference between a GTX 1070 TI 8GB built by Asus, as well as a GTX 1070 TI 8GB from MSI in terms of performance. They aren’t the same GPU, so they won’t work together.
If your graphics cards do not have a SLI port on their back, you will need to double-check that they are SLI-compatible before purchasing. For the GTX 10 series, Nvidia discontinued support for 3 and 4 card SLI.
Slots and Plugs
Your motherboard and power supply are next on the list. SLI compatibility is required for both of these.
In other words, you’ll need enough PCIe x16 slots on your motherboard and power connectors on your power supply to support the number of GPUs you want to use.
It’s a frequent fallacy that adding more graphics cards to a SLI setup will allow you to achieve quadruple or even quadruple the amount of video RAM. As each card has to access the same information at the same time, Nvidia SLI only uses the RAM from one card.
SLI Bridge Series
You can utilize a regular bridge if you have more powerful cards (like the GTX 1080), but you won’t be able to get the full performance out of them. This would be possible because of the high-bandwidth brides.
That’s all the information we need to understand SLI. Is it worth telling your wife that the kid would have to wait while you get a second and third GTX 1080 TI?
That’s when things start to become a little fuzzy. Despite Nvidia’s assertion that double the performance can be achieved with twice the cards, this rarely happens.
SLI profiles must be introduced by game creators because games don’t come pre-configured to take advantage of multiple video cards. This is a difficult and time-consuming task, thus most developers avoid it.
Unless you’re playing a AAA title, this means that performance might not improve at all (in fact, it might even go down slightly).
It’s impossible to predict exactly how well a certain game will perform. In other words, you should be aware of this before deciding to invest in the technology or in particular games. Just in case, here is a list of SLI-certified games.
Be careful that we don’t end up being the lone voice on this issue. Is SLI a viable option for you, or are you simply going to save up for a new graphics card??
What is the benefit of SLI?
Using SLI, you may enjoy smoother, more fluid, and all-around mind-blowing visuals when gaming. It’s a technology that lets you combine the power of multiple video cards into a single unit, rather than having each operate at their individual peak.
SLI allows you to connect up to four GPUs, but you must be mindful of the decreasing returns and a slew of potential issues with heat and bugs if you want to maximize your performance. As the number of GPUs increases, so does the potential for system failure and the amount of heat that must be dealt with.
Is NVLink better than SLI?
However, the question of whether one is better than the other is more difficult to answer.
SLI, on the other hand, has a master-slave connection at its core. There is a single card that governs and directs the other cards’ workload. Each card may read and write to the memory of every other card in NVLink, hence there is no such hierarchy.
This has a significant effect on the amount of RAM that can be used. Because of this, even if you SLI two 8GB GPU cards, you won’t have access to all 16GB of usable RAM because of the hierarchical relationship that exists between the cards.
It is possible to legitimately double the amount of memory you have accessible if you NVLink the cards, as there is no bound.
So in theory, NVLink is superior to SLI in terms of performance. A multi-FPU setup is only useful if it can be used in a gaming environment, and this has yet to be accomplished by game developers.
There are ways to improve. It’s possible that things will get better in the future, too. However, for the time being, there isn’t much of a difference between the two configurations.
Does SLI increase performance?
Yes, and no. In terms of performance and frame rate, it can be beneficial. If you want to take use of it, you’ll need games that do, and there aren’t many of them. Then there’s the possibility that SLI as a technology is on its way to widespread redundancy. A new rule in January 2021 mandates that all games that support SLI must explicitly support the mode, rather than being able to implement the feature in-driver as before.
Although it can theoretically boost performance, you need games to take use of it, and developers are unlikely to create such games because of a diminishing market demand.
Can you run 2 GPUs without SLI?
You can, but the real question is why you would.
SLI allows you to use both GPUs on one screen, however if you use two GPUs without a SLI, you’ll almost always require two monitors to connect to your system. Then you may watch a movie in one room and play a game in the other, creating a’multi-room’ experience. However, you won’t be using two GPUs to run two different graphic outputs, thus they won’t operate together. Which raises doubts about its validity.
Without a SLI, can you use many GPUs to increase the performance of a single graphical channel, such as a game? You can, but only if you use something like NVLink or a high-end software package like GNU/Linux. An NVLink is compromising on the spirit of the question because it’s a non-hierarchical variant of SLI. Multiple GPUs can be run on the same output using software that is available, which is quite similar to SLI.
Can you SLI 2 different brands?
Think about it this way… In this case, you can utilize two round pegs from either Nvidia or AMD, regardless of which company makes the square and round pegs. Alternately, you can use any two square pegs, regardless of the manufacturer. Using one round and one square peg will save you money, but you will be extremely dissatisfied.
The explanation for this is simple. It is the goal of SLI to connect cards of the same type – whether it’s internal architecture, core clocks or shaders. SLI doesn’t mind if the GPUs are built by different companies as long as they have the same basic structure. Even if two models from the same company are used, the internal consistency is lost or mismatched. the outcome? Malfunction.
Is this the same model, but from a different manufacturer? Positive feedback all around. Is there a different model, or a different brand? Negative review.
Does SLI bridge brand matter?
Most of the time, this question can be answered with a simple no. However, only with exceedingly high levels of customization can some unit builders claim to discern a brand-specific difference. Some builders suggest that if you use all of your graphics cards to their full capabilities, you’ll be able to reach higher frame rates per second. Flexible SLI bridges have also been argued to be superior, however this is primarily a matter of personal preference.
As a whole, the effectiveness of bridge brands varies so little that you may ignore it while planning your setup. In reality, it’s more likely to have an aesthetic impact on your setup than it is to have any quantitative effect on performance.
For those who want to play games that support SLI, keep in mind that the bulk of titles currently don’t. For those who can afford it and will benefit from improved FPS and higher resolutions, SLI is a viable option. Nonetheless.