What's new
  • Please do not post any links until you have 3 posts as they will automatically be rejected to prevent SPAM. Many words are also blocked due to being used in SPAM Messages. Thanks!

The NVIDIA GTX 960 Performance Review

SKYMTL

HardwareCanuck Review Editor
Staff member
Joined
Feb 26, 2007
Messages
12,857
Location
Montreal
Within the graphics card market the $199 to $250 price point has historically been a space where manufacturers have offered a near-perfect blend of price and performance for gamers who can’t justify spending $400 and more for an upgrade. This also happens to be where AMD and NVIDIA move a ton of volume so getting the formula right is a key component towards success or failure of a given generation. With the new $199 GTX 960, NVIDIA is hoping they have a card which appeals to the millions of buyers still sporting cards like the GTX 460, GTX 560 and GTX 660.

With the GM206-based GTX 960 intending to replace some of the most popular cards NVIDIA has ever produced, there are some huge expectations riding on it. However, it finds itself within a hotly contested segment that’s full of excellent alternatives. On AMD’s side there’s the newer $210 R9 285 which incorporates their Tonga architecture and a full stable of high end features like TrueAudio and Freesync from the higher end R9 290-series.

Meanwhile, we can’t discount the R9 280X from this equation either. It may be a rebranded HD 7970GHz Edition which means there’s a lack of TrueAudio and Freesync support but with a price that starts around the $220 mark and 3GB of memory, it could be a perfect solution for those looking for raw performance rather than a wide feature set and lower power consumption.


The GM206 core represents the Maxwell architecture in miniaturized form since it only has a pair of GPC modules compared to the quad GPC layout from the GM204 used on NVIDIA’s GTX 970 and GTX 980. However, the primary building blocks from the current flagship parts have been carried over en masse so within each GPC there is a quartet of Streaming Multiprocessors. Every one of these SMMs contains 128 CUDA cores alongside eight texture units.

Tertiary functions also receive a pretty drastic cut-down with only a pair of 64-bit memory controllers alongside two banks of 16 ROPs and 1024K of L2 cache. This, alongside the obvious transistor space needed for native DX12 support adds up to a die size of 227mm² and 2.94 billion transistors.


With 1024 CUDA cores, 32ROPS, 64TMUs and a 128-bit memory interface granting a narrow 112.16 GB/s of memory bandwidth, the GTX 960 certainly doesn’t look like an impressive card on paper. As a matter of fact, other than memory and core frequencies, its specifications are significantly eclipsed by its predecessor, the GTX 760.

Luckily there’s more to this card that what first meets the eye since it has been designed to do more with less. While it may be down in core count, Maxwell’s architecture has a number of efficiencies built into its processing pipelines which allows for much higher performance across the board. The memory interface has also received a significant upgrade in the form of enhanced compression algorithms which allows it to overcome some of the apparent bandwidth limitations. Accordingly to NVIDIA, this technology allows the 7Gbps, 128-bit interface to be effectively used at over 9Gbps granting an effective bandwidth of about 149 GB/s. This is still much less than what the GTX 760 offers but it should still allow the GTX 960 to overcome some of its memory handicap.

One of the real stars of this show is the GTX 960’s TDP which is pegged at just 120W, or 50W less than the GTX 760. This should make it very appealing for small form factor cases and could eventually lead to compact single slot cooling solutions as well. Core cutting only goes so far though since 25W is all that separates the GTX 960 from the GTX 970, a card that offers significantly more performance.

The ASUS STRIX OC card we will be looking at in this review is an interesting addition to the GTX 960 lineup since it offers much higher clock speeds for just $15 more than the reference design. In the standard Gaming Mode there’s a frequency uplift over 100MHz for the Base and Boost while OC Mode (accessed through ASUS’ GPU Tweak utility) pushes things a bit further. In this review, we’ll be using the OC Mode for comparison purposes.

One of the completely justifiable concerns here is that the GTX 960 may just not have the performance mustard to become a worthy replacement for NVIDIA’s GTX 760. This is likely one of the reasons NVIDIA is launch it into the $199 price point rather than at the $250 segment initially targeted by their 760. There’s also fairly yawning price and performance canyon between the GTX 970 and GTX 960, so there’s room for another SKU to bridge the gap if NVIDIA were inclined to go that route.

With NVIDIA firmly targeting the GTX 960 towards the 1080P crowd, its aspirations are fairly mundane but there are some extremely tempting options from AMD in this same bracket. The real question is whether or not this new card will really offer enough for gamers who waited out the GTX 760 in the hope Maxwell would offer them something more in the performance department.
 

SKYMTL

HardwareCanuck Review Editor
Staff member
Joined
Feb 26, 2007
Messages
12,857
Location
Montreal
A Closer Look at the ASUS GTX 960 STRIX OC

A Closer Look at the ASUS GTX 960 STRIX OC



ASUS’ GTX 960 STRIX OC is an amazingly compact card given the large designs of their GTX 970 and GTX 980 versions. It incorporates a diminutive version of the excellent DirectCU heatsink design with direct contact heatpipes alongside fans which are engineered to completely power off until the core hits roughly 65°C. In our testing this meant silent operation in most low-power applications like Starcraft and DOTA.

As you will see in an upcoming 7-way roundup, the STRIX also happens to be one of the shortest GTX 960 cards currently available with a length of just 8.5”.


The DirectCU II heatsink design is relatively straightforward with a quartet of heatpipes which make direct contact with the core. This facilitates heat transfer and enhances cooling performance, particularly when overclocking since the GM206 core doesn’t actually put out a lot of heat.


According to NVIDIA, a GTX 960’s maximum Power Limit will be largely determined by its power input connector layout. In ASUS’ case they have installed a single 6-pin which should grant up to 10% extra power headroom which should be more than enough to max out the minimal amount of voltage overhead NVIDIA allows on these cards.


Unlike the GTX 980 and GTX 970, the GTX 960 can only be used in SLI with one other card. No triple or quad SLI support has been included, hence it receives a single SLI connector.


Around back ASUS has installed a full length secondary heat-spreader which covers memory modules and a few tertiary components.


Underneath the heatsink ASUS has included a 5-phase PWM with ASUS’ Super Alloy Power components. This includes silent chokes that reduce coil whine, a dedicated SAP CAP to enhance overclocking headroom Super Alloy Capacitors for a longer lifespan and the Super Alloy MOS for an increased voltage threshold. We can also see just how small the PCB is, which should cascade down into version that target the small form factor market.


The rear-facing connectors are pretty basic with a trio of DisplayPort 1.2 ports, a single DVI output and one HDMI 2.0.
 

SKYMTL

HardwareCanuck Review Editor
Staff member
Joined
Feb 26, 2007
Messages
12,857
Location
Montreal
Test System & Setup

Main Test System

Processor: Intel i7 4930K @ 4.7GHz
Memory: G.Skill Trident 16GB @ 2133MHz 10-10-12-29-1T
Motherboard: ASUS P9X79-E WS
Cooling: NH-U14S
SSD: 2x Kingston HyperX 3K 480GB
Power Supply: Corsair AX1200
Monitor: Dell U2713HM (1440P) / ASUS PQ321Q (4K)
OS: Windows 8.1 Professional


Drivers:
AMD Catalyst Omega (public release)
NVIDIA 347.25 Beta


*Notes:

- All games tested have been patched to their latest version

- The OS has had all the latest hotfixes and updates installed

- All scores you see are the averages after 2 benchmark runs

All IQ settings were adjusted in-game and all GPU control panels were set to use application settings


The Methodology of Frame Testing, Distilled


How do you benchmark an onscreen experience? That question has plagued graphics card evaluations for years. While framerates give an accurate measurement of raw performance , there’s a lot more going on behind the scenes which a basic frames per second measurement by FRAPS or a similar application just can’t show. A good example of this is how “stuttering” can occur but may not be picked up by typical min/max/average benchmarking.

Before we go on, a basic explanation of FRAPS’ frames per second benchmarking method is important. FRAPS determines FPS rates by simply logging and averaging out how many frames are rendered within a single second. The average framerate measurement is taken by dividing the total number of rendered frames by the length of the benchmark being run. For example, if a 60 second sequence is used and the GPU renders 4,000 frames over the course of that time, the average result will be 66.67FPS. The minimum and maximum values meanwhile are simply two data points representing single second intervals which took the longest and shortest amount of time to render. Combining these values together gives an accurate, albeit very narrow snapshot of graphics subsystem performance and it isn’t quite representative of what you’ll actually see on the screen.

FCAT on the other hand has the capability to log onscreen average framerates for each second of a benchmark sequence, resulting in the “FPS over time” graphs. It does this by simply logging the reported framerate result once per second. However, in real world applications, a single second is actually a long period of time, meaning the human eye can pick up on onscreen deviations much quicker than this method can actually report them. So what can actually happens within each second of time? A whole lot since each second of gameplay time can consist of dozens or even hundreds (if your graphics card is fast enough) of frames. This brings us to frame time testing and where the Frame Time Analysis Tool gets factored into this equation.

Frame times simply represent the length of time (in milliseconds) it takes the graphics card to render and display each individual frame. Measuring the interval between frames allows for a detailed millisecond by millisecond evaluation of frame times rather than averaging things out over a full second. The larger the amount of time, the longer each frame takes to render. This detailed reporting just isn’t possible with standard benchmark methods.

We are now using FCAT for ALL benchmark results, other than 4K.
 

SKYMTL

HardwareCanuck Review Editor
Staff member
Joined
Feb 26, 2007
Messages
12,857
Location
Montreal
Alien: Isolation / Battlefield 4

Alien: Isolation





Battlefield 4



 

SKYMTL

HardwareCanuck Review Editor
Staff member
Joined
Feb 26, 2007
Messages
12,857
Location
Montreal
Dragon Age: Inquisition / Far Cry 4

Dragon Age: Inquisition





Far Cry 4



 

SKYMTL

HardwareCanuck Review Editor
Staff member
Joined
Feb 26, 2007
Messages
12,857
Location
Montreal
Metro: Last Light / Middle Earth: Shadow of Mordor

Metro: Last Light


* Please note that AMD's driver has issues in the level we are using as a benchmark. However, this problem does not seem to affect other areas of the game. As such, their results have been included for reference purposes only and ARE NOT factored into this review's conclusion.




Middle Earth: Shadow of Mordor



 

SKYMTL

HardwareCanuck Review Editor
Staff member
Joined
Feb 26, 2007
Messages
12,857
Location
Montreal
Clock Speed Stability Under the Microsope

Clock Speed Stability Under the Microsope


We have seen in the past that NVIDIA’s Maxwell architecture can utilize its GPU Boost feature to regulate clock speeds if temperatures reach a certain level. While ASUS’ card does have the ability to keep its fans off until the core hits 65°C, we have seen in the past that the STRIX series’ fans don’t kick themselves on until a bit late in the game. Supposedly the algorithm has been improved since we first tested it on ASUS’ GTX 970.


The ASUS STRIX remained at an incredibly low temperature throughout testing but we did notice its fan did kick itself on at 55°C rather than ASUS’ stated threshold of 65°C. This points to their algorithm now including overall ASIC TDP rather than just raw core temperature readings.


As with other cards in NVIDIA’s lineup the GTX 960 can boost itself above and beyond its stated maximum clock speeds and that’s exactly what ASUS has accomplished here. While the STRIX’s Boost frequency is officially rated at 1317MHz in its OC Mode, given the additional thermal headroom it actually hits 1416MHz on a continual basis.


While we don’t have any other GTX 960 to compare it to, ASUS’ version is able to deliver consistent framerates throughout testing.
 

SKYMTL

HardwareCanuck Review Editor
Staff member
Joined
Feb 26, 2007
Messages
12,857
Location
Montreal
Thermal Imaging / Acoustics / Power Consumption

Thermal Imaging



ASUS STRIX series cards are typically quite good in evenly spreading any built up heat over a large area and the GTX 960 version is no different. After more than an hour of running, all of the primary areas remain moderately warm and, other than small hot spots where exhaust air comes in contact with the display outputs, there are absolutely no points of excess heat.


Acoustical Testing


What you see below are the baseline idle dB(A) results attained for a relatively quiet open-case system (specs are in the Methodology section) sans GPU along with the attained results for each individual card in idle and load scenarios. The meter we use has been calibrated and is placed at seated ear-level exactly 12” away from the GPU’s fan. For the load scenarios, Hitman Absolution is used in order to generate a constant load on the GPU(s) over the course of 15 minutes.


The entire engineering process behind the STRIX’s DirectCU II cooler focused on delivering an optimal blend of cooling potential and a low noise profile. Hence, what we have is one of the quietest cards on the market which barely needs to spin its fans to an audible level and actually becomes completely silent in most cases. If you are building a PC for an area which requires a minimal acoustical profile, this is the one to buy.


System Power Consumption


For this test we hooked up our power supply to a UPM power meter that will log the power consumption of the whole system twice every second. In order to stress the GPU as much as possible we used 15 minutes of Unigine Valley running on a loop while letting the card sit at a stable Windows desktop for 15 minutes to determine the peak idle power consumption.


NVIDIA’s Maxwell architecture is known for its efficiency and the GTX 960 obviously carries on that tradition. Even though the ASUS card is overclocked to a significant amount, it still provides a substantial power savings over the GTX 760 and every other card in its price range.
 

SKYMTL

HardwareCanuck Review Editor
Staff member
Joined
Feb 26, 2007
Messages
12,857
Location
Montreal
Overclocking Results

Overclocking Results


NVIDIA claims there’s a good amount of overclocking headroom built into the GTX 960 and their default frequencies are quite conservative. ASUS has partially taken advantage of this my launching their card with an included “OC Mode” that automatically increases the STRIX OC’s Boost frequency to around 1317MHz. They also have a Silent button in their GPU Tweak utility that lowers clock speeds and keep the fans in a suspended mode in many applications. We’ve already seen that excess thermal headroom allows this card to go even further on its own (to about 1416MHz) so what does that mean for actual overclocking?

Before we discuss the actual frequencies we achieved, let’s talk about limitations. According to NVIDIA, their board partners have the ability to build whatever Power Limit they want into these cards provided the core is provided with enough input power and it detects sufficient PWM capacity. ASUS has taken a slightly conservative approach by limiting their card to 15% of additional Power Limit headroom which isn’t all that bad. Voltage meanwhile is capped at a maximum of +50mV, that that is what ended up providing a ceiling for our overclocking aspirations.


So, with the GPU voltage set to 1.213V and the Power Limit pushing 115% we set out to see what the GTX 960 STRIX OC had left in its tank. It turns out, there was quite a bit.

We were able to achieve another 92MHz of Boost which equated to the core running at a staggering constant 1510MHz while the memory plateaued at 7776MHz. The core frequency itself is extremely impressive when you consider a reference GTX 960 operates at a Boost speed of 1178Mhz but the memory was simply par for the course. With that being said, the amount of extra performance you can eke out of this card (or at least this sample) can be put to good use towards minimizing its limitations but in bandwidth-limited titles like Shadow of Mordor, expect limited real-world returns.


 
Top