Who has the better 'current tech'?

[ChosenUno]

you seriously have to stop taking everything so literal. I mean package as in CPU/GPU-Ram-NB-SB. Think of a laptop MoBo it can now be quite a bit smaller, talking about 14nm not the old 28nm- 14nm adds a great deal of possibilities.

Besides the only metric most of you are speaking of is FPS. Honestly it doesn't matter a great deal who has what or does what. But what I look at is the possibility going forward. AMD has proven longer legs by a huge margin which could arguably be considered the better tech, especially for the lions share of owners whereas most don't upgrade at 6 month intervals. It also could be argued that the games themselves influence this perception and negate any real facts. Consider GW and how it will ALWAYS lend itself to Nvidia hardware, this will skew the end result and lead to a false positive in this argument.

But again all this is moot when the great deal of you refrain from being reasonable and looking at it from all perspectives.

Again with the GameWorks BS. Can you actually use some facts? GW favours Nvidia when [H] shows otherwise?

Longevity because they don't have the cash to splurge. GCN is much less efficient than Maxwell and arguably even Kepler. If they didn't switch to HBM then we wouldn't even have this argument. And even then it's still not close efficiency-wise. How's that better tech when you use more power for the same result than your competitor?

Man, you oughta apply your own reasonings to yourself since you're not arguing with facts.

Laptop MoBo being smaller is a moot point, since AFAIK there are no laptops today with discrete graphics that doesn't use MXM. Where's the Fury laptop? Nvidia managed a full 980 in a laptop, where's the Nano in a laptop?

Going forward is another null and void argument. Pascal and Polaris will be different architectures on different manufacturing process. Are you sure Polaris will be much, much better than Pascal? If we project from today's numbers, Pascal will whoop Polaris' ass. If you have numbers that say otherwise, feel free to post them.

 

[razor1]

you seriously have to stop taking everything so literal. I mean package as in CPU/GPU-Ram-NB-SB. Think of a laptop MoBo it can now be quite a bit smaller, talking about 14nm not the old 28nm- 14nm adds a great deal of possibilities.

People should use the correct words , Package is used in chip manufacturing, which doesn't include the entire card.

What you are talking about is form factor not package.

Besides the only metric most of you are speaking of is FPS. Honestly it doesn't matter a great deal who has what or does what. But what I look at is the possibility going forward. AMD has proven longer legs by a huge margin which could arguably be considered the better tech, especially for the lions share of owners whereas most don't upgrade at 6 month intervals. It also could be argued that the games themselves influence this perception and negate any real facts. Consider GW and how it will ALWAYS lend itself to Nvidia hardware, this will skew the end result and lead to a false positive in this argument.

You want to talk about FPS, performance/watt, performance/mm2, and for the the gtx 980 and 980ti in their receptive markets have the performance/ $ too all of these are on nV's side right now. Now if you want to think about it this alittle bit, what is the wattage difference between the gtx 970 vs. the gtx 980 vs the Nano, now you have gtx 970 that can be put onto a card size that is similiar to the nano, the Nano uses more power that the gtx 980, So logic will dictate, since the gtx 980 only uses 20 watts more than the gtx 970 a small form factor is doable for this card since the ram amounts don't change from the gtx 970 to the gtx 980. You want to talk logic the person that wrote that post about the gtx 980 capable of being made in the smaller form factor, was absolutely correct, but the necessity to do it is it there? Nope, because the Nano was priced to high for anything tangible.

Gameworks is bogus, what you have listened to that video that was posted in another thread, all of the techreport stuff doesn't add up, Crysis 2, first off wasn't even a gameworks title, tessellation was done by Crytek. Secondly, the water is not culled during wireframe since there is no occluders, occluding the water so it has to be rendered when the engine is in wire frame mode. Witcher 3 even at early stages had hairworks active, pretty much 3 years before release. The amount of BS that people spread when they don't know what they are talking about is abundantly great.

But again all this is moot when the great deal of you refrain from being reasonable and looking at it from all perspectives.

read above.

 

[JustReason]

Again with the GameWorks BS. Can you actually use some facts? GW favours Nvidia when [H] shows otherwise?

Longevity because they don't have the cash to splurge. GCN is much less efficient than Maxwell and arguably even Kepler. If they didn't switch to HBM then we wouldn't even have this argument. And even then it's still not close efficiency-wise. How's that better tech when you use more power for the same result than your competitor?

Man, you oughta apply your own reasonings to yourself since you're not arguing with facts.

Laptop MoBo being smaller is a moot point, since AFAIK there are no laptops today with discrete graphics that doesn't use MXM. Where's the Fury laptop? Nvidia managed a full 980 in a laptop, where's the Nano in a laptop?

Going forward is another null and void argument. Pascal and Polaris will be different architectures on different manufacturing process. Are you sure Polaris will be much, much better than Pascal? If we project from today's numbers, Pascal will whoop Polaris' ass. If you have numbers that say otherwise, feel free to post them.

You are terrible at arguments. Worst on this site by a country mile. Efficiency? Where did I speak of efficiency. Really you cant without heavy consideration of other aspects ie: compute being disabled in Kepler.

Your argument on laptops... HAHA. My laptop point did not even mention discreet as I said APU laptop. To make it this easy for you to understand: APU = CPU + GPU ON A SINGLE PACKAGE... NO DISCEET GPU NEEDED.

Polaris and Pascal? How about the fact Pascal may still only be a Maxwell derivative dependent on when the architecture was actually finalized. And even basing on todays GPUs even now there are numerous sites after updated drivers that show at >=1440p AMD FuryX is on par and greater than the 980Ti. But none of that really matters as it depends on what day of the week and which site. I NEVER SPOKE OF FUTURE DISCREET GPUS ONLY WHAT HBM BRINGS TO OVERALL SIZE AND THAT POSSIBILITY IN LAPTOPS.

Do me a favor and refrain from responding to my posts. Your lack of knowledge and level of it are far below mine and a great deal of even the lowest on any site that I do not deem to waste any more of my time on you.

 

[JustReason]

People should use the correct words , Package is used in chip manufacturing, which doesn't include the entire card.

What you are talking about is form factor not package.
You want to talk about FPS, performance/watt, performance/mm2, and for the the gtx 980 and 980ti in their receptive markets have the performance/ $ too all of these are on nV's side right now. Now if you want to think about it this alittle bit, what is the wattage difference between the gtx 970 vs. the gtx 980 vs the Nano, now you have gtx 970 that can be put onto a card size that is similiar to the nano, the Nano uses more power that the gtx 980, So logic will dictate, since the gtx 980 only uses 20 watts more than the gtx 970 a small form factor is doable for this card since the ram amounts don't change from the gtx 970 to the gtx 980. You want to talk logic the person that wrote that post about the gtx 980 capable of being made in the smaller form factor, was absolutely correct, but the necessity to do it is it there? Nope, because the Nano was priced to high for anything tangible.

Gameworks is bogus, what you have listened to that video that was posted in another thread, all of the techreport stuff doesn't add up, Crysis 2, first off wasn't even a gameworks title, tessellation was done by Crytek. Secondly, the water is not culled during wireframe since there is no occluders, occluding the water so it has to be rendered when the engine is in wire frame mode. Witcher 3 even at early stages had hairworks active, pretty much 3 years before release. The amount of BS that people spread when they don't know what they are talking about is abundantly great.
read above.

not a lot of time... wasn't speaking of that video, didn't watch it don't plan to. My mention of GW isd in line with facts as I have spoke before not conspiracy but expectations within the market. Moving on... 980Ti is strong but not a price/performance leader and the 980 is now where near being it. Generally midrange say 960-970 380-390x(including 2xx series) is where it is at and seems the 390(X) is it for the most part.

 

[razor1]

You are terrible at arguments. Worst on this site by a country mile. Efficiency? Where did I speak of efficiency. Really you cant without heavy consideration of other aspects ie: compute being disabled in Kepler.

Your argument on laptops... HAHA. My laptop point did not even mention discreet as I said APU laptop. To make it this easy for you to understand: APU = CPU + GPU ON A SINGLE PACKAGE... NO DISCEET GPU NEEDED.

Polaris and Pascal? How about the fact Pascal may still only be a Maxwell derivative dependent on when the architecture was actually finalized. And even basing on todays GPUs even now there are numerous sites after updated drivers that show at >=1440p AMD FuryX is on par and greater than the 980Ti. But none of that really matters as it depends on what day of the week and which site. I NEVER SPOKE OF FUTURE DISCREET GPUS ONLY WHAT HBM BRINGS TO OVERALL SIZE AND THAT POSSIBILITY IN LAPTOPS.

Do me a favor and refrain from responding to my posts. Your lack of knowledge and level of it are far below mine and a great deal of even the lowest on any site that I do not deem to waste any more of my time on you.

JustReason, I think you are way off base here, you are the one that brought in the metrics of performance/mm2, that is efficiency.

Compute being disable in Kepler? Think you got that wrong, I think you mean Maxwell?

HBM only brings one thing for us today, and we are talking about today, not tomorrow, it brings us nothing in the realworld. Its nice to have in the future but not right now, there is no tangible benefits for the end users, unless they need a card the size of the nano and with its performance.

 

[razor1]

not a lot of time... wasn't speaking of that video, didn't watch it don't plan to. My mention of GW isd in line with facts as I have spoke before not conspiracy but expectations within the market. Moving on... 980Ti is strong but not a price/performance leader and the 980 is now where near being it. Generally midrange say 960-970 380-390x(including 2xx series) is where it is at and seems the 390(X) is it for the most part.

That is why is stated respective markets.

 

[Dayaks]

JustReason, I think you are way off base here, you are the one that brought in the metrics of performance/mm2, that is efficiency.

Compute being disable in Kepler? Think you got that wrong, I think you mean Maxwell?

HBM only brings one thing for us today, and we are talking about today, not tomorrow, it brings us nothing in the realworld. Its nice to have in the future but not right now, there is no tangible benefits for the end users, unless they need a card the size of the nano and with its performance.

That's my take too. nVidia has tech that actually improves the experience now. AMD always had tomorrow tech, a tomorrow that never actually comes and they bleed market share in the mean time. They had some useful improvements, like Crossfire over the PCIe bus, but weren't able to keep the drivers up to speed enough.

I was critical of HBM having no real world improvements and being a supply/cost/low VRAM issue before it even launched. Any idiot could figure that part out... except AMD.

 

[Anarchist4000]

That's my take too. nVidia has tech that actually improves the experience now. AMD always had tomorrow tech, a tomorrow that never actually comes and they bleed market share in the mean time. They had some useful improvements, like Crossfire over the PCIe bus, but weren't able to keep the drivers up to speed enough.

I was critical of HBM having no real world improvements and being a supply/cost/low VRAM issue before it even launched. Any idiot could figure that part out... except AMD.

The problem with this is that if Nvidia had the same technology, you could have tomorrow's tech today. As a gamer would you want newer features and faster graphics or just settle with what works now? If for instance Nvidia looked further ahead, how much better would the games of today be? If Nvidia adopted something similar to GCN when it rolled out, we could have had DX12/Vulkan years ago. One of the big advertised features was ease of porting from consoles to PC. The biggest reason for the change is probably the reduced latency. If it wasn't for the latency needs of VR we'd probably still be using DX11.

As for HBM, I personally find it much more interesting than just the bandwidth. The close proximity to the GPU should offer lower latency, smaller packages(it's stacked, GDDR5 would be like trying to fit 32 memory chips onto a card), and reduce the form factor of cards. The smaller cards are not only easier to fit in a case, but cooling becomes substantially easier. Just speculation here, but a Zen APU with HBM could all be on a single chip that sits into a socket. For many gamers that could be significant in a number of ways. Even as a single card with 4gb the cards seem to compete reasonably well with larger capacity Nvidia cards. For a mid range product HBM1 should be more than sufficient.

 

[Criticalhitkoala]

As a gamer would you want newer features and faster graphics or just settle with what works now? If for instance Nvidia looked further ahead, how much better would the games of today be?

This is an interesting arguement imho. How exactly are games today worst than they should be?

Also unlike consoles where there's a definite gap in hardware differences between gen 3 and gen 4 of the xbox/ps4, there is so many different generations of pc's and video cards that I think games now actually do a rather good job at being not that bad. Yeah you have your occasional issue like Batman, and the initial fallout 4 problems, but in 2015 I downloaded and played 53 games from Steam, and only 2 maybe had performance issues that were only issues because I was picky as all get out.

To answer your question about, faster graphics is nice. But I can happily settle for what works now because being fancy as shit isn't the only thing that counts. Fallout 4 takes zero advantage of me having a medium top tier graphic and cpu solution and arguably it looks like 2012 graphics, but it's fun as hell. Same as Smite and various other games.

To repeat my point of view from earlier in this post. I totally don't get people who keep talking about to utopia future that AMD will provide like they are the only ones who have a solution. I'm pretty sure Intel and Nvidia are probably already sitting on 2~3 generations of solutions ready to compete. And also if AMD does win for a few months, there's probably already something in the pipeline ready to be released (Like the 980ti right when the fury was announced).

 

[ChosenUno]

You are terrible at arguments. Worst on this site by a country mile. Efficiency? Where did I speak of efficiency. Really you cant without heavy consideration of other aspects ie: compute being disabled in Kepler.

First , that has to be the worst rationale ever, "if they cared". Doesn't matter if they do or not, the point of Nano having the highest power to performance ratio is undeniable.

No it seems you cant be bothered with anything other than FPS metrics. You have no vision. Look at it this way: HBM adds so much to the possibilities. The Nano is by far without a doubt the Performance to mm^2 leader and by a mile.

So first you talk about power to performance ratio, which is power efficiency. Next you bring in performance to mm^2 ratio, which is also an efficiency metric.

Your argument on laptops... HAHA. My laptop point did not even mention discreet as I said APU laptop. To make it this easy for you to understand: APU = CPU + GPU ON A SINGLE PACKAGE... NO DISCEET GPU NEEDED.

Why mention that in a conversation about current tech and specifically R9 Nano, which YOU brought into the discussion?

*snip* But none of that really matters as it depends on what day of the week and which site. I NEVER SPOKE OF FUTURE DISCREET GPUS ONLY WHAT HBM BRINGS TO OVERALL SIZE AND THAT POSSIBILITY IN LAPTOPS. *snip*

So after arguing about a very random metric, console sales, which Nvidia doesn't participate in, I refuted with smartphone/tablet sales, which AMD doesn't participate in, and you then move on to Nano being a performance/watt and performance/mm2 leader. When shown that logically Nvidia can do the same or better than what AMD did with the Nano, start blabbering about Nano package size and then to future APUs with HBM and future laptops, while the conversation topic is CURRENT TECH.

Seriously, what are you on about? Good lord, talk about rambling on.

 

[n=1]

Well given the amount of physical PCB space required for 12 GDDR5 chips and a 384-bit bus, I think on a strictly performance/mm² metric, the Nano should at least tie with the 980 Ti if not edge it out.

 

[Creig]

That's my take too. nVidia has tech that actually improves the experience now. AMD always had tomorrow tech, a tomorrow that never actually comes

That's where you're wrong. Yes, AMD takes the long range view while Nvidia looks at the present. Which is why Nvidia generally has higher performing flagship products than AMD. However, the result of this is that cards like the 7970 and 290X have aged much better than their Nvidia counterparts. That's the "tomorrow that eventually arrives" benefit of AMD vs Nvidia.

If you don't mind forking out the cash for a new video card every year or two, buy Nvidia. If you want your card to remain relevant for 2-3 years (or even longer), buy AMD.

Given the cost of high end video cards these days, I much prefer AMD's approach.

 

[razor1]

That's where you're wrong. Yes, AMD takes the long range view while Nvidia looks at the present. Which is why Nvidia generally has higher performing flagship products than AMD. However, the result of this is that cards like the 7970 and 290X have aged much better than their Nvidia counterparts. That's the "tomorrow that eventually arrives" benefit of AMD vs Nvidia.

If you don't mind forking out the cash for a new video card every year or two, buy Nvidia. If you want your card to remain relevant for 2-3 years (or even longer), buy AMD.

Given the cost of high end video cards these days, I much prefer AMD's approach.

That's personal preference, I tend to upgrade when I can't play games at the highest settings for games coming out at that time, so for me, I almost always upgrade per generation. People that have money to spend 500 bucks or more for a graphics card, aren't the same people that purchase 200 bucks for a graphics card. Longevity of technology is pretty much non existent, technology is a depreciating asset there will always be something better that comes out, to bank on longevity is not something you should be considering, its hit or miss. The g80 was a great card and had longevity on its side. As did the 9700 pro. But no one could have predicted they would have lasted as long as they did at the time they were released since there were market pressures that drove the longevity of them.

There were many cards from both vendors that had short life spans too, the fx series from nV, the rv3x0 series from AMD, the r600 series from AMD, the 8500 series from ATI....

 

[Anarchist4000]

This is an interesting arguement imho. How exactly are games today worst than they should be?

Latency, multi-adapter setups, and portability would all seem worthy candidates of improvement. Ambiguity of driver implementations I'd say is by far the biggest one. A good API shouldn't require constant driver updates from the IHVs to fix issues. The devs originally pushing for Mantle's development I'd say are a good argument for that being the case.

To answer your question about, faster graphics is nice. But I can happily settle for what works now because being fancy as shit isn't the only thing that counts. Fallout 4 takes zero advantage of me having a medium top tier graphic and cpu solution and arguably it looks like 2012 graphics, but it's fun as hell. Same as Smite and various other games.

I won't argue against gameplay being more important than pretty graphics. My concern is what other features could they have implemented that would have added to the experience? Physx could be a candidate. The issue there is more proprietary, but how would games be different with more robust physics acceleration widely available? Sound processing is a personal annoyance. Before Microsoft decided software sound was good for everyone and all the 3D effects and hardware acceleration disappeared. Only now with VR does it seem like anyone is starting to care again.

To repeat my point of view from earlier in this post. I totally don't get people who keep talking about to utopia future that AMD will provide like they are the only ones who have a solution. I'm pretty sure Intel and Nvidia are probably already sitting on 2~3 generations of solutions ready to compete. And also if AMD does win for a few months, there's probably already something in the pipeline ready to be released (Like the 980ti right when the fury was announced).

The problem with this is that AMD often leads in regards to progress on hardware. Smaller nodes, DX9, tessellation, async/hardware scheduling, etc. I'm a hardware person, I like to see cool new features. My issue with Nvidia here is they always seem to be an anchor with solid implementations a year later. I'd much rather see them introducing new features and capabilities.

The performance argument I still don't think can be made as we need to see a DX11 Nvidia vs DX12 AMD comparison as that lines up with the design goals of each card. DX-A vs DX-B comparisons, not one is newer so it's obviously better.

 

[n=1]

That's personal preference, I tend to upgrade when I can't play games at the highest settings for games coming out at that time, so for me, I almost always upgrade per generation. People that have money to spend 500 bucks or more for a graphics card, aren't the same people that purchase 200 bucks for a graphics card. Longevity of technology is pretty much non existent, technology is a depreciating asset there will always be something better that comes out, to bank on longevity is not something you should be considering, its hit or miss. The g80 was a great card and had longevity on its side. As did the 9700 pro. But no one could have predicted they would have lasted as long as they did at the time they were released since there were market pressures that drove the longevity of them.

There were many cards from both vendors that had short life spans too, the fx series from nV, the rv3x0 series from AMD, the r600 series from AMD, the 8500 series from ATI....

Aren't we fast approaching the limits of silicon? I'd like to think going forward, release cadences will become more and more stretched out until we move to non-silicon based technologies.

 

[razor1]

yeah we are, but lets see what nanotubes and what else replaces silicon wafers we might still have similar advantages as current node drops once traditional silicon wafers are replaced.

 

[Anarchist4000]

There's a whole 3rd dimension of silicon to explore yet. Probably why AMD said power efficiency in the future will be extremely important. Bring on the half inch think GPUs!

 

[Creig]

Longevity of technology is pretty much non existent, technology is a depreciating asset there will always be something better that comes out, to bank on longevity is not something you should be considering, its hit or miss.

These days, hardware longevity is actually becoming a reality. Look at current CPU stagnation. It used to be that we upgraded our CPUs/motherboards nearly as often as our video cards. Nowadays there are still people running high end video cards with overclocked 2500K processors and are doing just fine. Those CPUs are now five years old and counting.

Both the 7970 and 290X have shown extreme resistance to obsolescence for a GPU. The 7970 is now four years old while the 290X is over two years old. Both cards are still putting out respectable numbers in games and will probably continue to do so for some time yet to come since the GCN architecture within the 7970/290X will live on in AMD's upcoming Polaris video cards.

 

[razor1]

These days, hardware longevity is actually becoming a reality. Look at current CPU stagnation. It used to be that we upgraded our CPUs/motherboards nearly as often as our video cards. Nowadays there are still people running high end video cards with overclocked 2500K processors and are doing just fine. Those CPUs are now five years old and counting.

Both the 7970 and 290X have shown extreme resistance to obsolescence for a GPU. The 7970 is now four years old while the 290X is over two years old. Both cards are still putting out respectable numbers in games and will probably continue to do so for some time yet to come since the GCN architecture within the 7970/290X will live on in AMD's upcoming Polaris video cards.

While it seems that way, they won't, different generations of games push cards differently, while we will need more of each facet of performance from GPU's, some parts will be pushed more than others, and this is what newer generation GPU's will be better at.

GCN has horrid triangle throughput, that is its Achilles Tendon, if we start looking at what new games are pushing for base polygon counts, that will show up even with the lowest tessellation factors, its shows up at x8 factor right now in some games, if base polycounts increase two fold, which they are, like in Arkham Knight it will start showing up at x4.

Its pretty obvious AMD's Fiji, is capable of more shader performance than current generation titles are using, and they tend to have better performance when its the bottleneck vs. competitors cards, so you will see nV's cards have a performance disparity when new games push that more too. What is good for today won't be good for tomorrow, which one lasts longer its a crap shot if we are to base buying a card soon after release. In hind sight is all great but no one will or can predict that because the dev's are looking at things differently.

 

[KickAssCop]

NVidia has better tech.

 

[Creig]

While it seems that way, they won't, different generations of games push cards differently, while we will need more of each facet of performance from GPU's, some parts will be pushed more than others, and this is what newer generation GPU's will be better at.

GCN has horrid triangle throughput, that is its Achilles Tendon, if we start looking at what new games are pushing for base polygon counts, that will show up even with the lowest tessellation factors, its shows up at x8 factor right now in some games, if base polycounts increase two fold, which they are, like in Arkham Knight it will start showing up at x4.

Its pretty obvious AMD's Fiji, is capable of more shader performance than current generation titles are using, and they tend to have better performance when its the bottleneck vs. competitors cards, so you will see nV's cards have a performance disparity when new games push that more too. What is good for today won't be good for tomorrow, which one lasts longer its a crap shot if we are to base buying a card soon after release. In hind sight is all great but no one will or can predict that because the dev's are looking at things differently.

That's your prediction. Four years worth of benchmarks says GCN is a good architecture. Whether or not that will continue into the future is anybody's guess. But I haven't seen anything that indicates we'll see a sudden shift in programming that will render GCN obsolete. If anything, with the introduction of DX12 and Vulkan, GCN will actually receive a boost in gaming power as time goes on.

It cannot be disputed that Tahiti/Hawaii have definitely aged better than Fermi/Kepler.

 

[razor1]

That's your prediction. Four years worth of benchmarks says GCN is a good architecture. Whether or not that will continue into the future is anybody's guess. But I haven't seen anything that indicates we'll see a sudden shift in programming that will render GCN obsolete. If anything, with the introduction of DX12 and Vulkan, GCN will actually receive a boost in gaming power as time goes on.

It cannot be disputed that Tahiti/Hawaii have definitely aged better than Fermi/Kepler.

the latest tomb raider didn't show that did it?

And its not programming that is the driving force, its art assets, two different things. Programming tends to find ways to use new features, and use them to the fullest possible extent, art assets have limitations based on hardware and programming, so if art assets push triangle through put more, then AMD is in trouble, but programming might push shader through put more as well, its a balancing act which way is better, well tell ya what bet on what you have now is a better approach than bet on the future of what games might have. And this isn't opinion, its how dev's think. They don't plan for what is out there next generation, they plan for what they have right now in their hands and what they can do with what they have.

AMD by losing so much marketshare, they have to understand geometry through put will be pushed more than what their cards are better suited for, pixel through put. We have seen this with the FX line when nV lost some marketshare but still had enough to maintain developer's interest, AMD cards with better, pretty much everything, games got held back by the FX line. Then go to the x1900xt launch, games started to push pixel shaders more, then the g80 came out, and it had what was needed, but the r600 didn't, and rv6x0 held games back because AMD had enough marketshare for dev's to pay attention to them.

Edit, this is easy for you to answer, lets say you were major owner of a game company and you have put in 15 million dollars to produce a game, will you focus on features that less that 20% of the people that might buy your game have? Or will you focus on 80+% have?

That is directly linked to how much possible sales you might get.

After that how much time would you want your team to spend on features of those less that 20% of the people have? Time, money resources?

 

[Creig]

the latest tomb raider didn't show that did it?

Please tell me you're not trying to use a game released YESTERDAY as proof of the superiority of one architecture over another...

 

[noko]

So the answer to this thread is - it depends.

AMD PUREHAIR is better useful tech (software implementation) of hardware than HairlessWorks.

Multiple workloads concurrent using compute and graphics AMD GPU's will do better - better tech in that aspect.

Serial fast and easier to program workloads - Nvidia has the better tech.

SLI/CFX - AMD has superior tech to communicate between the two cards, XDMA and shines as resolution goes up - Nvidia has better support for SLI and games (better software tech)

Memory compression - Nvidia - leading to superior performance per given memory bandwidth amount and bus width - superior cost savings

Memory design and pushing new standards - AMD - HBM DDR5, DDR4. . .

Now if you look at tech that is cheaper to make, has better profits combined with better marketing and developer support - Nvidia

Better tech in software? For API - AMD work with DX9 through DX12 - not to mention Vulkan. Then Nvidia gameworks stuff has some rather good stuff for Nvidia hardware and even AMD on some of it.

Anyways take whatever side you want or just look at both for what it is.

 

[razor1]

Please tell me you're not trying to use a game released YESTERDAY as proof of the superiority of one architecture over another...

I didn't say that did I I'm saying Keplar being held back or not having legs is not all that accurate. I did say in my first post this thread, both IHV's hardware are good, just that one was late and didn't have the numbers in all metrics that make it worth while specifically all metrics regarding power.

Please tell me, where I stated that, you seem to think that AMD has better technology, and that surely isn't the case that's why they didn't sell. You think everyone that buys graphics cards are dumb so stupid marketers can have their way with gimmicks? Maybe that's why Huddy and the other PR and Marketing guys at AMD fail so much? Because they are so bad that when they open their mouth people get a whiff of their bad breath and run?

If people were so dumb to begin with to believe nV's marketing, then what ever AMD's, marketing have stated should have the same affect even though how ludicrous some of it is.

 

[primetime]

Nvidia most certainly has a leg up in marketing lol..even people who cant spell computer will want nvidia titan's .....im talking people who game exclusively on consoles lol...what they going to do? Make love to the dam card

 

[Derfnofred]

This entire topic is phrased in such a way that it's non-falsifiable. So as much as we may want to say this or that, there's nothing we're going to pin down.

So I'll make my own standards, and frame it from looking at what I can buy today to play today's games (so this is obviously biased towards DX11):
1.) FPS/Watt across almost every tested game (I'd say all, but I might have missed one or three) goes to NVidia.
2.) FPS/die size likewise gets a nod for NVidia.
3.) FPS/$$ depends on where in the lineup you go. Top end goes towards NVidia.

AMD provides some cool packaging on some of its cards, which may benefit part of the market. If you nerd out on applied tech demoes for next-gen stuff (Fury!), then there's value there. You're obviously not getting a bad card! But, frankly, NVidia's hardware/software pipeline is better designed for the insults and vagaries of DX11.

FTR: I rock out a 280x and am pleased with the performance/$ it has given me. If it magically gets better with DX12 titles, then yay me! I'm not expecting any magic. When it comes down to buying my next card, I'll assess based on the best value proposition at that time (which for me is a perf/$, and looking at lightly used cards, different for someone else).

 

[JustReason]

So first you talk about power to performance ratio, which is power efficiency. Next you bring in performance to mm^2 ratio, which is also an efficiency metric.



Why mention that in a conversation about current tech and specifically R9 Nano, which YOU brought into the discussion?



So after arguing about a very random metric, console sales, which Nvidia doesn't participate in, I refuted with smartphone/tablet sales, which AMD doesn't participate in, and you then move on to Nano being a performance/watt and performance/mm2 leader. When shown that logically Nvidia can do the same or better than what AMD did with the Nano, start blabbering about Nano package size and then to future APUs with HBM and future laptops, while the conversation topic is CURRENT TECH.

Seriously, what are you on about? Good lord, talk about rambling on.

For everyone else here I apologize for having to make this post but it had to be done.

Spoiler: clarification

OK so apparently you cant read nor understand so I am gonna give you a refresher.

You said:

Well, if nVidia really cared about Nano's market they'd have released a product there. Didn't you see they managed to cram a full 980 into a laptop? What's that you're talking about performance/size?

CPU/GPU for console metric - what's that metric measuring? Are you sure you're not confusing metric with statistics?

I said:

First , that has to be the worst rationale ever, "if they cared". Doesn't matter if they do or not, the point of Nano having the highest power to performance ratio is undeniable.

CPU/GPU consoles: consider that steam machines are in competition with consoles and for the most part they are Intel/Nvidia consoles. In the sales Metric they are very very far behind to the point of being non-existent.

My first point was to the SFF market where the Nano leads easily (in context). Second point was to Console sales metric where I see nothing about my point that would be an issue.

Then you said:

Just like how AMD has a huge presence in the phone/tablet space, right? What's that? They don't compete in that market? Exactly. Nvidia doesn't want to enter that slice of the market, therefore they don't release a product specifically for it. How's that hard to understand?

You think Nvidia is unable to release a comparable SKU to the Nano, when the Fury X have worse perf/watt ratio compared to the 980Ti even though the Fury X is watercooled?

Ok. How about this metric: Smartphone/tablet sales. Nvidia is infinitely better at this metric compared to AMD. In the sales metric AMD is literally non-existent.

Then I said:

Then all you prove is you are a fool. You argue against the metric then use the same metric for your justification. It matters not whether Nvidia can or cant make a Nano competitor as they have none. Just like your DX12 argument above, Nvidias Nano competitor is nonexistent. Besides the Nano is quite impressive on its own. You are just hating to hate as you have no real substance with which to base your hate.

I never made the metric only explained it to you. Then the very metric you claim is stupid is what you use to make your point (self explanatory). The DX12 aside was in context to your response to another poster.

Then you said:

First of all, I did not bring in the console market argument. I just demonstrated how ridiculous it is.

As for the Nano, you're entitled to your opinions and me to mine. I don't think it's impressive at all, given how its bigger brothers perform against the competition. Given some logic which apparently you seem to lack, I can deduce the following:

A. I can see that Fury X, which is watercooled and has a 300W TDP when overclocked, is still barely on par with a stock clocked GTX 980Ti, which has a TDP of 250W.
B. Now if you reduce a Fury X's TDP to 175W, change the water cooler to a short air cooler, you get a Nano.
C. Since the Fury X had a bigger original TDP AND the advantage of water cooling, it's logical to believe that a GTX 980Ti undergoing A and B like the Fury X will have a higher performance at the same TDP than the Nano.

Follow me? Or is that too hard to understand?

Then I said:

No it seems you cant be bothered with anything other than FPS metrics. You have no vision. Look at it this way: HBM adds so much to the possibilities. The Nano is by far without a doubt the Performance to mm^2 leader and by a mile. Given HBM allows for far smaller packages this also means having an APU laptop where nearly everything other than connections are on the package. There is a lot about the Nano that is impressive to reasonable people.

And you have yet to acknowledge the facts: For that size there is nothing from anyone that can beat or match it.

Performance/watt efficiency you claim to know:
http://www.tomshardware.com/reviews/amd-radeon-r9-nano,4285-9.html#
So in the SFF market looks like the Nano is by far the winner. There was also a performance per mm ( size of the card not the GPU) but cant find that right now, but it should be obvious which what I was referring to. And the last line speaks to the whole thing which you have yet to acknowledge.

Then you said (referencing my post to someone else):

Again with the GameWorks BS. Can you actually use some facts? GW favours Nvidia when [H] shows otherwise?

Longevity because they don't have the cash to splurge. GCN is much less efficient than Maxwell and arguably even Kepler. If they didn't switch to HBM then we wouldn't even have this argument. And even then it's still not close efficiency-wise. How's that better tech when you use more power for the same result than your competitor?

Man, you oughta apply your own reasonings to yourself since you're not arguing with facts.

Laptop MoBo being smaller is a moot point, since AFAIK there are no laptops today with discrete graphics that doesn't use MXM. Where's the Fury laptop? Nvidia managed a full 980 in a laptop, where's the Nano in a laptop?

Going forward is another null and void argument. Pascal and Polaris will be different architectures on different manufacturing process. Are you sure Polaris will be much, much better than Pascal? If we project from today's numbers, Pascal will whoop Polaris' ass. If you have numbers that say otherwise, feel free to post them.

I realize I must post what he was referring to:

you seriously have to stop taking everything so literal. I mean package as in CPU/GPU-Ram-NB-SB. Think of a laptop MoBo it can now be quite a bit smaller, talking about 14nm not the old 28nm- 14nm adds a great deal of possibilities.

Besides the only metric most of you are speaking of is FPS. Honestly it doesn't matter a great deal who has what or does what. But what I look at is the possibility going forward. AMD has proven longer legs by a huge margin which could arguably be considered the better tech, especially for the lions share of owners whereas most don't upgrade at 6 month intervals. It also could be argued that the games themselves influence this perception and negate any real facts. Consider GW and how it will ALWAYS lend itself to Nvidia hardware, this will skew the end result and lead to a false positive in this argument.

But again all this is moot when the great deal of you refrain from being reasonable and looking at it from all perspectives.

Then I said:

You are terrible at arguments. Worst on this site by a country mile. Efficiency? Where did I speak of efficiency. Really you cant without heavy consideration of other aspects ie: compute being disabled in Kepler.

Your argument on laptops... HAHA. My laptop point did not even mention discreet as I said APU laptop. To make it this easy for you to understand: APU = CPU + GPU ON A SINGLE PACKAGE... NO DISCEET GPU NEEDED.

Polaris and Pascal? How about the fact Pascal may still only be a Maxwell derivative dependent on when the architecture was actually finalized. And even basing on todays GPUs even now there are numerous sites after updated drivers that show at >=1440p AMD FuryX is on par and greater than the 980Ti. But none of that really matters as it depends on what day of the week and which site. I NEVER SPOKE OF FUTURE DISCREET GPUS ONLY WHAT HBM BRINGS TO OVERALL SIZE AND THAT POSSIBILITY IN LAPTOPS.

Do me a favor and refrain from responding to my posts. Your lack of knowledge and level of it are far below mine and a great deal of even the lowest on any site that I do not deem to waste any more of my time on you.

Now you can see why I had to post what you referenced otherwise my post responding to you wouldn't make sense. In that post I never mentioned efficiency hence my response in this post. I think the APU point is too obvious to miss for others obviously. And it goes without saying that you didn't heed the last line.

So now we are caught up. If you didn't try so hard to skew the arguments and what was said this could have been so much easier. It isn't hard to tell my first post was explaining the two points you seemed to not understand. But then you had to go and make it some kind of flame war.

 

[ChosenUno]

*snip*

First of all:

• CPU/GPU in consoles sales metric is irrelevant in the conversation of better tech, specifically regarding video cards. I knowingly offered and emphasize another metric, sales of smartphone/tablet, which is as irrelevant as the console sales metric, to point out how ludicrous it is. Maybe I should've chosen car sales instead?
• Regarding the Nano's size, it wins only because there's no direct competition. I've already pointed out twice how it is feasible for Nvidia to make a competitive/better offering in that space. Why they don't do it, there are many possible financial reasons.
  In short: Shrink a 980Ti's PCB, cap TDP at 175W, lower clock speeds, more aggresive GPU Boost. Bam, you now have a Nano beater.
• Nano is not the leader of perf/watt by a mile. In the same link you posted:
  
  
  
  6/9 games the 970 has a better perf/watt ratio. How do you justify the Nano being the clear leader? Please don't pull hyperboles from nowhere without facts.

 

[JustReason]

I misread the graph and was going by what they wrote, My fault for rushing. So was wrong there.

As far as why or if or whatnot as far as Nvidia competing with the Nano in its segment, your statements make as little sense as the DX12 arguments you so lovingly hate. Doesn't matter if they could or not. The Nano leads its segment easily and comfortably, a fact which you seem to do all necessary to avoid admitting.

As far as the efficiency, I was basing the comment on power on it having a sole 8 pin connector compared to my 8+6 on my 290 and it beats my 290. I am certain of the Power to mm^2 aspect which is the only metric I am certain of as I saw it some time ago.

But on the whole my point about the Nano was not that it was better than everything else, generally don't like creating competing scenarios as I prefer to speak to its performance solely. However I did manage to find a quote from a review from someone I generally don't trust as he tends to be very ANTI-AMD, SKYTML (I think I spelled that right) from Hardware Canucks.

AMD faces falling a market share and to combat that, halo products like this one help prop up a brand’s image and get folks excited again. And there’s a lot here to be excited about since NVIDIA has absolutely nothing that can remotely compete against it in certain categories. This is truly a wolf in sheep’s clothing, and a surprising one at that.

So if he can state that then it has to be true.

 

[n=1]

[*]Regarding the Nano's size, it wins only because there's no direct competition. I've already pointed out twice how it is feasible for Nvidia to make a competitive/better offering in that space. Why they don't do it, there are many possible financial reasons.
In short: Shrink a 980Ti's PCB, cap TDP at 175W, lower clock speeds, more aggresive GPU Boost. Bam, you now have a Nano beater.

As I pointed out previously, due to the amount of GDDR5 chips required (12) and the 384-bit bus requirement, I highly doubt shrinking a 980 Ti's PCB down to Nano's level is feasible (or wouldn't be detrimental to performance).

If anything a 980 ITX version would be a far more feasible competitor.

 

[ChosenUno]

As I pointed out previously, due to the amount of GDDR5 chips required (12) and the 384-bit bus requirement, I highly doubt shrinking a 980 Ti's PCB down to Nano's level is feasible (or wouldn't be detrimental to performance).

If anything a 980 ITX version would be a far more feasible competitor.

True, I forgot just how complex the 980Ti PCB was compared to the 970/980.

 

[ChosenUno]

I misread the graph and was going by what they wrote, My fault for rushing. So was wrong there.

As far as why or if or whatnot as far as Nvidia competing with the Nano in its segment, your statements make as little sense as the DX12 arguments you so lovingly hate. Doesn't matter if they could or not. The Nano leads its segment easily and comfortably, a fact which you seem to do all necessary to avoid admitting.

As far as the efficiency, I was basing the comment on power on it having a sole 8 pin connector compared to my 8+6 on my 290 and it beats my 290. I am certain of the Power to mm^2 aspect which is the only metric I am certain of as I saw it some time ago.

But on the whole my point about the Nano was not that it was better than everything else, generally don't like creating competing scenarios as I prefer to speak to its performance solely. However I did manage to find a quote from a review from someone I generally don't trust as he tends to be very ANTI-AMD, SKYTML (I think I spelled that right) from Hardware Canucks.

So if he can state that then it has to be true.

Well, we'll see. IMO at its original price of $650 there was no need to respond to the Nano since it was way overpriced. However now that it's at $400 there's a much bigger chance there'll be a response, since it's now much closer to the 970 Minis.

 

[EQvet80]

Nvidia...

 

[n=1]

I wish pointless necros was a bannable offense

 

[noko]

AMD. . .

 

[{NG}Fidel]

I do.