AMD sent out a note yesterday with some interesting news about how the graphics card market fared in Q1 of 2016. First, let's get to the bad news: sales of new discrete graphics solutions, in both mobile and desktop, dropped by 10.2% quarter to quarter, a decrease that was slightly higher than expected. Though details weren't given in the announcement or data I have from Mercury Research, it seems likely that expectations of upcoming new GPUs from both NVIDIA and AMD contributed to the slowdown of sales on some level.
Despite the shrinking pie, AMD grabbed more of it in Q1 2016 than it had in Q4 of 2015, gaining on total market share by 3.2% for a total of 29.4%. That's a nice gain in a short few months but its still much lower than Radeon has been as recently as 2013. That 3.2% gain includes both notebook and desktop discrete GPUs, but let's break it down further.
| Q1'16 Desktop | Q1'16 Desktop Change | Q1'16 Mobile | Q1'16 Mobile Change | |
|---|---|---|---|---|
| AMD | 22.7% | +1.8% | 38.7% | +7.3% |
| NVIDIA (assumed) | ~77% | -1.8% | ~61% | -7.3% |
AMD's gain in the desktop graphics card market was 1.8%, up to 22.7% of the market, while the notebook discrete graphics share jumped an astounding 7.3% to 38.7% of the total market.
NVIDIA obviously still has a commanding lead in desktop add-in cards with more than 75% of the market, but Mercury Research believes that a renewed focus on driver development, virtual reality and the creation of the Radeon Technologies Group attributed to the increases in share for AMD.
Q3 of 2016 is where I think the future looks most interesting. Not only will NVIDIA's newly released GeForce GTX 1080 and upcoming GTX 1070 have time to settle in but the upcoming Polaris architecture based cards from AMD will have a chance to stretch their legs and attempt to continue pushing the needle in the upward direction.
People see that they fixed
People see that they fixed software, and going in right direction…they just need to keep doing that, and not late with new Arh.
That and the AMD stomping all
That and the AMD stomping all over NVIDIA in most newer games. That’s why I sold my 970 and got a 390 to hold me over until Vega and the 1080Ti are out.
the R9 400’s are just around
the R9 400’s are just around the corner.
If you go 390.. might as well go all the way with a 390X !
“… discrete graphics
“… discrete graphics solutions, in both mobile and desktop, dropped by 10.2% quarter to quarter, a decrease that was slightly higher than expected.”
And Nvidia thinks they’re going to sell a mid-range GPU for $699? LOL
It’s good to see AMD getting back some marketshare, I don’t personally care for Polaris but I’m not the majority of the market.
Although the 1080 is a pretty mediocre card for the price point, it is only a mid-range chip – so that pretty much guarantees the 1080 Ti/Titan P are going to be god damn monsters (hopefully not for $899 and $1499 respectively)
Hopefully Vega will be able to butt heads with the 1080 Ti/Titan.
I think upcoming AMD parts
I think upcoming AMD parts might be better for mobile, which is good since I need a new laptop. The mobile market is obviously more important than the desktop market going forward. I would like an AMD GPU or even a 14 nm AMD APU. I think AMD can increase market share significantly in the mobile market. Their APUs will have huge advantages over Intel and Nvidia solutions. Intel can’t provide a good GPU, and they seem to have given up on doing so with their recent change of focus on the enterprise market and the layoffs. Nvidia doesn’t have a good processor unless people are willing to accept their ARM cores. I wouldn’t buy an ARM based laptop yet. Something like the rumored PS4 Neo specs would make a great laptop APU. AMD could have relatively good financials given all of the console products and their possible upcoming mobile parts.
I wouldn’t buy Nvidia GPUs right now, but I also wouldn’t spend $600 or more on a GPU anyway. For reasonable price ranges, AMD still offers the best value. I also think Nvidia is behind on DX12 development; AMD has essentially been working on DX12 for years via their Mantle development. I also think that asynchronous compute will be more important going forward and the 1080 still doesn’t seem like it is going to have very usable asynchronous compute abilities. Good asynchronous compute will be available for all of the consoles in addition to all AMD parts going back to the initial GCN parts (7700 and above, I believe). This should be a large enough market segment such that developers will support it even if it is not usable on Nvidia parts.
Great for AMD, and that fully
Great for AMD, and that fully in hardware GPU async-compute(That even Nvidia is slowly improving on, not there yet though)! So let AMD get up to that 50% market share and get into an async-compute on the GPU arms race with Nvidia! Now for both AMD’s and Nvidia’s GPUs to get some PowerVR like dedicated on GPU hardware Ray Tracing unit options and things will be even better for doing graphics on the GPU without any need of help from any CPUs!
I’m hoping that AMD’s Radeon Technologies Group(RTG) Raja Koduri bakes even more CPU like functionality into the ACE units on AMD’s GCN based GPUs, and I can tell the CPU to GTFO of all graphics related workloads on my PCs/Laptops! CPU are the MOOKs of the Graphics processing world, GPUs rule!
The dedicated graphics card
The dedicated graphics card market is going to become increasingly more of a niche market. An APU like what is rumored to be going into the PS4 Neo would make an excellent mobile part. If you consider such an APU with GDDR5/X attached to it, you would not need a dedicated GPU except for the higher side of the midrange market. With HBM based APUs, the dedicated graphics card market will be pushed to an even higher end niche. Intel can not compete on the graphics front, and they seem to have given up on it to some extent, so I expect AMD to grab some mobile market share. Nvidia can not compete in the laptop APU market very well either due to their lack of an integrated CPU, other than their ARM cores. I would expect a lot of laptop makers to be switching to AMD APUs for the same reason that all of the recent consoles have been AMD based. No one else can offer a single chip solution that comes close in performance, although some ARM based solutions may be coming close in the future. It will probably be a while before they can beat a Zen based CPU and a Polaris based GPU though.
I still think Nvidia should
I still think Nvidia should find some way to grab up VIA for their x86 license 😉
Even if somehow they able to
Even if somehow they able to get it how many year will nvidia need to be competitive with intel offering? Looking at what happen to AMD nvidia probably don’t want to be the second ‘victim’. Also intel for their part did everything in their might to halt nvidia advancement in x86 world.
Why x86, when there is
Why x86, when there is OpenPower Power8/power9 licenses to be had, for both Nvidia and AMD/others, why why this x86 single ISA obsession! There is nothing special about the x86 ISA, in fact the x86 CISC ISA is not very good for mobile devices, where the custom ARM RISC ISA designs have better lower power usage metrics that will be including of AMD’s K12/Polaris ARM based APU SKUs very soon. The Power ISA(RISC) designs Power8/power9 based SKUs from IBM/OpenPower and third parties will have more server/HPC/workstation adoption rates for systems that can use the 8 CPU processor threads per core at 12 cores per die for some power8’s, and who knows what the power9 designs will bring. Google’s getting Power9s!
I see no reason why AMD should not/could not get a power8/Power9 license and have a third ISA to integrate its GPU/other IP with, and like Nvidia, profit from the third party OpenPower hardware ecosystem.
Somehow people have seriously fallen for Intel’s x86 be all end all marketing tactics over the years, when the only reason that x86 was adopted industry wide was dew to IBM’s forced cross-licensing agreement with Intel that required Intel to cross-license the x86 16/32 bit ISA with AMD and others at the very beginning of what was to become the PC market that we know today! The x86 ISA was never that great of and ISA, in fact the x86 ISA has a lot of legacy baggage included, and extended over the years!
AMD would be very wise to extend its expertise into that Power ISA based market also, and have that x86, and custom ARM ISA business also. The power8 design is one of the very widest order superscalar CPU designs out there, and AMD adding its Vega IP and GPU accelerator IP into the OpenPower based market mix would be very well received by IBM! As we know from history that IBM likes to have a second source of competitively bid for CPU, or GPU/other supplies at its disposal, just look at who was responsible for giving AMD/others a crack at that x86 16/32 bit CPU ISA parts supply chain business in the first place!
For the simple reason that
For the simple reason that devs will not port legacy software to it.
Also amd pushed 64 bit on it
Legacy windows cruft, and
Legacy windows cruft, and much of the server software already being Linux Kernel based, as is most of the mobile market. Only the consumer PC/laptop market has the x86 ISA legacy issues mostly for windows x86 based code! The Linux ecosystem will be the one Google will be using for their Power9s, and possibly with Nvidia GPU accelerators. There is plenty of Linux support for the Power8/Power9 based IBM, and OpenPower power ISA based third party power licensee market ecosystem, and Nvidia could probably easily go with some power8 based gaming systems and use 4 power8 core designs, that still would have 32 total processor threads and Steam OS(Debian based)/Vulkan will run just fine on any power based gaming platforms.
So any Linux OS based games/gaming engines could have their software ported over from x86 to power with just a compiler option to target the Power8/Power ISA, with very little in that way of extra work once the Linux driver stack is there, and because of Nvidia’s/IBM’s OpenPower supercomputer government contracts there will be plenty of support on Nvidia’s side for the Linux Kernel. The Vulkan graphics API will dwarf the DX12 graphics API in total cross market/cross platform support so there is where the development dollars will be there for Vulkan! Both Nvidia(already has jumped), and AMD/others should jump on that OpenPower market, as that is where some serious new money in the server/HPC/workstation business is going. AMD’s really needs to get some of that OpenPower GPU accelerator business, and it can still have its x86, and custom ARM business also. No reason why AMD can not profit off of 3 different ISAs, and put that single ISA dependency far behind!
Just go look at all the new Vulkan API updates and new commits happening weekly, and see where things are going for the Linux Kernel/Vulkan API, Android N is already there with the Vulkan graphics API! And Android lives mostly on the ARM ISA, and OpenPower systems, as well as the other Linux Kernel/BSD kernel based OSs.
no joke via is doing
no joke via is doing absolutely NOTHING with it
im have two gtx 780, but this
im have two gtx 780, but this time i will go for one AMD HBM2 graphics. i hope not fall in the green team bag again.
Help me Baby Jesus…
I wouldn’t rule out GDDR5x
I wouldn’t rule out GDDR5x yet. It seems to be relatively efficient on the 1080. My next card will probably be Polaris based with GDDR5 or 5x, but it would be nice if AMD could make a next generation part with more than 4 GB of HBM1 in between the current Fury and Vega with HBM2. We don’t need to jump to a new generation of HBM in one generation, but 4 GB doesn’t seem to be enough anymore. It could be acceptable to only have 4 GB if they update the compression schemes significantly and perhaps do some other memory capacity optimizations. It just might not be worth the development work though. Is the bigger Polaris part supposed to be HBM1?
Sadly, 1st-Gen HBM is limited
Sadly, 1st-Gen HBM is limited to 4GB. Sorry.
There are several ways they
There are several ways they could get around that if they needed to, but it would take a lot of work. It sounds like HBM2 may be available such that we can get Vega early, Such an early release implies to me that Vega is just a big Polaris with HBM, and probably will not contain any large design changes. It has been featured on slides as if it is a next generation GPU though. I guess we will have to just wait and see what happens.
The JEDEC 1st-Gen HBM
The JEDEC 1st-Gen HBM standard only specifies the 1024 bit interface to a single HBM stack, so nothing is stopping anyone from going to more than 4 HBM stacks except the extra costs of etching the extra traces, and placing the micro-bump connectors on an interposer’s silicon substrate.
So there is no hard limit HBM memory size limit on HBM GEN 1, it all depende on the maxium size of the available interposers, and that interposer market is probably gearing up for larger sizes at some future time depending on damand! HBM2 will be available soo enough, but that HPC/Server market will be getting at those, at first, rare HBM2 stocks, as the markups on the Pro/HPC/server grade GPU accelorators and AMD server/HPC APUs on an interposer with on an interposer module GPU accelorator are much greater than the markups on any consumer SKUs that may use HBM2.
Well AMDs “Computing and
Well AMDs “Computing and Graphics” division lost revenue from $470m to $460m in total, from the CFO commentary “GPU ASP decreased sequentially driven by lower consumer ASPs”, no mention of volume which means it’s roughly flat.
For those that don’t speak the lingo, ASP = average sales price. If you look at Steam’s hardware survay to see what cards are trending up from new sales AMDs high end is pretty absent too. So sure, in a shrinking market flat sales mean market share is up but I think it mostly means nVidia had a quite slow quarter on the low end, the GTX970/980 Ti held the high end well causing AMD to sell a cheaper mix of product. It’s not bad news, but it’s also not all that great news.
It’s easy for nVidia to charge premium prices on the high end like the $699 GTX1080 and operate on slimmer margins to squeeze AMD in their price brackets. Because AMD really needs every dollar they can find, still losing $100m/quarter and running out of assets to monetize. They really need to launch a product with a solid profit margin if they want to stay alive, market share doesn’t matter if you can’t pay the bills.
I see the same thing. AMD all
I see the same thing. AMD all but lost the >$200 GPU market.
And Polaris 10 doesn’t look good at all to regain anything on the desktop now that nvidia released Pascal.
And frankly, I have little hope vega will stand a chance against nvidia HBM2 solutions.
nvidia know the demand for the 1070 will be massive, and because its a die harvested part it make sense to release the 1080 first.
The 1070 will better AMD best card, so nvidia can dictate the price. I wouldn’t be surprised if its in the $499 region, with the GTX 980 at $399 and 970 at $299…
AMD needs to get back in the
AMD needs to get back in the Server/HPC/workstation business ASAP, there is not enough revenues in the Shrinking consumer OEM PC/laptop markets. Now the DIY home system builder market is a different story, but that is for PC builds only, as we are all forced to get our laptops from the OEM market! AMD has a very healthy semi-custom market presence, so those new revenues will help also!
Steam’s hardware survey will be very different in a year’s time owning to AMD’s continued Open Sourcing of its driver/middle-ware/other Mantle derived API influence, so revisit things in a year when Linux(Steam OS)/Vulkan have better market penetration, the closed ecosystem windows 10/DX12 market will have some temporary affect on Steam’s survey results too, as AMD’s async-compute lead will still be there in a year, Nvidia can not engineer its way out of its async-compute deficiency in its GPU SKUs in only a year, as that is not possible for any CPU/GPU re-engineering time-lines!
DX12/windows 10 and the Steam hardware survey is going to be affected by M$’s UWP final end plan, so there will be some serious unknowns in a year’s time, in spite of M$’s spin-doctoring otherwise about M$’s true UWP intents!
The increase in mobile is to
The increase in mobile is to be expected. Amd used to dominate the market share on mobile. It is really a sily decision to simply hand the market to nvidia back in 2012. Now they still in the red for each quarter and now they have to fight back for what they have lost.
These numbers showing The
These numbers showing The Radeon Group is doing things right. Keep it up!
AIB market in decline…
AIB
AIB market in decline…
AIB = add-in boards / discreet graphics cards
The attach rate of AIBs to desktop PCs has declined from a high of 63% in Q1 2008 to 37% [in Q4 2015].
1999 114 million units
…
2013 65 million units
2014 44 million units (2.9 million are enthusiast level AIBs
2015 50 million units (5.9 million are enthusiast level AIBs!)
The high-end gamer AIBs were the bright spot in the PC market in 2015, and especially in Q4’15.
Overall add-in board market decreased in Q4’15, AMD gained market share, while Nvidia lost share. Quarter-to-quarter AIBs shipments decreased -4.9% and -7.9% year-to-year.
Source: http://www.jonpeddie.com/publications/add-in-board-report
This is bad news for AMD from
This is bad news for AMD from my point of view, at least on the desktop. They released new desktop cards in Q1 with no new competition from Nvidia, and their share increase was almost negligible.
Fiji was a waste of time.
Fiji was a waste of time. They most likely lost money on this generation.
But the thing they did was make HBM possible for everyone else via lots of R&D sacrifices.
AMD seem to have decent HW, but the software stack driver is killing them.
I think if you where to give the Hawaii to nvidia engineers they would boost most if not all game benchmark by at least 15%.
Also what happen to all the work on the audio DSP processor in Hawaii and Fiji ? We are entering the world of VR where 3D sound is a pillar of the experience and not one place mention how AMD got an edge by having dedicated audio DSP in silicon…
And I see little to no changes since the “radeon group” was formed. AMD is just on the same it was 5 years ago, on a path to irrelevancy by making mistake after mistakes.
That’s why AMD needs to get
That’s why AMD needs to get back more of its HPC/Server/workstation market-share with its Zen/Vega APU on an Interposer SKUs for that market, as the consumer markets are in a state of decline, at least for the OEM made PCs/laptops! The AIB, and DYI markets will be going back up again as the newer 16nm/14nm GPU SKUs and Vulkan/DX12 Graphics APIs begin to be used! So the home gaming system builders market is going to be shifting over to the latest hardware, and the VR gaming market is just getting on its legs! Still the consumer markets are fickle, but the HPC/Server/workstation markets are going to be going through some steady growth, and as is the exascale computing market with U-Sam funding of the government exascale initiative!
There is also the Power/OpenPower market with IBM, ARM holdings style, licensing the power8/power9 designs out to third parties via OpenPower, and Nvidia is already there in that market with its Pascal, and Volta GPU accelerators on some government supercomputing design wins. There is nothing stopping AMD from getting some GPU accelerator business from any of the OpenPower licensees, and AMD getting some of that markets business, AMD has some IP of its own to compete with Nvidia’s NVLink, so AMD needs to look at that Power based market for future revenues also. Google is going to be a big user of Power9’s so AMD needs to get its GPU IP interfaced up with the power based CPU cores, to go along with AMD custom ARM(K12) and Zen x86 based offerings!
and AMD getting some of that
and AMD getting some of that markets business 192.168.1.1