SLI: Myth and Reality
Several weeks after the release of the GeForce 6
series, NVIDIA announced the rebirth of SLI. For the
uninitiated, the moniker stands for Scalable Link Interface
(and not Scan Line Interleaving from the ol' 3dfx days).
The concept of SLI is pretty simple - instead of using a
single card, you use two cards to render one screen by
dividing the screen into two parts. If that's not possible,
the two cards can take turns rendering the screen, ie. the
first card renders all odd frames and the second card
renders even frames. An intriguing idea and early test does
show some promise. However, as most reviewers found out
(later), SLI was anything but simple.Now, more than a year after it's introduction, SLI has garnered more and more attention, mostly from gamers who want to play graphics rich games such as F.E.A.R, Quake 4 and Call of Duty at the highest settings. Of course, the engineers and developers at NVIDIA haven't been sitting idle during that time. With every driver version, NVIDIA strive to make SLI easier and more compatible with games. With the promise of faster performance (and better visuals), SLI was fast becoming the choice of many games, it certainly was the choice for many Futuremark's 3DMark ORB users. Even NVIDIA's archrivals, ATI who scoffed SLI at it's launch, have followed NVIDIA's footsteps with Crossfire.
The arrival of Crossfire can be considered a boon for SLI users. Newer Forceware drivers allows you to mix cards from different manufacturers or even different series of cards, ie. a GeForce 7800GTX from ASUS and a GeForce 7800GTX from Gigabyte or any GeForce 7800GTX and a 7800GT. NVIDIA have even certified Intel's latest chipset - the 955 series - as SLI capable chipset. No such luck for AMD users, more likely because AMD does not make chipsets for their processors..
So, after a year, is SLI ready for more 'maintream' use? Granted, installation and use are easier than before, even requiring less hassle than ATI's Crossfire. What interest us more is whether or not SLI is a viable option, both from a performance and quality standpoint for gamers. Let us elaborate.
Understanding SLI
We won't be discussing much of the technical aspects of SLI here. There are already countless SLI articles dealing with that subject available on the Internet. Now, you might dismiss this as just another SLI article, but we beg to differ. In this article, we'll look at what does SLI really bring to the table for gamers. What you say? Well, in a nutshell, SLI brings two things to the table for gamers: performance and quality. By splitting the rendering load between two cards, you should be able to enjoy faster frame rates - half the load per card or twice the performance. At least in theory. This is obviously what we mean with performance. However, you can also force the use of higher sample antialiasing with SLI - 8x and 16x. The use of more samples over the standard 4x or 8xS AA (may) offer slightly improved visuals - very important for those looking for better visuals.Evaluating performance should be easy to do. All we have to do is run an SLI setup in both single and multi card mode. Well, that's exactly what we intend to do (as did other reviewers in their SLI articles). However, before we do that, let's look at this from a different perspective - in what system will you likely use SLI to gain performance? Using a pair of GeForce 6800GT or even the GeForce 7800GT / GTX may provide us with the highest frames rates under the sky but that wouldn't tell us much about SLI at all. Why? For one, these cards are already system limited with the Athlon 64 3500 or even the Athlon FX 53. The only way to make most benchmarks graphics limited is to crank up AA and AF. With these settings, we believe more readers using (or planning to) those kind of rigs would be more interested in quality rather than performance. For those looking for quality, we will run benchmarks with a pair of GeForce 7800GTX, focusing on performance with AA and AF. In addition to that, we will discuss SLI antialiasing - whether or not it's viable option and does it offer any significant improvement over the standard 4x. However, the results of that test will not be shown here, but on the next article, which hopefully will be ready next week. .
To test 'pure' SLI performance, we decided to use the slowest card with SLI support - a pair of GeForce 6600. These cards are basically the same as GeForce 6600GT, but they're equipped with a slower core and memory - the slower DDR2 (though still with 128 bit bus). This might suprise you, but think about it for a minute. This setup is basically the cheapest SLI you can buy and the single GeForce 6600 badly needs the extra performance SLI (may) offer. So in a way, we're testing SLI in a setup that will likely give us the most performance increase.
Examining Performance Factors
Let's look at these factors closely for a moment. Graphics cards are more than just simple graphics adapters. In a sense, they resemble general purpose processors both in transistor numbers and complexity. So, it might be easier to explain scalability of multi graphics card rendering solutions by looking at multiprocessor systems. Multiprocessors system have been around much longer than multi graphics card rendering solutions. However, even now most desktop PCs and notebooks mostly feature single processors. This will change in the future with dual and multi core solutions: they're basically two or more processors packed into a single die (with some creative engineering to maximize optimal performance). If you look at benchmark results from such systems, you'll find that for the most part these systems can either be slower or much faster than your average single processor desktop systems. Why is that?Load Balancing
The answer is load balancing. With dual and multi processor systems, we want these processors to be fully utilized (most of the time). Only then will we see a significant boost in performance. That's why multiprocessing benchmarks may run an application with several threads or even several applications at once - the purpose is to have enough task for each processor in the system. It is up to the developer (and operating system) how these threads and application are load balanced. A well balanced system will reap the most reward in multithreading environment or when running multithreaded applications.
However, not all applications are multithreaded. Look at the benchmarks, we can see that for the most part multiprocessor system will score something around 30 to 70% higher than single processor systems in a majority of benchmarks. Some very special applications may have and even higher scores. But in real life, 30 to 50 % faster performance is already very good for multiprocessor systems. Dual and multi core processors score slightly higher, mostly because these processors have less latency and even more improvements to maximize their multitasking / multithreading performance.
So, how does this relate to SLI? Since SLI is based on the same principle, it's also susceptible to same pitfalls as multiprocessor systems. An SLI rig will only offer significant increase in performance if the rendering task can be optimally load balanced between the cards. In games where the rendering load can be balanced effectively and efficiently, we may see at least 50 % increase in performance (or more). However, not all applications and games can be load balanced. Even those who do may not benefit that much. In this respect, you should never expect an SLI system to offer 100 % or even 50 % performance increase in every application. Just like multiprocessor systems, even a 30 to 50 % overall increase is good.
Latency
With multiprocessor systems (or even dual or multi core processors), it is up to the operating system (and the application's developer) to ensure how applications and threads are load balanced to reach optimal performance. With SLI, this important task is handled by the drivers. So, unlike single card solutions, the graphics card drivers must perform load balancing with SLI systems as an additional task in the rendering process. This means another step in the rendering process and that means more time to complete the whole process. Thankfully, load balacing can be done very quickly, so any slowdowns caused by load balancing can effectively be 'hidden' by the performance increase. Now, if you're paying attention, you can see that there is catch here - if the performance increase is minimal or even non existant, we might be getting lower frame rates with SLI systems. Why? Because the drivers must still perform load balancing and that's takes time - thus lower frame rates because each frame took more time to be rendered. With applications that can't be load balanced, it is better to turn SLI off (choosing the single card mode) completely.
Rendering (Dis)Parity
Whoa! What the heck is that? Rendering disparity basically occurs when load balancing is not, well, balanced effectively. Simply put, the load on one card is lighter / smaller than the other. For example, if we were to split the screen in half (like in scissor mode), there will be time when the load is heavier on the upper half and vice versa. When this happens, the half with the lighter load will be rendered faster than the other half. So, you'll be seeing the next frame on one half of the screen while the other is still displaying the previous frame. Nasty, isn't? Thankfully, load balancing in SLI is not 'dumb'. The driver will continously monitor the load and adjust the load balancing ratio to avoid such problems. However, this does mean slightly more work, both for the developers at NVIDIA and the drivers. That means long hours for the developers and (slightly) higher latency from the driver (in addition to just doing standard load balancing). With alternate frame rendering, such matters will not be (much of) a problem. That's why multi card SLI mode use this approach as the default.
Usage
Unlike math, 1 + 1 doesn't equal 2 with SLI. Yes, we've talked about load balancing, latency and disparity between two cards, but there is at least one other thing that makes it so. If you're familiar with the 3D rendering pipeline, you'll know that to render an image, the graphics card must process both polygons and textures. In an SLI setup, for the most part, only texture related tasks are load balanced. So, in a way, an SLI setup is really 1 + 1 = 1.5 since the vertex units (the 0.5 part) in the second card are hardly used. Now, before you scream bloody murder, it's not that important anyway. SLI is targeted for games, well, actually most if not all consumer level graphics cards are targeted for games. Games usually don't use that much polygon to really tax a single graphics card, much less two. As long as the geometry load is still within the capabilities of a single card, it doesn't really matter that we didn't load balanced the geometry load. However, if a game does use more polygons (or vertex shaders) more than one card can handle, it won't run faster with two cards. In this situation, one single but more capable or higher clocked card is preferred.
With these factors in mind, we think that users should never expect SLI as a replacement over a faster single card solution. SLI should be considered complementary - meaning it will allow you to have faster performance in some games, but not all games. For a more well rounded and 100 % effective solution, it will always be better to buy a faster single card. Rather than using two GeForce 6600GT, it's better to use a single GeForce 6800GT. Particularly since we should expect no more than an average 50 % increase in performance. Of course, that's theory anyway, so let's see if that theory is right. Let's take a look at what SLI has to offer in terms of performance.
Performance - GeForce 6600 SLI
Just as before, we have to use a system where performance differences can be easily shown. We reused our test setup from the last article, replacing the single hard drive with an RAID 0 array for RAID 0 testing. A reference clocked GeForce 7800GTX which is system limited with this setup, is again used, with a resolution of 1024 x 768, 32 bit - no AA, AF or any other settings are enabled. This also mean not enabling HDR on Splinter Cell: Chaos Theory and Serious Sam II, no Ultra Settings in Quake 4, or Soft Shadows in F.E.A.R. All other settings are maxed out - full detail.We'd like to thank both Tagan and Kingston for supplying with the additional power supply and 1 GB memory modules for this article.
Our test setup
AMD Athlon 64 3500+ socket 939
2 x 1024 MB Kingston KVR 3-3-3 PC3200 DDR-SDRAM
MSI K8N NForce 4 SLI motherboard
2 x Gigabyte GeForce 6600 DDR2 256 MB graphics card
Maxtor DiamondMaxPlus9 80 GBs Serial ATA 8 MB buffer
ASUS E-616 DVD-ROM
Tagan TG530-U15 530 watts ATX/BTX power supply
Windows XP Professional with Service Pack 2 installed
NVIDIA Forceware 81.98 reference driver
NVIDIA NForce 4 6.70 reference driver
Creative SoundBlaster Live! 24 bit 5.12.1.512 driver.
DirectX 9.0c
The results:
At first glance, SLI doesn't seem to offer any performance improvement. However, if we look closely at SLI's AA and AF results, we can see SLI indeed offer a performance increase over a single card. The gain is substantial but as we expected are no where near 100 % - it's more around 30 %. This increase can still be seen at 1600 x 1200, although the nominal frame rates still won't offer high enough frame rates for enjoyable gaming. That's not because of SLI, rather the graphics card of choice here - the GeForce 6600.
Homeworld 2 is another story all together. SLI couldn't offer you higher frame rates with this game, in fact we're getting much lower frame rates. If you really want to use SLI to play Homeworld 2, use single mode rendering or SLI antialiasing. Even the game's default profile embedded in the drivers made use of SLI antialiasing by default.
This game doesn't have a profile embedded in the drivers, so it serves as an excellent tool for examining SLI with run of the mill games. We know the scores for minimum fps seem strange, so we'll have to rely on average frame rates on this one. Just like with Call of Duty, SLI is only able to flex it's muscles with AA and AF enabled, here it succeded making 1280 x 1024 with 4x AA and 16x AF still playable. While this is good, we can see that for the most part SLI will not offer you higher frame rates without AA and AF. Not to mention the performance increase.is minimal, particularly at 1024 x 768 and it's only higher on higher resolutions where the nominal frame rates are too slow for comfortable gaming..
SLI may seem to cause minimal increase in performance in this game, but the difference in results are still within the standard deviation between runs. So, with this game SLI behaved pretty much the same way as with Homeworld 2. The only exception, thankfully, is the absence of performance penalty.
Splinter Cell - Caspian Oil Refinery, 1024 x 768, 32
bit
Single
Multi
Splinter Cell - Caspian Oil Refinery, 1280 x 1024, 32
bit
Single
Multi
Splinter Cell - Caspian Oil Refinery, 1600 x 1200, 32
bit
Single
Multi
Splinter Cell's Caspian Oil Refinery demo is a pretty good graphics benchmark. Unfortunately, the game won't work with AA, so these results are without AA and AF. This is the first game where we see SLI multi rendering mode is faster than single rendering by default. Of course, you can see for yourself the increase is nothing spectacular - it's merely around 2 fps faster.
These are old games, so its more than likely performance won't be much different with newer drivers. So far, we can see SLI does not have a good case to support it. Yes, it can provide higher frame rates, but only with AA and AF - you only get (very) minimal increase in performance with AA and AF turned off. Then there is also the compatibility problem - some game will run slower with SLI's multi rendering mode. If we were to draw a conclusion on these results alone, we can say that SLI is not worth the price - you're way better off with a single, faster graphics card. But let's look at newer games.
Keep in mind that F.E.A.R is one extremely heavy, graphically rich game. Seeing that much performance increase at 1024 x 768 is good news indeed. By adding another card, we can now reach the minimum 30 fps mark and experience 30% percent higher frame rates on average. But even with SLI, we won't be seeing high enough frame rates in AA and AF with two GeForce 6600. This is also true for higher resolutions. With high end cards such as the GeForce 7800GTX, we may get high enough frame rates for truly lag free gaming in F.E.A.R by using SLI.
Well, it's clear that you can only play this game comfortably at 1024 x 768 (or lower) without AA and AF on a GeForce 6600. Slapping in another card won't give you any performance increase whatsoever. So, even with newer games, SLI will not always give you higher frame rates.
Splinter Cell: Chaos Theory - Lighthouse, 1024 x
768
Single
Multi
Splinter Cell: Chaos Theory - Lighthouse, 1280 x
960
Single
Multi
Splinter Cell: Chaos Theory - Lighthouse, 1600 x
1200
Single
Multi
This game is really the success story for SLI. We're seeing a tremendous boost in performance, very close to 100% in some cases. That's both in default and with AA and AF. If only all new games share that trait. Let's see how the two GeForce 6600 cope with HDR rendering.
Splinter Cell: Chaos Theory - Lighthouse (HDR), 1024 x
768
Single
Multi
Splinter Cell: Chaos Theory - Lighthouse (HDR), 1280 x
960
Single
Multi
Splinter Cell: Chaos Theory - Lighthouse (HDR), 1600 x
1200
Single
Multi
Just like in default and with AA and AF, slapping in a second card really helped the system reach that minimum 30 fps mark. The increase in performance is also pretty constant across resolutions, so it's more than likely using a pair of higher clocked or more capable cards will net the same results as well.
Unfortunately, we can't show you Serious Sam II results with this setup. The results must've got lost somewhere on our hard drives (our bad). But these results are enough to draw some general conclusions. Just as we expected: it is hopeful thinking to expect 100% increase in performance with SLI. Of the games we tested here, those who gain with SLI mostly gain 30 % or less, with Splinter Cell: Chaos Theory being the exception rather than the norm. Even then, the increase is mostly with AA and AF - increases under default settings are minimal. SLI won't help your low end graphics card at all with some newer games (Quake 4). You simply have to use a faster card such as the GeForce 6800GT or the GeForce 7 series..
Temporary Conclusion:
In no way whatsoever should you consider SLI as a replacement for a single high end (or faster) graphics card. The results taken from using SLI with the two GeForce 6600 here are nowhere close to a single GeForce 6800GT. That's why we believe SLI should be considered as a complementary solution: SLI is only worth it for those looking for better visuals. By this we mean, SLI will only be worth the money if you want to play games with the settings set to full, with AA and AF and want higher frame rates. In a way, SLI is NVIDIA's answer to the GeForce 6 and 7 series tendency to have a very high AA performance penalty.Go to top