AMD Gains Significant Market Share in Q2 2016

Subject: Graphics Cards | August 24, 2016 - 10:34 AM |
Tagged: nvidia, market share, jpr, jon peddie, amd

As reported by both Mercury Research and now by Jon Peddie Research, in a graphics add-in card market that dropped dramatically in Q2 2016 in terms of total units shipped, AMD has gained significant market share against NVIDIA.

GPU Supplier Market share this QTR Market share last QTR Market share last year
AMD 29.9% 22.8% 18.0%
NVIDIA 70.0% 77.2% 81.9%
Total 100% 100% 100%

Source: Jon Peddie Research

Last year at this time, AMD was sitting at 18% market share in terms of units sold, an absolutely dismal result compared to NVIDIA's dominating 81.9%. Over the last couple of quarters we have seen AMD gain in this space, and keeping in mind that Q2 2016 does not include sales of AMD's new Polaris-based graphics cards like the Radeon RX 480, the jump to 29.9% is a big move for the company. As a result, NVIDIA falls back to 70% market share for the quarter, which is still a significant lead over the AMD.

Numbers like that shouldn't be taken lightly - for AMD to gain 7 points of market share in a single quarter indicates a substantial shift in the market. This includes all add-in cards: budget, mainstream, enthusiast and even workstation class products. One report I am received says that NVIDIA card sales specifically dropped off in Q2, though the exact reason why isn't known, and as a kind of defacto result, AMD gained sales share.

View Full Size

There are several other factors to watch with this data however. First, the quarterly drop in graphics card sales was -20% in Q2 when compared to Q1. That is well above the average seasonal Q1-Q2 drop, which JPR claims to be -9.7%. Much of this sell through decrease is likely due to consumers expecting releases of both NVIDIA Pascal GPUs and AMD Polaris GPUs, stalling sales as consumers delay their purchases. 

The NVIDIA GeForce GTX 1080 launched on May 17th and the GTX 1070 on May 29th. The company has made very bold claims about product sales of Pascal parts so I am honestly very surprised that the overall market would drop the way it did in Q2 and that NVIDIA would fall behind AMD as much as it has. Q3 2016 may be the defining time for both GPU vendors however as it will show the results of the work put into both new architectures and both new product lines. NVIDIA reported record profits recently so it will be interesting to see how that matches up to unit sales.

Video News

August 24, 2016 | 10:54 AM - Posted by Joseph Taylor (not verified)

The unfortunate dark side of this story is that a even larger part of a smaller market is still a decline in absolute sales.

August 24, 2016 | 11:05 AM - Posted by Anonymous (not verified)

This is just a fart compared to still nVidia. AMD is still far, far behind.

August 24, 2016 | 11:14 AM - Posted by Joseph Taylor (not verified)

AFAIK AMD/ATI marketshare has never been above 45% so this is more a reversion to the historical averages.

August 24, 2016 | 11:48 AM - Posted by Anonymous (not verified)

Might want to check your facts there. Not in the last decade, but before that they definitely did have a larger marketshare. NV4x got Nvidia back up to par, from G80 onwards they've had the upper hand. Before 2006 though? Very different story. AMD dominated '04/'05, before that they switched places a bunch of times. Mid to late '90s things were much murkier and I don't think there's any data for that period, but if I had to guess I'd say 3DFX was on top most of the time.

August 24, 2016 | 11:41 AM - Posted by Searching4Sasquatch (not verified)

NVIDIA was clearing out Maxwell parts for Pascal. The Pascal parts weren't in volume for much time at all. The next GPU marketshare snapshot should tell the real story. I have a feeling AMD fanboys are going to be crying again, especially when they see what Polaris didn't do.

August 24, 2016 | 02:32 PM - Posted by tatakai

keep dreaming. They were both clearing out product then.

August 24, 2016 | 11:43 AM - Posted by Anonymous (not verified)

That is the mainstream GPU market talking for AMD's new generation mainstream GPU offerings, and that's why AMD went for that market segment first. So currently Nvidia has the new generation high end/high cost market all to itself but at that price range is not where the majority of the purchases/revenues come from. AMD needs revenues more than profits right now, and has some fairly large debt payments starting in a little over 2 years so it's the revenues that will service that debt.

Don’t expect AMD to show much in the way of profits, but those all important revenues will be getting larger for AMD, and not only from Polaris, but from Zen as well starting in 2017. AMD is making all the right moves to claw back that GPU market share, and get back into the HPC/Server/workstation markets with Zen were those mad markups can be had, and AMD’s revenue situation can only get better in the high margin HPC/Server/workstation market.

And in the background to all this good news for AMD is AMD’s continued development of its new HPC/Server/Workstation APUs on an Interposer that will allow for AMD to pair 16/32 Zen cores with a large Greenland/Vega GPU accelerator die and HBM2 for an even more powerful and power saving line of APUs engineered for the HPC/Server/Workstation markets. Look for AMD to add some FPGA compute on the HBM2/newer stacks for even more in memory compute added to its GPU, and APU, HPC/Workstation/Server SKUs and even exascale SKUs for the supercomputer market.

August 29, 2016 | 02:40 PM - Posted by Anonymous Nvidia User (not verified)

The only reason they went for the "mainstream" cards is that they didn't have any big silicon close to being done. Not rumor they expect Vega enthusiast cards to be out by H1 2017.

Still rumors persist of possible small Vega in October but in AMD slide no such distinction was made.

The going after mainstream was largely marketing BS so they could keep their stock value up. It was their only option available. Rush out Polaris because Nvidia got Pascal out earlier than expected.

August 24, 2016 | 11:51 AM - Posted by Zaxx (not verified)

AMD coulda done even better if availability of their new gem (480) woulda been worth a damn. The 480s are just now starting to become easier to find but retailers are still price gouging the 480 like crazy as well as jacking up the price of the 470. A lot of buyers got sick of waiting and went for the 1060 6gb and 3gb instead. Lack of availability for this launch has costed amd a nice chunk of change as well as infuriated otherwise loyal buyers...esp with the resulting (and continuing) price gouging by retailers which is made possible by the lack of cards.

August 24, 2016 | 12:08 PM - Posted by Anonymous (not verified)

These Q2 numbers include a month of 1070 sales and a month and a half of 1080 sales and Nvidia still dropped over 9%. Polaris only had one day of sales in Q2.

This implies one of two things. Either yield issues definitely did impact Nvidia's launch as suspected by many or the Uber top end of the enthusiast market is not big enough to move the market share needle. Most likely a combination of the two.

The real battle takes place in Q3 between the 480,470 and 460 vs. the 1060. That is where the battle will be won or lost.

Keep in mind, the majority of monitors being used are 1080p or 1440p and most are at 60hz. So all that is needed for 90% of the market is a card that can get you to 60 fps at these two resolutions. In most cases, that can be obtained by a card below the 1070.

August 24, 2016 | 12:10 PM - Posted by Jann5s

Its really important, if AMD's market share drops too much you get the effect that game developers stop considering your hardware during development leading to even worse performance and thus less sales etc. I think 20% was already approaching this for some game developers so seeing it back to 30% makes me happy.

August 24, 2016 | 12:23 PM - Posted by Butthurt Beluga

Perhaps when Vega GPU family is released we might see AMD claw back another 10%?
2016-2017 are big years for AMD, and I think AMD is ready to show the world they still have some teeth with Polaris, Vega, and Zen.

On a software level as well, I get the feeling the "AMD has no/horrible drivers" is all but a distant memory now - Radeon Software Crimson and day one drivers have been amazing as of late.

I'm very jealous of the people who invested in AMD when they were at or even below $2 a share. I would not be surprised if Zen and Vega launch saw AMD stock skyrocket to $10 a share.

August 24, 2016 | 02:42 PM - Posted by renz (not verified)

well it depends on how competitive Vega will be. when AMD launch 290X they almost not gain anything in terms of market share. plus nvidia will not going to sit still as well. if AMD become threat enough they will respond to it to protect their market share. that's what happen back in 2010. nvidia initiate aggressive price war and in the end it is AMD that raise the white flag by the end of 2010.

August 24, 2016 | 01:07 PM - Posted by John H (not verified)

This is units shipped.. and Nvidia's mix of GPUs is likely (substantially) richer than AMD's. They have a long way to go to match Nvidia.

August 24, 2016 | 04:02 PM - Posted by Anonymous (not verified)

Nvidia's "Mix" of GPU SKUs reflects a segmentation scheme on Nvidia's part to segment on more levels to Milk for profits while disabling SLI on the GTX 1060, and removing more cores. Milk, milk, milk, and gimp, gimp, gimp the features to force Nvidia's customers to pay to play at 4k, 1440 and 1080 also. AMD’s GCN GPU SKUs have more compute resources and give more FP/Flops at lower clocks than any of Nvidia’s overpriced kit. The RX 480 has more FP compute at lower clocks with Nvidia gimping out the FP resources and raising the clocks to make up it’s GPUs hardware deficit!

The only thing richer about Nvidia is JHH’s wallet after the fools spend all their money, and hock some possessions to get the funds for Nvidia’s Overpriced SKUs.

August 24, 2016 | 04:37 PM - Posted by renz (not verified)

comparing FP performance between nvidia and AMD is useless. why Fury X not significantly faster than 980TI despite having massive advantage in terms of pure FLOPS vs 980Ti? people often comparing pascal and polaris clock but at the same time they forgetting about 1060 also has less shader count than RX480. there are reasons why AMD firepro accelerator are not really getting used in HPC hardware despite their hardware dominating the spec on paper.

August 29, 2016 | 02:47 PM - Posted by Anonymous Nvidia User (not verified)

If only more games were made with just compute in mind. Most people buy video cards to play games and not be good at general computing tasks. If we wanted another CPU we could buy a dual board.

Apparently compute isn't very good at doing VR which is an intensive graphics task. AMD is literally getting hammered on most titles coming out. Check hardocp if you want to see the truth.

August 24, 2016 | 01:17 PM - Posted by zmeul (not verified)

the numbers are wildly different from what Mercury Research published couple of days ago:

August 24, 2016 | 03:37 PM - Posted by Ophelos

I noticed that also. Let's hope Ryan updates the story with the real numbers. Instead of these marked down ones.

August 24, 2016 | 04:52 PM - Posted by bryf50 (not verified)

Mercury Research is reporting the shares as a four-quarter volume-weighted average to smooth the noise of seasonal inventory cycles and reveal ongoing share trends.

August 24, 2016 | 04:52 PM - Posted by bryf50 (not verified)

Mercury Research is reporting the shares as a four-quarter volume-weighted average to smooth the noise of seasonal inventory cycles and reveal ongoing share trends.

August 24, 2016 | 05:24 PM - Posted by Anonymous (not verified)

Yes, they differ this time just as they do every quarter. Both show AMD gaining share as both of them have shown for four quarters in a row.

What's interesting in the Q2 numbers is the fact that AMD gained share with virtually no help from Polaris as it came out after Q2 had ended while Nvidia had 30 days of 1080 sales and 15 days of 1070 sales.

AMD mentioned on their earnings call that their "legacy" GPUs sold better than they had anticipated leading up to the 480 release. This marketshare data bares that out.

August 24, 2016 | 01:37 PM - Posted by Anonymous (not verified)

I think the r9 390 and 380x sold well and we are seeing the results here. Q3 will show who was able to ship more units of 14nm vs 16nm cards, the 480 has been virtually sold out from day 1. The 1070 had always been in stock, we shall see.

August 24, 2016 | 05:41 PM - Posted by Anonymous (not verified)

During Q2 the 390 was ahead of the 970 in non Nvidia titles performance wise. As were the 380 and 380X ahead of the 960. This is the meat of the market where the volume is located. Makes perfect sense AMD would gain share as everybody in the enthusiast market were trying to buy a 1070 or 1080 that didn't exist in sufficient volumes to move the market s needle.

August 24, 2016 | 02:43 PM - Posted by renz (not verified)

true battle will be in Q3.

August 24, 2016 | 05:18 PM - Posted by Anonymous (not verified)

Oh No, That Green Goblin has nicked some code! Ohz Noz, 3.5 to 4, and the code was nicked and out the door, for some 5 Green Goblin finger discount savings, that will never be passed on to anyone but JHH's fat bank account!

"NVIDIA Ansel Post-FX Based on Stolen MasterEffect ReShade.fx Code?"

August 24, 2016 | 06:01 PM - Posted by John Pombrio

From the license.txt for MasterEffect:
1 Copyrights of the Master Effect exclusively belongs to author - Gilcher Pascal aka Marty McFly.
2 Master Effect (the SOFTWARE) is DonateWare application, which means you may or may not pay for this software to the author as donation.
3 If included in ENB presets, credit the author (Gilcher Pascal aka Marty McFly).
4 Software provided "AS IS", without warranty of any kind, use it on your own risk.
5 You may use and distribute software in commercial or non-commercial uses. For commercial use it is required to warn about using this software (in credits, on the box or other places). Commercial distribution of software as part of the games without author permission prohibited.
6 Author can change license agreement for new versions of the software.
7 All the rights, not described in this license agreement belongs to author.
8 Using the Master Effect means that user accept the terms of use, described by this license agreement.

August 24, 2016 | 09:41 PM - Posted by Anonymous (not verified)

Failure to give Props to the license Holder's By Name from Nvidia in the required manner after Nvidia used the code.

"Commercial distribution of software as part of the games without author permission prohibited."

That may cause Nvidia some trobles still!

August 24, 2016 | 06:26 PM - Posted by renz (not verified)

this thing has been resolved. some people try to make a big deal about it. move on.

August 24, 2016 | 06:33 PM - Posted by John Pombrio

If only this increase in market share actually helped AMD make a profit. In Q1, 2016, AMD's Graphics and Compute Division lost $70 million. In Q2, 2016 with MORE increase in market share, the division lost MORE money, $81 million dollar loss. Now AMD has decided to go sell only lower cost and lower profit cards for most of 2016 and is up against a much more robust family of higher cost and higher profit cards from NVidia. Hopefully AMD will start to see some improvement in their graphics division revenue with the 480 and 470 sales.

August 24, 2016 | 09:34 PM - Posted by Anonymous (not verified)

Only revenues matter to the lenders not profits and Revenues pay for loans, the R&D, and any salaries also! So AMD are selling in the consumer GPU market segment that produces the most revenues with which to pay the bills, AMD does not have much of the GPU Accelerator market, it does have a little more of the GPU Workstation market! And those lower profit cards sell in larger numbers than the more costly/overpriced Nvidia Kit! So AMD is doing what it set out to do and stay alive! And Now with Zen getting AMD back into the server market that will really be producing some better margins so AMD can get enough revenues to service all the debt that will come dew in a few years time.

That High Volume 150-300 dollar consumer GPU market produces the most unit sales figures and the most reasonable high volume unit sales with which to spread out the most logical/reasonable Amortization of R&D costs over the many more units sold so as not to make up too much of a percentage of each unit’s total cost. So AMD can price to sell in large enough volumes to make even more revenues relative to any flagship Consumer GPU SKUs that cost too much and are low volume so the amounts of R&D costs for amortization per flagship unit sold has to be much higher further inflating unit costs and deflating total unit volumes sold.

So why does AMD have to focus on only pleasing the one segment of the population with the least amount of money to spend, and help their fragile egos by producing a top tier flagship card for some bragging rights among the Knuckle Daggers, and other such persons that will most likely not have the money to afford to purchase any top tier SKU! With A quick look at the first Zen SKU for the desktop one can readily see that it’s made from half of a Zen 16 core server chip and that was very intentional on AMD’s part, as those 16/32 Zen core HPC/Server/Workstation SKUs are where the real revenues are for AMD to really prosper, so AMD can afford a consumer CPU/APU division in the first place.

So AMD is giving its CPU users more than just a bone, it’s giving as its first consumer Zen desktop CPU SKUs something that they would be whining for if AMD had instead gone for a lesser consumer 4 Zen core desktop SKU. And that Zen SKU should put AMD in a more solid competition with Intel on the Price/performance front for the High End desktop market and into the workstation/HPC/Server markets the soonest. That first consumer Zen SKUs is the direct result of AMD’s server/HPC/Workstation focus, and why not, Intel does the same to an end result of some great revenues and profits.

AMD’s primary focus is on increasing it’s revenues and market share it the mainstream GPU market and getting its Zen CPU SKUs back into the more revenue positive HPC/Server/Workstation market where the real companies make their bread and butter. A flagship GPU card from AMD will only be possible with the revenues that the real consumer mainstream market GPU revenue earner produces, so Polaris 10/11 is that volume revenue earner to keep AMD alive until Zen gets AMD’s foot back into the Door of the Real mad revenues earner the HPC/Server/Workstation market! Great strategic thinking AMD! Vega is coming but its has to be this way with Polaris/mainstream first for the revenues to keep AMD solvent, so Vega even can happen at all. And that half of a 16 core Zen HPC/server/Workstation consumer Zen SKU will be very appreciated because even a half of a 16 Zen core HPC/server/Workstation part’s Zen complement of Cores it still is right up there with a recent Intel E series 8 core part, and Zen does not have to beat Intel with the latest and greatest either, it's the performance/Dollar figure that gives AMD’s products more value in the end.

August 24, 2016 | 09:46 PM - Posted by Anonymous (not verified)

AMD momentum just keeps on building. Next quarter should see an even bigger market share gain for AMD with Polaris!

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

This question is for testing whether you are a human visitor and to prevent automated spam submissions.