Let’s just go ahead and get it over with, Nvidia’s GeForce GTX 780 Ti is a performance monster. The GPU is the full GK110 core (Keplar) unlocked, meaning all 15 of the SMXes are activated. In comparison this is a step beyond even the original Nvidia GeForce Titan, which featured just 14 SMXes. As impressive as the Titan was back in February of last year, due to one of its SMXes disabled, there was still lots of performance going untapped.
Things have changed substantially within the GPU market since the Titan’s debut. AMD’s release of the R9 290 and R9 290X, Nvidia once again had a fight on its hands. The R9 290X, despite having heat issues (which the reference cooler only made worse) exchanged blows with the Titan, and was considerably cheaper to boot. Nvidia’s GeForce 780 was within the R9 290 or 290X type of price range, but lacked the Titan’s or AMD’s level of performance. Therefore the Titan was the only card Nvidia had to battle AMD on the bleeding edge, and AMD had it soundly beaten on the price vs performance ratio. The Release of AMD’s R9 series also forced Nvidia to reduce the pricing, but that wasn’t enough.
Enter the GeForce GTX 780 Ti, the first fully enabled GK110 card. With the 15th SMX enabled (this providing 2880 Stream Processors vs the Titan’s 2688) and with a slight clock speed jump on the GPU, and additional memory bandwidth, Nvidia have managed to pull the performance crown back from AMD. That’s not to say that the 780 Ti is a ‘better’ card than the Titan in all areas, hence the reason the Titan still exists. The Titan has twice the RAM (6GB vs 3GB of GDDR5) and much better compute potential. Therefore, I suppose you could say that the Titan is for work and play, where as the GTX 780 Ti is built to be cheaper and more powerful for games. So in the end we have the GTX 780 Ti boasting 2880 Stream Processors, 240 Texture Units, 48 ROPS, and 3GB of GDDR5 RAM at 7000MHZ effective on a 384 Bit Memory Bus, providing 336GB/s of bandwidth. This is higher than even the R9 290X, which manages ‘just’ 320GB/s despite being on a 512 bit bus, due to the ram running at only 6GHZ effective.
EVGA’s GTX 780 Ti SC SuperClock ACX takes the basic 780 Ti specs, and throws a factory overclock into the mix. Its custom cooler allows EVGA’s GTX 780 Ti SC ACX to be clocked at 1006MHZ Core vs the reference design’s 875MHZ. The Boost 2.0 clock meanwhile has seen a corresponding increase to 1072MHZ vs the reference boards Boost of 928MHZ. As you’d expect, EVGA’s card requires the standard 8 and 6 pin power connectors found on GTX 780 Ti’s, drawing up to 250W of power (more if you decide to overclock the card more yourself).
EVGA 780 Ti SC | GTX 780 Ti | GTX Titan | GTX 780 | R9 290X | |
Stream Processors | 2880 | 2880 | 2688 | 2304 | 2816 |
Texture Units | 240 | 240 | 224 | 192 | 176 |
ROPs | 48 | 48 | 48 | 48 | 64 |
Core Clock | 1006 MHz | 876 MHz | 837MHz | 863MHz | 727MHz |
Boost Clock | 1072Mhz | 928 Mhz | 87 6Mhz | 900 Mhz | 1000MHz |
Memory Clock | 7GHz GDDR5 | 7GHz GDDR5 | 6GHz GDDR5 | 6GHz GDDR5 | 6GHz GDDR5 |
Memory Bus Width | 384-bit | 384-bit | 384-bit | 384-bit | 512-bit |
VRAM | 3GB | 3GB | 6GB | 3GB | 4GB |
FP64 | 1/24 FP32 | 1/24 FP32 | 1/3 FP32 | 1/24 FP32 | 1/8 FP32 |
TDP | 250W | 250W | 250W | 250W | 230W |
EVGA bundles isn’t the best best in terms of games, but there are plenty of goodies in the box. 1x 6 pin and 1x 8 Pin PCIE-E Power Adapter, a DVI to VGA Adapter, a few stickers, a driver disc, a cool poster and of course the usual assortment of warranty info, manuals. Oh yeah, and the card itself. You’ll also be given a code (I received mine via email which the retailer provides) for Assassin’s Creed 4 Black Flag. The non reference cooler should keep temps down, typically under load the card (using EVGA’s stock clock of 1006 / 1072 MHZ) rarely went past the low 60’s. Fan noise is fairly silent in normal operations, but is more noticeable if you manually crank the fan speed above the 70 percent mark. . There is a slight hint of coil whine, which is pretty much par for the course on higher end graphics cards. On my own card it was only audible if the side of the case was removed, game audio very low and you’re close to the system. Be wary if you’ve a smaller case, the card is lengthy at 10.5 inches, or 27 cm if you prefer. If you’re using in a smaller case then be sure you’ve got the clearance for the card.
Test Setup
The test rig we’re using for the benchmark was an Intel Haswell 4770K running at 4.4GHZ, with 16GB of DDR3 RAM, clocked at 1750MHZ and paired with an SSD. Nvidia drivers are their latest versions as of time of writing, 332.21. Windows 8.1 64-bit is used, and all games are their latest patched version from either Steam, Uplay or Origin, respectively. All benchmarks were run on their highest settings at either 1920×1080 or 2560×1440 and V-Sync was of course disabled for each benchmark. We’re looking at the average frame rate throughout.
2560×1440 (average) | EVGA 780 Ti SC | GTX 780 Ti | GTX Titan | GTX 680 | R9 290X |
Crysis 3 + FXAA | 49 FPS | 45 FPS | 44 | 31 | 42 |
Metro: Last Light | 55 FPS | 52 FPS | 47 | 34 | 51 |
Tomb Raider FXAA | 63 | 58 | 52 | 35 | 53 |
Sleeping Dogs | 47 | 45 | 40 | 31 | 40 |
Bioshock Infinite | 81 | 72 | 65 | 46 | 63 |
Crysis 3:
Crysis 3 Benchmark was conducted with everything of course set to the highest settings. The GTX 780 Ti literally rips Crysis 3 to pieces at 1080P, using FXAA frame rates continued to hit between the mid 60’s to mid 70’s in the “Welcome to the Jungle” level, despite all of the chaos happening on screen. Ramping things up, we to tax the GPU a little more by using 1080P at 8X MSAA. The frame rate of course takes a hit, reaching the mid 40’s to low 50’s depending on how busy things get on screen.
Moving on to 2560x1440P and FXAA however and we stay at the 40 to 50 mark, the performance impact being similar to enabling MSAAx8. So, what happens if we do the unthinkable and try out 8X MSAA and 2560×1440? Despite the frame rate being somewhat less optimal, hitting between the mid 20’s to low 30’s, it’s impressive that Crysis 3 is running at this frame rate at all. When you consider that previous generations single card leader, Nvidia’s GTX 680 manages to barely break 30FPS at 1440P with only FXAA, the results of the GTX 780 Ti are even more impressive. Of course, EVGA’s model manages to eek out a few frames per second more than the reference model thanks to its higher clock speed, between 4 – 6FPS to be precise.
Metro Last Light:
Metro Last Light from 4A Games features an extremely taxing game engine, fully utilizing the features of DX11. Starting things out with 1440P with SSAA and Nvidia’s Hardware Physx Disabled, you’re going to find that you’re hitting the 55 frames per second mark using the built in Metro Last Light benchmark. This gives the EVGA 780 Ti SC around a 4 frame margin over the reference card once again. You’ll also be seeing a 5 to 6 frame advantage over AMD’s R9 290X. SSAA takes a massive impact on frame rate at 1440P, bringing the frame rate to only 33FPS. It’s likely because of this you’ll not be wanting to use SSAA and instead force a more efficient Anti-Aliasing method in Nvidia control panel.
Running at 1080P with SSAA and Nvidia’s Hardware Physx, along with all other settings at their highest we manage to hit an average frame rate of 45 FPS with the built in Benchmark. Disabling Hardware Physx and we see a slight gain of 5 FPS, taking us to 50FPS total. It’s clear that for owners of 1080P displays, the better option is to force the more efficient methods of Anti-Aliasing and of course, enable Nvidia’s Hardware Physx. It adds a lot of atmosphere to titles which support it.
Tomb Raider
Despite Definitive Edition just seeing a release on Sony’s Playstation 4 and Microsoft’s Xbox One, the PC version is the undisputed kind of resolution, frame rate and features tessellation, which somewhat makes up for the improved TressFX and lighting seen on the consoles. With Tomb Raider we’re seeing a frame rate of 63FPS for the average during the built in Benchmark, using the “Ultimate” preset with FXAA, Tessellation and of course TressFX enabled at 2560×1440. The same familiar story of the EVGA SuperClock enjoying a slight advantage over the reference GTX 780 Ti continues, with the GTX 780 Ti reference being 5 frames per second slower on average. Meanwhile the R9 290x is pretty competitive, reaching a frame rate of 52. It’s worth noting that if you’re an owner of the GTX 680 in this testing, you’ll be lucky to reach beyond the mid 30’s.
Bioshock Infinite:
Bioshock Infinite is run with FXAA, DX11, Ultra Quality with DDOD enabled. Even at 1440P, EVGA’s GTX 780 Ti Superclock domplately dominants the Unreal Engine 3 game. Managing 81 frames per second, the EVGA 780 Ti SuperClock ACX pulls ahead almost 10 FPS over the standard reference unit. Nvidia’s Titan and AMD R9 290X are outclassed here in terms of raw performance. Clearly the Unreal Engine 3 isn’t pushing modern high end graphics cards hard enough and you’re left with the option to pile on the Anti-Aliasing. If you’re the owner if the GTX 680, 46 FPS is still respectable enough, but is still almost 30FPS slower than the EVGA GTX 780 Ti.