What will be AMD'S next Move?

[Silverforce11]

As to huge amount of draw-calls in turn based Civ(another terrible Mantle candidate),
first thats ridiculous programming at best, intentional crippling move at worst.

A Mantle discussion again? Why not.. bit slow here anyway..

Do you ever scroll around the global map in your turn based game? I'm 100% sure smooth 60 fps is appreciated. Then there's also this matter of heavy CPU usage for AI calculations, which currently stutter like mad due to the CPU being crunched. With Mantle, the overhead for graphics is reduced so that you can still have smooth high fps graphics during AI turns.

It would be a completely different gaming experience.

 

[f1sherman]

smooth camera panning appreciated - yes.
huge number of draw calls needed in TBS - hardly

AI calculations? Please... this is not Deep Blue vs Kasparov. I can do those calculations on my fingers.
Also, what intensive AI calculations are done while panning camera??

But yeah, I see what you're saying.
Less CPU overhead is always appreciated, even if the cause of it is bad programming.
And I won't complain one bit, once Civ Mantle is out

 

[Silverforce11]

You truly speak like one with no appreciation for good AI, which is at the heart of a strategy game.

It is extremely complex and indeed the major bottleneck. But if you are gonna spout "I can do those calculations on my fingers" then you won't ever appreciate what Mantle can achieve for strategy games. You'll just have to wait and see.

 

[RussianSensation]

Some people wont Admit though but Nvidia can play a waiting game even 6 to 7 months if they want and it wont hurt there sale but AMD cannot because there position is not strong like Nvidia in the market so my advice is that u can just Hype ur next product even if it is coming next year so some people can get there hopes up again.

You mean how 5870 crushed 285, then 480 beat 5870, then how 7970 crushed 580 and 680 beat a 925mhz 7970? It has nothing to do with hyping but how the GPU industry works. No one denies at the awesome value of 970 but 980 at $549 only beats 290X by 15-22%. You honestly believe AMD cannot beat that card ever?

I know some people here are quick to who claim that AMD needs a 350-400W liquid cool card but can't comprehend that AMD can strategically delay their next flagship to 1H 2015 because they might just get it out on 20nm node. Alternatively, waiting another 6 months could allow AMD to produce a 500mm2+ die due to even lower 28nm wafer prices. These choices are better than rushing a 10% higher clocked 290X and getting burned in reviews. Also, it gives AMD another 6 months to work on PowerTune/internal ASIC efficiency as from Tom's Hardware analysis it seems NV figured out that Maxwell is hardly more efficient than Kelper at max load which suggests it's NV's effective power tune technology and ability to turn off non-gaming functioning units in Maxwell which brought the real breakthrough in performance/watt in games. AMD needs to visit this aspect and if they need 5-6 months to get it correctly, then hopefully it's not just 6 months wasted.

On the point of LC, liquid cooling adds a premium aspect to the card, not some kind of a negative as some people here like to spin it as. Think about how many people here buy water blocks and now you'll get a quiet reference card with LC. I've been wanting this for 5 years from AMD/NV. 295X2 proves that LC is superior to air cooling as it runs cooler and quieter than the much lower power consuming but prone to throttling Titan Z. And we also know that despite how good a blower is, after-market 970/980 cards show a 15-20C decline in temperatures and superior noise levels to a reference NV blower.

A LC card does 3 things right away:

1) Solves temperature issues as hardly any air cooled heatsink can compete with an aftermarket 120mm LC solution. Even 600W Gigabyte Windforce 3X will probably not keep up with a single 1x120mm LC.

2) Solves noise issues as LC results in a 25-35C drop in temperatures and as a bonus handles overclocking loads much better due to stronger heat dissipation vs. air cooled solution as long as VRM temps are under control.

3) You get to fully exhaust 250-300W of power out of your case. That means you can buy 2 LC cards and again exhaust almost all the heat out of your case. All of a sudden the size of the case is not as much of a limitation. Right now you either have to get loud and inefficient for overclocking reference blowers from either camp OR have to get resort to 2 open air cooled cards that requires a medium sized case at least. Right now I can't install 2 MSI Gaming 970s in a spanking new Lian Li V359 MicroATX case. Since these newer smaller cases can fit 2x 240 mm radiators, suddenly I can put 2 reference LC cards with 2x120mm blocks on one side of this case and a 1x240 mm CPU cooler like Swiftech X220 on the other. Problem solved!! All of a sudden flagship CF/SLI is doable in a MicroATX without me having to spend $100s of dollars on custom water cooling loops.

Moving to LC for reference cards, if executed correctly, is an epic win since I will be finally able to go dual GPU and X99 beast mode OC in MicroATX case like the 359 without having to worry of thermal throttle or loud reference blower noise.

Obviously NV will bring Maxwell flagship but after Fiji, AMD has Bermuda. So the competition is only getting started from both despite your insinuations that AMD is more or less done for. I think what you missed is how your triple 780Tis couldn't even beat 3x290s at 4K and multi-monitor gaming but you still bought NV for thousands of dollars more ($1,200 for triple 290s vs. $2,100 for triple 780Tis), which honestly makes me question if you even care what AMD will bring in 2015.

 

[CrazyElf]

But you know what people in this thread who claim doom and gloom for AMD don't get? If I am upgrading for only a 40% increase from 7970 OC to 970 OC or from dual 7970s OC to dual 970s OC, that means I will be just as likely to 'waste' $$$ for another 40% increase after 970/970s. In other words in 1.5 years I will be upgrading again. On the other hand, if I want 2x the performance, 970 and 980 fail at giving me that at 7970's price nearly 3 years later, which leaves me waiting to 2015.

In the first case it means even if I switch to NV, in 1.5 years AMD and NV are competing for another upgrade from me OR 970/980 did nothing to make me move away from 7970s.

My point is the biggest incentive to upgrade are next gen PC games. Otherwise you'll just capture the hardcore segment in the otherwise shrinking dGPU market. Therefore, the only way to claim death for AMD is if they will never release anything better than 290X. Who believes that?

At the moment, there isn't much that really demands a lot of upgrading in terms of games. Many games are shallow console ports as well.

Performance wise, only people who need to upgrade are:
1. Multi-monitor setups
2. 4k gamers
3. Potentially certain mods

But in terms of raw performance, the 980 and 970 do not offer much over Kepler (780Ti) and Hawaii (290 or 290X).

Anyone who is upgrading for a 980 is a hardcore buyer who "must" have the newest and coolest.

Under such a situation, a better option might be to add another GPU to the setup, taking advantage of the price adjustments.

There is one other issue right now. The performance per generation is not what it once was. Even hardcore is going to have to rationalize upgrading in the future.

There are different kind of buyers.

The enthusiast that just buy whatever is new. This one was the one that bought Tahiti at launch prices, then switched to 670/80s. This one gives almost 50/50 of their attention to either company.

The loyal fan that only buys one GPU maker. Like Russian said, the NV packs quite a lot more in this group that AMD, whose fans are actually perf/price fans most of the time (the same people that is buying 970's at that killer price)

And the dude that buys every major performance improvement from either company, usually buys only 1 gpu from each manufacturing process and stick with it until it is no longer suitable for their gaming requeriments (if the budget allows).

Performance is important for all 3 of them. Perf/watt is only really important for the second group when it favours their GPU brand of choice (just see this thread to confirm this) and it might be a factor for the third group, mosty because of budget reasons (PSU reqs, etc). Perf/price is only important for the third group mostly, the first group coming second, the middle group just doesnt care about perf/price and are just willing to let go a killer oportunity in perf/price just because it wasnt made by their GPU brand of choice

I would argue that a disproportionate number of the buyers appear to be in the second category - Nvidia buyers especially.

They seem to want to rationalize their choice of GPU to us. :whiste: Some will insist that the other side is "certainly doomed".

Power-efficiency is an example of that. Power I think only becomes an issue if your GPU gets to the point where it begins to throttle. Granted this does seem to happen with Crossfire Hawaii chips and sometimes 780Ti SLI chips on air. Otherwise, the only thing noticeable will be a somewhat higher monthly electricity bill, and perhaps a warmer room during gaming sessions. In the winter, that might actually be a good thing, but it's bad in the summer.

I think I am a mixture of 1 and 3 to be honest.

The smart thing to do, imo:

1. Wait for the "big" Maxwell and AMD's chips to come out.
2. Look at performance, price, and as a distant third, power consumption.
3. Consider other factors, such as multi-GPU scaling, your long term plans (ex: are you planning to water cool, which can affect other things, such as your purchasing custom PCBs). I also like to research who has the best PCB. DirectFET Mosfets for example with many phases are well worth the money, does not use Elpida VRAM (overclocks poorly), and a few other things. I am willing to pay a premium for a good third party PCB and a good cooling solution.
4. Buy whomever has the best overall solution in terms of value for the money.
5. Upgrade carefully based on your research.
6. Sometimes there are great deals too, like the mining collapse which led to a lot of Hawaii chips flooding the market. The ASICs are taking over mining.

I think that the third category of upgrade person is going to have to wait more and more for upgrades in the future now that performance per generation is stagnating.

I wonder what will happen in the future especially because 28nm was the cheapest in cost per transistor.

 

[RussianSensation]

Power-efficiency is an example of that. Power I think only becomes an issue if your GPU gets to the point where it begins to throttle.

Ya, I don't remember even 1 NV user on our forum complaining non-stop that a reference 780Ti uses 269W and EVGA Classy 780Ti uses 286W at load (even more than Fermi 480!).

They throw around the power consumption metric when it suits them and ignore it if NV is ahead to justify a 15-20% increase in performance despite major NV price tax like 480 vs. 5870 or 780Ti vs. after-market 290/290s.

If AMD released a card 20% faster than 980 and it used 100W more power, almost no one who wants a flagship $500+ card would trade 20% of performance for a 100W drop in power on an enthusiast overclocked i5/i7/X99 rig to begin with. It's not as if 180-185W 980 uses is somehow a 60W card.

I think that the third category of upgrade person is going to have to wait more and more for upgrades in the future now that performance per generation is stagnating.

I wonder what will happen in the future especially because 28nm was the cheapest in cost per transistor.

It's a good list you created. I wouldn't tell people who want to play with a 970/970s to wait another 6-12 months because upgrading can be fun and you can resell your old hardware while it has value. At the same time I agree with your points that there just aren't enough games that require us to upgrade unless you have multiple monitors or 4K. I really don't get the side-grade from 290/290X/780 to 970. It makes no sense to me unless you can do so with minimal outlay in cash after selling your old card.

Certainly this is not going to be the last major wave of PC upgrades. A lot of PC gamers will upgrade once Skylake and BW-E launches as well. In the grand scheme of things, R9 290 $399 embarrassed NV's $1K titan and $650 780 but 970 that barely beats a 290 only undercut a 1 year old 290 by $70!

This is just another double standard of NV users that I keep pointing out that they almost always fail to acknowledge that when AMD brings out new cards, it produces major movements in performance/$. 970 is only a big deal because it's the first card to do this from NV since GTX460. But when AMD brings major changes in price/performance, it's as if NV users don't even notice because let's face it they wouldn't even take a 290X for $200 vs. $700 780Ti :whiste:

 

[RussianSensation]

But yeah, I see what you're saying.
Less CPU overhead is always appreciated, even if the cause of it is bad programming.
And I won't complain one bit, once Civ Mantle is out

Seriously? It's a known fact that strategy games are some of the most CPU intensive and sensitive to CPU scaling. Strategy games not FPS games are the primary reason why we want high IPC cores. Most FPS games hardly scale from i7 920 @ 4.0ghz vs. i7 4790K @ 4.4Ghz. Just look at BFG's review of 2500K vs. 4790K where he predominantly tests FPS games and they hardly show a difference. Yet in a game like Civ 5 or Starcraft 2, the CPU dependence is huge. Civ Beyond Earth is one of the best candidates for Mantle since you want to alleviate the CPU bottleneck in a Civ game.

Mantle is going to allow those with lower end CPUs to enjoy games like Civ without resorting to spending $300+ on a Core i7. But yes if you have a Core i7 2600K OC or faster, then Mantle doesn't matter as much.

 

[CrazyElf]

It's a good list you created. I wouldn't tell people who want to play with a 970/970s to wait another 6-12 months because upgrading can be fun and you can resell your old hardware while it has value. At the same time I agree with your points that there just aren't enough games that require us to upgrade unless you have multiple monitors or 4K. I really don't get the side-grade from 290/290X/780 to 970. It makes no sense to me unless you can do so with minimal outlay in cash after selling your old card. Certainly this is not going to be the last major wave of PC upgrades. A lot of PC gamers will upgrade once Skylake and BW-E launches as well. In the grand scheme of things, R9 290 $399 embarrassed NV's $1K titan and $650 780 but 970 that barely beats a 290 only undercut a 1 year old 290 by $70!

Strictly speaking from a price to performance standpoint, if you own a 290, 290X, Titan, 780, or 780Ti, the best option right now is to buy other people's used cards, considering they are getting the 970 and 980.

I think there's also a chance that the remaining mining cards (290s, 290Xs, and perhaps a few other models like the 7970/7950/280X/280) might be forced down even more in price by the emergence of the 970 and 980.

Only drawback I can see to the suggestion above is if you don't have a PSU big enough to take it.

Edit: If water cooling, buy used waterblocks as well. Heck, some enthusiasts will no doubt want to get rid of their waterblocks for their old GPUs as well, so you might get a deal there.

But as I said, personally, the big turnoff is the lack of games that can demand the hardware.

What we need:
1. More games that really push the envelope in graphics and gameplay
2. Good 4k monitors; what's really needed is an affordable IPS SST 4k monitor with a good response rate (possible in 2015)

Then, we might see a better justification for upgrades.

 

[RussianSensation]

What we need:
1. More games that really push the envelope in graphics and gameplay

Or more poorly optimized Ubisoft/EA/Activision "next gen" console ports with 2010-2012 graphics that still can't beat Crysis 3 but require us to get 4GB GPUs to hit 60 fps at 1080P.

I mean just imagine how demanding GTA V will be and how crappy it will look stock. :whiste:

2. Good 4k monitors; what's really needed is an affordable IPSSST 4k monitor with a good response rate (possible in 2015)

I don't see that happening until 2016-2017 at which point 290/970 is a low-end card.

 

[raghu78]

AMD's response should be strong as they must have foreseen the GTX 980 and GTX 970 when Nvidia launched GTX 750 Ti. I think AMD will have a response in Q1 2015 based on a next gen improved GCN 2.0 architecture.

AMD's APU roadmap shows a few features which are quite interesting

http://www.anandtech.com/show/7974/...hitecture-a10-micro-6700t-performance-preview

The inter frame power gating feature is one which could be present in GCN 2.0. Since AMD APUs derive their GPU architecture from their discrete GPUs and typically lag by a generation its quite logical that GCN 2.0 has inter frame power gating.

AMD's GCN 2.0 could also sport a more flexible scalar unit for significant performance gains

http://people.engr.ncsu.edu/hzhou/ipdps14.pdf

http://www4.ncsu.edu/~yyang14/

Future compute shader concurrent multithreading

http://forums.anandtech.com/showpost.php?p=36259210&postcount=188

AMD is likely to move to a completely compute based graphics pipeline and do away with fixed function graphics hardware.

Finally AMD is likely to introduce HBM with their next gen flagship GPUs. AMD has traditionally been the driver of graphics memory standards and is the first to market. AMD got GDDR5 out on HD 4870 around 2 years before Nvidia did with the GTX 480.

 

[3DVagabond]

Some people wont Admit though but Nvidia can play a waiting game even 6 to 7 months if they want and it wont hurt there sale but AMD cannot because there position is not strong like Nvidia in the market so my advice is that u can just Hype ur next product even if it is coming next year so some people can get there hopes up again.

No doubt nVidia has more loyal fans, but if AMD were to answer GM204 and up the performance metric, nVidia couldn't afford to go 6-7 months without responding without suffering. By then you'll have more Mantle games with Mantle out of beta and likely FreeSync monitors. nVidia would bleed market share severely.

 

[Silverforce11]

Seriously? It's a known fact that strategy games are some of the most CPU intensive and sensitive to CPU scaling.

It's a very well know fact indeed. Which is why when someone claims Mantle won't benefit strategy or turn based strategy (my favourite genre of games), I immediately know they have no clue what they are talking about and disregard their opinions entirely.

The fact is the people working on Civ Beyond Earth cannot say enough good things about Mantle.

 

[3DVagabond]

It's a very well know fact indeed. Which is why when someone claims Mantle won't benefit strategy or turn based strategy (my favourite genre of games), I immediately know they have no clue what they are talking about and disregard their opinions entirely.

The fact is the people working on Civ Beyond Earth cannot say enough good things about Mantle.

Has it really taken you this long to realize that?

 

[RussianSensation]

It's a very well know fact indeed. Which is why when someone claims Mantle won't benefit strategy or turn based strategy (my favourite genre of games), I immediately know they have no clue what they are talking about and disregard their opinions entirely.

The fact is the people working on Civ Beyond Earth cannot say enough good things about Mantle.

Well in fairness maybe he didn't know that turn-based strategy is just as CPU sensitive as real time strategy.

Anyway, on another note, in the short term the price cut of R9 290 to $299 and 290X to $399 is not going to cut it when 970 is $330-370. And now 780 is dropping to $310 to $320 with a $60 BL: P game, while 780Ti overclocked is dropping to $430 to $450.

AMD needs to suck it up and drop R9 285 to $179, R9 280X to $219, R9 290 to $269 and R9 290X to $299. I realize these 780/780Ti prices are just fire sale on 780Ti but now that gamers know these prices exist for a given level of performance, only the uninformed are going to buy a 290X at $400 and R9 290 at $300.

Since 290X can't even win by more than 1-2% at 4K vs. a 970, lack of HDMI 2.0/DSR and 100W more power, 290X cannot possibly cost more than $330. Maybe AMD can discontinue the 290X and just make make more 290s available to be able to target the $269 price.

 

[Erenhardt]

I hope mantle will help them make AI not suck for the change...

There is FC4 and new Dragon Age commng soon, other than there seem to be lack of blockbuster titles that could benefit new GPU release.

 

[krumme]

AMD needs to suck it up and drop R9 285 to $179, R9 280X to $219, R9 290 to $269 and R9 290X to $299. I realize these 780/780Ti prices are just fire sale on 780Ti but now that gamers know these prices exist for a given level of performance, only the uninformed are going to buy a 290X at $400 and R9 290 at $300.

Agree - its very simple. Either AMD drops prices or they will not sell cards.

There is only one time in history when they didnt adopt and that was at 680 introduction. I dont think RR still have this stupid idea that their brand, whatever can keep prices up. Last time was a major fuck up, and they had to adopt prices even further down, and it came way to late and ended with less marketshare and even lower margin per card.

This time there is absolutely no excuse for such a mistake, and as we have discussed AMD even have a cost advantage to NV here, - contrary to the 680 introduction - so there is notreason not to lower prices if they want to turn a profit - and do it now to the prices you have shown.

 

[ShintaiDK]

I hope mantle will help them make AI not suck for the change...

How would Mantle help with that, since the game needs to work on more than Mantle.

Plus we already saw the dark secret of Mantle. Game patches for new uarchs and other issues. Thats not going to work in the long run.

 

[ShintaiDK]

Does AMD even need much of a price cut? A small correction would be in order I guess, but Mantle allows AMD to carry a premium for their cards and its warranted. If Mantle allows for a better gaming experience in upcoming titles like Civ: Beyond Earth with Radeon graphics cards then they should carry a premium because of it. I think some people are underestimating just how good Mantle is. That's straight from developers too. What if Mantle makes a noticeable difference in Civilization: Beyond Earth. What will those millions of Civ players want for hardware? Mantle gives Radeon graphics significant added value IMO.

4 games or so after a year isnt exactly something that sells. Specially not when some of the games have reached anything but expected sales and unfixed bugs. And Civ BE doesnt look to change that.

Also Mantle supported games already need game patches for new uarch to work correct, something that havent been released yet for GCN 1.2.

But again, even MS dropped the bomb of reality. DX11.3 is there for a reason. Simply because only a relatively small group is able to handle the complexity of DX12 and Mantle.

 

[krumme]

Maybe AMD can discontinue the 290X and just make make more 290s available to be able to target the $269 price.

They dont have to. After nearly a year its safe to say they have very good yields on the 290 - and its depreciated a good deal. They can just sell most as 290x at whatever price is nessesary be it 259 or 289.

mrmt notion earlier of cash flow is also important here, so they need to adapt in this very week. The rest is a pipe dream - this is reality.

 

[ShintaiDK]

Its a shame the release today was nothing more than 285 in India. But the hope for a 285X release died some

AMD really needs a 285X out now. And not cheaping out with 5.5Ghz GDDR5 either. But full 7Ghz is needed to avoid the worst ripples.

And then they need to advance any bigger GCN1.2 designs they have. Because I doubt GCN2.0 is anything but near.

 

[sontin]

4 games or so after a year isnt exactly something that sells. Specially not when some of the games have reached anything but expected sales and unfixed bugs. And Civ BE doesnt look to change that.

And a GTX970 is faster in Thief 4 and BF 4 than a 290. ^_^

 

[desprado]

How would Mantle help with that, since the game needs to work on more than Mantle.

Plus we already saw the dark secret of Mantle. Game patches for new uarchs and other issues. Thats not going to work in the long run.

Remember 2 man works with in 2 months time that sound so easy but it is was PR and Bs both combination.Sniper Elite 3 Mantle support will come on Nov 2014 meaning 4.5 months work and same goes with BF4.

 

[desprado]

And a GTX970 is faster in Thief 4 and BF 4 than a 290. ^_^

it is 30% faster not just fast even if it use mantle.

 

[RussianSensation]

There is FC4 and new Dragon Age commng soon, other than there seem to be lack of blockbuster titles that could benefit new GPU release.

Assassin's Creed Unity Nov 11th - AC IV is very demanding at 1600P and this could be even more demanding

Project CARS - November 18 - probably a must have title for racing fans

Don't forget the the next COD given how unoptimized last COD game that with crap graphics brought performance to below 60 fps on a 770 and Titan couldn't even hit 50 fps at 1600p. :whiste:

And a GTX970 is faster in Thief 4 and BF 4 than a 290.
......it is 30% faster not just fast even if it use mantle.

Well we knew your posts are full of NV praise and AMD bashing but now you are just making stuff up.

http://www.guru3d.com/articles_pages/nvidia_geforce_gtx_970_and_980_reference_review,12.html

http://www.guru3d.com/articles_pages/nvidia_geforce_gtx_970_and_980_reference_review,15.html

A 980 is 19% faster than 290X in BF4 so how can a 970 be possibly 30% faster than a 290 in BF4?

You know what I didn't hear you from though is how 290X in CF is 33-40% faster than 780TI SLI at 4K in Watch Dogs, but instead you go about fabricating Thief and BF4 advantages of 970 over 290.

You know what else I didn't hear from you is how much $ you will lose on resale of your 780TI Tri-SLI? 780Ti now sells for $430. So while someone enjoyed R9 290 performance at $400 for 1 year ($350-360 for the last 5 months) and now 970 beats that card by 8-10% for $70 less, at least someone with 2-3x R9 290s doesn't feel like they flushed $1000+ into the toilet after 970 came out. Most of this is due to competition from AMD and AMD-powered PS4/XB1 consoles. Yet, you just can't stop talking about how awesome 970 is vs. 1-year old 290 but being bent over by NV for a year is swept under the rug....because NV iz ze best.

And I bet if R9 390X beats 980 in 5 months, you'll keep claiming that 390X is a GM200/210 competitor while R9 290X competes with the 980, right?

And side-note, during the 3-4 following R9 290/290X launch, those cards actually paid for themselves through scrypt/coin mining. Thus, I doubt that many 1 year old R9 290/290X owners are too upset right now, unlike 780Ti / Titan / Titan Black owners.

 

[desprado]

Assassin's Creed Unity Nov 11th - AC IV is very demanding at 1600P and this could be even more demanding

Project CARS - November 18 - probably a must have title for racing fans

Don't forget the the next COD given how unoptimized last COD game that with crap graphics brought performance to below 60 fps on a 770 and Titan couldn't even hit 50 fps at 1600p. :whiste:

But know i am getting more than 140fps on 1440p with MSAA 4x.

U forgot Shadow of Mordor,Ryse,Lords of fallen and Aline Isolation.