At the risk of sounding like a broken record, the biggest story in the GPU industry over the last year has been over what isn’t as opposed to what is. What isn’t happening is that after nearly 3 years of the leading edge manufacturing node for GPUs at TSMC being their 28nm process, it isn’t being replaced any time soon. As of this fall TSMC has 20nm up and running, but only for SoC-class devices such as Qualcomm Snapdragons and Apple’s A8. Consequently if you’re making something big and powerful like a GPU, all signs point to an unprecedented 4th year of 28nm being the leading node.

We start off with this tidbit because it’s important to understand the manufacturing situation in order to frame everything that follows. In years past TSMC would produce a new node every 2 years, and farther back yet there would even be half-nodes in between those 2 years. This meant that every 1-2 years GPU manufacturers could take advantage of Moore’s Law and pack in more hardware into a chip of the same size, rapidly increasing their performance. Given the embarrassingly parallel nature of graphics rendering, it’s this cadence in manufacturing improvements that has driven so much of the advancement of GPUs for so long.

With 28nm however that 2 year cadence has stalled, and this has driven GPU manufacturers into an interesting and really unprecedented corner. They can’t merely rest on their laurels for the 4 years between 28nm and the next node – their continuing existence means having new products every cycle – so they instead must find new ways to develop new products. They must iterate on their designs and technology so that now more than ever it’s their designs driving progress and not improvements in manufacturing technology.

What this means is that for consumers and technology enthusiasts alike we are venturing into something of an uncharted territory. With no real precedent to draw from we can only guess what AMD and NVIDIA will do to maintain the pace of innovation in the face of manufacturing stagnation. This makes this a frustrating time – who doesn’t miss GPUs doubling in performance every 2 years – but also an interesting one. How will AMD and NVIDIA solve the problem they face and bring newer, better products to the market? We don’t know, and not knowing the answer leaves us open to be surprised.

Out of NVIDIA the answer to that has come in two parts this year. NVIDIA’s Kepler architecture, first introduced in 2012, has just about reached its retirement age. NVIDIA continues to develop new architectures on roughly a 2 year cycle, so new manufacturing process or not they have something ready to go. And that something is Maxwell.

GTX 750 Ti: First Generation Maxwell

At the start of this year we saw the first half of the Maxwell architecture in the form of the GeForce GTX 750 and GTX 750 Ti. Based on the first generation Maxwell GM107 GPU, NVIDIA did something we still can hardly believe and managed to pull off a trifecta of improvements over Kepler. GTX 750 Ti was significantly faster than its predecessor, it was denser than its predecessor (though larger overall), and perhaps most importantly consumed less power than its predecessor. In GM107 NVIDIA was able to significantly improve their performance and reduce their power consumption at the same time, all on the same 28nm manufacturing node we’ve come to know since 2012. For NVIDIA this was a major accomplishment, and to this day competitor AMD doesn’t have a real answer to GM107’s energy efficiency.

However GM107 was only the start of the story. In deviating from their typical strategy of launching high-end GPU first – either a 100/110 or 104 GPU – NVIDIA told us up front that while they were launching in the low end first because that made the most sense for them, they would be following up on GM107 later this year with what at the time was being called “second generation Maxwell”. Now 7 months later and true to their word, NVIDIA is back in the spotlight with the first of the second generation Maxwell GPUs, GM204.

GM204 itself follows up on the GM107 with everything we loved about the first Maxwell GPUs and yet with more. “Second generation” in this case is not just a description of the second wave of Maxwell GPUs, but in fact is a technically accurate description of the Maxwell 2 architecture. As we’ll see in our deep dive into the architecture, Maxwell 2 has learned some new tricks compared to Maxwell 1 that make it an even more potent processor, and further extends the functionality of the family.

NVIDIA GPU Specification Comparison
  GTX 980 GTX 970 (Corrected) GTX 780 Ti GTX 770
CUDA Cores 2048 1664 2880 1536
Texture Units 128 104 240 128
ROPs 64 56 48 32
Core Clock 1126MHz 1050MHz 875MHz 1046MHz
Boost Clock 1216MHz 1178MHz 928Mhz 1085MHz
Memory Clock 7GHz GDDR5 7GHz GDDR5 7GHz GDDR5 7GHz GDDR5
Memory Bus Width 256-bit 256-bit 384-bit 256-bit
FP64 1/32 FP32 1/32 FP32 1/24 FP32 1/24 FP32
TDP 165W 145W 250W 230W
GPU GM204 GM204 GK110 GK104
Transistor Count 5.2B 5.2B 7.1B 3.5B
Manufacturing Process TSMC 28nm TSMC 28nm TSMC 28nm TSMC 28nm
Launch Date 09/18/14 09/18/14 11/07/13 05/30/13
Launch Price $549 $329 $699 $399

Today’s launch will see GM204 placed into two video cards, the GeForce GTX 980 and GeForce GTX 970. We’ll dive into the specs of each in a bit, but from an NVIDIA product standpoint these two parts are the immediate successors to the GTX 780/780Ti and GTX 770 respectively.  As was the case with GTX 780 and GTX 680 before it, these latest parts are designed and positioned to offer a respectable but by no means massive performance gain over the GTX 700 series. NVIDIA’s target for the upgrade market continues to be owners of cards 2-3 years old – so the GTX 600 and GTX 500 series – where the accumulation of performance and feature enhancements over the years adds up to the kind of 70%+ performance improvement most buyers are looking for.

At the very high end the GTX 980 will be unrivaled. It is roughly 10% faster than GTX 780 Ti and consumes almost 1/3rd less power for that performance. This is enough to keep the single-GPU performance crown solidly in NVIDIA’s hands, maintaining a 10-20% lead over AMD’s flagship Radeon R9 290X. Meanwhile GTX 970 should fare similarly as well, however as our sample is having compatibility issues that we haven’t been able to resolve in time, that is a discussion we will need to have another day.

NVIDIA will be placing the MSRP on the GTX 980 at $549 and the GTX 970 at $329. Depending on what you’re using as a baseline, this is either a $50 increase over the last price of the GTX 780 and launch price of the GTX 680, or a roughly $100 price cut compared to the launch prices of the GTX 780 and GTX 780 Ti. Meanwhile GTX 970 is effectively a drop-in replacement for GTX 770, launching at the price that GTX 770 has held for so long. We should see both GPUs at the usual places, though at present neither Newegg nor Amazon is showing any inventory yet – likely thanks to the odd time of launch as this coincides with NVIDIA's Game24 event – but you can check on GTX 980 and GTX 970 tomorrow.

Fall 2014 GPU Pricing Comparison
Radeon R9 295X2 $1000  
  $550 GeForce GTX 980
Radeon R9 290X $500  
Radeon R9 290 $400  
  $330 GeForce GTX 970
Radeon R9 280X $280  
Radeon R9 285 $250  
Radeon R9 280 $220 GeForce GTX 760

Finally, on a housekeeping note today’s article will be part of a series of articles on the GTX 980 series. As NVIDIA has only given us about half a week to look at GTX 980, we are splitting up our coverage to work within the time constraints. Today we will be covering GTX 980 and the Maxwell 2 architecture, including its construction, features, and the resulting GM204 GPU. Next week we will be looking at GTX 980 SLI performance, PCIe bandwidth, and a deeper look at the image quality aspects of NVIDIA’s newest anti-aliasing technologies, Dynamic Super Resolution and Multi-Frame sampled Anti-Aliasing. Finally, we will also be taking a look at the GTX 970 next week once we have a compatible sample. So stay tuned for the rest of our coverage on the Maxwell 2 family.

Maxwell 1 Architecture: The Story So Far


View All Comments

  • garadante - Sunday, September 21, 2014 - link

    What might be interesting is doing a comparison of video cards for a specific framerate target to (ideally, perhaps it wouldn't actually work like this?) standardize the CPU usage and thus CPU power usage across greatly differing cards. And then measure the power consumed by each card. In this way, couldn't you get a better example of Reply
  • garadante - Sunday, September 21, 2014 - link

    Whoops, hit tab twice and it somehow posted my comment. Continued:

    couldn't you get a better example of the power efficiency for a particular card and then meaningful comparisons between different cards? I see lots of people mentioning how the 980 seems to be drawing far more watts than it's rated TDP (and I'd really like someone credible to come in and state how heat dissipated and energy consumed are related. I swear they're the exact same number as any energy consumed by transistors would, after everything, be released as heat, but many people disagree here in the comments and I'd like a final say). Nvidia can slap whatever TDP they want on it and it can be justified by some marketing mumbo jumbo. Intel uses their SDPs, Nvidia using a 165 watt TDP seems highly suspect. And please, please use a nonreference 290X in your reviews, at least for a comparison standpoint. Hasn't it been proven that having cooling that isn't garbage and runs the GPU closer to high 60s/low 70s can lower power consumption (due to leakage?) something on the order of 20+ watts with the 290X? Yes there's justification in using reference products but lets face it, the only people who buy reference 290s/290Xs were either launch buyers or people who don't know better (there's the blower argument but really, better case exhaust fans and nonreference cooling destroys that argument).

    So basically I want to see real, meaningful comparisons of efficiencies for different cards at some specific framerate target to standardize CPU usage. Perhaps even monitoring CPU usage over the course of the test and reporting average, minimum, peak usage? Even using monitoring software to measure CPU power consumption in watts (as I'm fairly sure there are reasonably accurate ways of doing this already, as I know CoreTemp reports it as its probably just voltage*amperage, but correct me if I'm wrong) and reported again average, minimum, peak usage would be handy. It would be nice to see if Maxwell is really twice as energy efficient as GCN1.1 or if it's actually much closer. If it's much closer all these naysayers prophesizing AMD's doom are in for a rude awakening. I wouldn't put it past Nvidia to use marketing language to portray artificially low TDPs.
  • silverblue - Sunday, September 21, 2014 - link

    Apparently, compute tasks push the power usage way up; stick with gaming and it shouldn't. Reply
  • fm123 - Friday, September 26, 2014 - link

    Don't confuse TDP with power consumption, they are not the same thing. TDP is for designing the thermal solution to maintain the chip temperature. If there is more headroom in the chip temperature, then the system can operate faster, consuming more power.

    "Intel defines TDP as follows: The upper point of the thermal profile consists of the Thermal Design Power (TDP) and the associated Tcase value. Thermal Design Power (TDP) should be used for processor thermal solution design targets. TDP is not the maximum power that the processor can dissipate. TDP is measured at maximum TCASE"
  • NeatOman - Sunday, September 21, 2014 - link

    I just realized that the GTX 980 has a TDP of 165 watts, my Corsair CX430 watt PSU is almost overkill!, that's nuts. That's even enough room to give the whole system a very good stable overclock. Right now i have a pair of HD 7850's @ stock speed and a FX-8320 @ 4.5Ghz, good thing the Corsair puts out over 430 watts perfectly clean :) Reply
  • Nfarce - Sunday, September 21, 2014 - link

    While a good power supply, you are leaving yourself little headroom with 430W. I'm surprised you are getting away with it with two 7850s and not experiencing system crashes. Reply
  • ET - Sunday, September 21, 2014 - link

    The 980 is an impressive feat of engineering. Fewer transistors, fewer compute units, less power and better performance... NVIDIA has done a good job here. I hope that AMD has some good improvements of its own under its sleeve. Reply
  • garadante - Sunday, September 21, 2014 - link

    One thing to remember is they probably save a -ton- of die area/transistors by giving it only what, 1/32 double precision rate? I wonder how competitive in terms of transistors/area an AMD GPU would be if they gutted double precision compute and went for a narrower, faster memory controller. Reply
  • Farwalker2u - Sunday, September 21, 2014 - link

    I am looking forward to your review of the GTX 970 once you have a compatible sample in hand.
    I would like to see the results of the Folding @Home benchmarks. It seems that this site is the only one that consistently use that benchmark in its reviews.

    As a "Folder" I'd like to see any indication that the GTX 970, at a cost of $330 and drawing less watts than a GTX 780; may out produce both the 780 ($420 - $470) and the 780Ti ($600). I will be studying the Folding @ Home: Explicit, Single Precision chart which contains the test results of the GTX 970.
  • Wolfpup - Monday, September 22, 2014 - link

    Wow, this is impressive stuff. 10% more performance from 2/3 the power? That'll be great for desktops, but of course even better for notebooks. Very impressed they could pulll off that kind of leap on the same process!

    They've already managed to significantly bump up the top end mobile part from GTX 680 -> 880, but within a year or so I bet they can go quite a bit higher still.

    Oh well, it was nice having a top of the line mobile GPU for a while LOL

    If 28nm hit in 2012 though, doesn't that make 2015 its third year? At least 28nm seems to be a really good process, vs all the issues with 90/65nm, etc., since we're stuck on it so long.

    Isn't this Moore's Law hitting the constraints of physical reality though? We're taking longer and longer to get to progressively smaller shrinks in die size, it seems like...

    Oh well, 22nm's been great with Intel and 28's been great with everyone else!

Log in

Don't have an account? Sign up now