⋮    ⋮    ⋮  

Analysis: Emerging Trends in the PC and Hardware Industry for 2017 – What Do Intel, NVIDIA and AMD Have in Store?


2016 is drawing to a close and boy what an eventful year this has been. The PC industry has seen some major technological shifts that were long overdue and the time has come for us to go over some of the major ones and figure out what 2017 holds for us. They say ‘all is data’ and this particularly holds true for the PC industry where almost everything happens on a progressive basis, meaning you can spot where a particular technology or trend is headed if you are vigilant about it.

If you’ve yet to read Adrian’s assessment on 2016/17 finance in tech as well as major global economic talking points for 2017, it’s worth a read as this article focuses quite specifically on hardware and there are of course broader themes which he discusses that could significantly affect the industry. You can find that article here. Without any further ado, the top trends in the PC and hardware industry for 2016 and what they mean for 2017:

Graphics processing technology will continue to push forward at a fast pace driven by increasing process maturity

I think it is safe to say that probably the biggest and most impactful change in mainstream gaming technology was the process shift from 28nm to 14nm. Intel had already introduced 14nm some time ago, but the CPU side of things doesn’t really see phenomenal gains over node shrinks. GPUs on the other hand, thrive on node shrinks and can post seemingly impossible performance and power efficiency gains over a die shrink and gaming technology is fundamentally driven by the capability of GPUs.

This is why when the big chips on the 16nm FinFET node finally rolled out from TSMC (and 14nm FinFET from GloFo), it was very much a big deal. An industry that had been stuck on the 28nm node for quite a few years got a much-needed breath of fresh air and the results, well, you can see them for yourself.

A comparison between 300mm² and 450mm² wafers - the shift to the latter will result in increased economies of scale.

NVIDIA managed to post performance gains of more than 50% thanks to clock speed improvements as well as the on-die economies that come with a node shrink. AMD made an absolute killing by providing updated APUs to the console manufacturers as well as targeting the sweet spot for a mainstream budget but powerful GPU with the Polaris based RX 480. Needless to say, we have not seen generational performance gains like this for a long time.

So what will 2017 hold for the gaming technology industry? Well, we are certainly not going to get a node shrink again (TSMC will shift to the 10nm process, but that won’t trickle into the gaming circle just yet), that is for sure. But we will be seeing a much more mature 14nm/16nm node which will bring with it a flurry of performance and power efficiency gains thanks to our good buddy: physics. The current die limit (under the optimization constraints for a 300 mm2 wafer) for our processes is roughly 600mm2 and the mainstream GPUs we have seen currently are still quite a ways off from this very limit (this is more true for AMD than it is for NVIDIA).

  • Expect significant performance gains driven by bigger die sizes up to ~70% for high-end GPUs for NVIDIA (as compared to the GTX 1080) and up to a 100% for AMD (as compared to the RX 480).
  • Expect slight clock speed improvements.
  • NVIDIA will have the option to shift to HBM2 memory and will probably introduce the Volta architecture.
  • AMD will introduce the Vega architecture and will have the option to roll out bigger GPUs than Vega in the same year.

This means that with 2017, will come a perfected 14nm/16nm node with inherently stronger GPUs. NVIDIA will be launching its 1080 Ti graphics card in 2017 and will introduce Volta GPUs close to the year end. AMD will be introducing its Vega based graphics card and has the option to introduce bigger GPUs as well with the availability of the mature node. As far as the memory tech goes, NVIDIA will have the option to finally make the shift to HBM2, something it has been avoiding for quite some time now

The VR ecosystem will continue to grow at a slow but steady pace and could finally become mainstream in 2017

We can’t talk about trends in the PC and hardware industry without talking about virtual reality. Hailed as the next big thing, the VR ship left the port but never arrived at its destination. Met with growing pains of a very large magnitude the manufacturers in question seemed to have underestimated the lack of a functioning ecosystem and the high cost of entry for potential customers. The problem can be broken down into four distinct parts: 1) the VR headset is extremely expensive, 2) the hardware required to power the VR headset is relatively expensive, 3) there is a lack of a well-developed ecosystem for VR and a serious shortage of killer apps.

A side by side comparison of the two primary PC VR headsets: HTC Vive and the Oculus Rift.

Knowing how much very large companies including Facebook and console manufacturers like Sony and Microsoft have invested in the VR industry, we can be sure that the boom will eventually go off (given the right circumstances) – the only question is when. This is also where our expectation of 2017 begins:

  • Expect announcements of newer, more affordable VR headsets from Oculus and HTC.
  • Expect the VR Total Addressable Market to grow significantly as the GPU cycle dictates the cheap availability of VR-capable GPUs for all gamers.
  • The VR ecosystem and app development will continue to grow at a slow rate.
  • This is the single most critical factor, though: will we finally see the release of killer apps for VR in 2017? Ones that don’t feel like a tech demo for VR capability but are actual games built for VR from the ground up? In other words, 2017 could very well be the year VR goes mainstream.

The reason why we think that VR could potentially head for mainstream success in 2017 is because all of the other pesky problems are already taken care off. You can now pick up GPUs that are pretty cheap that will easily handle VR and considering the next upgrade cycle for DX12 is already in full force this can no longer be considered a detrimental factor. Secondly, like I already mentioned, I expect Oculus and HTC to release more affordable versions of their headsets which will eliminate the last real remaining problem for VR. After that, it's really all up to the developers.

AMD will make a comeback driven by Ryzen in the CPU market

One of the easiest to predict trends of 2017 however, will be AMD’s comeback in the CPU market. With its current offerings of FX processors miserably outclassed by Intel offerings, the Ryzen processor will bring the company back to the forefront and will easily tear into Intel's dominance on processors. The result?

The 'Ryzen' render output that the AMD Ryzen demo used to benchmark performance of the processor.
  • Expect AMD CPU market share to increase (after a hiatus of many years) and consequently Intel’s to decrease.
  • We are also expecting Intel to answer with a flurry of price cuts for its processors when the time comes putting its processors in a comparable price-to-performance bracket when compared to Ryzen. This will depend entirely on whether AMD chooses to price Ryzen using the value philosophy. Intel could also choose to take the higher (but very dangerous) path of claiming to have a premium product which they charge a premium for – a route successfully exploited by NVIDIA in the past.
  • Expect some render farms to adopt AMD’s Ryzen processor to benefit from its high IPC and power efficiency and cost effectiveness.

AMD Ryzen is likely to single-handedly put the company back on the map for its offerings of x86 processors and this won’t just be about the gaming industry. Ryzen will be able to offer very competitive performance in the server and enterprise market as well. Unlike games which are usually not optimized to handle a large amount of threads, the server and enterprise applications of this processor will offer much higher value to any buyer. So to sum it up, Intel’s temporary free run is over and will be facing real competition from AMD this time around.

Keep in mind however, that while the performance will help it make inroads towards disrupting the market, the turnaround in market share for the enterprise/server segment will not necessarily be immediately apparent. This is because these segments usually work on medium to long term contracts and usually have longer sales cycles than the consumer market. Data centres which provide their clients with guarantees of like for like hardware replacements for example will be difficult to dislodge from existing supplier agreements.

Of course, if you are reading this you also want to know about AMD's share price position. AMD stock has appreciated massively in 2016, partly due to some savvy debt restructuring as well as some once off deals which helped get the books closer to being balanced. It seems likely given recent Ryzen and Vega information that the company could well continue its turnaround in 2017 and see further share price gains, although investors would do well to keep in mind that the stock is a volatile one at the best of times and is currently priced with a significant turnaround in the near future. If you think Ryzen and Vega are likely to fail, it probably isn't the stock for you, but we think further gains (barring global economic headwinds) are likely. It's also possible for the management to turn the company around once and for all. All the indicators are green and the time couldn't be more right for this to happen.

NVIDIA is heading towards blue-chip status

The company has already posted phenomenal capital gains this year and the trend is expected to continue into 2017. There are a couple of reasons why this is true. NVIDIA is one of the two major GPU manufacturers and the world has a new obsession: Deep Learning.

A slide from NVIDIA's TX1 Jetson module press kit.

Almost all major automation is now being done using the deep learning approach and as any computer enthusiast would tell you, the highly parallel task of training DNNs is best done on a GPU. So far the ecosystem is painted green with NVIDIA's CuDNN a favored tool that is used by many companies (AMD is also on the verge of breaking into this market).

So not only is it going to be leading in the gaming technology sector if past performance holds, but it will remain one of the biggest (if not the biggest) supplier of the underlying hardware necessary to achieve a deep learning approach to just about any Big Data application. And that's not all, as the new, sole provider of the brain behind Tesla's AutoPilot, NVIDIA has successfully managed to diversify its income streams (as much as a fabless chip design house can at any rate) to a reasonable degree. If Tesla's approach to AutoPilot pays off, and there is every reason to believe that it will, then other companies will follow suit.

Furthermore, the management's recent actions, which include the termination of $63 Million warrants hint at a company that is quickly maturing into a very stable financial entity. To put it bluntly, its showing all the signs of heading towards blue-chip status.

NVIDIA also has a reasonable debt pile maturing in the coming years. Its strong earnings and good credit rating don't worry the bond markets too much so it should be able to refinance and roll the debt that it needs to without too much difficulty, but do keep in mind that with interest rates increasing (already 25 basis points this year, more likely to come barring broader economic headwinds) mean that it may have to pay slightly higher interest than it is currently, but it won't be anything the company can't afford.

Intel will face growing competition, difficulty in node shrinks and release a platform that will start disrupting its industry by the end of 2017

Speaking about Intel, the company has been working on its 3D point based Optane platform that is expected to arrive in 2017 as well.  So you can expect the memory and storage solutions from Optane to seep into the market next year. That said, however, since they will initially be priced and targeted at the high-end enthusiast market, we don’t expect them to catch on and become mainstream till 2018.

Intel Optane storage solutions will be considerably faster than conventional SSDs.

It will, however, result in SSD prices dropping. SSDs which rely on the current process node, benefit just as much from the increasing economies of scale as any GPU. This is why we expect SSD prices to continue on the downward trend as they become faster and more affordable for the average consumer.

Intel will, however:

  • Introduce Kaby Lake refresh in 2017.
  • Shift to 10nm for risk production of initial wafers.
  • Optane platform to be introduced in 2017.
  • SSD prices to drop as a result of higher economies due to smaller node sizes and competition by Optane.

The company extended Moore's Law even further in 2016, shifting from its usual Tick Tock cadence to the new PAO (Process, Architecture, and Optimization) cadence which basically adds another 'Tock' at the end. This allows the company to gain some much-needed breathing space as the fight against physics to shrink nodes becomes increasingly tougher. The company has already started up its 10nm factory and will continue to push forward, entering the risk production stage by the end of 2017 if all goes to plan. As always I would like to point out that you cannot compare TSMC's 10nm node to Intel's since the latter is usually far better specced and truer to a real 10nm node (more details here).

Intel's share price faces a relatively uncertain future as the new paradigm sets in where Intel is spending huge amounts on R&D leaving it less able to fend off any competition from AMD. The grapevine has been buzzing with some pretty disturbing stories of bad management decisions at Intel but at this time we are not sure of their accuracy so will refrain from mentioning them. One thing is for sure however, Intel's lost monopoly is definitely not a bullish indicator and will probably result in negative pressure on the company's share price, although earnings are likely to remain in reasonable shape for the short to medium term.

Bonus: The PC industry will finally die remain stable

It would be really cool if all the PC industry doom and gloom forecasters finally see their prediction come to pass. Unfortunately, there is still no sign of that happening. Yes, in 2016, the mainstream industry has seen casual users shift to notebooks and mobility but the high-end market has actually gotten much larger this year as proven by the unprecedented demand and sales of the GTX 1070/GTX 1080. What is actually happening is that the industry is redshifting.

As the casual users leave for the mobility side of things, the dollar attach rate in the higher end segment continues to increase, easily offsetting any deficit from their low-end counterparts. In other words, the death of the PC industry has been greatly exaggerated (as always) and the DIY/PC gaming market has been offsetting any losses seen on the OEM, casual side of things.

Well, that is all for this particular iteration of our roundup.


The author has no position in any of the stocks mentioned. WCCF TECH INC has a disclosure and ethics policy.