NVIDIA’s Maxwell has already made waves in the mobile market with the GM107 core which graced the second generation GTX 860M. In many ways that unassuming core heralded subsequent iterations of the architecture including the desktop-centric GTX 980 and GTX 970. Now this core design is getting a reintroduction to the notebook space with the GTX 980M and GTX 970M.
While the GM107 represented a so-called “first generation” Maxwell design, that initial version of the architecture has since been refined. Now there is even more emphasis upon performance per watt and some additional latent features have been added as well. As a result, the GTX 980M and GTX 970M may be in the same lineup as the GTX 860M and use the same generalized core design but their actual capabilities have been drastically augmented.
For those wondering, the quick movement away from the GTX 880M series (they were initially launched in March) is due to a number of different reasons. First and foremost, the notebook gaming market is still expanding at a rapid pace and gamers are constantly searching for the best possible solution that combines performance and the capability to actually push playable framerates when they’re using battery power. While the Kepler architecture was extremely efficient for its time, Maxwell’s primary focus is to enhance efficiency, making it a perfect fit for this demanding segment. In many ways the rebranded high end 800M-series cards were a stop-gap solution put in place until NVIDIA’s notebook product stack could go exclusively with the newer architecture.
Due to the serious focus on overall power consumption relative to performance, the GTX 980M and GTX 970M can do significantly more within a given TDP envelope. While this metric is important in the desktop space, notebook GPUs live and die by their compatibility with the slim and light chassis that personify today’s laptops. If a given architecture runs too hot and requires too much power, it will need to be throttled back for mobile usage.
Maxwell changes this equation by offering lower wattage parts without sacrificing much in the way of performance. Its internal communications have been given a thorough facelift, the die’s layout has been changed and a new caching hierarchy has been implemented. As a result, a core with a similar amount of die space and CUDA cores as the previous generation can offer significantly more performance while also operating at notebook-accepted TDP levels. In many ways this achievement is the Holy Grail for mobile-bound GPUs.
With all of this in mind, NVIDIA has also been gradually optimizing their core designs so the desktop and notebook parts can provide reasonably similar performance metrics. For example, while the Fermi-based GTX 480M only offered 40% of the desktop SKU’s throughput (and we all know how hot-running that particular architecture was), the GTX 980M should be able to offer up to 75% of its bigger brother’s capability. Unfortunately the limitations found in mobile processors will lead to a much larger gap but on paper at least, the GTX 900-series has some serious gaming chops.
Other than the aforementioned second generation GTX 860M, Maxwell’s presence in the notebook market will initially take to form of two different cores: the GTX 980M and GTX 970M. Both of these are based upon a heavily modified GM204 core with a number of its Streaming Multiprocessors disabled in an effort to reduce power consumption and heat production. If you want to know more about the GM204 and what makes this architecture tick, look no further than our in-depth analysis.
The GTX 980M sits at the product stack’s very top twelve enabled SMs which nets it 1536 cores and 96 texture units. These are paired up with a full allotment of ROPs, a newly optimized 256-bit GDDR5 memory bus and up to 4GB of memory capacity. For those keeping track at home, these specifications nearly equal those offered by the desktop GTX 970, though clock speeds will be lower to keep it within a given set of TDP boundaries.
The GTX 970M has two more of its SMs disabled, resulting in a product that features 1280 cores and 80 TMUs but there are some other differences here as well. To further reduce the system requirements NVIDIA has also cut off a single ROP partition which has a trickle-down effect upon the memory controller bandwidth (which is down to a 192-bit interface) and has the capability to house up to 3GB of memory.
Like with all other notebook GPU’s, the listed frequencies for Maxwell cores can be considered a variable metric which is completely based upon NVIDIA’s Boost algorithms. Should a notebook design provide sufficient cooling and an adequate power supply, these graphics processors won’t have any issue hitting their stated specifications. On the other hand, certain systems may be constrained in their cooling capabilities which will necessitate lower clock speeds.
Needless to say, Maxwell’s improvement over the Kepler-based GTX 680M is impressive according to NVIDIA’s internal numbers. The possibility of up to double the performance in optimal conditions will likely have users of previous-generation notebooks looking to upgrade, especially with the large number of games being realeased this holiday season.
In realistic terms, neither the GTX 980M nor GTX 970M are supposed to offer a massive performance uplift over their 800M-series predecessors. We expect there to be an approximate 20-25% framerate increase but most of the benefits of moving towards the newer architecture will be battery life-focused. In comparison to AMD’s current generation of mobile graphics cards, it won’t even be close regardless of your perspective and that could pose a real problem for the long-term viability of the notebook-based Radeon lineup.
Past the obvious performance benefits, Maxwell has several noteworthy features as well, many of which we have already tested. For example, Battery Boost is a way for NVIDIA’s mobile architecture to achieve optimal playable framerates while reducing power consumption. Through the use of algorithms integrated within GeForce Experience, frame rates are capped at a user-determined level so the GPU doesn’t have to work overly hard when massive performance isn’t necessary. According to NVIDIA's own internal numbers, the GTX 980M is able to achieve completely playable framerates while still offering over 90 minutes of battery life. If those figures hold up, this could be a game changer.
Two other recent additions to NVIDIA’s long list of add-ons are GameStream and ShadowPlay. The ability to use GameStream abilities between the Maxwell-equipped notebook and a compatible SHIELD device may not be a major selling factor but it can drastically expand the reach of a portable gaming station. ShadowPlay meanwhile allows a gamer to record or stream their every move while utilizing a minimum of system resources since the Maxwell GPU has a dedicated hardware encode / decode module.
So where does all of this leave the competition within the notebook space? Nowhere near as close as we’ve come to expect from the desktop market. AMD has been busy spinning their wheels with a continued focus on APUs and their high end mobile lineup still based on the nearly three year old Pitcairn cores.
This unfortunate situation has presented an opportunity for NVIDIA and with Maxwell they’re about to move the goalposts even further afield by giving laptop manufacturers what they want. Naturally from day one they’ve racked up well over a dozen design wins on gaming notebooks from the likes of Alienware, ASUS, Acer, MSI, Origin and plenty of others. Will this be an insurmountable hurdle for AMD’s own lineup? Likely not but NVIDIA still has a massive head start.
While the GM107 represented a so-called “first generation” Maxwell design, that initial version of the architecture has since been refined. Now there is even more emphasis upon performance per watt and some additional latent features have been added as well. As a result, the GTX 980M and GTX 970M may be in the same lineup as the GTX 860M and use the same generalized core design but their actual capabilities have been drastically augmented.
For those wondering, the quick movement away from the GTX 880M series (they were initially launched in March) is due to a number of different reasons. First and foremost, the notebook gaming market is still expanding at a rapid pace and gamers are constantly searching for the best possible solution that combines performance and the capability to actually push playable framerates when they’re using battery power. While the Kepler architecture was extremely efficient for its time, Maxwell’s primary focus is to enhance efficiency, making it a perfect fit for this demanding segment. In many ways the rebranded high end 800M-series cards were a stop-gap solution put in place until NVIDIA’s notebook product stack could go exclusively with the newer architecture.
Due to the serious focus on overall power consumption relative to performance, the GTX 980M and GTX 970M can do significantly more within a given TDP envelope. While this metric is important in the desktop space, notebook GPUs live and die by their compatibility with the slim and light chassis that personify today’s laptops. If a given architecture runs too hot and requires too much power, it will need to be throttled back for mobile usage.
Maxwell changes this equation by offering lower wattage parts without sacrificing much in the way of performance. Its internal communications have been given a thorough facelift, the die’s layout has been changed and a new caching hierarchy has been implemented. As a result, a core with a similar amount of die space and CUDA cores as the previous generation can offer significantly more performance while also operating at notebook-accepted TDP levels. In many ways this achievement is the Holy Grail for mobile-bound GPUs.
With all of this in mind, NVIDIA has also been gradually optimizing their core designs so the desktop and notebook parts can provide reasonably similar performance metrics. For example, while the Fermi-based GTX 480M only offered 40% of the desktop SKU’s throughput (and we all know how hot-running that particular architecture was), the GTX 980M should be able to offer up to 75% of its bigger brother’s capability. Unfortunately the limitations found in mobile processors will lead to a much larger gap but on paper at least, the GTX 900-series has some serious gaming chops.
Other than the aforementioned second generation GTX 860M, Maxwell’s presence in the notebook market will initially take to form of two different cores: the GTX 980M and GTX 970M. Both of these are based upon a heavily modified GM204 core with a number of its Streaming Multiprocessors disabled in an effort to reduce power consumption and heat production. If you want to know more about the GM204 and what makes this architecture tick, look no further than our in-depth analysis.
The GTX 980M sits at the product stack’s very top twelve enabled SMs which nets it 1536 cores and 96 texture units. These are paired up with a full allotment of ROPs, a newly optimized 256-bit GDDR5 memory bus and up to 4GB of memory capacity. For those keeping track at home, these specifications nearly equal those offered by the desktop GTX 970, though clock speeds will be lower to keep it within a given set of TDP boundaries.
The GTX 970M has two more of its SMs disabled, resulting in a product that features 1280 cores and 80 TMUs but there are some other differences here as well. To further reduce the system requirements NVIDIA has also cut off a single ROP partition which has a trickle-down effect upon the memory controller bandwidth (which is down to a 192-bit interface) and has the capability to house up to 3GB of memory.
Like with all other notebook GPU’s, the listed frequencies for Maxwell cores can be considered a variable metric which is completely based upon NVIDIA’s Boost algorithms. Should a notebook design provide sufficient cooling and an adequate power supply, these graphics processors won’t have any issue hitting their stated specifications. On the other hand, certain systems may be constrained in their cooling capabilities which will necessitate lower clock speeds.
Needless to say, Maxwell’s improvement over the Kepler-based GTX 680M is impressive according to NVIDIA’s internal numbers. The possibility of up to double the performance in optimal conditions will likely have users of previous-generation notebooks looking to upgrade, especially with the large number of games being realeased this holiday season.
In realistic terms, neither the GTX 980M nor GTX 970M are supposed to offer a massive performance uplift over their 800M-series predecessors. We expect there to be an approximate 20-25% framerate increase but most of the benefits of moving towards the newer architecture will be battery life-focused. In comparison to AMD’s current generation of mobile graphics cards, it won’t even be close regardless of your perspective and that could pose a real problem for the long-term viability of the notebook-based Radeon lineup.
Past the obvious performance benefits, Maxwell has several noteworthy features as well, many of which we have already tested. For example, Battery Boost is a way for NVIDIA’s mobile architecture to achieve optimal playable framerates while reducing power consumption. Through the use of algorithms integrated within GeForce Experience, frame rates are capped at a user-determined level so the GPU doesn’t have to work overly hard when massive performance isn’t necessary. According to NVIDIA's own internal numbers, the GTX 980M is able to achieve completely playable framerates while still offering over 90 minutes of battery life. If those figures hold up, this could be a game changer.
Two other recent additions to NVIDIA’s long list of add-ons are GameStream and ShadowPlay. The ability to use GameStream abilities between the Maxwell-equipped notebook and a compatible SHIELD device may not be a major selling factor but it can drastically expand the reach of a portable gaming station. ShadowPlay meanwhile allows a gamer to record or stream their every move while utilizing a minimum of system resources since the Maxwell GPU has a dedicated hardware encode / decode module.
So where does all of this leave the competition within the notebook space? Nowhere near as close as we’ve come to expect from the desktop market. AMD has been busy spinning their wheels with a continued focus on APUs and their high end mobile lineup still based on the nearly three year old Pitcairn cores.
This unfortunate situation has presented an opportunity for NVIDIA and with Maxwell they’re about to move the goalposts even further afield by giving laptop manufacturers what they want. Naturally from day one they’ve racked up well over a dozen design wins on gaming notebooks from the likes of Alienware, ASUS, Acer, MSI, Origin and plenty of others. Will this be an insurmountable hurdle for AMD’s own lineup? Likely not but NVIDIA still has a massive head start.
Last edited:
