NVIDIA GeForce GTX 760 SLI Review

With NVIDIA dominating the high end GPU market, we take a look at the latest mid-range GeForce card based on the GK104 core – the NVIDIA GTX760…

Product on Review: GeForce GTX 760 2GB GDDR5
Manufacturer and Sponsor: NVIDIA
Price: £200-£230

The recent flurry of graphics cards from NVIDIA have certainly raised some eyebrows. The GTX TITAN, while being somewhat of an exclusive model, reserved for those more affluent of enthusiasts set the benchmark for others to follow. A little later and we get a new family flagship card in the form of the GTX 780 that mirrors previous NVIDIA cards which lead a family of GPUs. Then came it’s little brother, re-badged but revved up GTX 680 in the form of the GTX 770. This trio of cards have made life very difficult for AMD who are still clinging to their once fabled HD7970. Today, NVIDIA have adjusted their cross-hairs to the mid-range market and are assaulting this area with the GTX 760.

The GTX 760, like the GTX770 has some inherited components from the overflowing NVIDIA parts bin, namely the GK104 core codenamed Kepler. Kepler astounded us all with its tremendous power to performance ratio making it perhaps the most efficient GPU core available at the time. The core on the GTX 760 is however not the full fat GK104 we see in both the GTX 770 and GTX 680 models, instead this version has had 2 SMX units trimmed from the available 8 leaving 6 units remaining.

NVIDIA told us the GTX760 will be replacing the out-going GTX 660 Ti which pitches it directly against the AMD HD7950 GPU. This is a brave move indeed as the HD7950 is a card most would place in the high end sector rather than mid-range. We would however assume, given the average 2 year upgrade path of most gamers that the likely customers for the GTX 760 would be from those who currently own a GTX 560.

It is priced aggressively at £209 so if NVIDIA’s claims are to be believed, AMD should begin to worry because at this price it weighs in significantly cheaper than the HD7950.

Faster AND cheaper? This is not the NVIDIA we are familiar with!

About NVIDIA

Founded in 1993, NVIDIA has continuously reinvented itself to delight users and shape the industry. From our beginnings in PC graphics, we expanded into professional graphics to become the standard bearer in visual computing. We later harnessed the parallel computing capabilities of the GPU to advance high-performance computing. Our move into mobile put us at the center of one of the industry’s fastest-growing segments. With the invention of the virtual GPU, we’re accelerating cloud computing for consumers and enterprises.

Specification

Based on the same GK104 found on the GTX680, the GTX760 contains 1152 CUDA cores at a clockspeed of 980Mhz. This clockspeed coupled with GPU Boost 2.0 will allow the card to typically boost overclock to 1033MHz. The GTX760 also comes equipped with 2GB of GDDR5 and features four 64-bit memory controllers (256-bit). This memory is clocked to 6008Mhz (effective) however if previous results are anything to go by, this tasty memory will overclock much further than reference speeds so expect to see some good improvements in our overclocking section.

Above is an 8 Core Kepler. NVIDIA tell us that 2 of the cores would be disabled either by disabling 1 graphics processing cluster all together or just one core from two of the clusters.

Kepler Core
The GTX760 is equipped with a GK104 core is fabricated on the 28nm process with every component was designed with power efficiency in mind to ensure the GPU gave the best performance-per-watt possible.

While the 28nm manufacturing process holds the bulk of the power saving features, the way the new architecture works is also key to furthering Kepler’s power reduction. SMX, an evolution of the SM from Fermi is NVIDIA’s new streaming multiprocessor:

Each SMX runs at the graphics clock speed rather than 2x that speed as previously but because the GK104 has 1536 CUDA cores (8xSMX), over Fermi’s (GF110) 512 CUDA cores (16xSM), Kepler allows the GTX760 to operate at twice the performance per watt measurement compared to the GTX560.

To feed each SMX, each unit has four warp schedulers, each capable of firing off two instructions per warp, every clock. With a redesigned scheduling function, the GTX760 is a lot less complex and much more streamlined allowing the compiler to determine which instructions will be issued and can provide this information direct to the hardware block rather than going around the houses using Multi-port decode and a register scoreboard along with the dependency check before the information gets issued. This method is both quicker and more importantly, more power efficient.

Perhaps the biggest change from Fermi to Kepler was the omission of the old Shader Clock. The shader clock was getting tired as it was part of the old Tesla architecture which was implemented as an area optimisation. Because of the way Kepler executes instructions at a much more streamlined manner and at a higher clock rate, fewer copies of the execution unit need be made.

Another improvement over Fermi with the introduction of SMX is the redesigned Polymorph engine (2.0). The Polymorph engine takes card of the tessellation workload of the GTX770. The design of the Polymorph engine is to ensure that the tessellation we see on screen has little (as possible) impact on the rendering performance. What you may find bizarre though is that NVIDIA have cut the Polymorph engine count in half to 8 from Fermi’s 16 yet claim it is almost twice as quick. This is thanks to the improved core an memory clockpeed.

With both a higher core clockspeed based on older (but still very fast) GK104 core and of course that blisteringly fast memory, the GTX760 looks set to replace not only the GTX660Ti, but also offer a viable, yet cheaper alternative to the HD7950 GPU.

GPU Boost 2.0

GPU Boost caused a bit of a stir when it was first introduced because it did away with the traditional ‘absolute’ method of dictating MHz. Rather it worked by overclocking the card by different means. To achieve an overclock using GPU Boost, one increased the MHz and Voltage as per usual however the card would then ‘Boost’ when placed under load. So rather than having a fixed clockpseed, the GPU adjusts the clockspeed depending on load, ensuring the power is there when you need it rather than it being full on all of the time, thus saving power consumption and heat production by having the card overclock dynamically.

GPU Boost 2.0 is an evolution of this idea. The techs at NVIDIA have determined that GPU power was unnecessarily limiting performance when GPU temperatures was low. Instead of boosting GPU performance based on a power target, it is now based upon temperature with this temperature threshold being 80 degrees Celsius.

Because the boosting is now determined by temperature, noise levels are also reduced as NVIDIA claim the GTX cards deliver a better acoustic profile.

So to summarise, overclocking the card using GPU boost 2.0 means adjusting the temperature threshold, for example from 80c to 85c. The GPU clockspeed will then gradually increase until it reaches this new temperature and will be dynamically adjusted on-the-fly to stay within this threshold. Because of this it is much more difficult to claim one GPU is faster than another without having both working under precisely the same operating conditions.

It therefore should go without saying (but we will say it anyway) that the colder your card runs, the faster it should theoretically operate as a colder card allows overvolting and thus even higher clockspeeds.

GPU Boost 2.0 also features overvoltage control. NVIDIA make no bones that enabling this function will damage the long term operability of the graphics card and as with AMD’s overdrive, you have to acknowledge this fact before you can unlock the cards voltage control. We applaud NVIDIA for this because they have acknowledged that enthusiasts, like us here at Vortez, like to push hardware to the max and without voltage control, the whole process becomes so much more uninteresting and much more difficult.

So, because GTX760’s boost clock and voltage level are now tied to the GPU’s temperature, the voltage can be dialled higher which in turn will allow the GPU to attain higher clockspeed values. Obviously, increasingly board voltage, as with most other products, will gradually degrade the GPU through electron migration causing a premature death of the GPU. Voltage support will be offered on a board partner by partner basis however, we would be extremely surprised if it wasn’t enabled in every GTX760’s VBIOS.

We will see how GPU Boost 2.0 performs in our overclocking section later in the review.

GeForce Experience

GeForce Experience is a new application from NVIDIA that optimizes your PC in two key ways. First, it maximizes your game performance and game compatibility by automatically downloading the latest GeForce Game Ready drivers. Second, GeForce Experience intelligently optimizes graphics settings for all your favorite games based on your hardware configuration.

Some enthusiasts will view (excuse the pun) altering settings to get the very best from a game one of the ‘fun’ elements of owning a powerful graphics card. Others just want to play their game of choice without the endless testing and benchmarking to determine what settings work best. Many don’t have the inclination or indeed the time to learn what effect each setting has upon the games visual elements.It soon becomes a bind when upon a new driver release those settings which were great before can be bettered (or lowered) for a better gaming experience. FXAA, TSAA, MMAA or simply AA. Low, medium, high, ultra texture settings along with overall visual settings can be all too confusing when configuring a games visuals and while it would be easy to just crank everything up to maximum and hoping for the best, this all too often results in choppy, unplayable gaming ruining the gaming experience.

NVIDIA’s GeForce Experience removes much of the guess work when configuring a game thereby giving users the means to jump straight into the game with the best optimised settings for there hardware.

This handy utility will also keep your GPU updated with the latest driver that can be set to automatically download, ready to be installed as soon as they are available and also it gives you access to the beta drivers.

How does it work?
GeForce experience requires an internet connection which, once connected to the NVIDIA Cloud datacenter will download optimal game settings that are tailored to your PC based upon CPU, GPU and monitor resolution.

Low end hardware will have the visuals reduced to improve framerates while high end hardware will inevitably have higher settings assigned. This is all done with a click of a button which ensures that the game not only looks great but perhaps more importantly, plays smoothly.

Testing of the settings are done by NVIDIA themselves who benchmark the game, set a framerate target to ensure smooth gameplay then adjust the settings to ensure both framerate and smooth gameplay are optimised.

In our own tests, the utility worked well and is great for those who want to jump straight into the game with the minimal of fuss and furore of adjusting GPU settings.

Our one complaint is that the supported games list requires expanding. While most of the popular games are supported, many, especially some older but very popular titles are not.

First Look

The GTX760 features are fairly anonymous looking cooler atop of a black PCB. No doubt board partners will sprinkle their own magic on the reference design in the way of stickers to improve the fairly nondescript appearance.

We are happy to see NVIDIA have once again utilised a black power circuit board. As you can see from the shot above, the PCB is actually much shorter than the cooler. The whole card measures just 24cm however this could quite easily be shortened with a redesigned heatsink assembly bringing the length of the card to the PCB’s 17cm.

Like it’s brethren, the GTX760 features the GEFORCE GTX moniker emblazoned across the leading edge of the card. This time however it is simply screen printed and does not illuminate as with the GTX TITAN cooling solution.

The card requires two 6-pin PCIe power cables with a PSU capable of delivering 500W (recommended by NVIDIA).

Thankfully, the GTX760 is multi SLI capable meaning this card can be matched up to more than one other GTX760.

The I/O area is a standard affair featuring DVI-D, DVI-I, HDMI 1.4a and a full size DisplayPort. The remaining area is taken up with vents to allow hot air to be exhausted out of the PC enclosure.

Closer Look

Removing the cooler was a cinch with just 10 Phillips style screws to remove. The heatsink is actually quite small compared to the real estate afforded by the cooling shroud and as you can see, the main heatsink only covers the core, not the memory or VRM areas.

Surprisingly, the core is not fixed in anyway to the shroud, rather it is simply sandwiched between the card and shroud and only becomes fixed when all the components are screed together. With GPU boost dependent on temperature, it was surprising to see such a basic design. Hopefully this will not have an affect on our overclocking results!

The card has the VRM at opposite ends to the power ports which is not the most efficient design. As you can see, the VRM area has it’s own passive heatsink attached via push-pins. There are a total of 8 empty GDDR5 ram slots (4 on each side of the PCB) which hints at the potential of 4GB models becoming available in the future.

The VRM is made up of a 4+2 phase design with each GPU YAGEO R22 choke coupled to 3 MOSFETs. The remaining two R33 chokes have a single MOSFET each and are reserved for memory and PLL operations. The whole PCB carries solid state capacitors which is a familiar sight now with all modern graphics cards.

The memory is high quality HYNIX carrying the product code of H5GQ2H24AFR R0C so should be good for some overclocking fun.

The GK104 core is an excellent choice as it is both power efficient and (relatively) cool running. With 6 SMX units it will be interesting to see how it compares to the HD7950 which on paper appears to be the more powerful option.

Test Setup

How we test graphics cards

Testing a graphics card is an extremely lengthy process. We sometimes see reviews where a graphic card has been tested using out of date equipment, out of date games and out of date drivers. While the latter can be excused due to the sheer amount of driver updates making a thorough review impossible we will however notify the reader when we have used a different driver to the normal for example a pre-release driver. For the most part however we will always use the same hardware and where possible the same drivers unless a newer driver rectifies an issue we have encountered or significantly improves performance to such an extent that continuing with an outdated driver is not feasible.

Frame Time Analysis
We have also seen the emergence of frame time analysis. We covered this in a separate article here: GPU Reviews: Why Frame Time Analysis is important so we won’t go into the reasoning again but to summarise, our testing comprises both FPS testing and perhaps arguably more importantly, frame time analysis. We also look at overall performance by equating averages across the games and finally before concluding our review we take a look at value for money.

HardwareWith the latest Intel chipset release we decided to update our ageing X58 setup to the new X79 chipset. Not only that but we will be using the flagship Core i7-3960X Extreme Edition processor clocked to 4.5GHz to ensure we have no CPU bottlenecks. With 16GB of ram on board there should be minimal paging so the benchmarks we run today will hopefully give a true reflection of the graphics card’s performance.

High End Graphics cards on test:

NVIDIA GTX760 2GB GDDR5
(Forceware 320.39)
ZOTAC GTX760 2GB GDDR5
(Forceware 320.39)
NVIDIA GTX760 2GB GDDR5 SLI
(Forceware 320.39)
Sapphire Dual-X HD7950 3GB GDDR5 (Catalyst 13.6 beta)
Club 3D HD7970 royalAce 3GB GDDR5 (Catalyst 13.6 beta)
Gainward Phantom GTX 770 2GB GDDR5 (Forceware 320.18)
NVIDIA GeForce GTX 770 2GB GDDR5 (Forceware 320.18)
NVIDIA GeForce GTX 780 3GB GDDR5 (Forceware 320.18)
NVIDIA GeForce GTX 690 4GB GDDR5 (Forceware 314.22)
NVIDIA GeForce GTX TITAN 6GB GDDR5 (Forceware 314.22)
NVIDIA GeForce GTX TITAN SLI 12GB(2x6GB) GDDR5 (Forceware 314.22)
NVIDIA GTX680 2GB GDDR5 (Forceware 314.22)
AMD HD7970 3GB GDDR5 (AMD CAT 13.3 beta)

Motherboard: ASUS P9X79 Pro
CPU: Intel Core i7-3960X Extreme Edition @ 4.5GHz
RAM – 16GB (4x4GB) Corsair Vengeance (Red) DDR3 @ 1866MHz 9-10-9-27
Power Supply – Thortech Thunderbolt Plus 1200W
Hard Drive – Kingston HyperX 240GB
Cooler – Corsair H100

Benchmarks
If you are a regular reader of our reviews you will know we like to test the latest hardware with the latest games and benchmarks on the market. Here are the games we have chosen:

Battlefield 3
Crysis 3
Far Cry 3
BioShock Infinite
DiRT: Showdown
The Elder Scrolls: Skyrim

We also take into consideration the benchmarkers out there so have included 3 of the more popular synthetic benchmarks available:

Unigine: Valley
Unigine: Heaven 4
Futuremark 3DMark Firestrike

Let’s see how today’s graphics card performed…

Temperature, Acoustics and Power Consumption

Power Consumption

To test for power consumption we use the included iPower Meter found with our Thortech PSU. As system load figures obviously differ from idle it is extremely difficult to get an accurate figure of system draw and then simply take this figure away from the overall amount because each game will generate varying levels of CPU and hard drive activity. Dynamic power adjustments which fluctuate in game will also affect accurate power figures thus rendering any such power consumption calculations redundant at worst and approximate at best.

We will therefore take our system readings averaged over a benchmark run of Tessmark (100% load) and after 20 minutes of being idle in Windows 7. These can then be used as a comparison to other graphics cards to determine how much more or less power you can expect a GPU to consume.

NVIDIA recommend a PSU capable of 500W. We however believe you could possibly get by with a smaller PSU given that both our GTX760 samples pulled much less than this. It is however worth bearing in mind that you will need twin 6-pin PCIe power cables which are unlikely to come with a low specification unit.

Temperature

To test temperatures we measured idle temperatures after booting windows, letting all applications finish loading and ran a few benchmarks. Once the benchmarks were complete we left the card to reach a cooling plateau where we then took the idle temperatures. For the load tests, we would normally run Furmark for 20 minutes, taking the absolute maximum temperature attained however we found that this throttled the card and resulted in spurious results. So we set Heaven running continuously for 20 minutes and used this as a temperature result as the card did not throttle with this application.

The new range of GPUs from NVIDIA will keeping upping the clockspeed until either a power or temperature target is reached – usually the latter so with a threshold of 80c set on the reference card this is where the temperature maxxed out.

Acoustics

The card is neither silent nor extremely noisy. With such a hollow shroud, the noise is heightened somewhat when the card is put under load. It certainly is not screeching loud, far from it but it is noticeable above our case fans when running under load conditions for a while. Thankfully there was no capacitor squeal.

Let’s see how the card performs in our suite of benchmarks…

Futuremark 3DMark Fire Strike (Synthetic)

Fire Strike is the new showcase DirectX 11 benchmark designed for high-performance gaming PCs. It is Futuremark’s most ambitious and technical benchmark ever, featuring real-time graphics rendered with detail and complexity far beyond what is found in other benchmarks and games today.

*Note* SLI results omitted due to inconsistent results with driver 320.39.

Unigine: Heaven 4 (Synthetic)

Heaven Benchmark with its current version 4.0 is a GPU-intensive benchmark that hammers graphics cards to the limits. This powerful tool can be effectively used to determine the stability of a GPU under extremely stressful conditions, as well as check the cooling system’s potential under maximum heat output. It provides completely unbiased results and generates true in-game rendering workloads across all platforms.

Unigine: Valley (Synthetic)

Valley Benchmark is a new GPU stress-testing tool from the developers of the very popular and highly acclaimed Heaven Benchmark. The forest-covered valley surrounded by vast mountains amazes with its scale from a bird’s-eye view and is extremely detailed down to every leaf and flower petal. This non-synthetic benchmark powered by the state-of-the art UNIGINE Engine showcases a comprehensive set of cutting-edge graphics technologies with a dynamic environment and fully interactive modes available to the end user.

Battlefield 3 (FPS)

Battlefield 3 leaps ahead of its time with the power of Frostbite™ 2, the next installment of DICE’s cutting-edge game engine. This state-of-the-art technology is the foundation on which Battlefield 3 is built, delivering enhanced visual quality, a grand sense of scale, massive destruction, dynamic audio and incredibly lifelike character animations. As bullets whiz by, walls crumble, and explosions throw you to the ground, the battlefield feels more alive and interactive than ever before. In Battlefield 3, players step into the role of the elite U.S. Marines where they will experience heart-pounding missions across diverse locations including Paris, Tehran and New York.

Average and minimum frames per second were measured using Fraps during the parking lot scene on Operation Sword Breaker. High and Ultra presets were loaded at both resolutions.

Crysis 3 (FPS)

Crysis 3 from Electronic Arts is the fourth instalment of the Crysis franchise and sequel to Crysis 2. It uses the CryEngine 3 gaming engine which is perhaps the most advanced gaming engine to date and therefore the most demanding. The meme ‘Can it run Crysis’ has never been so apparent…

Far Cry 3 (FPS)

Far Cry 3 is a tropical survival shooter video game from Ubisoft Montreal and Massive studios. It’s open world design has similar traits to it’s forbear Far Cry 2 but it does it all so much better. With stunning visuals and settings to bring a PC to it’s knee’s, this game is so much more than the ‘just another console port’ it was reported to be.

BioShock Infinite (FPS)

Set in 1912 during the growth of American exceptionalism, BioShock Infinite is a thrid person shooter developed by Irrational games and published by 2K Games. It is the third instalment of the BioShock franchise. the game utilises the Unreal Engine 3, modifying it with their own lighting engine to present a visually stunning experience like no other.

DiRT: Showdown (FPS)

DiRT Showdown is the new arcade racing game from the team that brought you the award-winning DiRT series, uncaged in 2012. Pick up and play controls combine with electrifying events, frenzied crowds and stunning graphics to deliver high octane, dive in and drive thrills from event one.

The Elder Scrolls: Skyrim (FPS)

Skyrim is not a direct successor to the very popular Oblivion. Instead it takes place 200 years after the events of Oblivion. The non-linear playing that made the Elder Scrolls such a popular series does however make a return along with the beautiful scenery the encompasses the world of Skyrim.

Average and minimum frames per second were measured using Fraps during the drop down passage to Riverwood. V-sync was disabled to allow fluctuating (higher the 60) FPS for these runs. It is however recommended that v-sync is left enabled during normal gameplay due to the games dynamics.

Value For Money

We make a simple calculation for this set of results. For each game we add all of the resulting FPS totals together and then divide by the four sets to give us an overall performance figure. While the figures alone do not give you an accurate picture of how the card will perform in any given scenario, they do tell you, when averaged, which GPU is the more capable across all of the benchmarks and settings tested.

Gaming Experience (Frame Time Analysis)

Here we take a look at both average (black) and 99th Percentile (coloured) response times of the graphics cards. It is highly unlikely a user will notice any micro stutter (also incorrectly referred to as lag but the effect can be interpreted to be similar) at anything below 30ms but every gamers perception differs.

Comment

The graphs above may appear very daunting however, what you want to see is a smooth line (not fuzzy) with very few peaks and troughs. While you are unlikely to notice peaks and troughs below 30ms, any peak above 30ms may be seen as a stutter during gameplay. This will however depend greatly on your own perceptions. Multi card setups are more likely to show frame lag due to the nature of cross communicating to render frames.

Average and 99th Percentile response times (Frame Time Analysis)

Overclocking (GPU BOOST 2.0)

For overclocking our NVIDIA cards we will be using EVGA’s Precision too (ver 4.0) as this gives us all of the options we require.

Above is a screenshot of GPU-Z showing the reference speeds and reference boost clocks

Above is a screenshot of the actual overclock we achieved via Precision. With both power and temperature target sliders maxxed we hit a wall at around +140MHz on the core so we reduced this by 5MHz to ensure rock solid stability. We have more success with the memory speed, adding 549MHz which is a tremendous result.

It is possible to add a single increment of voltage (12mv) however this made no difference to the overclock of our sample.

After settling on our stable clockspeed which was verified by multiple runs of Unigine Heaven and 3DMark we benchmarked the overclock and compared it to stock results with Unigine Heaven 4.0 to evaluate the benefits of overclocking the GPU.

Conclusion

In fairness to the GTX760 we should have compared it against a ruck of mid-range examples however we wanted to show the benefits of SLI and for that reason we threw it in the mix with the best GPU solutions currently on the market so the FPS graphs should be viewed with this in mind.

NVIDIA told us that the GTX 760 would compare favourably to the AMD HD7950 which we thought was a very bold statement. This claim however is true as the GTX 760 is the better card under most scenarios. At 1920×1080, the GTX 760 and HD7950 exchanged victories however the overall performance at this resolution goes to the GTX760 by a split decision. This victory would be further exemplified were we to pitch the GTX 760 against a HD7950 without Boost technology and without a factory overclock.

When you add in the price factor, the GTX760 becomes even more desirable because on average it is some £30 cheaper than a HD7950, even one with Boost enabled. Tipping the scales at just over the £200 mark, the GTX 760 is priced fairly and it has to be because while the card may be slightly better at 1920×1080, the 1GB of extra memory available to the HD7950 makes it a better option at the highest resolutions with filtering applied. This is shown in our overall results where all resolutions and settings are bundled together to show the faster card overall. To this end, were the NVIDIA GTX 760 more expensive than its AMD counterpart and you wanted to game at 2560×1600 with additional filtering, we would have to recommend the HD7950.

Our frame time analysis showed that the NVIDIA cards offered smoother gameplay too with the AMD card having a few too many spikes which are likely to be interpreted as stuttering on screen.

What we haven’t mentioned yet is SLI performance and for good reason because here is where we feel the GTX 760 offers the best value against performance. It is the TITAN’s equal, betters a GTX 780 yet cost much less than both so if you only intend on gaming at 1920×1080, we so no reason why you would opt for single ultra high end card when the GTX 760 SLI setup is the better option and here is where it offers very good value for money.

We pondered on giving the card our Gold award however, because it lacks the memory to perform equally well at higher resolutions we thought silver was more fitting. Were it on SLI alone, it would get Gold however, we sacrificed this for our value award as the smart gamer who is looking for flagship performance at a reasonable price, a GTX 760 setup simply makes a hell of a lot of sense.

To summarise:
If you are looking for a mid-range card that offers good value and performance to boot then the GTX 760 is worth considering. If however you are yearning for a cheap SLI setup then the GTX 760 is a flagship killer that can hold its own with the very best GPUs out there….and then some!

Pros
+ Good Value
+ Good FPS performance
+ Excellent SLI capability

Cons
– Struggles at the highest resolutions
– Basic cooling