MegaTechPC Presents: SLI GeForce GTX Titan versus CF 7970!
(Please note that the article's conclusion can be found in the second post HERE)
I know, I know. Titan has been out for a month and there have been countless reviews, comparisons, benchmarks, and everything else in between inundating the fair members here at OCN. So why post a comparison that is late to the game and where the conclusions are fairly obvious? Well, I figured we needed a different perspective on this subject. An everyman's comparison that captures the vantage point of the typical OCN member. Well, the typical OCN member that water cools every major component in his rig anyway. At this time the 7970 is AMD's flagship and, having bested the GTX 680 in recent months, must now face the new King of the GPU's; Nvidia's GTX Titan!
For the past four weeks I have been working on this test day and night, whenever I could get the time between my job, my wife, and my 2 year old daughter. I have gone as far as sanely possible in my quest to make this comparison as fair and consistent as possible. The only variables are the video cards themselves. Each setup was tested on a fresh install of Windows 8 Pro with exactly the same hardware used. The testing was performed with both stock and overclocked cards all under water and at both 1080p and 1440p. I tested single cards as well as dual and carefully plotted the results in graphical form using FRAPS, Excel, and synthetic benches. The latest drivers available at the time of testing were used for both setups (13.2 beta 7 for the 7970's and 314.09 for the Titans). I may not be a professional hardware site like Anand or Guru3d but I have done the best I could to get consistent results that you can rely on. I am also going to be filming a video version of this comparison for my Youtube channel (MegaTechPC) and will post that here as well. Let's crack on shall we?
Test Setup:
For all testing the overclocks for each of the GPU setups are as follows:
Sapphire R7970 OC - 1225MHz Core / 1754MHz Memory
EVGA GeForce GTX Titan SC - 1100MHz - 1176MHz / 3506MHz Memory
Here is a list of all the hardware used in this comparison:
CPU - Intel Core i7 Extreme 3960X - 5GHz
Memory - Patriot Viper Xtreme 4 x 4GB RAM - 2133MHz
Motherboard - Asus Rampage IV Extreme
PSU - Corsair AX1200W
SSD - Corsair Force 3 240GB
HDD - Hitachi Deskstar 1TB
Case - Silverstone Temjin TJ11
OS - Microsoft Windows 8 Pro
Monitors - 2 x Achieva Shimian IPSB 2560x1440
Here is a list of the water cooling components:
CPU Block - EK Supreme HF
Motherboard Blocks - EK R4E Full Board Kit
GPU Blocks - EK FC7970 / FC Titan
Radiators - 2 x Hardware Labs Black Ice GTX (560mm and 240mm)
Water Pump - Alphacool VPP-655 with Bitspower Pump Top and Mod Kit
Reservoir - Bitspower 150mm Cylinder res
Fittings - Bitspower Silver Shining
Tubing - Durelene 1/2" x 3/4"
Synthetic Benchmarks:
3DMark Vantage
3DMark11
3DMark Firestrike
Unigine Heaven 4.0
Unigine Valley 1.0
Aliens vs Predator Benchmark
Metro2033 Benchmark
I know that synthetic benches aren't very popular around here due to the fact that AMD and Nvidia optimize their drivers specifically for them and the results aren't necessarily applicable to actual gaming but as a bencher I personally love these synthetics and consider them a good starting point for any video card comparison. For the 3DMark testing I ran each setup under the Performance and Extreme presets and recorded the total score, GPU score, and combined scores. For the Unigine testing I used the highest possible graphical settings (including extreme tessellation and maximum anti-aliasing) at both 1080p and 1440p, then recorded the average, maximum and minimum FPS. AvP runs by default at 1080p so I recorded a 100 second run with FRAPS both at stock clocks and overclocked on each setup. Finally, Metro2033's built-in benchmark was used with all settings to maximum (4xMSAA, Very High Quality, DOF enabled) but Advanced PhysX turned off. I recorded average, maximum and minimum FPS as well.
3DMARK VANTAGE
Starting off with 3DMark Vantage you can really see the power of the new TItans in comparison to the 7970's, which scored pretty well in their own right. The single Titan had a nearly 25% higher Graphics score than the 7970 while the SLI Titans were nearly 15% higher. Combine two Titans with a 5GHz 3960X and 2133MHz memory and you get a 66,601 PScore in Vantage! The Titan advantage opens up even more with higher resolutions when you run the Xtreme Preset.
Individual Vantage screens:
3DMARK11
Moving on to 3DMark11 we see an even more massive advantage for the Titans in SLI beating the CF 7970's in Graphics Score by 46%! Single card scores were also 41% higher for Titan than the 7970 could muster even though the latest beta 7 drivers allowed a single 7970 to score nearly 12k points. As we saw with Vantage, increasing resolution to the Xtreme Preset merely allows the Titans to continue to press their advantage with an overwhelming 57% advantage in Graphics Score in SLI and a 54% advantage for the single cards.
3DMark11 Screens:
3DMARK FIRESTRIKE
The newest 3DMark benchmark was also the most difficult to dial in the Titans. While +129MHz core was fine in 3DMark11 and Vantage, it caused crashes in Firestrike and I had to lower the clocks all the way down to +70MHz. This was a harbinger of things to come with the Titans as GPU Boost 2.0 changes everything about overclocking, mostly for the worse. The 7970's meanwhile are set-it-and-forget-it easy, clocking to 1225MHz and 1754Mhz core and memory with ease in all tests. Here we see an advantage of the single Titan over the 7970 of nearly 25% but the SLI scaling is not as good as Cross Fire as dual card scores were only around 17% better (The 7970's scaled at 95% while the second Titan only improved scores 83% over the single). Once again upping the test to Xtreme allowed the Titans to pull away from the 7970's due to the sheer power of the GK110 GPU.
3DMark Firestrike Screens:
UNIGINE HEAVEN 4.0
Now that the Futuremark tests are complete we can move on to the Unigine tests starting with Heaven 4.0, which is basically the same as the previous iterations of the bench but with some tweaks that really push even current-generation cards to their limits. Running the test in the Ultra Preset with Extreme Tessellation and maximum anti aliasing (8x) even the mighty Titan struggles, especially at 1440p. Still, its the 7970 that gets its clock cleaned in this bench at both 1080p and 1440p, losing to the Titan by 56% and 53% respectively. Things aren't much better for the AMD cards in Cross Fire either as SLI Titans crush them at 1080p by 50% and by 48% at 1440p. This is a massive performance gap between the cards when you consider that both are already overclocked significantly.
Unigine Heaven 4.0 Screens:
UNIGINE VALLEY 1.0
Unigine's newest benchmark is called Valley 1.0 and its a gorgeous rendering of a lush, mountainous terrain complete with shifting times of day and varying weather conditions (the music is really soothing too). Though not as GPU intensive as Heaven, Valley is still a AAA benchmark and taxes both setups well. The Titan performs well here as throttling is not much of a factor while the 7970's chug right along at their constant clocks. At 1080p the SLI Titans are a force to be reckoned with, putting up FPS numbers rivaling triple GTX 680 scores. In this particular test I saw a 31% increase in FPS over the CF 7970's while the single card advantage was significantly greater at 44%. At 1440p the single Titan is 42% faster than the single 7970 while the SLI Titans manage an impressive 88.2 FPS which is 35% faster than the CF 7970's 64.9 FPS.
Unigine Valley 1.0 Screens:
ALIENS VERSUS PREDATOR BENCHMARK
I know this is an older bench and the options are rather slim for it but I decided to include it as it does give some real world flavor to the results. I ran the bench at both stock and overclocked speeds on all of the setups and used a 100 second FRAPS run for scoring. Interestingly I noticed in this bench particularly that overclocking the 7970's yields far more impressive results than do the Titans. This is to be expected given the dynamic nature of the Titan's GPU Boost, meaning that even at stock they are actually overclocked. Since the 7970's hold their clocks at stock and at OC, the gains are higher (in this bench overclocking the CF 7970's added 28 FPS to the total, or a 22% increase). That said, the Titans performed exceptionally well, hitting nearly 200 FPS average in this bench (a 25% advantage over the OC'd 7970's).
METRO 2033 BENCHMARK
The last of my synthetic benchmarks for this comparison is the Metro 2033 built-in benchmark. This particular bench is quirky and the minimum and maximum FPS numbers are all over the place, making the data a little useless. Still, the average FPS numbers scale as expected with the single and dual Titan setups scoring significant victories over the AMD cards. The settings for this test were fully maxed out minus the Advanced PhysX option considering that the AMD cards lack that ability. Otherwise everything was set to Very High with 4xMSAA, 16x Anisotropy and both DX11 features enabled (Tessellation and DOF). Metro remains one of the most graphics intensive DX11 titles out there and the results bear that out with SLI Titans just hitting 70 FPS at 1440p. This was a 43% advantage over CF 7970's and is similar to the 40% advantage the single Titan held over the single 7970. At 1080p the 7970 creeps up closer to the Titan hitting 48 FPS with a single card which was just 27% slower than the single Titan. SLI Titans managed a 24% advantage over CF 7970's at 1080p.
Metro 2033 Screens:
Gaming Benchmarks
Battlefield 3
Crysis
Crysis 3
Medal of Honor Warfighter
Far Cry 3
For the actual game play benchmarks I used FRAPS to record a 2 minute play through of each game including average, max, and min FPS. I also recorded frametimes of both setups in each game which I will be going over later. Each game has been fully maxed out graphically and tested at both 1080p and 1440p. I did a stock run for each setup at 1440p but only graphed overclocked runs as that is the really pertinent data anyway (I can post stock numbers if anybody really wants to see them). I must say that this part of the testing was by far the toughest as I had to find a 2-minute section of each game that was fairly repeatable with a good representation of the game engine and various effects. In the video version of this comparison I will show exactly where these run-throughs took place and discuss my benching methodology for each game. With that being said, let's continue shall we?
BATTLEFIELD 3
Battlefield 3 remains one of the most popular PC games out there well over a year after its release and for good reason as it has incredible graphics and effects thanks to its Frostbite 2.0 game engine as well as an immersive and team-based MP component. For this test I will be focusing on the SP campaign which has slightly better visuals than the MP and obviously eliminates unpredictable variables inherent in MP. The FRAPS run was performed at the very beginning of "Operation Guillotine" where you run down the hill avoiding mortar blasts until setting off your own mortar and jumping the wall to get down to the river on the other side.
As we can see in the 1080p graph there appears to be a 201 FPS limit in BF3 and that caused some interesting issues with the SLI Titan setup. At first I assumed that there was a CPU bottleneck situation here as the CF 7970's actually beat the Titans by about 5% at 1080p. In single card runs the Titan is 20% faster so I can only assume that the FPS cap in BF3 is somehow negatively affecting the SLI Titans, probably due to GPU Boost 2.0. The 7970's have no such issues and performed very well at 1080p. At 1440p neither setup was able to hit the 201 FPS cap (though the Titans got to 198 FPS) so we don't see the strange behavior we saw at 1080p and the SLI Titans win by 15%. For the record, either card is sufficient to max BF3 at 1440p with over 60 FPS!
Game Settings:
CRYSIS
"But can it run Crysis?"
Yes. Yes indeed, both of these setups can max out Crysis more than 5 years after its release but that just shows how far ahead of its time Crysis was when it released. Even the two fastest single-GPU video cards on the planet in 2013 (7970 and Titan) can only manage 76 FPS and 87 FPS respectively in this title when overclocked at 1080p! At 1440p neither card can maintain a 60 FPS average and that is on a game released in 2007!
The FRAPS run I used was at the very beginning of the game just before you kill the first KPA soldier standing on the rock. Run down the beach, kill the other two KPA soldiers, take down a couple of trees and go meet up with Jester to complete the bench. With all settings on maximum and at 1440p this game still looks breathtaking and both of the dual-card setups run it butter-smooth with the Titans edging out the 7970's by 11 FPS (or 12%). Like was seen in BF3, the first two real world benchmarks seem to tighten things up between the 7970 and the Titan more so than the synthetics did. In fact, at 1080p the CF 7970's actually beat the Titans by 3 FPS ( 2%) making Crysis the second game that performed better with dual 7970's than with dual Titans! Looking at the single card numbers this looks to be an SLI scaling issue with the Titan launch drivers as the single Titan is 15% faster than the single 7970 at 1080p. Simply put, the 7970's more mature drivers scale much better in Crysis in CF (68%) than do the Titans (only 44%), at least at 1080p.
Game Settings
CRYSIS 3
Crysis 3 is the newest game I have included in this comparison and its also the most graphically demanding. In fact, even the mighty Titan was barely able to pull off playable frame rates at 1440p fully maxed out in this game (29 FPS) while the 7970 was almost a slide show at just 21 FPS! If you want to play Crysis 3 at maximum settings and at 1440p you better be packing a dual card setup and even then don't expect 60 FPS as the SLI Titans will only get you to 55 (a 41% improvement over the 7970's 39 FPS). At 1080p obviously things become more sane with both single cards managing playable frame rates, though the Titan is a significant 45% faster and a whopping 53% faster in SLI! Nvidia's SLI scaling it would seem is doing just fine in Crysis 3 at 81% and clearly is superior to AMD's CF. In fact I would say that the Titan is a vastly superior video card in Crysis 3 than the 7970 is.
The FRAPS run I used was at the beginning of the game as I haven't had any time to play through so far (in fact I bought the game just for this comparison and haven't done any actual gaming with it yet). After Psycho hands you the gun, follow him down the ramp and do a few 360-degree pans while continuing up the stairs and into the control room. What little I have seen of the game in my testing still looks simply astonishing and is already easily the best looking PC game I have ever seen, especially at 1440p. Crytek definitely did the original Crysis justice with Crysis 3, at least graphically, and like the original game it brings current generation video hardware to its knees.
Game Settings:
FAR CRY 3
Far Cry 3 is a game that I have struggled with since it launched late last year. At the time I was running CF 7970's in my main rig and for some reason the game would never run smooth as well as having ridiculous input lag (even in the menus). I was getting around 25-30 FPS with CF enabled and this went on through several driver updates. It was so bad that I gave up playing the game on my sig rig and installed it on my dedicated folder (with a single 580 Lightning) just so I could play it. Fast forward to this comparison test and a fresh install of Win8 on my SSD and the game suddenly worked normally enough that I was able to actually get consistent FRAPS runs with it. I have no idea why it gave me so much trouble on my old Windows 8 install but since the game worked fine on the new one I included it in my comparison. I'm glad I did too because this game is nearly as graphically demanding as Crysis 3 and gave both setups a good work out.
The test run for this game begins after finding the mushrooms to cure Daisy for Dr. Earnhart. Starting in your save house in Amanaki Village you simply walk down the lane, across the road and into the bamboo area, looking around you at the lush jungle vegetation. After crossing another road and climbing down a steep rocky embankment the two minutes are upand its time to check out the results. At 1440p we see a similar story as we saw in Crysis 3 in that the single 7970 is simply not playable with an average of 24 FPS and a minimum of 16 FPS. The Titan is barely playable with minimums dipping to 25 FPS and an average of 34 FPS which is 42% faster than the 7970. Once a second 7970 is added to the mix frame rates become much better at 46 FPS though still well behind that of the SLI Titans which nearly make it to the magical 60 FPS number. Interestingly in my testing the CF 7970's had a higher minimum frame rate (34 FPS) than the SLI Titans did (29 FPS) but this was not noticeable during testing.
Both of the cards are sufficient to run FC3 at maximum settings at 1080p though the single Titan is some 36% faster than the single 7970. The dual card testing at 1080p revealed slightly better CF scaling than SLI with a second 7970 improving performance by nearly 90% while the second Titan yielded only an 80% improvement over the single card. The minimum frame rate discrepancy we saw at 1440p did not occur at 1080p and minimums scaled more or less as expected.
Game Settings:
MEDAL OF HONOR WARFIGHTER
Running on the Frostbite 2.0 game engine, I expected MOHW to be very similar in performance to Battlefield 3 and was mostly correct in that assumption. Again there appears to be some sort of frame rate cap in this game of around 200 FPS and both dual card setups were able to reach that limit. Interestingly though I didn't experience any of the odd performance discrepancies in MOHW as I did in BF3 and the Titans held a significant advantage over the 7970's of 15% at 1080p. Both cards are powerful enough to run MOHW smoothly at both 1080p and 1440p, though the Titan was able to keep minimums at or above 60 FPS even at 1440p! Overall the Titan is around 35% faster than the 7970 in this game.
For the FRAPS run I used the opening sequence where you start off popping the terrorist in the back of the head after sneaking onto the beach. After setting the charge and watching the entire crane structure collapse around you, you dispatch of the remaining terrorists (a couple of which are conveniently burning to death already) and the run is complete. While I have noted mostly negative opinions of this title here on OCN I have nothing but praise for the visuals at least. The game looks pretty close to BF3 graphically and that's no bad thing. As far as game play goes, I haven't gotten enough playing time into it yet to really have a concrete opinion so far. I will say that I enjoyed the first MOH on PC quite a lot.
Game Settings:
Unfortunately this marks the end of the game testing as I wasn't able to get a copy of Tomb Raider before breaking down the 7970's. I would also liked to have tested Skyrim or Witcher 2 but I simply ran out of time as the testing is immensely time consuming (I've literally been testing since the beginning of March!). With the inclusion of Crysis 3 and FC3 I do believe that these results give a good idea of how these setups will fare in the coming months as more and more graphically demanding titles (such as BF4) release.
See final conclusion below!
The Testing Rig:
(Please note that the article's conclusion can be found in the second post HERE)
I know, I know. Titan has been out for a month and there have been countless reviews, comparisons, benchmarks, and everything else in between inundating the fair members here at OCN. So why post a comparison that is late to the game and where the conclusions are fairly obvious? Well, I figured we needed a different perspective on this subject. An everyman's comparison that captures the vantage point of the typical OCN member. Well, the typical OCN member that water cools every major component in his rig anyway. At this time the 7970 is AMD's flagship and, having bested the GTX 680 in recent months, must now face the new King of the GPU's; Nvidia's GTX Titan!
For the past four weeks I have been working on this test day and night, whenever I could get the time between my job, my wife, and my 2 year old daughter. I have gone as far as sanely possible in my quest to make this comparison as fair and consistent as possible. The only variables are the video cards themselves. Each setup was tested on a fresh install of Windows 8 Pro with exactly the same hardware used. The testing was performed with both stock and overclocked cards all under water and at both 1080p and 1440p. I tested single cards as well as dual and carefully plotted the results in graphical form using FRAPS, Excel, and synthetic benches. The latest drivers available at the time of testing were used for both setups (13.2 beta 7 for the 7970's and 314.09 for the Titans). I may not be a professional hardware site like Anand or Guru3d but I have done the best I could to get consistent results that you can rely on. I am also going to be filming a video version of this comparison for my Youtube channel (MegaTechPC) and will post that here as well. Let's crack on shall we?
ALT=""][URL=http://www.overclock.net/content/type/61/id/1381518/width/500/height/1000%5B/IMG]http://www.overclock.net/content/type/61/id/1381518/width/500/height/1000[/IMG[/URL]]
[URL=http://www.overclock.net/content/type/61/id/1381519/][IMG alt=""]http://www.overclock.net/content/type/61/id/1381519/width/500/height/1000/flags/LL[/URL]
[URL=http://www.overclock.net/content/type/61/id/1381519/][IMG alt=""]http://www.overclock.net/content/type/61/id/1381519/width/500/height/1000/flags/LL[/URL]
For all testing the overclocks for each of the GPU setups are as follows:
Sapphire R7970 OC - 1225MHz Core / 1754MHz Memory
EVGA GeForce GTX Titan SC - 1100MHz - 1176MHz / 3506MHz Memory
Here is a list of all the hardware used in this comparison:
CPU - Intel Core i7 Extreme 3960X - 5GHz
Memory - Patriot Viper Xtreme 4 x 4GB RAM - 2133MHz
Motherboard - Asus Rampage IV Extreme
PSU - Corsair AX1200W
SSD - Corsair Force 3 240GB
HDD - Hitachi Deskstar 1TB
Case - Silverstone Temjin TJ11
OS - Microsoft Windows 8 Pro
Monitors - 2 x Achieva Shimian IPSB 2560x1440
Here is a list of the water cooling components:
CPU Block - EK Supreme HF
Motherboard Blocks - EK R4E Full Board Kit
GPU Blocks - EK FC7970 / FC Titan
Radiators - 2 x Hardware Labs Black Ice GTX (560mm and 240mm)
Water Pump - Alphacool VPP-655 with Bitspower Pump Top and Mod Kit
Reservoir - Bitspower 150mm Cylinder res
Fittings - Bitspower Silver Shining
Tubing - Durelene 1/2" x 3/4"
Synthetic Benchmarks:
3DMark Vantage
3DMark11
3DMark Firestrike
Unigine Heaven 4.0
Unigine Valley 1.0
Aliens vs Predator Benchmark
Metro2033 Benchmark
I know that synthetic benches aren't very popular around here due to the fact that AMD and Nvidia optimize their drivers specifically for them and the results aren't necessarily applicable to actual gaming but as a bencher I personally love these synthetics and consider them a good starting point for any video card comparison. For the 3DMark testing I ran each setup under the Performance and Extreme presets and recorded the total score, GPU score, and combined scores. For the Unigine testing I used the highest possible graphical settings (including extreme tessellation and maximum anti-aliasing) at both 1080p and 1440p, then recorded the average, maximum and minimum FPS. AvP runs by default at 1080p so I recorded a 100 second run with FRAPS both at stock clocks and overclocked on each setup. Finally, Metro2033's built-in benchmark was used with all settings to maximum (4xMSAA, Very High Quality, DOF enabled) but Advanced PhysX turned off. I recorded average, maximum and minimum FPS as well.
3DMARK VANTAGE
Starting off with 3DMark Vantage you can really see the power of the new TItans in comparison to the 7970's, which scored pretty well in their own right. The single Titan had a nearly 25% higher Graphics score than the 7970 while the SLI Titans were nearly 15% higher. Combine two Titans with a 5GHz 3960X and 2133MHz memory and you get a 66,601 PScore in Vantage! The Titan advantage opens up even more with higher resolutions when you run the Xtreme Preset.
Individual Vantage screens:
3DMARK11
Moving on to 3DMark11 we see an even more massive advantage for the Titans in SLI beating the CF 7970's in Graphics Score by 46%! Single card scores were also 41% higher for Titan than the 7970 could muster even though the latest beta 7 drivers allowed a single 7970 to score nearly 12k points. As we saw with Vantage, increasing resolution to the Xtreme Preset merely allows the Titans to continue to press their advantage with an overwhelming 57% advantage in Graphics Score in SLI and a 54% advantage for the single cards.
3DMark11 Screens:
3DMARK FIRESTRIKE
The newest 3DMark benchmark was also the most difficult to dial in the Titans. While +129MHz core was fine in 3DMark11 and Vantage, it caused crashes in Firestrike and I had to lower the clocks all the way down to +70MHz. This was a harbinger of things to come with the Titans as GPU Boost 2.0 changes everything about overclocking, mostly for the worse. The 7970's meanwhile are set-it-and-forget-it easy, clocking to 1225MHz and 1754Mhz core and memory with ease in all tests. Here we see an advantage of the single Titan over the 7970 of nearly 25% but the SLI scaling is not as good as Cross Fire as dual card scores were only around 17% better (The 7970's scaled at 95% while the second Titan only improved scores 83% over the single). Once again upping the test to Xtreme allowed the Titans to pull away from the 7970's due to the sheer power of the GK110 GPU.
3DMark Firestrike Screens:
UNIGINE HEAVEN 4.0
Now that the Futuremark tests are complete we can move on to the Unigine tests starting with Heaven 4.0, which is basically the same as the previous iterations of the bench but with some tweaks that really push even current-generation cards to their limits. Running the test in the Ultra Preset with Extreme Tessellation and maximum anti aliasing (8x) even the mighty Titan struggles, especially at 1440p. Still, its the 7970 that gets its clock cleaned in this bench at both 1080p and 1440p, losing to the Titan by 56% and 53% respectively. Things aren't much better for the AMD cards in Cross Fire either as SLI Titans crush them at 1080p by 50% and by 48% at 1440p. This is a massive performance gap between the cards when you consider that both are already overclocked significantly.
Unigine Heaven 4.0 Screens:
UNIGINE VALLEY 1.0
Unigine's newest benchmark is called Valley 1.0 and its a gorgeous rendering of a lush, mountainous terrain complete with shifting times of day and varying weather conditions (the music is really soothing too). Though not as GPU intensive as Heaven, Valley is still a AAA benchmark and taxes both setups well. The Titan performs well here as throttling is not much of a factor while the 7970's chug right along at their constant clocks. At 1080p the SLI Titans are a force to be reckoned with, putting up FPS numbers rivaling triple GTX 680 scores. In this particular test I saw a 31% increase in FPS over the CF 7970's while the single card advantage was significantly greater at 44%. At 1440p the single Titan is 42% faster than the single 7970 while the SLI Titans manage an impressive 88.2 FPS which is 35% faster than the CF 7970's 64.9 FPS.
Unigine Valley 1.0 Screens:
ALIENS VERSUS PREDATOR BENCHMARK
I know this is an older bench and the options are rather slim for it but I decided to include it as it does give some real world flavor to the results. I ran the bench at both stock and overclocked speeds on all of the setups and used a 100 second FRAPS run for scoring. Interestingly I noticed in this bench particularly that overclocking the 7970's yields far more impressive results than do the Titans. This is to be expected given the dynamic nature of the Titan's GPU Boost, meaning that even at stock they are actually overclocked. Since the 7970's hold their clocks at stock and at OC, the gains are higher (in this bench overclocking the CF 7970's added 28 FPS to the total, or a 22% increase). That said, the Titans performed exceptionally well, hitting nearly 200 FPS average in this bench (a 25% advantage over the OC'd 7970's).
METRO 2033 BENCHMARK
The last of my synthetic benchmarks for this comparison is the Metro 2033 built-in benchmark. This particular bench is quirky and the minimum and maximum FPS numbers are all over the place, making the data a little useless. Still, the average FPS numbers scale as expected with the single and dual Titan setups scoring significant victories over the AMD cards. The settings for this test were fully maxed out minus the Advanced PhysX option considering that the AMD cards lack that ability. Otherwise everything was set to Very High with 4xMSAA, 16x Anisotropy and both DX11 features enabled (Tessellation and DOF). Metro remains one of the most graphics intensive DX11 titles out there and the results bear that out with SLI Titans just hitting 70 FPS at 1440p. This was a 43% advantage over CF 7970's and is similar to the 40% advantage the single Titan held over the single 7970. At 1080p the 7970 creeps up closer to the Titan hitting 48 FPS with a single card which was just 27% slower than the single Titan. SLI Titans managed a 24% advantage over CF 7970's at 1080p.
Metro 2033 Screens:
Gaming Benchmarks
Battlefield 3
Crysis
Crysis 3
Medal of Honor Warfighter
Far Cry 3
For the actual game play benchmarks I used FRAPS to record a 2 minute play through of each game including average, max, and min FPS. I also recorded frametimes of both setups in each game which I will be going over later. Each game has been fully maxed out graphically and tested at both 1080p and 1440p. I did a stock run for each setup at 1440p but only graphed overclocked runs as that is the really pertinent data anyway (I can post stock numbers if anybody really wants to see them). I must say that this part of the testing was by far the toughest as I had to find a 2-minute section of each game that was fairly repeatable with a good representation of the game engine and various effects. In the video version of this comparison I will show exactly where these run-throughs took place and discuss my benching methodology for each game. With that being said, let's continue shall we?
BATTLEFIELD 3
Battlefield 3 remains one of the most popular PC games out there well over a year after its release and for good reason as it has incredible graphics and effects thanks to its Frostbite 2.0 game engine as well as an immersive and team-based MP component. For this test I will be focusing on the SP campaign which has slightly better visuals than the MP and obviously eliminates unpredictable variables inherent in MP. The FRAPS run was performed at the very beginning of "Operation Guillotine" where you run down the hill avoiding mortar blasts until setting off your own mortar and jumping the wall to get down to the river on the other side.
As we can see in the 1080p graph there appears to be a 201 FPS limit in BF3 and that caused some interesting issues with the SLI Titan setup. At first I assumed that there was a CPU bottleneck situation here as the CF 7970's actually beat the Titans by about 5% at 1080p. In single card runs the Titan is 20% faster so I can only assume that the FPS cap in BF3 is somehow negatively affecting the SLI Titans, probably due to GPU Boost 2.0. The 7970's have no such issues and performed very well at 1080p. At 1440p neither setup was able to hit the 201 FPS cap (though the Titans got to 198 FPS) so we don't see the strange behavior we saw at 1080p and the SLI Titans win by 15%. For the record, either card is sufficient to max BF3 at 1440p with over 60 FPS!
Game Settings:
CRYSIS
"But can it run Crysis?"
Yes. Yes indeed, both of these setups can max out Crysis more than 5 years after its release but that just shows how far ahead of its time Crysis was when it released. Even the two fastest single-GPU video cards on the planet in 2013 (7970 and Titan) can only manage 76 FPS and 87 FPS respectively in this title when overclocked at 1080p! At 1440p neither card can maintain a 60 FPS average and that is on a game released in 2007!
The FRAPS run I used was at the very beginning of the game just before you kill the first KPA soldier standing on the rock. Run down the beach, kill the other two KPA soldiers, take down a couple of trees and go meet up with Jester to complete the bench. With all settings on maximum and at 1440p this game still looks breathtaking and both of the dual-card setups run it butter-smooth with the Titans edging out the 7970's by 11 FPS (or 12%). Like was seen in BF3, the first two real world benchmarks seem to tighten things up between the 7970 and the Titan more so than the synthetics did. In fact, at 1080p the CF 7970's actually beat the Titans by 3 FPS ( 2%) making Crysis the second game that performed better with dual 7970's than with dual Titans! Looking at the single card numbers this looks to be an SLI scaling issue with the Titan launch drivers as the single Titan is 15% faster than the single 7970 at 1080p. Simply put, the 7970's more mature drivers scale much better in Crysis in CF (68%) than do the Titans (only 44%), at least at 1080p.
Game Settings
CRYSIS 3
Crysis 3 is the newest game I have included in this comparison and its also the most graphically demanding. In fact, even the mighty Titan was barely able to pull off playable frame rates at 1440p fully maxed out in this game (29 FPS) while the 7970 was almost a slide show at just 21 FPS! If you want to play Crysis 3 at maximum settings and at 1440p you better be packing a dual card setup and even then don't expect 60 FPS as the SLI Titans will only get you to 55 (a 41% improvement over the 7970's 39 FPS). At 1080p obviously things become more sane with both single cards managing playable frame rates, though the Titan is a significant 45% faster and a whopping 53% faster in SLI! Nvidia's SLI scaling it would seem is doing just fine in Crysis 3 at 81% and clearly is superior to AMD's CF. In fact I would say that the Titan is a vastly superior video card in Crysis 3 than the 7970 is.
The FRAPS run I used was at the beginning of the game as I haven't had any time to play through so far (in fact I bought the game just for this comparison and haven't done any actual gaming with it yet). After Psycho hands you the gun, follow him down the ramp and do a few 360-degree pans while continuing up the stairs and into the control room. What little I have seen of the game in my testing still looks simply astonishing and is already easily the best looking PC game I have ever seen, especially at 1440p. Crytek definitely did the original Crysis justice with Crysis 3, at least graphically, and like the original game it brings current generation video hardware to its knees.
Game Settings:
FAR CRY 3
Far Cry 3 is a game that I have struggled with since it launched late last year. At the time I was running CF 7970's in my main rig and for some reason the game would never run smooth as well as having ridiculous input lag (even in the menus). I was getting around 25-30 FPS with CF enabled and this went on through several driver updates. It was so bad that I gave up playing the game on my sig rig and installed it on my dedicated folder (with a single 580 Lightning) just so I could play it. Fast forward to this comparison test and a fresh install of Win8 on my SSD and the game suddenly worked normally enough that I was able to actually get consistent FRAPS runs with it. I have no idea why it gave me so much trouble on my old Windows 8 install but since the game worked fine on the new one I included it in my comparison. I'm glad I did too because this game is nearly as graphically demanding as Crysis 3 and gave both setups a good work out.
The test run for this game begins after finding the mushrooms to cure Daisy for Dr. Earnhart. Starting in your save house in Amanaki Village you simply walk down the lane, across the road and into the bamboo area, looking around you at the lush jungle vegetation. After crossing another road and climbing down a steep rocky embankment the two minutes are upand its time to check out the results. At 1440p we see a similar story as we saw in Crysis 3 in that the single 7970 is simply not playable with an average of 24 FPS and a minimum of 16 FPS. The Titan is barely playable with minimums dipping to 25 FPS and an average of 34 FPS which is 42% faster than the 7970. Once a second 7970 is added to the mix frame rates become much better at 46 FPS though still well behind that of the SLI Titans which nearly make it to the magical 60 FPS number. Interestingly in my testing the CF 7970's had a higher minimum frame rate (34 FPS) than the SLI Titans did (29 FPS) but this was not noticeable during testing.
Both of the cards are sufficient to run FC3 at maximum settings at 1080p though the single Titan is some 36% faster than the single 7970. The dual card testing at 1080p revealed slightly better CF scaling than SLI with a second 7970 improving performance by nearly 90% while the second Titan yielded only an 80% improvement over the single card. The minimum frame rate discrepancy we saw at 1440p did not occur at 1080p and minimums scaled more or less as expected.
Game Settings:
MEDAL OF HONOR WARFIGHTER
Running on the Frostbite 2.0 game engine, I expected MOHW to be very similar in performance to Battlefield 3 and was mostly correct in that assumption. Again there appears to be some sort of frame rate cap in this game of around 200 FPS and both dual card setups were able to reach that limit. Interestingly though I didn't experience any of the odd performance discrepancies in MOHW as I did in BF3 and the Titans held a significant advantage over the 7970's of 15% at 1080p. Both cards are powerful enough to run MOHW smoothly at both 1080p and 1440p, though the Titan was able to keep minimums at or above 60 FPS even at 1440p! Overall the Titan is around 35% faster than the 7970 in this game.
For the FRAPS run I used the opening sequence where you start off popping the terrorist in the back of the head after sneaking onto the beach. After setting the charge and watching the entire crane structure collapse around you, you dispatch of the remaining terrorists (a couple of which are conveniently burning to death already) and the run is complete. While I have noted mostly negative opinions of this title here on OCN I have nothing but praise for the visuals at least. The game looks pretty close to BF3 graphically and that's no bad thing. As far as game play goes, I haven't gotten enough playing time into it yet to really have a concrete opinion so far. I will say that I enjoyed the first MOH on PC quite a lot.
Game Settings:
Unfortunately this marks the end of the game testing as I wasn't able to get a copy of Tomb Raider before breaking down the 7970's. I would also liked to have tested Skyrim or Witcher 2 but I simply ran out of time as the testing is immensely time consuming (I've literally been testing since the beginning of March!). With the inclusion of Crysis 3 and FC3 I do believe that these results give a good idea of how these setups will fare in the coming months as more and more graphically demanding titles (such as BF4) release.
See final conclusion below!
The Testing Rig:
