Maxwell 1 Architecture: The Story So Far

Before we dive into the story and architecture of Maxwell 2, we’d like to spend a bit of time recapping what we’ve seen so far with Maxwell 1 and the GM107 GPU. While both GPUs are distinctly Maxwell, Maxwell 2 is essentially a second, more feature-packed version of Maxwell. Meanwhile it retains all of the base optimizations that went into Maxwell 1, implemented on a larger scale for a larger GPU.

Beginning with the Maxwell family of architectures, NVIDIA embarked on a “mobile first” design strategy for GPUs, marking a significant change in NVIDIA’s product design philosophy. As of Maxwell NVIDIA’s top-down philosophy that saw the launch of high-end desktop class GPUs come first has come to an end, and as NVIDIA has chosen to embrace power efficiency and mobile-friendly designs as the foundation of their GPU architectures, this has led to them going mobile first on Maxwell. With Maxwell NVIDIA has made the complete transition from top to bottom, and are now designing GPUs bottom-up instead of top-down.

By going mobile first NVIDIA is aiming to address several design considerations at all once. First and foremost is the fact that NVIDIA is heavily staking the future of their company in mobile, and that means they need GPU designs suitable for such a market. This mobile first view is primarily focused on SoC-class products – the Tegra family – but really it even extends to mobile PC form factors such as laptops, where discrete GPUs can play an important role but are going to have strict thermal requirements. By designing GPUs around mobile first, NVIDIA starts with a design that is already suitable for Tegra and then can scale it up as necessary for laptop and desktop GeForce products. Graphics is – as we like to say – embarrassingly parallel, so if you can build one small module then it’s relatively easy to scale up performance by building chips with more modules and tying them together. This is the mobile first philosophy.

What this means is that NVIDIA is focused on power efficiency more than ever before. The SoC market is brutal for both the demands placed on the hardware and for the competitive nature of that market, and given the fact that SoCs are so heavily constrained by thermal and power considerations, every bit of power saved can be reinvested in additional performance. This in turn calls for a GPU that is especially power efficient, as it is efficiency that will win the market for NVIDIA.

Maxwell then is an interesting take on NVIDIA’s designs that does not radically alter NVIDIA’s architecture, but has had every accommodation made to improve energy efficiency. The result is a Kepler-like architecture with a number of small design tweaks that improve efficiency in some manner. As NVIDIA tells it, there is no single aspect of Maxwell that is disproportionally responsible for NVIDIA’s energy improvements, but rather it is the culmination of these small changes. Through these changes NVIDIA has been able to come close to doubling their performance per watt versus Kepler, which is nothing short of amazing given the fact that all of this is being done on the same 28nm process as Kepler.

Starting with the Maxwell 1 SMM, NVIDIA has adjusted their streaming multiprocessor layout to achieve better efficiency. Whereas the Kepler SMX was for all practical purposes a large, flat design with 4 warp schedulers and 15 different execution blocks, the SMM has been heavily partitioned. Physically each SMM is still one contiguous unit, not really all that different from an SMX. But logically the execution blocks which each warp scheduler can access have been greatly curtailed.

The end result is that in an SMX the 4 warp schedulers would share most of their execution resources and work out which warp was on which execution resource for any given cycle. But on an SMM, the warp schedulers are removed from each other and given complete dominion over a far smaller collection of execution resources. No longer do warp schedulers have to share FP32 CUDA cores, special function units, or load/store units, as each of those is replicated across each partition. Only texture units and FP64 CUDA cores are shared.

Among the changes NVIDIA made to reduce power consumption, this is among the greatest. Shared resources, though extremely useful when you have the workloads to fill them, do have drawbacks. They’re wasting space and power if not fed, the crossbar to connect all of them is not particularly cheap on a power or area basis, and there is additional scheduling overhead from having to coordinate the actions of those warp schedulers. By forgoing the shared resources NVIDIA loses out on some of the performance benefits from the design, but what they gain in power and space efficiency more than makes up for it.

NVIDIA still isn’t sharing hard numbers on SMM power efficiency, but for space efficiency a single 128 CUDA core SMM can deliver 90% of the performance of a 192 CUDA core SMX at a much smaller size.

Moving on, along with the SMM layout changes NVIDIA has also made a number of small tweaks to improve the IPC of the GPU. The scheduler has been rewritten to avoid stalls and otherwise behave more intelligently. Furthermore by achieving higher utilization of their existing hardware, NVIDIA doesn’t need as many functional units to hit their desired performance targets, which in turn saves on space and ultimately power consumption.

NVIDIA has also been focused on memory efficiency, both for performance and power reasons, resulting in the L2 cache size been greatly increased. NVIDIA has from 256KB in GK107 to 2MB on GM107, and from 512KB on GK104 to the same 2MB on GM204. This cache size increase reduces the amount of traffic that needs to cross the memory bus, reducing both the power spent on the memory bus and improving overall performance.

Increasing the amount of cache always represents an interesting tradeoff since cache is something of a known quantity and is rather dense, but it’s only useful if there are memory stalls or other memory operations that it can cover. Consequently we often see cache implemented in relation to whether there are any other optimizations available. In some cases it makes more sense to use the transistors to build more functional units, and in other cases it makes sense to build the cache. The use of 2MB of L2 cache in both GM107 and GM204 – despite the big differences in ROP count and memory bus size – indicates that NVIDIA’s settling on 2MB as their new sweet spot for consumer graphics GPUs.

Finally there’s the lowest of low level optimizations, which is transistor level optimizations. These optimizations are something of a secret sauce for NVIDIA, but they tell us they’ve gone through at the transistor level to squeeze out additional energy efficiency as they could find it. Given that TSMC 28nm is now a very mature process with well understood abilities and quirks, NVIDIA should be able to design and build their circuits to a tighter tolerance now than they would have been able to when working on GK107 and GK104 over 2 years ago.

The NVIDIA GeForce GTX 980 Review Maxwell 2 Architecture: Introducing GM204
Comments Locked

274 Comments

View All Comments

  • mesahusa - Tuesday, September 23, 2014 - link

    Nvidia and AMD havent moved to 22 because they don't have the funding. Intel has tens of billions to blow away in R&D. Broadwells going to be released in 2015, and its 14nm.
  • Hrel - Monday, September 22, 2014 - link

    In light of Nvidia not even trying to reduce manufacturing nodes It would be really nice to see them go on the offensive in the price war. $300 for the GTX980, everything lower from there. Probably not now, but like, spring 2015, that'd be great! Make good and sure to wipe out all the hold outs (like myself) keeping their old cards because they still play everything they play on 1080p. Kinda, get all your customers caught up on hardware in the same upgrade time frame.

    Then when they finally do drop nodes they can focus on making every card they sell run games at 8K resolution.
  • Nfarce - Monday, September 22, 2014 - link

    Hate to break the news to you, but if you want to game at high level (above 1080p), you need to pay at high level. There is nothing new about that in the entire history of PC enthusiast building and gaming either for those of us who remember making the "huge leap" from a 15" 1024x768 resolution CRT monitor to a whopping 19" 1600x1200 CRT monitor. At least not in the 20 years since I've been involved with it anyway.

    Besides all that, that's why GPU makers offer cards for different budgets. If you can't afford their top tier products, you can't afford to game top tier. Period and end of discussion.
  • tuxRoller - Monday, September 22, 2014 - link

    It seems as though the big improvement nvidia has made is to enable cpu-level scheduling/dvfs granularity into their chip. However, once all cores are engaged it ends up using as much power as its predecessor (see tomshardware).
    What I really want to know is how much of this due to purely driver-level changes.
  • yhselp - Tuesday, September 23, 2014 - link

    Exceptional design. The sad thing is that NVIDIA will take forever to release a 30 SMM Maxwell GPU and once it finally does, it would cost a ton; even later on when they release a "budget" version for an unreasonable price of around $650 it would be too late - the great performance potential of today wouldn't be so great tomorrow. Striving for and building amazing GPUs is the right way forward, not empowering the people with them is a crime. Whatever happened to $500 flagship products?
  • Rhodie - Wednesday, September 24, 2014 - link

    Just got a GTX970, and only latest Nvidia drivers will install for 9xx series cards it seems. Unfortunately the latest drivers totally screw up some programs that use CUDA, seem to hide its
    presence from programs lile Xillisoft Video Convertor Ultimate:-/ No response of course from either Nvidia or Xillisoft regarding the problem. Wonder how many other programs the drivers break?
  • garadante - Thursday, September 25, 2014 - link

    Geeze. Anandtech, do an updated best value graphics card list because since the launch of the 970/980 retailers are giving some serious price cuts to 770/780/780 Ti's. Newegg has a 780 for less than $300 after rebate and just a hair over $300 before rebate. I'm seeing 780 Ti's for ~$430 and 770s for ~$240. I am amazed to see price cuts this deep since I haven't seen them the last several generations and considering how overpriced these cards were. But while supplies last and prices hold/drop, this completely flips price/performance on it's head. I feel bad recommending an AMD 290 Tri-X to a friend a couple months back now. xD
  • garadante - Thursday, September 25, 2014 - link

    Please do an updates best value graphics card* Where are my manners! D:
  • jman9295 - Friday, September 26, 2014 - link

    Newegg has an Asus DirectCU II GTX 780 selling in the $290 range after a mail in rebate, promo code and discount. It also comes with a pre-order copy of the new Borderlands game. That has to be the best value to performance GPU out right now. It is almost a full $100 less than the cheapest non-reference R9 290 on newegg and $40 less than the cheapest reference R9 290 which is crazy since this same Asus GTX 780 was selling for over $550 just last month with no free games (and still is on Amazon for some reason).
  • mixer4x - Thursday, September 25, 2014 - link

    I feel bad for having just bought the 290 tri-x just a month ago! =(
    I bought it because you never know when the new cards will be released and how much they will cost. Unfortunately, the new cards came out too soon!

Log in

Don't have an account? Sign up now