AMD Radeon R9 390X, R9 380 And R7 370 Tested

[cdrkf]

What's interesting is the 980ti isn't better at 4k though, so I guess nvidia are shader limited at 4k, whilst amd are rop limited. I think what's apparent is both companies are wrestling with the physical limits of 28nm.

 

[somebodyspecial]

4K is the job of then next gen after a die shrink, not 28nm and even then they'll only make me happy probably SOME of the time.

I do not think making you happy figures anywhere on AMD's agenda.

GPU improvement has been mostly stagnant for the past 3-4 years from being stuck at 28nm and the way I look at Fury, I see it like an open-beta for HBM: HBM will become necessary to get the most out of 16nm when it becomes available. Having it out in the wild a year before 16nm hits gives AMD that much more time to review their future HBM(2) and architecture plans based on field results. Fury might be a questionable success but that's alright since it was apparently never meant to shatter frame rate records in the first place with its 64 ROPs limiting it to the same pixel fill rate as the R9-390X.

In principle, the 980Ti's 96GP/s pixel fill rate (vs 67GP/s for Fury) make it much better suited at driving 4k, albeit at possibly lower details.

You should always be worried about making your customers happy as a business (as many as possible, period - and I'm in the 95% or so running 1440p or less). A test project that makes no money was not what they needed this year. They needed tried and true GDDR5 which has plenty of bandwidth today, is cheap and should have given more rops, more dedicated to the GAMING side etc, as WINNING should have been the goal period as your market share slips yearly. If winning isn't your goal when on the edge of bankruptcy, fire your management...LOL. Not even sure we'd need anything more than faster GDDR5 at 16nm. NV clearly gets a ton from GPU overclocking still (20% more from 980ti when overclocked) and they could have gone to 512bit next, while currently they sit at 384. Heck they went back to 256 with 980, due to better compression algorithms etc.

Stagnant for 3-4 years? LOL
http://www.guru3d.com/articles_pages/gigabyte_geforce_gtx_980_ti_g1_gaming_soc_review,15.html
Even 780ti is 1/2 speed of 980ti and GTX 680 came march 2012. Just over 3yrs ago, and is handily beaten by 780ti right?

http://www.anandtech.com/show/5699/nvidia-geforce-gtx-680-review

http://www.anandtech.com/show/9306/the-nvidia-geforce-gtx-980-ti-review/6
A better example since anandtech has both 680 and 980ti shown here. 680 scores 22.5fps 1440p, while 980ti scores 85.4. Almost 4x faster, and you call it stagnant. Never mind if I went back a full 4yrs (it wouldn't be 680 then), and 580 is in that list to at 15fps...I guess we have MAJOR differences in progress. That is just one example but you should get the point. Min fps shows same 12fps for 680, and even more spread with 980ti 56fps (over 4x faster!). Can't show 4K because 680 can't do it (more mem due to progress there counts as part of the end product too), otherwise it might get worse and I'm not saying 980ti is 4k able...LOL (it don't count if I'm turning crap off the devs wanted me to see). Even when you come to a game like civ5 it's 2x faster than 680, and nearly bottlenecked by cpu. I think the worst they get is 2x faster in any of the games. Also note when 28nm started they barely got 4.3Billion transistors in them (radeon 7970 Dec 2011 gpu $550 and the same 384bit bus...LOL), today they have 8.9Billion. Now a small part of that is 5.5ghz mem on those (7970) vs. todays 7ghz, but we're still talking mostly gpu that got them to 2-4x faster in todays games. There have been a lot of major modifications that add up to a multiple like that, so not you seem confused.

A year early on HBM1 means nothing when going to a new die shrink, finfet+ and HBM2 all at once. This experience wasn't needed, (probably useless when taking in all 3 changes mentioned, might be different SLIGHTLY if it was HBM1 again but it isn't) and just wasted cost and appears to have limited production to where they can't even get a sample to toms for more than a day, maximum pc said they had a single card for 4 of their sister sites and had no time to even share it (LOL) etc. How much is any experience worth if it steals even more market share from you for another year, and makes LOSSES on top of that problem? I don't need experience that badly at the expense of everything else...LOL. IF I'm AMD, I need sales, and profitable parts in those sales, and R&D on the GPU CORE itself to WIN vs. the other guy without WATER, who clearly has shown there is ample bandwidth in GDDR5 even at 384bit bus for cards this gen. IF HBM was a fix for a problem that REALLY existed (as in it suddenly vaulted AMD's fury to 50% faster than NV because bandwidth REALLY was an issue), I would be all for it. But that isn't the case, thus again management is stupid and NV wins again doing the smarter moves. IE, GDDR5. Previously NO CONSOLE (went mobile instead for last 5 years, now getting somewhere with it). Better drivers (and day 1 for all major games), WHQL drivers monthly, Gsync (freesync sorta works) etc etc.. Pascal will have 3x mem bandwidth of maxwell, but it won't be 3x faster in any game IMHO. I could be wrong but I would not bet on it. I'll be shocked if it's 2x maxwell2 in any game as I don't think NV will risk max sized dies on the new process first. I might not say that if there was mass evidence we are memory bottlenecked NOW, but we are not clearly. IE, when NV went to 28nm 680 they went up only .5B transistors and 294mm^2 die. In the 680 article one of the many changes mentioned is the fact that NV managed to get a CHIP & board made that was capable of 6ghz operation. Imagine what it took to get 7ghz going. From the 680 article above:
"Perhaps the icing on the cake for NVIDIA though is how many revisions it took them to get to 6GHz: one. NVIDIA was able to get 6GHz on the very first revision of GK104, which after Fermi’s lackluster performance is a remarkable turn of events."

It's not as simple as just slapping memory on it because it's faster, and suddenly it works. It takes engineering to get it all to work faster, bigger, better prefetching etc etc etc...Look at the massive chip re-org just from 580 to 680 in that article. Many things different in there that are tougher, from GPC's to polymorph 2.0 etc etc. They may seem like small changes alone to you, but in the grand scheme of things all together from 680-980ti it gets you 2-4x perf on the same process node. DX feature level on 680 was 11.0 IIRC, look at 980ti. Support for dx12 has some significant changes correct?

http://www.legitreviews.com/geforce-gtx-980-ti-dx12-feature-level-and-tier-details_164782
The differences between support for dx11.0 and 12.1 are pretty good. That siggraph 2015 vid is pretty impressive.

A bad move is a bad move. I don't think HBM2 (let alone hbm1) would have been necessary until the end of 16nm when they're cramming FAR more than 8.9B transistors in gpus. People can run GDDR up to 8ghz, so you could probably easily ship faster stuff stock and also in NV's case could go from 384 to 512b bus while doing it. Unless people started doing something stupid like adding 8K benchmarks, 512B and 7.5ghz-8ghz memory (next gen GDDR5 probably goes higher than 8ghz Oced as it shrinks, gets even more optimized etc) would be more than enough for 4K at least at the beginning of 16nm. Looking at what they've done in 3.5yrs cramming 4B in then and now 8.9B (7.9B in NV's 980ti case) I see your point. 3-4yrs later in the process yeah, I agree, but no sign you are near correct for the first rev with NV having so much room to play with and not even tapping out 384bit bus as they get exactly the 20% they oc the gpu in most games. Also note the 980ti is almost 2x faster than 690 in most stuff and does it with far less watts. I'd call that progress too. You can say it was because of process advancements, less mem chips, some design, etc etc, but it all adds up to FAR better perf than 3-4yrs ago. The fact that NV was able to go back to 256bit bus with 980 vs. 780ti and BEAT it should say something about gpu, and they did it with ~80 less watts (and 2B less transistors at $150 less price). No matter how I look at it, I say they've made some pretty nice improvements in the last 3-4yrs EASY. Even the stupid compute benchmarks show this at anandtech for 980ti vs. 680 (2-4x faster).

The worst was sony vegas in compute but that's a terrible example as Adobe uses cuda and would fare far better IMHO. There is a good reason anandtech uses it instead of Adobe which would be so easy to compare NV/Cuda vs. AMD/OpenCL (or opengl, whichever is faster for them), and it is because AMD cards like vegas despite AMD claiming a few years ago get ready for massive premiere perf coming...OK, when? They shows slides showing 45% IIRC but we've heard no peep since so anandtech avoids it and toms too for that matter (and don't respond to why they do it in the forums either...hmmpf). Not saying toms is biased here, just saying they're avoiding showing you how bad things really are (hey, AMD needs all the help they can get). Not divulging easily had info isn't as bad as saying it isn't there. 😉 I think this is the same reason toms never tests a CUDA plugin in any of their pro apps vs. OpenCL versions on AMD (which most popular pro apps have one OR MORE for each side and toms has tested many of these apps).

Even something as simple as slapping an IMC on AMD's chips years ago that allowed victory (until Intel got one too) is CPU progress to me. Along with other things it all added up to FASTER chips. Just having two competitors hashing it out year after year helps, but it's still BETTER CHIPS pretty much year after year. Maxwell 2 is nothing short of amazing compared to 680's gpu (3.5B vs. 980ti 7.9B transistors etc, same 28nm). It's not easy to make a ~600mm^2 die work. There are many factors going into the design that allow that to actually make money, while not lighting on fire...LOL. Don't forget you can put much better visuals on newer models too. You go ahead and take a 680 home if you want, but I'll take a 980ti to go thanks Faster is faster, no matter how it happened. Even 2x the transistors doesn't explain the 4x advances in some stuff. Never mind the fact that you won't be looking at the same graphics on a 680 vs. 980ti soon with DX12.1 etc stuff. You can say they're just re-arranging or adding some more of x or whatever, but if it scores higher and can look better, it's advancement. In any case, we have vastly different ideas of what gpu progress is I guess, and I'll leave it at that.

 

[InvalidError]

IF I'm AMD, I need sales, and profitable parts in those sales, and R&D on the GPU CORE itself to WIN vs. the other guy without WATER, who clearly has shown there is ample bandwidth in GDDR5 even at 384bit bus for cards this gen. IF HBM was a fix for a problem that REALLY existed (as in it suddenly vaulted AMD's fury to 50% faster than NV because bandwidth REALLY was an issue), I would be all for it.

GDDR5 even at 512bits and 8GT/s won't cut it at the higher end on 16nm with GPUs having 3-4X the throughput per given die area, neither will quad-stack HBM1. It took a new generation of texture compression to enable the DRAM interface reduction and these 20-30% memory bandwidth efficiency improvements only occur about once every five years.

Having 45% more shaders and 60% more bandwidth means little when the pixel bandwidth - the rate at which texture and shader outputs can be combined to generate finished pixels - remains unchanged compared to the 290X/390X. Fury manages to keep what few ROPs it has more consistently busy, but likely needed a handful of extra ROPs to start letting loose.

Most of the R&D money behind Fury was going to get spent on it regardless of the outcome, might as well get some useful data and revenue from AMD/ATI enthusiasts out of it if they can since it is the only fundamentally new thing they have released in the past couple of years and there won't be anything else fundamentally new until next year either. It is just a shame they did not see fit to increase the number of ROPs by at least eight along the way.

 

[ern88]

So, would you guys recommend the R9 390x or the GTX 980 for a upgrade from the HD7950?

 

[de5_Roy]

So, would you guys recommend the R9 390x or the GTX 980 for a upgrade from the HD7950?

wait and see how the little fijis turn out, and if nvidia does drop prices. then decide.

 

[somebodyspecial]

What's interesting is the 980ti isn't better at 4k though, so I guess nvidia are shader limited at 4k, whilst amd are rop limited. I think what's apparent is both companies are wrestling with the physical limits of 28nm.

Depends on the games you look at. Both have their victories at 4K. Look at more than one site and you'll find there are many 4K victories for NV. IE BF4, Bioshock Infinite, Grid2, Grid Autosport, Total War Rome 2, GTA5, Project Cars, Batman Arkham Origins (huge here, 81fps to 64 for OC on both) etc. Far Cry, Metro LL, Middle Earth SOM, Hitman Abosolution for AMD (among others for both sides, but much of 4K is 30fps or less even with details down) etc. I wonder how Metro LL REDUX would be though with more taxing engine inside. Witcher goes AMD until you turn hairworks on...LOL. But tessellation is 2x faster on Maxwell according to Beyond3D tests so how many games will that show up on in the future? Also note at many places you see them talk about the experience being better on NV (like toms mentioning HITCHES in in witcher 3, techreport comments 980 itself is smoother in 4k than furyx in their experience and their charts show it too).
IE from toms article on furyx: "Frame time variance is naturally higher at 4K, though we do record more spikes from the Fury X that manifest as noticeable on-screen hitches."
Hitman was another where frame times were all over the map for fury x vs. 980ti. Maybe they can fix it with drivers, but YMMV here with no WHQL driver since Dec8,2014 while NV has one monthly (and a win10 one before the OS ships..LOL).

Thief and Tomb Raider show up for AMD also, but I'd also note both well below 30fps so who cares. I won't be playing a game at 26 or 23fps.
"Most of the game isn’t this demanding."...That means nothing to me. I don't want to see any spot under 30, and I'd like quite a bit higher min in ANY benchmark just to avoid these types of situations. Most, implies more than once it IS demanding. Also some of the wins are fairly big (GTA5 15%, BF4 11%, etc).

Overclocking throws pretty much everything to Nvidia:
http://www.legitreviews.com/amd-radeon-r9-fury-x-video-card-review_167134/4
Many times furyx OC can't catch regular 980ti here. Even Far Cry 4 and Metro LL (which seem to love AMD) goes NV when both OCed. Thief goes massively NV here too. Even not OCe'd thief went to NV at legit (reverse of toms scores), but because they have a card that regularly runs 1076 where toms used ref straight up. Not sure why you'd by 1ghz when Zotac is selling their card on newegg with 1050mhz for $649 anyway and for $680 you get another 100mhz over legit's card (msi gaming card at 1176).

Only sapphire is selling Fury X for $650 at Amazon or Newegg and it's on a ships 1-3 week wait, while XFX at amazon is $680 and 1-3 months wait, while Visiontek is $699 2-4 weeks. Seems to me AMD has an availability problem. It looks like only XFX got them in (only card with reviews and only 4) and hence the longer wait I guess. The 1-3 month thing is not exactly inspiring. A quick look at newegg shows 1 card with 1 review (the 2nd review there the guy admits he doesn't own it...LOL). That guy is NOT a verified owner either so maybe ZERO reviews/owners of MSI? Nothing in stock. So how many have they shipped? After looking again, none of the 4 amazon reviewers (one hates it) are verified either. So the people that reviewed on amazon didn't buy it there, and the one guy on newegg didn't buy at newegg (and amazon doesn't even list the MSI as a part to sell, so where did this guy get it?). I think most get the point 😉 When I read the first run might only be 30K I laughed and thought probably not true, but now I'm wondering if that was AMD being hopeful about the 30K. I mean the first titan only had 100K first run (then sold so fast that massively upped quickly), but 30K should produce far more reviews and at least a SINGLE verified buyer at newegg/amazon (two huge players) I'd think after 5 days. Maybe not.

 

[InvalidError]

Seems to me AMD has an availability problem.

I do not think the Fury was ever intended for mass availability at all in the first place.

At 600sqmm, that's 120ish usable dies per 300mm wafer before you factor in defects. HMB1 is also a limited production volume item from Hynix since AMD is the only production customer for it.

 

[somebodyspecial]

I'm kind of confused by the driver used for NV cards. May 31 shows 353.06 WHQL and 353.30 WHQL June22. When exactly did you guys benchmark this card with a 352.90 beta? Or was there some reason you chose those over newer ones? I knew people tested products a bit before NDA but over a month early?

http://www.extremetech.com/gaming/208874-amds-radeon-fury-x-previewing-performance-power-consumption-and-4k-scaling/3
Extremetech says Metro LL can go up 25% for 353.30 drivers but they didn't get time to test it. They had very short time (gee, shocker, maximum pc seemed to insinuate it was done on purpose by AMD), very few benchmarks too so not sure if the driver affects others. I hope you use the newer drivers when you get your card for further testing since it would seem Nvidia is going after at least one game that appears to be a problem for them and some sites show it already reversed (likely different card or possibly not stock, or settings or something - IIRC another site had the same driver but Metro LL clear win for NV).

http://www.hardwarecanucks.com/forum/hardware-canucks-reviews/69682-amd-r9-fury-x-review-fiji-arrives-18.html
Maybe even Middle Earth is a toss up. Is toms using FXAA and ultra textures or not? 10% loss in 4K here.

Another I don't think I mentioned before techpowerup. Also note wccftech now saying pump whine went to users in first production units of FuryX. I guess AMD might have fudged a bit on that only in reviewers model info 😉
http://wccftech.com/amd-radeon-fury-x-reportedly-suffering-buzzing-coil-whine/
"AMD’s Antal Tungler has confirmed that the problem exists in early production units. However a fix (for the pump whine) has been applied by Cooler Master USA and it is hoped that the problem has been resolved for future R9 Fury X units."

Buyer beware. This kind of reminds me of retail 290x units that were not supposed to clock down but toms & others showed swings down into the 700's. Between beta drivers issues, lacking perf vs. 980ti (only really a duel at 4K and only then until you OC), whine shipped to users, coil whine shown at wccftech too from user vids (won't be fixed by cooler master fix, that is choke related not pump whine so how many have that?), lack of stock or even cards for reviewers etc, I say they failed another launch. But maybe just because they hyped everything so much (the cooler at 500w yet almost no overclocking using it, 980ti losing in a dozen games slide etc).

 

[Arabian Knight]

So, would you guys recommend the R9 390x or the GTX 980 for a upgrade from the HD7950?

GTX 980 ti (and not the 980) .... and wait for price drop.

 

[GObonzo]

seems to be a lot of fanboys again spouting how AMD can't compare and should be wary of the future. seriously doubt AMD is hurting or will be for some time since these "nex-gen" consoles are using their hardware. that's quite a bit more profit than just selling GPUs to us.

 

[ubercake]

seems to be a lot of fanboys again spouting how AMD can't compare and should be wary of the future. seriously doubt AMD is hurting or will be for some time since these "nex-gen" consoles are using their hardware. that's quite a bit more profit than just selling GPUs to us.

AMD is doing pretty badly and it's unfortunate:
http://money.cnn.com/quote/quote.html?symb=AMD

Also, there was speculation that Microsoft may buy AMD, but here's a good article regarding why they won't that will also give you an idea about the current state of AMD:
Why Microsoft Corporation Won't Buy Advanced Micro Devices Inc.

 

[InvalidError]

seriously doubt AMD is hurting or will be for some time since these "nex-gen" consoles are using their hardware. that's quite a bit more profit than just selling GPUs to us.

Console SoCs are low margin, low volume sales - if you put the PS4 and XBO sales together, we're talking only 10-15M chips per year with a gross margin in the neighborhood of $10 per chip. A company the size of AMD needs much more than 150M$/year in gross profit to stay afloat.

 

[mapesdhs]

seems to be a lot of fanboys again spouting how AMD can't compare and should be wary of the future. ...

Fanboys or not, they're right.

... seriously doubt AMD is hurting or will be for some time since these "nex-gen" consoles are using their hardware. that's quite a bit more profit than just selling GPUs to us.

Wow, boy do you ever have that wrong. As NOA explained when I visited them in 1995 re the N64, profits are made on the games, not the base consoles. InvalidError nailed it there.

Ian.

 

[Arabian Knight]

seems to be a lot of fanboys again spouting how AMD can't compare and should be wary of the future. ...

Fanboys or not, they're right.

... seriously doubt AMD is hurting or will be for some time since these "nex-gen" consoles are using their hardware. that's quite a bit more profit than just selling GPUs to us.

Wow, boy do you ever have that wrong. As NOA explained when I visited them in 1995 re the N64, profits are made on the games, not the base consoles. InvalidError nailed it there.

Ian.

Nintendo is the only company that profits from consoles and handhelds. and thats why their hardware allways sucks in comparison of same era consoles.

look at the Wii U ... my Phone is faster than it.

 

[mapesdhs]

A.R., I only mentioned Nintendo because that's a specific example I know about from personal experience of being somewhat involved at the time (I had the first ever web page on the N64, or Ultra64 as it was called before launch), but the same concept applies (the N64 isn't like the handhelds in that sense). How their main console devices compare to other models is not relevant; key point is, for conventional console designs, the money is in the game sales, not the console units. That's why SGI ended up with a sour taste from the design deal, the price drops meant they made very little in the end, though today we all gain from the legacy (many NV/AMD key people over the years have been ex-SGI, some still are).

Ian.

 

[jcchg]

There is a minor mistake on "Gaming Power Consumption": it's written 380X on the header instead of 380.

 

[somebodyspecial]

seems to be a lot of fanboys again spouting how AMD can't compare and should be wary of the future. seriously doubt AMD is hurting or will be for some time since these "nex-gen" consoles are using their hardware. that's quite a bit more profit than just selling GPUs to us.

Those consoles have barely double digit margins (according to AMD still) and don't even make enough in a year to cover their interest on their debt (that takes about 200mil+). They currently make the chip for ~$100-110 and make $10-12. AMD will only make REAL money on those deals if they get to multiple die shrinks late in this gens life (yrs 4-7) where margins get much better. Unfortunately all of mobile will kill them before they get to that and with HBM etc helping out might actually be able to REALLY do 1080p gaming at 10nm which makes consoles totally moot in ~2017 merely half through their life. That's only if consoles already aren't pointless far before then due to stuff like shield tv (think rev2 I guess, or any new console with this xmas 14nm versions from samsung and forward) amping up android gaming yearly.

When the unreal4/unity5 etc engine stuff starts hitting (and how easy it is to port from these to PC or PC to mobile) I would suspect we'll start to see some real quality on android a LOT more than today, and even today they have a LOT of stuff I want to play that is nothing like angry birds, minecraft etc. Consoles just don't sell enough volume, which is why NV rightly passed and AMD should have too! The console decision (robbing R&D from CPU/GPU for 5yrs) along with paying 3x the price for ATI killed them pretty much. Note MS lost 3B on the last gen and Sony lost 4B. That might be less this time due to shooting so low, but that is exactly why they'll die this time as mobile & socs grow so fast now in perf. Someone will put a heatsink/fan on one and put it in a box running it at 50+watts at some point here. I'm guessing it will happen with HBM in the next gen as that will really change the power inside a box (talking perf here). You could be looking at a REAL 1080P capable box for $300 with far more abilities on top too.

Consoles are dead for Nin/Sony/MS. That ship is sailed and gets worse once Vulkan hits making DX12 tough to sell, which is why MS is even giving the OS away this time. They may even end up with their OS being worthless for all but people who use it for pro apps and such if DX12/win10 don't do some real damage QUICKLY before ARM/Android/SteamOS (assuming ported to ARM soon)/linux get great gaming and simple porting for devs to any of these which they will eat up like candy vs. making console/DX12 PC games for a very small crowd with the high cost of porting everywhere else. Consoles will be last in the thought process vs. 2B units on mobile that will be quite capable at 14nm and beyond. Considering how fast Valve has ported over 1200 games to linux, ARM won't be a hard to jump from there for a SteamOS port. Vulkan will be the wave of the future for easy porting to anywhere along with highly sophisticated engines like Unreal4 making it as easy as possible too.

Note AMD has lost 6Billion in the last 12yrs, and 7B in the last 15yrs. They will lose money for all this year and 1/2 of next also which will probably end up in another FY loss for 2016 unless ZEN is BETTER than Intel and they can pump them out like mad between multiple fabs. The fabs part was an issue they couldn't overcome the last time they had a CPU king. But that will be different this time [IMHO] IF it's a winner as pumping them out won't be such an issue with possibly GF+TSMC and maybe even samsung if they copy GF process (both share any ideas now) to pump out higher watt larger chips like cpu/gpu. At this point it looks like only ZEN can help them back to profits and it can't come soon enough. Fury and current apu's will do nothing for their bottom lines as they have no pricing power with no perf kings (HBM causing shortage, if not the gpu itself, and both cause high price & low margins anyway). AMD should have just went GDDR5 for Fury cards and they might have been much better off, volume wise and able to price correctly while making profits. But still they don't sell enough of them to make money if CPU isn't helping the gpu side and even worse when cpu/apu is actually robbing profits (what little there is from console/gpu if any after interest on debt). I say this as a person who has been buying or selling their stock for 2 decades+ (jeez I'm getting old) and also owned a PC business for 8yrs so I know a little about margins on this stuff. They need to start charging more for products period. It's up to them to figure out how to do that and make money, preferably due to having a WINNER in something, with actual pricing power (maybe ZEN).

 

[cdrkf]

seems to be a lot of fanboys again spouting how AMD can't compare and should be wary of the future. seriously doubt AMD is hurting or will be for some time since these "nex-gen" consoles are using their hardware. that's quite a bit more profit than just selling GPUs to us.

Those consoles have barely double digit margins (according to AMD still) and don't even make enough in a year to cover their interest on their debt (that takes about 200mil+). They currently make the chip for ~$100-110 and make $10-12. AMD will only make REAL money on those deals if they get to multiple die shrinks late in this gens life (yrs 4-7) where margins get much better. Unfortunately all of mobile will kill them before they get to that and with HBM etc helping out might actually be able to REALLY do 1080p gaming at 10nm which makes consoles totally moot in ~2017 merely half through their life. That's only if consoles already aren't pointless far before then due to stuff like shield tv (think rev2 I guess, or any new console with this xmas 14nm versions from samsung and forward) amping up android gaming yearly.

When the unreal4/unity5 etc engine stuff starts hitting (and how easy it is to port from these to PC or PC to mobile) I would suspect we'll start to see some real quality on android a LOT more than today, and even today they have a LOT of stuff I want to play that is nothing like angry birds, minecraft etc. Consoles just don't sell enough volume, which is why NV rightly passed and AMD should have too! The console decision (robbing R&D from CPU/GPU for 5yrs) along with paying 3x the price for ATI killed them pretty much. Note MS lost 3B on the last gen and Sony lost 4B. That might be less this time due to shooting so low, but that is exactly why they'll die this time as mobile & socs grow so fast now in perf. Someone will put a heatsink/fan on one and put it in a box running it at 50+watts at some point here. I'm guessing it will happen with HBM in the next gen as that will really change the power inside a box (talking perf here). You could be looking at a REAL 1080P capable box for $300 with far more abilities on top too.

Consoles are dead for Nin/Sony/MS. That ship is sailed and gets worse once Vulkan hits making DX12 tough to sell, which is why MS is even giving the OS away this time. They may even end up with their OS being worthless for all but people who use it for pro apps and such if DX12/win10 don't do some real damage QUICKLY before ARM/Android/SteamOS (assuming ported to ARM soon)/linux get great gaming and simple porting for devs to any of these which they will eat up like candy vs. making console/DX12 PC games for a very small crowd with the high cost of porting everywhere else. Consoles will be last in the thought process vs. 2B units on mobile that will be quite capable at 14nm and beyond. Considering how fast Valve has ported over 1200 games to linux, ARM won't be a hard to jump from there for a SteamOS port. Vulkan will be the wave of the future for easy porting to anywhere along with highly sophisticated engines like Unreal4 making it as easy as possible too.

Note AMD has lost 6Billion in the last 12yrs, and 7B in the last 15yrs. They will lose money for all this year and 1/2 of next also which will probably end up in another FY loss for 2016 unless ZEN is BETTER than Intel and they can pump them out like mad between multiple fabs. The fabs part was an issue they couldn't overcome the last time they had a CPU king. But that will be different this time [IMHO] IF it's a winner as pumping them out won't be such an issue with possibly GF+TSMC and maybe even samsung if they copy GF process (both share any ideas now) to pump out higher watt larger chips like cpu/gpu. At this point it looks like only ZEN can help them back to profits and it can't come soon enough. Fury and current apu's will do nothing for their bottom lines as they have no pricing power with no perf kings (HBM causing shortage, if not the gpu itself, and both cause high price & low margins anyway). AMD should have just went GDDR5 for Fury cards and they might have been much better off, volume wise and able to price correctly while making profits. But still they don't sell enough of them to make money if CPU isn't helping the gpu side and even worse when cpu/apu is actually robbing profits (what little there is from console/gpu if any after interest on debt). I say this as a person who has been buying or selling their stock for 2 decades+ (jeez I'm getting old) and also owned a PC business for 8yrs so I know a little about margins on this stuff. They need to start charging more for products period. It's up to them to figure out how to do that and make money, preferably due to having a WINNER in something, with actual pricing power (maybe ZEN).

Consoles are doing very very well this gen:
http://venturebeat.com/2015/05/05/ps4-overtakes-ps2-as-the-u-k-s-fastest-selling-sony-console/

Anyone who thinks that 'console market is tiny, is about to die off and has no future' really need to *look into the facts* 😛

The things that make consoles sucessful in my opinion- fixed hardware that is easy to develop for, easy of use (no fiddling with drivers / settings) and big title games to keep people interested. The current gen does all 3, and as a result is selling like hot cakes.

PC gaming has been ahead of consoles for years but it made little difference. I also don't see mobiles / tablets overtaking consoles (yes the graphics are getting up to scratch, but touch screen controls are horrid). The only other option is that as you say an android based console comes out that starts to steal away market share from the big names. The counter to that though is there are a wide range of exclusive titles only available on the big platforms and those wont get ported to an android based competitor no matter how capable the hardware.

I think something like the Shield console could work longer term, but it's going to take a while for it to get established and at the end of the day it's in of itself another console. Also this business of 'Microsoft and Sony both lost money on last gen' isn't true. They lost money at first but *both* companies made an overall profit on the last gen consoles by the end of their life spans. This time around they're not loosing money on them at all, which suggests they might be thinking of upgrading them more quickly (in the event a serious threat of a more powerful platform emerges).

I agree, consoles on their own aren't going to secure AMD's future, however the reason nVidia didn't work on them has *nothing* to do with them not wanting to, and *everything* to do with the fact that there is only 1 company in the world that has both an x86 licence *and* high end GPU tech (AMD). There was no other choice for an integrated x86 based solution.

 

[somebodyspecial]

seems to be a lot of fanboys again spouting how AMD can't compare and should be wary of the future. seriously doubt AMD is hurting or will be for some time since these "nex-gen" consoles are using their hardware. that's quite a bit more profit than just selling GPUs to us.

Those consoles have barely double digit margins (according to AMD still) and don't even make enough in a year to cover their interest on their debt (that takes about 200mil+). They currently make the chip for ~$100-110 and make $10-12. AMD will only make REAL money on those deals if they get to multiple die shrinks late in this gens life (yrs 4-7) where margins get much better. Unfortunately all of mobile will kill them before they get to that and with HBM etc helping out might actually be able to REALLY do 1080p gaming at 10nm which makes consoles totally moot in ~2017 merely half through their life. That's only if consoles already aren't pointless far before then due to stuff like shield tv (think rev2 I guess, or any new console with this xmas 14nm versions from samsung and forward) amping up android gaming yearly.

When the unreal4/unity5 etc engine stuff starts hitting (and how easy it is to port from these to PC or PC to mobile) I would suspect we'll start to see some real quality on android a LOT more than today, and even today they have a LOT of stuff I want to play that is nothing like angry birds, minecraft etc. Consoles just don't sell enough volume, which is why NV rightly passed and AMD should have too! The console decision (robbing R&D from CPU/GPU for 5yrs) along with paying 3x the price for ATI killed them pretty much. Note MS lost 3B on the last gen and Sony lost 4B. That might be less this time due to shooting so low, but that is exactly why they'll die this time as mobile & socs grow so fast now in perf. Someone will put a heatsink/fan on one and put it in a box running it at 50+watts at some point here. I'm guessing it will happen with HBM in the next gen as that will really change the power inside a box (talking perf here). You could be looking at a REAL 1080P capable box for $300 with far more abilities on top too.

Consoles are dead for Nin/Sony/MS. That ship is sailed and gets worse once Vulkan hits making DX12 tough to sell, which is why MS is even giving the OS away this time. They may even end up with their OS being worthless for all but people who use it for pro apps and such if DX12/win10 don't do some real damage QUICKLY before ARM/Android/SteamOS (assuming ported to ARM soon)/linux get great gaming and simple porting for devs to any of these which they will eat up like candy vs. making console/DX12 PC games for a very small crowd with the high cost of porting everywhere else. Consoles will be last in the thought process vs. 2B units on mobile that will be quite capable at 14nm and beyond. Considering how fast Valve has ported over 1200 games to linux, ARM won't be a hard to jump from there for a SteamOS port. Vulkan will be the wave of the future for easy porting to anywhere along with highly sophisticated engines like Unreal4 making it as easy as possible too.

Note AMD has lost 6Billion in the last 12yrs, and 7B in the last 15yrs. They will lose money for all this year and 1/2 of next also which will probably end up in another FY loss for 2016 unless ZEN is BETTER than Intel and they can pump them out like mad between multiple fabs. The fabs part was an issue they couldn't overcome the last time they had a CPU king. But that will be different this time [IMHO] IF it's a winner as pumping them out won't be such an issue with possibly GF+TSMC and maybe even samsung if they copy GF process (both share any ideas now) to pump out higher watt larger chips like cpu/gpu. At this point it looks like only ZEN can help them back to profits and it can't come soon enough. Fury and current apu's will do nothing for their bottom lines as they have no pricing power with no perf kings (HBM causing shortage, if not the gpu itself, and both cause high price & low margins anyway). AMD should have just went GDDR5 for Fury cards and they might have been much better off, volume wise and able to price correctly while making profits. But still they don't sell enough of them to make money if CPU isn't helping the gpu side and even worse when cpu/apu is actually robbing profits (what little there is from console/gpu if any after interest on debt). I say this as a person who has been buying or selling their stock for 2 decades+ (jeez I'm getting old) and also owned a PC business for 8yrs so I know a little about margins on this stuff. They need to start charging more for products period. It's up to them to figure out how to do that and make money, preferably due to having a WINNER in something, with actual pricing power (maybe ZEN).

Consoles are doing very very well this gen:
http://venturebeat.com/2015/05/05/ps4-overtakes-ps2-as-the-u-k-s-fastest-selling-sony-console/

Anyone who thinks that 'console market is tiny, is about to die off and has no future' really need to *look into the facts* 😛

The things that make consoles sucessful in my opinion- fixed hardware that is easy to develop for, easy of use (no fiddling with drivers / settings) and big title games to keep people interested. The current gen does all 3, and as a result is selling like hot cakes.

PC gaming has been ahead of consoles for years but it made little difference. I also don't see mobiles / tablets overtaking consoles (yes the graphics are getting up to scratch, but touch screen controls are horrid). The only other option is that as you say an android based console comes out that starts to steal away market share from the big names. The counter to that though is there are a wide range of exclusive titles only available on the big platforms and those wont get ported to an android based competitor no matter how capable the hardware.

I think something like the Shield console could work longer term, but it's going to take a while for it to get established and at the end of the day it's in of itself another console. Also this business of 'Microsoft and Sony both lost money on last gen' isn't true. They lost money at first but *both* companies made an overall profit on the last gen consoles by the end of their life spans. This time around they're not loosing money on them at all, which suggests they might be thinking of upgrading them more quickly (in the event a serious threat of a more powerful platform emerges).

I agree, consoles on their own aren't going to secure AMD's future, however the reason nVidia didn't work on them has *nothing* to do with them not wanting to, and *everything* to do with the fact that there is only 1 company in the world that has both an x86 licence *and* high end GPU tech (AMD). There was no other choice for an integrated x86 based solution.

You need to read financial reports for MSFT and SONY. Faster selling than ps2 means nothing. I never said NV didn't WANT to do them, just that Jen said doing them would ROB from R&D needed on CORE products so they WISELY passed on them. Which we see has clearly played out on AMD's CPU (gave up race until now with ZEN) and GPU (market share tanked, NV owns 77%, AMD launch issues 290x/furyx etc).

http://games.softpedia.com/blog/Microsoft-Reports-400-298-Million-Loss-on-the-Xbox-One-454321.shtml
400mil loss so far with 11.7mil units sold for MS. See the last decade of their financial reports for the entertainment divisions losses. Same can be done for Sony.

http://www.escapistmagazine.com/news/view/129463-Microsoft-Loses-2-Billion-Per-Year-On-Xbox-Analyst-Says
2-2.5B loss per year after removing the billions from android fees.

http://www.gamesindustry.biz/articles/2013-11-07-huge-xbox-losses-hidden-by-patent-royalties-says-analyst
The actual article end of 2013.

http://www.vg247.com/2013/01/07/xbox-360-and-ps3-losses-total-8-billion-ex-sony-employee-paints-grim-future/
Ben's article doesn't have the chart anymore unfortunately (at kotaku, on googledocs so maybe they killed it?..LOL).

I could go on, but you should get the point. Elop has even said he'd dump the division given the opportunity, but now he's gone as of last month.

https://webcache.googleusercontent.com/search?q=cache:F4WYE2lnDigJ:www.forbes.com/sites/adamhartung/2014/02/18/microsoft-should-give-xbox-one-to-nintendo/+&cd=3&hl=en&ct=clnk&gl=us
Regarding Sony's side:
"The new sales leader is Sony, but that should give no one reason to cheer. Sony lost money for 4 straight years (2008-2012), and was barely able to squeak out a 2013 profit only because it took a massive $4.6B 2012 loss which cleared the way to show something slightly better than break-even. Now S&P has downgraded Sony’s debt to near junk status."

"Whether Sony will make money on PS4 in 2014 is far from proven. Especially since it sells for $100/unit (20%) less than Xbox One – which compresses margins. What investors (and customers) can expect is an ongoing price war between Nintendo, Sony and Microsoft to attract sales. A competition which historically has left all competitors with losses – even when they win the market share war."

It gets worse, but you get the main point. Note now they have a price war against Android devices, iOS etc via mobile and of course more consoles yearly from that side. $60 games are a REALLY tough sell vs. $2-20 on android and that side sells 2BILLION units yearly now. If they don't get to have the CASUAL years of 4-7 (hardcore have been proven to be the buyers in yr1-3), even at lower cost hardware this is a loss.

Who would you dev for? 2Billion units next year that have X1 or above gpus, or 20-30mil consoles that have $60 game price tags? Especially if you have to wait 5-7yrs to get good profits from unit sales of your game. Only the big guys can wait for that. A $10 game is much easier to sell than $60 right (insert any number for that $10). See GDC 2013/2014/2015 surveys for where devs went already. PC 52%, mobile 50% and far less for consoles (percent of people making something for these things).

The world is bigger than the UK and 2mil units is a joke. That is just 2% of mobile units sold YEARLY now and even if you just sell a $10 game to 2% of mobile next year (when at 14nm everyone should have X1 level gpu, faster than xbox360/ps3), you sell 2mil x $10, for 20mil revenue, of which google, apple, take their 1/3 leaving a dev with ~13mil. This is a profit maker even for games like Torchlights, Trine's, Divinity Original Sin, Pillars of Eternity, Shroud of Avatar, Wasteland 2, etc etc etc...None of those were made for more than 10mil, heck Witcher1-2 were made for $7Mil each (not including consoles, which pushed it to ~15mil or so). So you can make really great stuff (if not EA etc, huge overhead) and make profits on mobile. You have to have a pretty big hit on consoles to make money until in the later years of their cycle when they have 80-100mil units out there on each one.

The whole point of the new consoles is gamepads (including your xbox gamepads which work). Same for many android devices. They realize gaming needs to be adapted to a gamepad and output to TV/Monitor etc. You are not limited to the tablet or phone. NV has game mapping software, so you can just load up what you need per game if desired. New IP will be made with a gamepad hookup in mind, and likely with output to tv in mind too. Add to that stuff like GRID and you have a compelling sales pitch from now forward. You gain nothing if you have to upgrade xbox1/ps4 at 3yrs, as devs will ignore the platform with <50mil units to sell to (and right now far lower for each side). IF they put out a new on, current owners and devs making stuff for those will be mightily angry. They are still deving for xbox360/ps3 due to 100mil each on the streets (some times not including ps4/xbox1).

Let me know how UK sales look next year, after hardcore are done (who just got access...LOL). Your own link shows my point already:
"With mobile gaming revenue expected to exceed console gaming revenue this year, it’s important for Sony and Microsoft — who’ve both invested heavily in their current hardware platforms — to build and sustain sales momentum."



Note PC gaming revenue already exceeds console revenue now. Really, start reading financial reports and understand exactly what is going on. The devil is in the financial details of each console player. While you're at it check Nintendo too. Their reports show they need to quit hardware and make android/pc games to stop bleeding. You also need to consider the cost of the actual console is NOT the whole bill. Marketing, R&D to make it, etc adds up and must be recovered too. Again, I could go on, but if you're not reading balance sheets, FY reports etc, what is the point? Your single 2mil units sold in UK data point means nothing to me. 😉 Profits are the only thing that counts here and that is even worse given the massive mobile competition (consoles based on socs too) with cheap games and massive units sold yearly for devs to aim at for profits that don't take 7yrs to get, which can easily bankrupt a small dev. The real race is about to heat up as everyone moves to 14nm for xmas and beyond, and devs have a pretty good idea of what they're working with hardware wise on the entire group.

Like I said they can basically aim at X1 for any next xmas game and most can play them easily by then (for any dev starting development now on a new game or already in dev). There will probably be a few hundred mil units with X1 for xmas this year too. NV releasing a 14nm Finfet xmas soc from samsung fabs, qcom & apple having the same also coming up shortly. Between that group they will have a few hundred mil units using X1+ gpu levels by mid year 2016 and worst case, a year later another 2Billion doing it (not a long wait for more sales). Surely a console is coming with HBM2 at some point and hopefully larger box with ~100w, which could really change android side (shield tv rev2 next xmas? 2016) especially as they add more good games (ported or new IP) up to xmas 2016. If NV wins the lawsuit, things get even easier for devs assuming everyone has to license NV off the shelf GPU IP going forward. You'd end up with devs working with the exact tech on desktops as on mobile then. There is no "good" way I see xbox1/ps4 having a good time going forward profits wise.

AMD Another 181mil loss today, gpu revenue down 54%. Just a note, as I've been reading Intel/AMD financial reports from today/yesterday. Intel's quarter was crap too, but hiding it with tax shifts etc. 10nm delayed into H2 2017, so 3yrs on 14nm, tick tock 1yr off schedule now. Just a note for stock buffs.

 

[somebodyspecial]

One more point, since PC sales weren't down 54% (see Intel's quarterly report), I'd say NV stole more share probably crossing 80% discrete now or pretty close. I should have snipped the other post, now that I see how long it was...LOL. Oh well. Also, I expect to see exclusives for android too, or at least exclusive for some time period, then released for PC (as a counter to pure CONSOLE exclusives for xbox1/ps4, which there are few anyway this gen).

 

[InvalidError]

I never said NV didn't WANT to do them, just that Jen said doing them would ROB from R&D needed on CORE products so they WISELY passed on them. Which we see has clearly played out on AMD's CPU (gave up race until now with ZEN) and GPU (market share tanked, NV owns 77%, AMD launch issues 290x/furyx etc).

I would not put so much stock behind console SoCs severely hurting AMD's GPU and CPU R&D since they are basically souped-up APUs with some client-specific extras. The bulk of the design is off-the-shelf and proven in existing products, the only significant nugget of R&D is customer extras and checking that the integrated design behaves as expected, most of which also done using existing testbenches.

And some of what they learned along the way likely got backported into plans for their other products to make future products work better.

This is a bit like people laughing at Intel for "wasting" billions of dollars in the mobile segment where they are having trouble gaining traction. Even if the mobile R&D fails to produce successful consumer products, a fair amount of the results will still be useful for desktop, server and other products.

 

[10tacle]

Regarding consoles vs. PC, disturbingly, I am seeing a trend of young kids using their tablets more for gaming than their Playstations and Xboxes. Maybe it's just anecdotal with my experience, but between friend's kids and my own niece and nephew and extended relative's kids, when they are together they play tablets on sofas while the poor consoles sit there alone and un-played. I asked one of them why he prefers to play Minecraft on his tablet instead of on his XBone. He simply stated that it's "less hassle" which I assume means all he has to do is swipe the screen on the tablet instead of having to get up, insert a game DVD, turn on the box, and use a controller. Pretty sad.

 

[teknobug]

Regarding consoles vs. PC, disturbingly, I am seeing a trend of young kids using their tablets more for gaming than their Playstations and Xboxes. Maybe it's just anecdotal with my experience, but between friend's kids and my own niece and nephew and extended relative's kids, when they are together they play tablets on sofas while the poor consoles sit there alone and un-played. I asked one of them why he prefers to play Minecraft on his tablet instead of on his XBone. He simply stated that it's "less hassle" which I assume means all he has to do is swipe the screen on the tablet instead of having to get up, insert a game DVD, turn on the box, and use a controller. Pretty sad.

Mobile gaming is gaining A LOT of ground, I personally cannot stand mobile gaming, it's not for me. Even when I'm on a long Greyhound bus ride, I can't play a mobile game on my phone or table for very long.

 

[logainofhades]

I can't play a mobile game on my phone or table for very long.

I can, depending on the game. I am a bit of an old school Final Fantasy fan, and Final Fantasy Record Keeper can keep me occupied for quite awhile, when I have nothing better to do. Keep meaning to install the NES and SNES emulators, on my tablet, so I can play the old FF games. The like the old dragon warriors and breath of fire games as well.

 

[ubercake]

I can't play a mobile game on my phone or table for very long.

I can, depending on the game. I am a bit of an old school Final Fantasy fan, and Final Fantasy Record Keeper can keep me occupied for quite awhile, when I have nothing better to do. Keep meaning to install the NES and SNES emulators, on my tablet, so I can play the old FF games. The like the old dragon warriors and breath of fire games as well.

I like the good old dice, board and card games on the tablet as well.