GeForce GTX 680 SLI
On March 22nd, 2012 NVIDIA debuted its next generation GeForce GTX 680 video card, based on the Kepler GPU. While there will be more in the Kepler family later on, currently the GeForce GTX 680 represents NVIDIA's flagship video card at $499. This provides a competitive price point to AMD's Radeon HD 7970 flagship video card at $549. While there is a $50 price difference, we found that GeForce GTX 680 more than keeps up with Radeon HD 7970. In fact, at 1080p the GTX 680 outperforms Radeon HD 7970 and provides a better gameplay experience in our single-video card testing.
We looked at performance at 2560x1600 and 5760x1200 with a single-video card previously. What we didn't have a chance to look at, until now, was how two GeForce GTX 680 video card perform in SLI, and the kind of gameplay experience we are provided. This evaluation focuses on GeForce GTX 680 SLI performance compared to Radeon HD 7970 CrossFireX performance and we threw in a GTX 580 SLI for comparisons’ sake.
There is nothing "new" when it comes to the GTX 680's SLI support. In the past generation you could setup NV Surround with two video cards, and you can do that again. Setup is extremely easy, with only one bridge connector required. With two GTX 680 video cards you'll need 4x 6-pin connectors in total, whereas with Radeon HD 7970 CrossFireX you'll need 2x 8-pin and 2x 6-pin connectors.
To setup our triple-display configuration for testing in NV Surround we utilized all the DVI connectors. We plugged two displays into the primary card, and the third display into the secondary card, all on DVI. NV Surround is a snap to setup in the control panel, and it only takes a few seconds, you don't even need to reboot. Both AMD's and NVIDIA's setup is simple, quick, and both worked without issue.
To make our SLI solution work we are using one NVIDIA GeForce GTX 680 reference video card and one GALAXY GeForce GTX 680 video card, which is a reference design at this time. (Galaxy, will be selling its own card very soon.) Both cards we have run at the same 1006MHz baseclock. With the nature of GPU Boost, both video cards can run at independent clock speeds. This means that GPU Boost can be boosting each GPU separately as it needs to for the best performance, the clock speeds don't have to match up. This provides the best performance at all times based on GPU utilization, power, temp and other things. We will be testing on three Dell 2408WFP displays at 5760x1200.
We evaluate what each video card configuration can supply us in terms of a playable gaming experience while supplying the best culmination of resolution and "eye candy" graphical settings. We focus on quality and immersion of the gameplay experience rather than how many frames per second the card can get in a canned benchmark or prerecorded timedemo situation that often do not represent real gameplay like you would experience at home. Then we will follow with apples-to-apples testing in with minimum, maximum, and average framerates.
Test System Setup
We will be using a ASUS P8P67 WS Revolution motherboard, an Intel Core i7 2600K Overclocked at 4.8GHz, and 8GB of Corsair RAM.
For the AMD Radeon HD 7970 we are using AMD supplied drivers, which are version 8.95.5 (Cat 12.3 Beta) drivers. These are superior to Cat 12.2 WHQL, and actually contain some code that will be in 12.4. We are using CAP 12.2 CAP 1 which is the latest CAP.
For the NVIDIA GeForce GTX 580 we are using ForceWare 296.10.
For the NVIDIA GeForce GTX 680 we are using the brand new 301.10 WHQL drivers.
Mass Effect 3
Released on March 6th, 2012, Mass Effect 3 is the third installment in the widely popular Mass Effect video game series. This action role-playing game is a primarily a third person shooter developed by BioWare and published by Electronic Arts.
We are using the full downloaded version of Mass Effect 3 from Origin with the latest patch update. It is a DX9 game that uses the Unreal 3 Game Engine. Our run-through covers the entire mission at Lesuss Monastery.
Highest Playable Settings - Multi-Display Multi-Card
(Click Graph for Larger Image)
Both NVIDIA and AMD released new drivers and application profiles for Mass Effect 3 at this game's launch to support SLI and CrossFireX. It was good to see CrossFireX support from AMD on game launch day. We therefore had no issues running SLI or CrossFireX in this game. The only thing we did have to modify was to set the appropriate resolution on NV Surround GTX 580 SLI and GTX 680 SLI. The game detected 5760x1200 just fine under the Radeon HD 7970 CrossFireX platform, but under the NVIDIA cards the game only went up to 1920x1200. We had to manually set 5760x1200 in the INI file for those configurations. When we did so, the game ran just fine in 5760x1200. However, the menu's were enlarged, and a bit hard to navigate since those did not scale properly with the aspect ratio.
Unfortunately this game has few graphics options to manipulate. The Anti-Aliasing option enabled was NVIDIA's FXAA which is built into the game. Other than that, dynamic shadows were enabled. Performance at the highest in-game settings was phenomenal even on GTX 580 SLI at the highest resolution supported by our configuration, 5760x1200. The new GeForce GTX 680 SLI is technically the fastest solution here, averaging 170.7 FPS. The new GeForce GTX 680 SLI performed exactly 13.2% faster than AMD Radeon HD 7970, which averaged 150.7 FPS. GeForce GTX 680 SLI improves upon GTX 580 SLI performance by 35% in this game. Though GTX 680 SLI is faster, when even your slowest configuration is averaging above 100 FPS and not even dropping below 77 FPS in the game, you can't call any configuration slow or a loser in this game.
This game just doesn't push PC graphics cards, even at NV Surround or Eyefinity resolutions. There is simply no need for two video cards to run this game smoothly, as one video card is all that's really needed to push it at 5760x1200 as we've previously tested. The gameplay experience between all three configurations is exactly the same in this game, and no bottlenecks were experienced.
Batman: Arkham City
We are using the full downloaded version of Batman: Arkham City from Steam with the latest patch update. To learn what this game has to offer in terms of graphics options and testing procedure please read our Batman: Arkham City Performance Review and our Batman: Arkham City DirectX 11 Performance Review. We will be running Batman with DX11 enabled along with MVSS and HBAO and we will use tessellation levels that are playable. In all the tests below, all of the features are enabled, and we adjusted resolution, AA, and tessellation as needed to find the highest playable settings.
*Note, a patch was just recently released for this game, and we are using this patch in this evaluation.
Highest Playable Settings - Multi-Display Multi-Card
(Click Graph for Larger Image)
Batman posed much more of a graphics challenge for our video cards and allowed the newest generation of video cards to shine and deliver the best gameplay experience. First let's talk about where we were with the last generation GeForce GTX 580 SLI in this game at 5760x1200. We found that with GTX 580 SLI we could not enable any MSAA setting, either 2X or 4X at 5760x1200. When we enabled an MSAA setting performance was too sluggish with GTX 580 SLI. Therefore, FXAA was the highest AA setting that was playable, which doesn't strain memory.
We were never able to run with any level of PhysX with GTX 580 SLI at 5760x1200. Even if we disabled tessellation, we were not able to use normal or high PhysX with GTX 580 SLI at 5760x1200. Performance was simply too slow. PhysX drained performance quite a bit on 580 GTX SLI at this resolution in this game. However, without that, we were able to have high tessellation enabled with GTX 580 SLI, which always gave it an advantage over the AMD competition.
Now we come to the new kid on the block, GeForce GTX 680 SLI. For the first time we were able to use normal PhysX at 5760x1200 with tessellation on. We tried to push PhysX up to high, but even on 680 SLI it bottlenecked performance even with just FXAA enabled. However, with normal PhysX enabled we found performance was smooth and playable with high tessellation and FXAA.
This wasn't the only playable setting we found available with GeForce GTX 680 SLI however. We found that if we disabled PhysX, that this game was playable with 4X MSAA and high tessellation on GTX 680 SLI! Yep, 2X MSAA and 4X MSAA were smooth and playable on GTX 680 SLI with high tessellation to boot. Therefore, you have a choice with 680 GTX SLI. If you chose to run with normal PhysX, you'll have to run at FXAA for the best performance. However, if PhysX is of no interest to you, you can disable it and run with 4X MSAA instead of FXAA. All of this along with high tessellation and the highest in-game settings on 680 GTX SLI. We have an apples-to-apples graph below that shows 4X MSAA performance.
The AMD Radeon HD 7970 CrossFireX configuration did not perform so well in this game. It was clear from the start that even with two 7970 cards installed, tessellation was causing a heavy burden on performance. We tried our best to play this game with high tessellation and FXAA, but the performance just wasn't there to support it. We even knocked tessellation down to normal mode, and while it was right on the edge of being playable, overall it was still a drag on performance. Only by disabling tessellation was 7970 CFX playable at 5760x1200 with FXAA and highest in-game settings.
Therefore, there is a large difference in the gameplay experience between GTX 680 SLI and Radeon HD 7970 CFX in this game. With GTX 680 SLI you can have high tessellation enabled in all instances, along with either 4X MSAA or normal PhysX. There just isn't any comparison; GTX 680 SLI smokes 7970 CFX in this game.
4X MSAA Comparison
We mentioned above that 4X MSAA was playable on GTX 680 SLI, as long as PhysX was turned off. In the graph below we compare all three configurations at 5760x1200 4X MSAA and high tessellation with the highest in-game settings.
First note that GeForce GTX 680 SLI is playable at this setting shown, but the other two configurations are not. At 4X MSAA GTX 680 SLI is averaging near 60 FPS, and doesn't drop below 35 FPS. You can see quite clearly that at 4X AA and high tessellation 7970 CFX struggles quite a bit in this game, averaging 40 FPS, but its minimum is down to 17 FPS and it spends a lot of time under 30 FPS, or dancing around it. Even GTX 580 SLI is faster than 7970 CFX here at 4X AA and high tessellation.
Keep these facts in mind and you'll see it doesn't make sense, yet it is fact. GTX 580 video cards only have 1.5GB of VRAM per GPU, and a much lower memory bandwidth than 7970 video cards. The new GTX 680 has 2GB of VRAM per GPU, and still a lesser memory bandwidth than 7970. Radeon HD 7970 has 3GB of VRAM per GPU and a much higher memory bandwidth. Yet, in this setting of 5760x1200 with 4X MSAA enabled, where you'd think memory capacity and performance would greatly affect performance, the two lesser capacity and lesser memory bandwidth solutions are faster.
We also wanted to compare the exact same settings at FXAA instead of MSAA, to lessen the strain of memory causing any bottlenecks. Again, this is 5760x1200 with FXAA and high tessellation.
The results speak for themselves; GTX 680 SLI is 39% faster than Radeon HD 7970 CFX. It is clear in the tessellated part of the test GTX 680 SLI holds up performance much better. Even GTX 580 SLI is faster than HD 7970 CFX here.
It is worth mentioning that NVIDIA was a "development partner" on this particular title, so surely it has had time to greatly optimize its coding for this game. Still this does not change the outcome of 680 and 580 being better Batman:Arkham City solutions.
We are using the downloaded version of Battlefield 3 via Origin. This game supports two different antialiasing methods, traditional MSAA via the "Deferred AA" option and shader based FXAA via the "Post AA" option. We evaluate the highest playable gameplay settings and take into account performance and image quality, striving for the best. In that regard, we will use whatever the appropriate AA setting, or combination of AA settings, are best for the video cards being used. Our testing procedure and evaluation of graphics features in BF3 are explained in our Battlefield 3 Gameplay Performance and IQ evaluation. We are using the latest patch for this game.
Highest Playable Settings - Multi-Display Multi-Card
(Click Graph for Larger Image)
Battlefield 3 is a great game to test at NV Surround and Eyefinity resolutions, not just because it looks great, but because this game also is a VRAM memory hog in our testing. We've found though that this game uses more VRAM in multiplayer mode, than it does in single player mode. The table and graph above reflect testing in the single player campaign. However, we did do some specific testing in multiplayer that we will talk about below in the next section.
The new GeForce GTX 680 SLI solution did great in BF3 and we had no trouble running at 5760x1200 with the highest ultra settings including HBAO turned on. In addition to that, we found FXAA performed brilliantly and there was performance to spare. We notched the MSAA level up to 2X MSAA on top of FXAA and were surprised to find this setting playable. We were averaging 50 FPS and never dropped below 30 FPS in the single player campaign.
The AMD Radeon HD 7970 CrossFireX solution equally provided a similar gameplay experience. We found that all ultra settings were playable along with HBAO. We found FXAA was playable and on top of that 2X MSAA as well. The average framerate was just slightly higher on 7970 CFX at 53.4 FPS, and the minimum slightly higher as well at 40 FPS.
The aging GTX 580 SLI did not do so well at 5760x1200. We found that we had to disable HBAO and use the lesser SSAO setting, which reduces the quality of ambient shadows in the game. There was no other way though, with HBAO enabled performance was extremely bottlenecked on GTX 580 SLI even with just FXAA. Once we reduced the quality to SSAO, then FXAA only was playable at 5760x1200. Therefore, the new GTX 680 SLI improves upon GTX 580 SLI by allowing HBAO and a higher AA setting, which at this resolution is impressive.
We performed specific testing in multiplayer on a 64 player server that was fully loaded to test performance and look at VRAM usage. We used MSI Afterburner to display real-time VRAM usage of each GPU and wrote down the maximum capacity the game was using on each solution. To get a consistent look at performance we used the same server, which had a low ping, and the map we tested on was Caspian Border in the Conquest Large game type, we used the same method described in our look at multiplayer performance in BF3. Since this is multiplayer we aimed to keep the average FPS around 60 if we could.
Keep in mind, the memory capacity on GTX 580 SLI is 1.5GB per GPU, and on GTX 680 SLI it is 2GB per GPU and on 7970 CFX it is 3GB per GPU. In SLI and CFX this is not combined, as each GPU must have a copy of the framebuffer, so only the total capacity of one GPU is your max VRAM capacity for SLI and CFX.
GeForce GTX 680 SLI - Using the highest in-game settings of ultra, along with HBAO and default motion blur, we found that the highest playable setting was FXAA at 5760x1200 on GTX 680 SLI. At 5760x1200 FXAA and the highest in-game settings we were averaging 60-70 FPS, and the lowest framerate we saw was 50 FPS. This setting allowed an incredibly smooth multiplayer experience.
When we tried to turn on 2X MSAA it brought the FPS down to 40 FPS average and at 4X AA the game was completely choppy and unplayable. What you can see from the table is that at FXAA we are already tapping the full potential of the VRAM capacity at 2012 MB, going to 2X or 4X only takes us even higher over the capacity.
Radeon HD 7970 CrossFireX - The memory capacity may seem odd at first, but keep in mind that these results are taken with the game running at 5760x1200 with ultra settings and HBAO enabled. In other words, the highest possible settings capable, which are not actually the highest playable settings on 7970 CFX as you will read below. We wanted to take these memory results at the same settings between all configurations so these would be comparable. Even with just FXAA enabled we are over the VRAM limit, 3541 MB was being used, when the cards only have 3GB per GPU. Increasing to 2X and 4X AA took us to near 5 GB of memory. This game can certainly use what's there, that's for sure. At these memory capacities, the game was not playable at the highest in-game settings on 7970 CFX.
In terms of performance with Radeon HD 7970 CFX we had to lower the ambient occlusion setting of HBAO down to SSAO in order to get smooth enough performance. With GTX 680 SLI we were able to keep ambient occlusion at the higher HBAO mode, but with 7970 CFX the performance wasn't there to give us a smooth gameplay experience with HBAO enabled. We had to disable HBAO and use SSAO instead. We also had to disable motion blur all together. By turning off motion blur and turning the AO quality down to SSAO, performance felt smooth enough to us. However, there are some hardcore players that may still find that not quite smooth enough. In that case, you could disable AO all together, and with that disabled completely it provides a large smoothness increase.
Once we did all that, then it was only playable at FXAA, just the same as GTX 680 SLI. When we increased to 2X and 4X MSAA, performance was choppy and unplayable. Even though 7970 CFX has more memory, and a higher memory bandwidth, that didn't do anything for us here in multiplayer over GTX 680 SLI. While both setups were at FXAA, GTX 680 SLI had the better experience since HBAO was enabled.
GeForce GTX 580 SLI - GeForce GTX 580 SLI was the most bottlenecked in multiplayer. At the highest in-game settings the memory capacity was at its limit with just FXAA, and at 2X AA it just couldn't go any higher. We had so much trouble that at 4X AA the game crashed, and just wouldn't run for us.
As for performance, we had to disable motion blur and disable ambient occlusion completely at 5760x1200, and only use FXAA. At these settings performance was averaging just 50 FPS and the minimum was around 40 FPS. This performance level might not even be fast enough for some hard core players, and in that case the only thing left to do is to lower the quality settings from ultra to high, which will give you the performance you need.
Multiplayer Summary - GTX 680 SLI offered the best multiplayer experience, despite it having less VRAM capacity and memory bandwidth. We were able to run with motion blur enabled and HBAO turned on at 5760x1200 with FXAA and averaged 60-70 FPS. This amount of performance is perfect for multiplayer, and with the highest in-game settings enabled the game looked great at multiplayer. AMD Radeon HD 7970 CrossFireX struggled for performance, even though it had more RAM and memory bandwidth. To get the game to feel smooth enough with enough performance we had to lower ambient occlusion and motion blur. GeForce GTX 680 SLI was the clear winner in multiplayer.
4X MSAA Comparison
Now we are back to the single player part of the game, and looking at 4X MSAA performance at 5760x1200 to see if there is bottlenecking. The GTX 580 SLI configuration was incapable of running at this setting, it crashed on us.
At 5760x1200 with 4X MSAA and the highest in-game settings Radeon HD 7970 CrossFireX was faster, and had a much better minimum framerate. The average framerate isn't massively different, but you can see how 680 SLI dips below 30 FPS quite a few times, while 7970 CFX does not. The behavior between single player and multiplayer in this game is very different.
Now we can look at performance with FXAA enabled, reducing memory bottlenecks.
Radeon HD 7970 CFX is slightly faster than 680 SLI, but not very much so. GTX 580 SLI is completely bottlenecked at this setting. Again, the single player result is opposite of multiplayer, where 680 SLI was vastly superior.
The Elder Scrolls V: Skyrim
We are using the full downloaded game of Skyrim from Steam. We have the latest 1.4 version patch applied as well. To learn about the graphics options in this DX9 game please read our Skyrim Gameplay Performance Review which also details our testing procedure. You can also check out our Skyrim AMD CrossFireX Performance and IQ Review. To further your Skyrim pleasure, we also have a Tweaking Skyrim Image Quality article.
For this evaluation, we are using the built in default Skyrim settings, no modifications other than to disable VSYNC. No arrows were taken to the knee during the testing of this game. Note that we have patch version 1.5 applied and we have the high resolution texture pack installed as well.
Highest Playable Settings - Multi-Display Multi-Card
(Click Graph for Larger Image)
Skyrim is another game where we thought we might see some memory bottlenecks, but we didn't at 5760x1200. With the new GeForce GTX 680 SLI we were able to play this game at settings even well beyond Radeon HD 7970 CrossFireX.
With GeForce GTX 680 SLI we had no trouble running this game with 8X MSAA in-game and FXAA turned on at the same time at 5760x1200. Performance was extremely high at this setting, and there was room to go even higher. The only option left to us was to enabled NVIDIA's Transparency Supersampling. We didn't think we would be able to because this requires a large amount of memory capacity and memory bandwidth at high resolutions like this. We were surprised, more than surprised, to find that not only 2X TR SSAA worked, but we could even go up to 4X TR SSAA! At 8X MAAA + FXAA + 4X TR SSAA we were averaging 65 FPS in this game. We could not do 8X TR SSAA, that was clearly the limit, but 4X TR SSAA is no small feat. It provides a level of image quality superior to what Radeon HD 7970 CFX was able to deliver.
With Radeon HD 7970 CrossFireX we also found 8X AA + FXAA to be playable, with some room to spare. We tried to enabled AMD's Adaptive Supersampling mode, but found it completely unplayable at 8X AA or 4X AA. The game was stuttering and slow. For whatever reason, AMD's Adaptive Supersampling has never behaved well with this game.
GeForce GTX 580 SLI didn't fare so well, we couldn't run at 8X MSAA + FXAA. The problem was that 8X AA was just a bit too much for the 580's framebuffer, and we got inconsistent and choppy performance. Once we dropped to 4X AA it smoothed out and therefore 4X AA + FXAA was completely playable.
In this apples-to-apples performance test we are comparing all video cards at 8X MSAA + FXAA at 5760x1200. GeForce GTX 680 SLI is 23% faster than Radeon HD 7970 CrossFireX at these settings.
Deus Ex: Human Revolution
We are using the full version of Deus Ex: Human Revolution from Steam, patched to the latest version This game runs in DX11, and supports Ambient Occlusion, Diffusion Depth of Field, and other effects, It also includes NVIDIA's Antialiasing method called FXAA, with three settings. This game does use a fair bit of tessellation on characters; when you are up close talking to them is where performance can drop as high levels of tessellation are used. We have this scenario tested in our evaluation.
For our run-through we are going to play single player campaign on the next-to-last level of Panchaea from the tower after talking to Hugh Darrow; to the elevator leading down to the final boss. We also stop and talk to William Taggart and David Sarif to get some tessellation strain on our cards. You can read more in our full Deus Ex: Human Revolution gameplay and performance review here.
Highest Playable Settings - Single Display
(Click Graph for Larger Image)
We had no trouble running each configuration at 5760x1200 with the highest in-game settings in Deus Ex. This is no small feat, as high SSAO and soft shadows cause a big drain on performance, along with tessellation. Still, with no other AA options other than FXAA, we found no bottlenecks on any configuration. GeForce GTX 680 SLI was 16% faster than Radeon HD 7970 CrossFireX in this game. If you look at the tessellation performance in the graph, between 281 seconds and 421 seconds you will find GTX 680 SLI performing much faster than Radeon HD 7970 CFX. Even GTX 580 SLI is able to provide a more consistent tessellation performance in that area. This once again points to NVIDIA's tessellation performance being superior.
We tested the power utilization at the wall of the entire system without a video card, and with each video card at idle and full load. For full load power and temperature testing we used real gaming, in this case every game we tested. The power supply used in testing is a Enermax MaxRevo 1350W. Our system is very lean with only one optical drive and one hard drive being powered. Total system wattage at idle without video card is 67W.
The power results blew our minds. First off, the idle power numbers truly surprised us. GeForce GTX 680 SLI had the lowest idle power utilization at 102W system power. This was after leaving the system alone at the desktop for 30 minutes. We had all three displays enabled and turned on. AMD Radeon HD 7970 CrossFireX idled at 127W system power. Again, this was after leaving it alone at the desktop for 30 minutes with the displays turned on. We noted that the second video card's fan was not spinning, indicating ZeroCore Power was in use on the new Radeon HD 7970 CFX. Yet, 680 SLI sat there using less power, with both fans running on both video cards. The highest idle wattage was GTX 580 SLI at 142W. Therefore, GTX 680 SLI wins hands down at idle power, improving vastly from GTX 580 SLI, and even besting AMD's 7970 GPU.
Equally impressive are the game's full load wattages. In every game the GeForce GTX 680 SLI configuration had the lowest full load wattage compared to Radeon HD 7970 CrossFireX. Actually, both were tied in Deus Ex, but in every other game GTX 680 SLI was more efficient at full load. This is impressive when you also consider GTX 680 SLI was faster in performance compared to 7970 CFX in every game we tested. GeForce GTX 680 SLI once again improves vastly over GTX 580 full load wattage numbers, and even improves over AMD's new Radeon HD 7970 GPU. There is no question; NVIDIA's new GTX 680 is more power efficient than AMD's Radeon HD 7970, period.
The biggest question in regards to performance and gameplay experience about GeForce GTX 680 SLI was if the 2GB of VRAM per GPU and lesser memory bandwidth compared to Radeon HD 7970 would be a hindrance. Our testing has clearly answered that question. In fact, in every game we tested, GTX 680 SLI offered a better gameplay experience compared to Radeon HD 7970 CrossFireX. We specifically tested at NV Surround and Eyefinity at the maximum resolution of our configuration at 5760x1200 to see if there would be any bottlenecks. We found that the new GeForce GTX 680 SLI has the performance where it counts.
In Mass Effect 3 GeForce GTX 680 SLI allowed the highest gameplay experience possible at 5760x1200 and performed faster than Radeon HD 7970 CFX. In Batman we saw one of the more extreme cases where GeForce GTX 680 SLI was superior to Radeon HD 7970 CFX. We were able to play at 5760x1200 with high tessellation and 4X AA or high tessellation and FXAA with normal PhysX. With Radeon HD 7970 CFX we had to disable tessellation entirely, and even then 4X MSAA was never a playable option.
Battlefield 3 was another example where the GeForce GTX 680 SLI provided a better experience in the multiplayer part of the game. We averaged 60-70 FPS at 5760x1200 with FXAA and ultra in-game settings with HBAO turned on. This means the highest possible in-game settings were turned on, and the average framerate was very high. We never saw the minimum framerate drop below 50 FPS. The Radeon HD 7970 CrossFireX on the other hand had to have ambient occlusion turned down to SSAO, with motion blur disabled, and only then was it mostly acceptable. In some cases, you may even find turning off ambient occlusion provides the best possible performance.
In Skyrim we also experienced a higher level of antialiasing possible with GeForce GTX 680 SLI. To our surprise we could turn on 8X MSAA + FXAA + 4X Transparency Supersampling at 5760x1600 and still get above 60 FPS average. With Radeon HD 7970 CFX we could play with 8X MSAA + FXAA, but we couldn't enable AMD's Adaptive Supersampling. In Deus Ex we again saw strong performance from GTX 680 SLI, outperforming Radeon HD 7970 CFX, especially in the tessellated scenes.
Memory Capacity and Bandwidth
We know exactly what you guys are thinking. The Radeon HD 7970 has 3GB of VRAM, the GeForce GTX 680 has 2GB; the Radeon HD 7970 has 264GB/sec of memory bandwidth and the GeForce GTX 680 has 192GB/sec of memory bandwidth. You'd expect Radeon HD 7970 CrossFireX to simply blow GeForce GTX 680 SLI out of the water at 5760x1200. The simple fact is, it does not, and in fact GeForce GTX 680 SLI provides a better gameplay experience with better performance. Amazing, but true. Obviously AMD’s driver engineers need to figure out how to utilize the hardware more efficiently, because at the moment, NVIDIA is taking AMD to school.
There is one interesting game to look at right now for video card memory usage, and that is Battlefield 3 multiplayer. There is a big difference between the VRAM usage in single player and multiplayer. When we cranked this game up in a 64 player server at the highest in-game settings we saw it get near to 5 GB of VRAM usage on Radeon HD 7970 at 4X AA at 5760x1200. This game seems certainly capable of maximizing VRAM usage on video cards in multiplayer in NV Surround or Eyefinity resolutions. It makes us really want to try out 4GB, or higher video cards. A couple of 4GB GTX 680 video cards are looking real interesting to us right now in this game, and from what rumors we have heard, Galaxy is very likely to make this happen for us.
SLI smoothness vs. CrossFireX smoothness
We don't know what other descriptive word to use, other than "smoothness" to describe the difference we feel between SLI and CrossFireX when we play games. We've expressed this difference in gameplay feeling between SLI and CrossFireX in the past, in other evaluations, and we have to bring it up again because it was very apparent during our testing of 680 SLI versus 7970 CFX.
We can't communicate to you "smoothness" in raw framerates and graphs. Smoothness, frame transition, and game responsiveness is the experience that is provided to you as you play. Perhaps it has more to do with "frametime" than it does with "framerate." To us it seems like SLI is "more playable" at lower framerates than CrossFireX is. For example, where we might find a game playable at 40 FPS average with SLI, when we test CrossFireX we find that 40 FPS doesn't feel as smooth and we have to target a higher average framerate, maybe 50 FPS, maybe 60 FPS for CrossFireX to feel like NVIDIA's SLI framerate of 40 FPS. Only real-world hands on gameplay can show you this, although we can communicate it in words to you. Even though this is a very subjective realm of reviewing GPUs, it is one we surely need to discuss with you.
The result of SLI feeling smoother than CrossFireX is that in real-world gameplay, we can get away with a bit lower FPS with SLI, whereas with CFX we have to aim a little higher for it to feel smooth. We do know that SLI performs some kind of driver algorithm to help smooth SLI framerates, and this could be why it feels so much better. Whatever the reason, to us, SLI feels smoother than CrossFireX.
Personally speaking here, when I was playing between GeForce GTX 680 SLI and Radeon HD 7970 CrossFireX, I felt GTX 680 SLI delivered the better experience in every single game. I will make a bold and personal statement; I'd prefer to play games on GTX 680 SLI than I would with Radeon HD 7970 CrossFireX after using both. For me, GTX 680 SLI simply provides a smoother gameplay experience. If I were building a new machine with multi-card in mind, SLI would go in my machine instead of CrossFireX. In fact, I'd probably be looking for those special Galaxy 4GB 680 cards coming down the pike. After gaming on both platforms, GTX 680 SLI was giving me smoother performance at 5760x1200 compared to 7970 CFX. This doesn't apply to single-GPU video cards, only between SLI and CrossFireX.
The Bottom Line
It all comes down to pricing and performance, and you'll find that on both price and performance GeForce GTX 680 SLI wins. GeForce GTX 680 MSRP is $499, and the Radeon HD 7970 is $549. When you throw two of these video cards into the mix the difference comes out to $100. Two GeForce GTX 680 cards will cost you $1000 and two Radeon HD 7970 cards will cost you $1100. When you are spending so much money on a graphics solution you want every dollar to count, and to save those dollars where possible.
In every game we tested GTX 680 SLI was faster, and provided gameplay options Radeon HD 7970 CFX did not. We were not bottlenecked at all at 5760x1200 with GTX 680 SLI. In fact, in BF3 Multiplayer GTX 680 SLI provided better performance, higher in-game settings, and a smoother experience, and this is a highly memory sensitive game. When we looked at power utilization we found GeForce GTX 680 SLI was more efficient in every single game, using less power, but giving us more performance.
With GeForce GTX 680 SLI you will be paying less money for a more efficient solution, getting more performance, a superior gameplay experience, and a smoother gaming feel than you will with AMD Radeon HD 7970 CrossFireX.
Thanks to Galaxy for providing us a second GTX 680 so we could complete our SLI testing. You can hit this HardForum post of mine to check stock on Galaxy GTX 680 cards. Galaxy is telling me as of this morning, Microcenter and Fry's are receiving cards today, with cards going in stock at Newegg, Amazon, and Tiger today or tomorrow. (All linked in my HardForum post above.)
Last Updated: Wednesday, March 28, 2012
Content of this thread is taken from the following: http://www.hardocp.com/article/2012/..._card_review/9