Should I expect Maxwell (980 Ti specifically) to be neglected like nuts like Kepler, Femi, Tesla, and so on?

[WangKe]

1.) More architectures means more work, more work means more time, more time optimising drivers for older cards means less time to come out with day 1 driver updates and better optimisations for the newest batch of cards.

 

2.) 780 Ti is EOL and on an architecture from early 2012 that was succeeded in late 2014 (3 years is the usual dropoff point for driver improvements, Kepler seems to be capping out). The 290 is not EOL and on an architecture that hasn't changed much from early 2012, and has not yet been succeeded by a new architecture. I can't find any graphs for it but I bet next year the improvements are going to start falling off for at least HD7XXX GCN cards, if not the rest of the cards on GCN too (If my thinking of improvements are more architecture based than individual card based are right, or at least slightly)

 

3.) They need to make a profit, that's why. I doubt anyone but AMD would complain if Nvidia dropped their prices for top end cards down that low and they were only short term investments, but unless the performance actually degrades after a year you could stick with it for a good few years.

 

4.) Fermi had good gains from drivers up until the usual dropoff point, you can refer to the video posted earlier for that. If there was a comparison for Kepler as well you'd most likely see the same, same for GCN as well.

1.) Uhm, what? 

2.) If the 200 series cards were based on a 2012 architecture, why is it keeping up with a 2014 architecture? Does that mean AMD made a superior GPU but had problems with the software for it? Or is it because AMD shows more care for its customers than the competition?

3.) There's a fine line between profit and screwing its customers.

4.) Then how is a weaker version of a 2013 card keeping up with a 2014 card in 2015? Not saying a 480 should still be as good as a 980, but a 780 Ti shouldn't drop off that bad to the point that a 970 can beat it. 2012 architecture, yes, but if AMD can do it with that, why not? I'm pretty sure Nvidia's filled to the brim with cash, I'm pretty sure they can beef up their software team or whatever that're in charge of driver updates.

[TheElt]

1.) Uhm, what? 

2.) If the 200 series cards were based on a 2012 architecture, why is it keeping up with a 2014 architecture? Does that mean AMD made a superior GPU but had problems with the software for it? Or is it because AMD shows more care for its customers than the competition?

3.) There's a fine line between profit and screwing its customers.

4.) Then how is a weaker version of a 2013 card keeping up with a 2014 card in 2015? Not saying a 480 should still be as good as a 980, but a 780 Ti shouldn't drop off that bad to the point that a 970 can beat it. 2012 architecture, yes, but if AMD can do it with that, why not? I'm pretty sure Nvidia's filled to the brim with cash, I'm pretty sure they can beef up their software team or whatever that're in charge of driver updates.

1 - It takes more time to optimise drivers for 3 different architectures than it does to optimise for 1 with minor changes between, time which Nvidia are spending on other things.

 

2 - Kepler was first released 2 months after GCN, the 600 series are Kepler based, so are the 700 series cards (Except 610, 620, 630, 710, 720, 730, 750, 750 Ti). They are both 2012 architectures. Neither company will show more care, I doubt AMD have the budget to and Nvidia don't have the time. It's just what they can do with the time and man power they have. Like I said, it's harder to make drivers for 3 architectures than it is for 1 with minor changes

 

3 - They're hardly screwing their customers. When Maxwell hits 3 years old I bet it will follow on with the trend of improvement pace slowing, as probably will GCN.

 

4 - 290 was released 31/10/13 and the 780 Ti was released 7/11/13. They traded blows with eachother when they were new and they still do now. The 970 is still on a new arch with driver improvements to come, Kepler has just hit the same point Fermi did in the video where the improvements become marginal. That would cost a lot of money, and take a while to hire people good at the job, money and time that can be spent on R&D for their next architecture that will prove more worthwhile for them

[WangKe]

1 - It takes more time to optimise drivers for 3 different architectures than it does to optimise for 1 with minor changes between, time which Nvidia are spending on other things.

 

2 - Kepler was first released 2 months after GCN, the 600 series are Kepler based, so are the 700 series cards (Except 610, 620, 630, 710, 720, 730, 750, 750 Ti). They are both 2012 architectures. Neither company will show more care, I doubt AMD have the budget to and Nvidia don't have the time. It's just what they can do with the time and man power they have. Like I said, it's harder to make drivers for 3 architectures than it is for 1 with minor changes

 

3 - They're hardly screwing their customers. When Maxwell hits 3 years old I bet it will follow on with the trend of improvement pace slowing, as probably will GCN.

 

4 - 290 was released 31/10/13 and the 780 Ti was released 7/11/13. They traded blows with eachother when they were new and they still do now. The 970 is still on a new arch with driver improvements to come, Kepler has just hit the same point Fermi did in the video where the improvements become marginal. That would cost a lot of money, and take a while to hire people good at the job, money and time that can be spent on R&D for their next architecture that will prove more worthwhile for them

1.) What do you mean 3 different architectures? If you mean Pascal with Kepler and Maxwell, why neglect Kepler while Maxwell is still around? Pascal ain't even out yet. I'm not saying Fermi should be treated better, nor should it even be included in the "3 different architectures" thing, but come on... Kepler shouldn't entirely be behind Maxwell by that much.

2.) Again, tell me how is this possible? It looks like AMD cares even "without the budget".

3.) Maxwell's barely gonna be two years old, and its gonna get the same treatment. Does Nvidia expect everyone to upgrade from a 980 Ti to a 1080 Ti or non Ti?

4.) The 290X was released on October 24, 2013 and the 780 Ti was released on November 7, 2013, while the 290 was released two days prior to the 780 Ti. The 290X traded blows with the 780 Ti, despite the 780 Ti beating it, but again... how is this possible?

[Lays]

1.) What do you mean 3 different architectures? If you mean Pascal with Kepler and Maxwell, why neglect Kepler while Maxwell is still around? Pascal ain't even out yet. I'm not saying Fermi should be treated better, nor should it even be included in the "3 different architectures" thing, but come on... Kepler shouldn't entirely be behind Maxwell by that much.

2.) Again, tell me how is this possible? It looks like AMD cares even "without the budget".

3.) Maxwell's barely gonna be two years old, and its gonna get the same treatment. Does Nvidia expect everyone to upgrade from a 980 Ti to a 1080 Ti or non Ti?

4.) The 290X was released on October 24, 2013 and the 780 Ti was released on November 7, 2013, while the 290 was released two days prior to the 780 Ti. The 290X traded blows with the 780 Ti, despite the 780 Ti beating it, but again... how is this possible?

 

 

What do you mean "kepler shouldn't entirely be behind maxwell by that much?"

 

Looking at the graphs, it's a few percent here and there...

 

You need to calm down with the conspiracy theories, the past few generations from AMD basically all use extremely similar architectures, if they optimize for only a few of their cards, basically every card benefits because of the extreme similarities.

The graphs at 1080p are super similar, 780 ti and 290x are extremely similar in the graph, and the 780 is just a tiny bit behind the 290, which isn't a surprise either.

 

The only graphs where the 290x and 290 are closer / ahead of 780 ti is at 1440p/4k, which having 512 bit buses makes sense. 

 

It's also possible "back in the day" when the 290 was further behind the 780 ti, that the drivers back then weren't as good as Nvidias, and they've just merely closed the gap.

[TheElt]

1.) What do you mean 3 different architectures? If you mean Pascal with Kepler and Maxwell, why neglect Kepler while Maxwell is still around? Pascal ain't even out yet. I'm not saying Fermi should be treated better, nor should it even be included in the "3 different architectures" thing, but come on... Kepler shouldn't entirely be behind Maxwell by that much.

2.) Again, tell me how is this possible? It looks like AMD cares even "without the budget".

3.) Maxwell's barely gonna be two years old, and its gonna get the same treatment. Does Nvidia expect everyone to upgrade from a 980 Ti to a 1080 Ti or non Ti?

4.) The 290X was released on October 24, 2013 and the 780 Ti was released on November 7, 2013, while the 290 was released two days prior to the 780 Ti. The 290X traded blows with the 780 Ti, despite the 780 Ti beating it, but again... how is this possible?

1 - I'm going on about the architectures Nvidia currently have ongoing support for, which is Fermi, Kepler and Maxwell, they are the 3 architectures they need to make drivers for, until Pascal releases and they most likely drop Fermi support. Maxwell changed up the performance at each tier, like the 700 series did to the 600 series (770 is faster than a 680, 760 is slower most cases than a 670, but faster than a 660 Ti), so this time the 970 is equivalent to the 780/780 Ti, and the 960 is slower than a 770 most cases, but faster than a 760).

 

2 - Explain what? That 3 years on a card meant to compete still does? Hey look, benchmarks that show the 290 behind the 780Ti, it's obvious that Nvidia care more than AMD do! Explain to me how it's possible!

 

3 - When GCN gets another year older, feel free to do a driver comparison with your 7970 going from launch to late 2016, I bet you'll see fps increase per driver drop when it hits 4 years old.

 

4 - Hey look, 290X trading with a 780 Ti! Your source says a 5% difference, bring at least 3 different sets of benchmarks from 3 different sources (5 or more ideally) and couple them together to get an average result for each, and do a difference from that. Or give me another overall result that shows the same as the one you keep saying

[CierroMierdo]

1.) What do you mean 3 different architectures? If you mean Pascal with Kepler and Maxwell, why neglect Kepler while Maxwell is still around? Pascal ain't even out yet. I'm not saying Fermi should be treated better, nor should it even be included in the "3 different architectures" thing, but come on... Kepler shouldn't entirely be behind Maxwell by that much.

2.) Again, tell me how is this possible? It looks like AMD cares even "without the budget".

3.) Maxwell's barely gonna be two years old, and its gonna get the same treatment. Does Nvidia expect everyone to upgrade from a 980 Ti to a 1080 Ti or non Ti?

4.) The 290X was released on October 24, 2013 and the 780 Ti was released on November 7, 2013, while the 290 was released two days prior to the 780 Ti. The 290X traded blows with the 780 Ti, despite the 780 Ti beating it, but again... how is this possible?

I can see your point, but it ain't that large of a gap. It would make sense if Nvidia somewhat redistributed their efforts on a 20-30-40 kind of three way split: 20 for Fermi, 30 for Kepler, and 40 for Maxwell. Based on what I see on the driver updates, not the 20-20-60 kind of thing that I've been observing lately, I personally think the 780 Ti could handle 60fps recording with shadowplay on resolutions higher than 1080p. There are a crapton of games that are rather easy to drive that 4k 60fps recording (not 1080p recording despite the game's running on 4K) can be done even if the game's on maxed settings. Making feature exclusives are annoying.

Seems like AMD's making an effort, but I wouldn't count Nvidia out of it.

 

What do you mean "kepler shouldn't entirely be behind maxwell by that much?"

 

Looking at the graphs, it's a few percent here and there...

 

You need to calm down with the conspiracy theories, the past few generations from AMD basically all use extremely similar architectures, if they optimize for only a few of their cards, basically every card benefits because of the extreme similarities.

The graphs at 1080p are super similar, 780 ti and 290x are extremely similar in the graph, and the 780 is just a tiny bit behind the 290, which isn't a surprise either.

 

The only graphs where the 290x and 290 are closer / ahead of 780 ti is at 1440p/4k, which having 512 bit buses makes sense. 

 

It's also possible "back in the day" when the 290 was further behind the 780 ti, that the drivers back then weren't as good as Nvidias, and they've just merely closed the gap.

It makes sense that a 290X would and should beat the 780 Ti at those resolutions. It has the bit bus and the extra 1gb VRAM to do it, while having a fair lot of stream processors along with it. I don't know why it took them a long time to do it. Maybe issues with developers? I don't know. But a 384 bit bus with 3gb VRAM has a small gap of different compared to a card that has 256 bit bus with 4 gigs of VRAM, unless someone can explain to me how relevant the bit bus is, and how it corelates with VRAM, with or without the drivers in question, that would be great.

 

1 - I'm going on about the architectures Nvidia currently have ongoing support for, which is Fermi, Kepler and Maxwell, they are the 3 architectures they need to make drivers for, until Pascal releases and they most likely drop Fermi support. Maxwell changed up the performance at each tier, like the 700 series did to the 600 series (770 is faster than a 680, 760 is slower most cases than a 670, but faster than a 660 Ti), so this time the 970 is equivalent to the 780/780 Ti, and the 960 is slower than a 770 most cases, but faster than a 760).

 

2 - Explain what? That 3 years on a card meant to compete still does? Hey look, benchmarks that show the 290 behind the 780Ti, it's obvious that Nvidia care more than AMD do! Explain to me how it's possible!

 

3 - When GCN gets another year older, feel free to do a driver comparison with your 7970 going from launch to late 2016, I bet you'll see fps increase per driver drop when it hits 4 years old.

 

4 - Hey look, 290X trading with a 780 Ti! Your source says a 5% difference, bring at least 3 different sets of benchmarks from 3 different sources (5 or more ideally) and couple them together to get an average result for each, and do a difference from that. Or give me another overall result that shows the same as the one you keep saying

1- IMHO, they should've ditched Fermi when Maxwell came out. That way, allocation of manpower would've been for Kepler, Maxwell, and for the upcoming Pascal. Would make sense at some point.

2- What about 2015 benchmarks? Thats 2014

3- When Greenland comes out, AMD should the rest of the XXXX cards. I honestly don't get why AMD or Nvidia have to allocate that much manpower and/or resources for products that are getting real old. By that time, consumers would be buying new GPU's.

4-I honestly don't see what you're proving to OP with 2014 benchmarks. Last time I checked, Crimson came in this year, days ago actually, and it seems that the new drivers, even prior to Crimson, have done alot of things for AMD on a positive note. So, yeah, I wanna see newer benchmarks.

Oh, no wait, hold on, saw it. Yeah, the 780 Ti pretty much as an advantage.

[Lays]

q

 

1- IMHO, they should've ditched Fermi when Maxwell came out. That way, allocation of manpower would've been for Kepler, Maxwell, and for the upcoming Pascal.

2- What about 2015 benchmarks? Thats 2014

3- When Greenland comes out, AMD should the rest of the XXXX cards. I honestly don't get why AMD or Nvidia have to allocate that much manpower and/or resources for products that are getting real old. By that time, consumers would be buying new GPU's.

4-I honestly don't see what you're proving to OP with 2014 benchmarks. Last time I checked, Crimson came in this year, days ago actually, and it seems that the new drivers, even prior to Crimson, have done alot of things for AMD on a positive note. So, yeah, I wanna see newer benchmarks.

 

 

If they ditched fermi we'd see shit tons of threads & posts about people moaning about that instead lol

 

There is no way to fix these people's opinions, they are biased or just looking for something to complain about

[CierroMierdo]

If they ditched fermi we'd see shit tons of threads & posts about people moaning about that instead lol

 

There is no way to fix these people's opinions, they are biased or just looking for something to complain about

I don't know. I personally think there's a crapton of corruption around, ranging from measely groups of friends, to corporate, and more commonly governments, AMD and Nvidia might as well just go full ass. There's gonna be people bitching about things here and there, so why not just do something as dickish as ditching Fermi for Nvidia's case and AMD go ditch 7000 series while the 200 series doesn't suffer from that to the point that it somewhat becomes a norm for people to upgrade every two, not three or more, years than that. But hey, the market is always unstable no matter what.

Maybe because AMD's been doing acceptable things nowadays while Nvidia's been having some scandals ranging from VRAMgate to whatever's going on with Nvidia right now on why there's hate on Nvidia on even reasons such as this. But oh well, seeing things like this happen to the point that an individual's form of leisure is also getting screwed with to a degree. I can somewhat sympathize with OP, but hey... if I was Nvidia and I had a strong following of fanboys to the point that whatever I do can't be wrong almost no matter what in terms of the business, I may as well should've just not gave a fuck about Kepler while I'm at it and they'd still go buy a Maxwell card. Thankfully, Nvidia's fanboys aren't totally that fucked up, or am I missing something?

But hey, thats just me saying "if you're gonna half-ass it, why not full-ass or no-ass it?"

[TheElt]

1- IMHO, they should've ditched Fermi when Maxwell came out. That way, allocation of manpower would've been for Kepler, Maxwell, and for the upcoming Pascal. Would make sense at some point.

2- What about 2015 benchmarks? Thats 2014

3- When Greenland comes out, AMD should the rest of the XXXX cards. I honestly don't get why AMD or Nvidia have to allocate that much manpower and/or resources for products that are getting real old. By that time, consumers would be buying new GPU's.

4-I honestly don't see what you're proving to OP with 2014 benchmarks. Last time I checked, Crimson came in this year, days ago actually, and it seems that the new drivers, even prior to Crimson, have done alot of things for AMD on a positive note. So, yeah, I wanna see newer benchmarks.

Oh, no wait, hold on, saw it. Yeah, the 780 Ti pretty much as an advantage.

1 - Yeah, it would have given them more development time for Kepler, and Fermi is past the maturity time for performance improvements, but like Lays said there'd be butthurt about ending support earlier than any architecture before, and there are still a fair few people running Fermi cards.

 

2 - There's no date for OP's results, the latest review from TPU on a 290 was 2014. Anandtech only have the 290X in their 2015 list, and at a rough guess I'd say the 290X comes out maybe a few % faster than the 780 Ti overall.

 

3 - They're still widely used cards, some of my friends still use the 7XXX cards and they're cheap 2nd hand for budget PCs. They all seem to have a cycle that makes everyone happy, a compromise that helps in places and not in others.

 

4 - Mainly parroting it so much to prove a point that one number from one place shouldn't be used as an end-all result, and that it varies from place to place.

[Lord_Karango17]

1.) What do you mean 3 different architectures? If you mean Pascal with Kepler and Maxwell, why neglect Kepler while Maxwell is still around? Pascal ain't even out yet. I'm not saying Fermi should be treated better, nor should it even be included in the "3 different architectures" thing, but come on... Kepler shouldn't entirely be behind Maxwell by that much.

2.) Again, tell me how is this possible? It looks like AMD cares even "without the budget".

3.) Maxwell's barely gonna be two years old, and its gonna get the same treatment. Does Nvidia expect everyone to upgrade from a 980 Ti to a 1080 Ti or non Ti?

4.) The 290X was released on October 24, 2013 and the 780 Ti was released on November 7, 2013, while the 290 was released two days prior to the 780 Ti. The 290X traded blows with the 780 Ti, despite the 780 Ti beating it, but again... how is this possible?

It isnt. My card beats a lot of the 970s out there in synthetics.

 

Kepler most definitely isn't dead, and getting no support from nvidia, but shouldn't newer architectures having good performance increases be a good thing?

[CierroMierdo]

1 - Yeah, it would have given them more development time for Kepler, and Fermi is past the maturity time for performance improvements, but like Lays said there'd be butthurt about ending support earlier than any architecture before, and there are still a fair few people running Fermi cards.

 

2 - There's no date for OP's results, the latest review from TPU on a 290 was 2014. Anandtech only have the 290X in their 2015 list, and at a rough guess I'd say the 290X comes out maybe a few % faster than the 780 Ti overall.

 

3 - They're still widely used cards, some of my friends still use the 7XXX cards and they're cheap 2nd hand for budget PCs. They all seem to have a cycle that makes everyone happy, a compromise that helps in places and not in others.

 

4 - Mainly parroting it so much to prove a point that one number from one place shouldn't be used as an end-all result, and that it varies from place to place.

1.) Welp, I'm not rich or anything, just working a 9-5 shift and all, but if the guys can a afford a Fermi GPU, two or three years worth of saving ain't too bad to save up for a new GPU. If I had a 580, I'd save up to buy a 980 or a 980 Ti and wait for whats coming out after Pascal and the refresh after that. As a consumer, I wouldn't mind Fermi being left in the dust. But yeah, not everyone's gonna have the same opinion on that matter as mine.

2.) I checked OP's link about it. It was posted on November 27, 2015. Seems recent, to me.

3.) Though I normally hate on the second hand market (as if silicone lottery wasn't bad enough, I could've been sold a component that has taken so much heat that its lifespan could be measured in months... I remember buying a second hand 4790k and it died out on stock voltage and stock everything in just 3 months. Had to buy a brand new 4790k and it still serves me well today. Its even overclocked), I don't really want it to end. I mean, I know some people are legit with what they sell as second-hand and have taken good care about it under their ownership prior to selling.

4.) Ehh, fair enough.

It isnt. My card beats a lot of the 970s out there in synthetics.

 

Kepler most definitely isn't dead, and getting no support from nvidia, but shouldn't newer architectures having good performance increases be a good thing?

Newer architectures having good performance increases are a good thing, but I think what OP's trying to say that the gap shouldn't be scary frightening. If it was Fermi VS Kepler, I wouldn't mind Fermi losing flat out as Kepler had the grunt to begin with, while Maxwell seemed like something that can be skipped entirely. IMHO, Maxwell seemed more like a software-sensitive line (by "sensitive" I mean like how Anakin Skywalker/Darth Vader is force sensitive, which explains why he's really good with the force) while doing with arguably less (prior to the 980 Ti and Titan X) in terms of, uhm, cuda cores, bit bus, TDP, and something else?

 

While, yes, Maxwell did perform better from the start to today, but if I owned a 780 Ti and a 970 was THAT close to it... I would've felt real bad. I mean, Maxwell cards OC like no tomorrow. It appears that silicone lottery can't fuck with Maxwell cards (I haven't heard or seen a 970 or a 980 get OC's lower than 14++), and despite that (correct me if I'm wrong) Maxwell's overclock scaling is allegedly crap compared to Kepler, it still ran away with it. If my 780 Ti was a loser in the Silicone lottery, I would flip my shit if I saw someone next to me with a 970.

If they wanted Maxwell to look real good, they should've axed Fermi or put it on some kind of legacy support or whatever so that atleast Maxwell would look that good.

[Lord_Karango17]

Newer architectures having good performance increases are a good thing, but I think what OP's trying to say that the gap shouldn't be scary frightening. If it was Fermi VS Kepler, I wouldn't mind Fermi losing flat out as Kepler had the grunt to begin with, while Maxwell seemed like something that can be skipped entirely. IMHO, Maxwell seemed more like a software-sensitive line (by "sensitive" I mean like how Anakin Skywalker/Darth Vader is force sensitive, which explains why he's really good with the force) while doing with arguably less (prior to the 980 Ti and Titan X) in terms of, uhm, cuda cores, bit bus, TDP, and something else?

 

While, yes, Maxwell did perform better from the start to today, but if I owned a 780 Ti and a 970 was THAT close to it... I would've felt real bad. I mean, Maxwell cards OC like no tomorrow. It appears that silicone lottery can't fuck with Maxwell cards (I haven't heard or seen a 970 or a 980 get OC's lower than 14++), and despite that (correct me if I'm wrong) Maxwell's overclock scaling is allegedly crap compared to Kepler, it still ran away with it. If my 780 Ti was a loser in the Silicone lottery, I would flip my shit if I saw someone next to me with a 970.

If they wanted Maxwell to look real good, they should've axed Fermi or put it on some kind of legacy support or whatever so that atleast Maxwell would look that good.

Yeah, I see what you mean... It does shock me a little when the 970 is only a few frames behind my card in games, where there is a pretty large gap in synthetics.

 

I guess my not minding the gap, is partially because I got my card for less than the price of a 970 (the retailer was getting rid of inventory is what I'm guessing), and every time I see the 970 close in benchmarks I don't feel bad because I did not spend more than the 970.

[CierroMierdo]

Yeah, I see what you mean... It does shock me a little when the 970 is only a few frames behind my card in games, where there is a pretty large gap in synthetics.

 

I guess my not minding the gap, is partially because I got my card for less than the price of a 970 (the retailer was getting rid of inventory is what I'm guessing), and every time I see the 970 close in benchmarks I don't feel bad because I did not spend more than the 970.

Thus why I somewhat sympathize with OP, but not so much. But yeah, a 970 getting that close... if I had a 480 and I didn't know Maxwell was gonna be that good and I bought something like a 780Ti on release, I'd feel ripped off. Thats a good chunk of cash I could've spent on buying games, some really fancy food, maybe a monitor, or a headphone. Maybe enduring crap settings to maintain a certain FPS wouldn't be so bad if I knew the 970 was gonna be that close to the Ti. But oh well, I'm farily happy with my 390. Its pretty good.