discussion NVIDIA GeForce GTX 980 Performance Review

[SKYMTL]

Clock Speed Stability on Maxwell

Clock Speed Stability on Maxwell

Both NVIDIA and AMD utilize an intricate web of hardware and software to manage TDP on their cards. In NVIDIA’s case their GPU Boost technology allows for the highest possible frequencies provided heat and power consumption remain below predefined thresholds.

As you might expect, this approach can also lead to lower than expected performance if a card either runs too hot or the built-in algorithms detect higher than expected power consumption and core load. AMD ran face first into the throttling issue with their R9 290X and it’s a situation NVIDIA wants to avoid with Maxwell.

The way GPU Boost handles its balancing act hasn’t changed but given GM204’s inherent efficiency, it has been given a bit more leeway. This has resulted in higher frequency potential than previous generations but once again those capabilities are tied directly to power draw and, more importantly, temperatures.

In this instance we used a continual in-game load to illustrate how the GTX 980 handles itself during a typical usage scenario and the results are quite interesting. While the core hits 76°C and stays there for a few minutes, it seems like Maxwell’s Power Limit was reached about midway through the test. Once that happened, temperatures were allowed to level off at 75°C.

Part of that temperature management obviously involved scaling back clock speeds a bit. The core went from operating at 1244MHz (which is actually 24MHz above NVIDIA’s stated Boost Clock) to 1227MHz and remained there even after an hour of constant load. We never experienced a situation where the Boost frequency would scale back to a point lower than 1223MHz but once again, the possibility of speeds fluctuation between Base and Boost is there provided a situation puts enough stress on the core.

On paper at least it is relatively easy to see where the clock speed scaling kicks in but under no circumstance does it affect onscreen performance. At most we are looking at a 1FPS reduction but even then, the Boost algorithms still strive to retain every shred of potential.

All in all the GTX 980 was an extremely stable gaming companion throughout testing with only a few instances where it didn’t perform up to expectations. Naturally, custom cooling or slightly higher fan speed RPMs would completely eliminate every bit of Boost speed reduction and perhaps even contribute to enhanced performance.

 

[SKYMTL]

Thermal Imaging / Acoustics / Power Consumption

Thermal Imaging

ITX users are probably looking that these pictures and drooling. Custom cooled card with massive 80mm fans might run extremely cool and quiet but for ambient temperatures, nothing beats a blower-style setup. The GTX 980 exhausts nearly all of its heat outside the case which results in lower ambient temperatures, though at the sacrifice of some cooling potential. This alongside its low power consumption numbers may make it a perfect choice for anyone who is looking for a high powered SFF system.

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 GTX 980 is one of the quietest running reference cards around which should be expected considering the minimal amount of heat its core puts out. With that being said, after several minutes of gaming temperatures do get up to around the 75°C mark and are maintained through a slight upshift in the fans’ rotational speeds.

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.

Please note that after extensive testing, we have found that simply plugging in a power meter to a wall outlet or UPS will NOT give you accurate power consumption numbers due to slight changes in the input voltage. Thus we use a Tripp-Lite 1800W line conditioner between the 120V outlet and the power meter.

This was pretty much a foregone conclusion; NVIDIA has massaged the 28nm manufacturing process to a point where these new chips to the point where they can accomplish some incredible things. Not only does the GTX 980 consume substantially less power than the GTX 780 Ti (let’s not even get into where AMD’s R9 290X sits here…) but also beats the GTX 680.

 

[SKYMTL]

Overclocking Results

Overclocking Results

After the experiences everyone had with overclocking on Kepler and constantly having to contend with artificially constrained limits, many were hoping for a change with Maxwell. While the dreaded Voltage and Power Limits are still there and will work in tandem to drag down an overclock, NVIDIA has allowed for a tiny bit more leeway this time around.

While still in its early beta phases, EVGA’s GTX 980 supporting Precision allows for 86mV of additional voltage while the Power Limit can be boosted to 125%. Both of those represent ever so slight improvements over what was offered with NVIDIA’s GTX 7xx series, some of which offered just 20% of additional power and less voltage adjustments.

Unfortunately due to a late arrival of the card we weren’t able to fine tune the overclock but the results were still promising. By maxing out power the Power and Voltage limits and increasing fan speed to a constant 55% the GTX 980 had no problem running at 1468MHz / 7916MHz with complete stability.

 

[SKYMTL]

Conclusion; The Maxwell Game Changer

Conclusion; The Maxwell Game Changer

So here we are at the end of another NVIDIA launch review and one thing is quite evident: the GTX 980 has done a nearly perfect job in emulating the GTX 680’s high points and adds more than a few highlights of its own. Like its spiritual predecessor, the GTX 980 is a highly disruptive product that will cause repercussions throughout the current GPU market and rewrites the book when it comes to performance per watt. Considering the technology involved, with the Maxwell architecture NVIDIA has created one of the most inspiring PC components to be released in the last two years.

Inspiring may sound like a heavy-handed word to use for a graphics card but there’s a lot to admire here. In order to achieve perf per watt improvements over Fermi, Kepler partially relied upon its 28nm manufacturing process but this time around NVIDIA didn’t have the luxury of a more efficient process node. Instead the Maxwell architecture has been designed from the ground up to maximize on-die communication efficiency and that focus has worked wonders. Despite using a 5.2 billion transistor chip rated at just 165W, the GTX 980 is able to remain ahead of NVIDIA’s own 250W TDP, 7.1 billion transistor GK110-based GTX 780 Ti in most situations.

By sticking with a tried and true 28nm manufacturing infra structure and allowing for a good amount of component commonality between Maxwell and Kepler cards, NVIDIA has insured costs remain low while facilitating production of custom AIB products. These savings have a trickle-down effect upon the GTX 980’s price which, at $550, is a screaming bargain given its performance and efficiency levels relative to the outgoing GTX 780 Ti and AMD’s R9 290X. Granted, $550 is a whole lot of money but it’s good to see some new value-minded blood injected into a segment that was recently introduced to the $1000 TITAN Black and $3000 TITAN Z. The GTX 970 will likely one-up this metric given its even lower starting cost but thanks to Canadian Customs you'll have to wait until next week to read our review of that one.

Much of the GTX 980’s appeal versus the GTX 780 Ti will come from its lower price point and vastly improved power consumption numbers since performance between the two cards isn’t noticeably different. On paper the GTX 980 was able to edge out a slim win and would likely go on to lead through most games. However, in texture-heavy scenarios even the GM204 core’s higher frequencies, more efficient caching and enhanced memory handling can’t make up for its lower Texture Unit count. Both Hitman Absolution and Call of Duty: Ghosts showed this to great effect as NVIDIA’s newest addition either lost or ended up in a literal tie against its predecessor.

While GTX 780 Ti users won’t see much reason to upgrade unless they want to pack the GTX 980’s lower TDP into a compact ITX setup, the same can’t be said for anyone sporting a GTX 680. By skipping Kepler’s second generation products, they can now buy a card that boasts substantially better across-the-board performance metrics. If you spent $500 for the GTX 680 when it first came out throwing down $550 for a GTX 980 seems like a no-brainer after more than two years of solid gaming framerates.

The GTX 980 must be causing a serious case of deja-vu for AMD. Back when the GTX 680 was launched, their Tahiti cards looked too expensive, too slow, too loud and too hot….and the situation hasn’t really changed this time around. In every almost every game the GTX 980 easily outmuscles their flagship R9 290X, consumes significantly less power and costs just $50 more. So what does this mean for AMD? They’re currently stuck with a power hungry, hot running architecture that is still quite competitive from a performance standpoint but some significant price cuts are desperately needed.

NVIDIA has taken their fair share of criticism for the strictly restrained overclocking limits on Kepler and we were hoping for more from the GTX 980. Both the Power and Voltage Limit maximums haven’t really been changed by all that much but they should still allow for more overhead provided you have better cooling or are willing to boost reference fan speeds. With that being said, overclocking was decent, though nothing spectacular. Expect board partners’ designs to help with this but the meager 87mV of voltage headroom likely won’t budge until the Classified, Matrix, Lightning and Super Overclock editions are launched sometime in the coming months.

It is extremely hard –if not impossible- to find any faults with the GTX 980. This card is able to deliver class-leading framerates at the very highest detail settings, it runs quietly and cool, power consumption is spectacular relative to performance and, when compared against yesterday’s high-end graphics cards, it is much more affordable. There are some lingering questions about how the GM204 core’s lower TMU count will affect performance in texture-heavy environments but if current games are any indication, the impact will be largely minimal. Regardless of that minor concern, like the GTX 680, this is a market-changing graphics card that’s backstopped by an impressive GPU architecture.

The only potential issue may boil down to availability. Even though we’ve heard reports of widespread shipments to channel partners, with all of these benefits and a price of $550, retailers may still have a tough time keeping the GTX 980 in stock.