[Techno-Kitchen] GTX1060 3GB has major VRAM bottlenecks in Forza Horizon 3 & Mirror's Edge Catalyst

[antihelten]

What you’re searching for is especially hard to find in this comparison since a 460 768MB also has reduced pixel throughput and memory bandwidth. It’ll always be slower, and even if the separation seems to grow larger can you really say it’s from VRAM and not those other two differences? Can you explain why our comparisons are invalid?

But even with equal otherwise cards that only have different VRAM, good luck find benchmarks cross mixing settings. You’re liable to find an all Ultra test in which both cards are unplayable and this can draw a false conclusion that “the card is too slow to use more than xGB of VRAM”.

In reality, cross mixing settings is the best approach, and less VRAM often means a lower mixing average, mostly through textures. With my 2GB 675M Fermi that’s comparable to 460 256-bit (it’s actually a very under clocked 560 Ti to the point where it performs similar) I often turned down shadows and lighting but kept textures up which still resulted in relatively high VRAM usage (although less than all max of course).

It's unclear what the smoking gun to convince you is. The (nearly) equal card finding playable settings exceeding 768MB should be evidence enough for you, unless you feel that passing the VRAM barrier doesn't result in penalties (stutter!).

Sure the specs of the 460 768MB differed slightly from the 460 1GB in areas other than VRAM, but the 1060 3GB differs vastly more than this compared to the RX 470/480 4GB. So if it's difficult to draw conclusions about the 460 768MB vs. the 460 1GB, then by comparison it must be impossible to draw conclusions about the 1060 3GB vs. the RX 470/480 4GB. Obviously that isn't the case

As far as a valid comparison goes the spec difference between the 460 768MB and the 460 1GB could quite easily be compensated for. Simply take the difference in performance between the two at launch as the baseline and compare from there. At launch the 768MB was roughly 8% slower than the 1GB on average, so any significant deviation from this would likely be due to VRAM.

Point is though if evidence is as hard to locate as you guys claim, then why do you keep using the 460 768MB as an example of a VRAM limited card?

Remember that you were the one who brought up the 460 768MB as a VRAM limited card relative to it's 1GB competition, so the onus is on you to prove that it actually was limited, if you want people to take you seriously.

And no, showing that the 460 1GB suffers from VRAM issues shouldn't convince anyone, since it doesn't have anything to do with the issue being discussed. The issue isn't whether or not the 460 768MB can suffer from VRAM issues, of course it can, every card in existence can if you crank the settings high enough. The question is if the 460 768MB suffers from VRAM issues relative to its higher VRAM alternatives (the 460 1GB, or the 1GB 6850/6870 on the AMD side), and obviously if both cards are equally VRAM limited in a given game/settings, then it doesn't suffer relative to the 1GB versions.

All of the tests from this thread?

All of the tests illustrates that the 1060 3GB can suffer from VRAM limitations, they don't show that memory is the bottleneck in most cases however.

We just went through this, and it's fairly obvious that the 1060 3GB does not (currently) suffer from VRAM issues in the majority of games.

Remember the plural of anecdote is not data.

 

[OatisCampbell]

Well, everyone knows anyone buying in the 1060 3GB Price range runs DSR 4K with 8xMSAA. 🙄

I still don't get outside of "buy AMD, please - it's BETTER" is this an issue for anyone, particularly on this board. The people buying these cards probably have i3 or worse CPUs. I'm pretty sure they are more than happy with medium settings.

And before I get lumped into the "Pro-Nvidia," I didn't recommend this card to the people who asked me about it and I still won't. However, if someone buys it - welps, I'm sure they won't be crying on a forum they can't do 1080P Hyper Settings on MIrror's Edge 2.

What these benchmarks make me think is that the GTX1060 6GB is about the minimum any gamer would want to consider.

33% faster than 480 and they cost around $250? I'd think that will be a real problem for AMD as I don't think many computer gamers that read sites like this are looking for $165 cards anyway.

GPU is the most important part of a gaming rig besides display, not the place to cut corners.

Vega sometime soon please, getting tired of my 290.

 

[frozentundra123456]

What these benchmarks make me think is that the GTX1060 6GB is about the minimum any gamer would want to consider.

33% faster than 480 and they cost around $250? I'd think that will be a real problem for AMD as I don't think many computer gamers that read sites like this are looking for $165 cards anyway.

GPU is the most important part of a gaming rig besides display, not the place to cut corners.

Vega sometime soon please, getting tired of my 290.

1060 6gb is not 30% faster than 480. More like 10% or so I think, depending on the mix of games. In any case, there are a wide variety of PC gamers, and a lot of them don't have the high standards most on this forum have. I actually think the 1050Ti is probably the min level for decent 1080p gaming, and the sweet spot is the 470 4gb since they cut the price.

 

[crisium]

Sure the specs of the 460 768MB differed slightly from the 460 1GB in areas other than VRAM, but the 1060 3GB differs vastly more than this compared to the RX 470/480 4GB. So if it's difficult to draw conclusions about the 460 768MB vs. the 460 1GB, then by comparison it must be impossible to draw conclusions about the 1060 3GB vs. the RX 470/480 4GB. Obviously that isn't the case

As far as a valid comparison goes the spec difference between the 460 768MB and the 460 1GB could quite easily be compensated for. Simply take the difference in performance between the two at launch as the baseline and compare from there. At launch the 768MB was roughly 8% slower than the 1GB on average, so any significant deviation from this would likely be due to VRAM.

Point is though if evidence is as hard to locate as you guys claim, then why do you keep using the 460 768MB as an example of a VRAM limited card?

Remember that you were the one who brought up the 460 768MB as a VRAM limited card relative to it's 1GB competition, so the onus is on you to prove that it actually was limited, if you want people to take you seriously.

And no, showing that the 460 1GB suffers from VRAM issues shouldn't convince anyone, since it doesn't have anything to do with the issue being discussed. The issue isn't whether or not the 460 768MB can suffer from VRAM issues, of course it can, every card in existence can if you crank the settings high enough. The question is if the 460 768MB suffers from VRAM issues relative to its higher VRAM alternatives (the 460 1GB, or the 1GB 6850/6870 on the AMD side), and obviously if both cards are equally VRAM limited in a given game/settings, then it doesn't suffer relative to the 1GB versions.



All of the tests illustrates that the 1060 3GB can suffer from VRAM limitations, they don't show that memory is the bottleneck in most cases however.

We just went through this, and it's fairly obvious that the 1060 3GB does not (currently) suffer from VRAM issues in the majority of games.

Remember the plural of anecdote is not data.

I already addressed every point you made, anticipating. Including in what you quoted. My quote is already a response to your dismissive off track beat. Furthermore, two users using 460 equivalent for years means nothing to you. This speaks volumes. You're now entering a hard-liner stance, it is clear you will never budge.

 

[OatisCampbell]

1060 6gb is not 30% faster than 480. More like 10% or so I think, depending on the mix of games. In any case, there are a wide variety of PC gamers, and a lot of them don't have the high standards most on this forum have. I actually think the 1050Ti is probably the min level for decent 1080p gaming, and the sweet spot is the 470 4gb since they cut the price.

I was looking at the difference in minimums/480 fps = 31%. Should I be saying "the 480 performance is 77% of the 1060"?

Either way, that is a HUGE amount of performance difference for under $100.. There's a good 6GB 1060 on newegg today for $240 AMIR, that's $75 more than the really cheap 480. Basically the cost of a game and a good six pack.

The 1060 would definitely have greater longevity.

Anyway, like I said, waiting on Vega to see what to do next. Hopefully AMD brings new bang for buck champ, they don't have anything I'd sell my custom cooled 290 for now.

 

[zinfamous]

I was looking at the difference in minimums/480 fps = 31%. Should I be saying "the 480 performance is 77% of the 1060"?

Either way, that is a HUGE amount of performance difference for under $100.. There's a good 6GB 1060 on newegg today for $240 AMIR, that's $75 more than the really cheap 480. Basically the cost of a game and a good six pack.

The 1060 would definitely have greater longevity.

Anyway, like I said, waiting on Vega to see what to do next. Hopefully AMD brings new bang for buck champ, they don't have anything I'd sell my custom cooled 290 for now.

And yet this 480 is shaping up to be faster than any 1060 6gb released to date. At least, that's what the benchmarks are showing. THis thing is slightly edging out 390X too, I believe.

I don't know why, I'm not sure if anyone really knows why. I'm not saying this one model proves that 480 is just overall better than any 1060, but XFX has managed to make this thing outperform anything in its class.

Also, a week ago, you could get it on sale for about 250 and with various codes and rebates, people were grabbing it for ~$235. Probably won't be the last time it will be found at that price.

Like you, I'm also "waiting for Vega," but this XFX GTR 480 at that $235 price and easily available was very tempting, but I still want to spend that much to step up from my 280X, which is still managing to run the things I throw at it just fine.

I still don't get the 1060 will have "greater longevity" argument, considering the 2 successive generations of history we have with nVidia and their lack of support for their cards. Sure, one can say "you can't possibly know for a fact that 1060 will fall off a cliff compared to the 480!" Sure, of course you can't--but it would be foolish to ignore what we actually do know about this situation and what nVidia/AMD have done, repeatedly, with their products. On top of this, Pascal, while very good, seems to be following the Maxwell trend and DX12 really does underperform here compared to AMD. Those are great cards and I think it's fair to see a real improvement over Maxwell, but I don't think Pascal is going to be the "future proof" option that Volta might be. But then again, nVidia has essentially zero history of being concerned about future proofing. It just isn't their strategy (much to their investor's delight)

 

[happy medium]

And yet this 480 is shaping up to be faster than any 1060 6gb released to date. At least, that's what the benchmarks are showing. THis thing is slightly edging out 390X too, I believe.

I don't know why, I'm not sure if anyone really knows why. I'm not saying this one model proves that 480 is just overall better than any 1060, but XFX has managed to make this thing outperform anything in its class.

Also, a week ago, you could get it on sale for about 250 and with various codes and rebates, people were grabbing it for ~$235. Probably won't be the last time it will be found at that price.

Like you, I'm also "waiting for Vega," but this XFX GTR 480 at that $235 price and easily available was very tempting, but I still want to spend that much to step up from my 280X, which is still managing to run the things I throw at it just fine.

I still don't get the 1060 will have "greater longevity" argument, considering the 2 successive generations of history we have with nVidia and their lack of support for their cards. Sure, one can say "you can't possibly know for a fact that 1060 will fall off a cliff compared to the 480!" Sure, of course you can't--but it would be foolish to ignore what we actually do know about this situation and what nVidia/AMD have done, repeatedly, with their products. On top of this, Pascal, while very good, seems to be following the Maxwell trend and DX12 really does underperform here compared to AMD. Those are great cards and I think it's fair to see a real improvement over Maxwell, but I don't think Pascal is going to be the "future proof" option that Volta might be. But then again, nVidia has essentially zero history of being concerned about future proofing. It just isn't their strategy (much to their investor's delight)

This is the fastest 1060 , it will sustain 2100 boost and smoke that 480.
http://www.newegg.com/Product/Produ...4125904&cm_re=gtx_1060-_-14-125-904-_-Product

https://www.techpowerup.com/reviews/Gigabyte/GTX_1060_Xtreme_Gaming/29.html

 

[zinfamous]

This is the fastest 1060 , it will sustain 2100 boost and smoke that 480.
http://www.newegg.com/Product/Produ...4125904&cm_re=gtx_1060-_-14-125-904-_-Product

https://www.techpowerup.com/reviews/Gigabyte/GTX_1060_Xtreme_Gaming/29.html

would love to see the GTR in that benchmark. OC vs OC is completely irrelevant if it means nothing regarding actual performance. I've yet to see a single 1060 where the ~6% max OC significantly outperforms the comparable ~4-5% OC 480s.

In other words, need a tangible reason for "better OC" to be an actual metric that reflects performance.

 

[PhonakV30]

BF3 ? Really? Just choose latest game.not old game.

 

[IllogicalGlory]

would love to see the GTR in that benchmark. OC vs OC is completely irrelevant if it means nothing regarding actual performance. I've yet to see a single 1060 where the ~6% max OC significantly outperforms the comparable ~4-5% OC 480s.

In other words, need a tangible reason for "better OC" to be an actual metric that reflects performance.

Polaris cards get crushed in BF4 no matter what the clock speed, and Pascal seems especially good at BF3. However, just showing that benchmark is, of course, not the whole story. The lead an OC'ed 480 has in Doom is about the same as the OC'ed 1060 has in BF3, for example. On average, the 1060 is about 5-6% ahead at stock, and to be fair, it does overclock better. I do think that at this point, the 1060 6GB is still a better purchase at the same price.

 

[Mikeduffy]

BF3 in a test suite is exactly what's wrong with TPU - they test old games and assume that performance in this title is representative of 2016 games.

When looking at BF4, it's reasonable to think that GCN won't look as good in DX12 because of Mantle - they simply won't have the level of dx11 driver optimization in this title.

A far better approximation of real-world GPU perforance would be looking at AA/AAA games realeased withing the past 12 months - I dare someone to argue against this.

I'm just so sick of Nvidiots using benches from games that are no longer relevant in 2016!

Warning issued for inflammatory language.
-- stahlhart

 

[IllogicalGlory]

BF3 in a test suite is exactly what's wrong with TPU - they test old games and assume that performance in this title is representative of 2016 games.

When looking at BF4, it's reasonable to think that GCN won't look as good in DX12 because of Mantle - they simply won't have the level of dx11 driver optimization in this title.

A far better approximation of real-world GPU perforance would be looking at AA/AAA games realeased withing the past 12 months - I dare someone to argue against this.

I'm just so sick of Nvidiots using benches from games that are no longer relevant in 2016!

They have actually removed it, though very recently. Anno 2205 should probably removed at some point, probably Assassin's Creed Syndicate, No Man's Sky and Far Cry Primal as well. BF1, Gears of War 4 and Titanfall 2 could be a couple of good replacements. I wouldn't object to Hitman being taken out, as it heavily favors AMD, but it does have a fair number of active players.

 

[railven]

What these benchmarks make me think is that the GTX1060 6GB is about the minimum any gamer would want to consider.

33% faster than 480 and they cost around $250? I'd think that will be a real problem for AMD as I don't think many computer gamers that read sites like this are looking for $165 cards anyway.

GPU is the most important part of a gaming rig besides display, not the place to cut corners.

Vega sometime soon please, getting tired of my 290.

While I agree with that statement, often times at least in my experience - this is where most people try to cut corners next to the CPU.

When you look at options, moving from one tier in a GPU or CPU buying decision is roughly $60-100 minimum (this doesn't including swapping brands), I've seen people drop the GPU to squeeze in a faster CPU or a "name-brand" PSU.

For friends/family I've often covered the difference bumping them up on the CPU so they can transfer that over to the GPU.

 

[antihelten]

I already addressed every point you made, anticipating. Including in what you quoted. My quote is already a response to your dismissive off track beat. Furthermore, two users using 460 equivalent for years means nothing to you. This speaks volumes. You're now entering a hard-liner stance, it is clear you will never budge.

You anticipated nothing and offered nothing but non-sequitors and red herrings and moving of goalposts.

At the end of the day the fact remains that you claimed the 460 768MB was limited relative to 1GB cards, and yet you haven't been able to provide a single shred of evidence to support this.

I'm not entering a hard-liner stance, since my stance is exactly the same as it was to begin with, i.e. show me that a 768MB card is VRAM limited relative to a 1GB card. I suppose you could argue that my initial stance was a hard-liner stance, but seeing as my initial stance was simply about asking you for evidence to support your claims (which you still haven't provided), I would feel sorry for you if you actually felt that people take a hard-line stance simply by asking you for evidence for your claims.

The fact that you apparently refuse to understand the issue (it's not about a 768MB card being VRAM limited in general, but rather about it being VRAM limited specifically in relation to a 1GB card), speaks volumes.

Either way this is getting off topic, and it's clear that you're not willing to actually provide any evidence to back up your claim, so let's just leave it.

 

[crisium]

BF3 in a test suite is exactly what's wrong with TPU - they test old games and assume that performance in this title is representative of 2016 games.

It has been odd that TPU have kept BF3 so long. They used CoD 4 Modern Warfare from late 2007 to late 2011, and I was relieved when they updated it. But now we are stuck on 5 years of BF3. I suppose it's to create a sense of continuity for comparing cards over the years, but now I think it's approaching hyperbole to draw actionable results off of such an old game now. What's even funnier is that they removed BF3 from their normal benchmark, so it's exclusively tested for OC now.

Sweclockers tests OC results in multiple games, luckily.

 

[MongGrel]

My old ASUS R9 280 DCU II TOP has 3 Gb, I would not even think of buying a card with only 3 for as an upgrade these days personally.

 

[crisium]

Text.

I already explained why what you are looking for is virtually impossible to find. It's almost a parody (are you trolling now?) as I explicitly state that there are 3 differences between the 460 versions, and how even finding an increasing gap over time cannot necessarily be tied to the VRAM and not pixel throughput or memory bandwidth. And then your very next post, where you quote me saying this, is you ironically saying that if we find differences over time it _must be_ from the VRAM. If I accepted your flawed premise, actually, I could prove myself right.

1GB version is only 8.6% faster in Summer 2010:
https://tpucdn.com/reviews/Zotac/GeForce_GTX_460_1_GB/images/perfrel_1920.gif

But now, in Spring 2012 it's 11.5% faster:
https://www.techpowerup.com/reviews/ASUS/HD_7870_Direct_Cu_II/26.html

I guess that settles it then, by your metric. I'll stick to 3+ years of using what's virtually a 460 2GB and being able to use more than 1GB, never mind 768MB, at playable settings. But I'm glad we settled things by your ironclad standards that give zero credence to anything I say.

 

[antihelten]

I already explained why what you are looking for is virtually impossible to find. It's almost a parody (are you trolling now?)

If it's so impossible to find any evidence to support your claim, then why did you even make said claim in the first place? could it be that you're were the one trolling?

as I explicitly state that there are 3 differences between the 460 versions, and how even finding an increasing gap over time cannot necessarily be tied to the VRAM and not pixel throughput or memory bandwidth. And then your very next post, where you quote me saying this, is you ironically saying that if we find differences over time it _must be_ from the VRAM.

So now you have been reduced to outright lying I see. I never said that differences _must be_ from the VRAM. Quite the contrary, here is what I said:

"At launch the 768MB was roughly 8% slower than the 1GB on average, so any significant deviation from this would likely be due to VRAM."

Who's trolling now?

If I accepted your flawed premise, actually, I could prove myself right.

1GB version is only 8.6% faster in Summer 2010:
https://tpucdn.com/reviews/Zotac/GeForce_GTX_460_1_GB/images/perfrel_1920.gif

But now, in Spring 2012 it's 11.5% faster:
https://www.techpowerup.com/reviews/ASUS/HD_7870_Direct_Cu_II/26.html

I guess that settles it then, by your metric. I'll stick to 3+ years of using what's virtually a 460 2GB and being able to use more than 1GB, never mind 768MB, at playable settings. But I'm glad we settled things by your ironclad standards that give zero credence to anything I say.

So finally, after all this back and forth bickering we can conclude that the 460 768MB lost 2.6% performance over 2 years relative to the 1GB version.

So in other words losing 2.6% performance constitutes a "Huge red flag" in your opinion.

You are certainly free to bitch and moan about losing 2.6% performance and calling it a huge red flag, but when that is all you have to complain about, then it's obvious that you don't have a case.

 

[crisium]

bitch and moan

This is what happens when one user tries to have a discussion, and another tries to have an argument. Insults, and everything has to be about meeting your criteria (benchmarks). It's not about 2.6%, or anything in these low information benchmarks that don't illustrate the stutter from running out of VRAM. It's about additional settings that have to be lowered simply due to one unique bottleneck, the VRAM.

That IS a reliable way of judging VRAM requirements. I did not max out these games, of course I had to juggle settings for the right balance. But the settings I landed at resulted in over 768MB of usage - in BF4 I used just over 1GB actually. Thus if I had a lower VRAM, but similar performance card, I would have had to make MORE sacrifices.

This is the whole extra compromise that VRAM limited cards offer that I am talking about. I'm not sure what you are fishing for.

VRAM is bottleneck compromise that means similar cards, for similar money, or even slower cards, can enjoy better IQ. I find this a unique compromise that exists outside of settling for performance because it's your max price.

As easy as it is to say "turn down the settings", when you realize texture settings from lowest to highest are often only single digit % difference, it seems like the most artificial handicap in the world.

I'm discussing the uniqueness of VRAM penalties, and you're off in insult land trying to fit me into your goal post of benchmarks that don't even illustrate the issue. You're welcome to pivot back to an actual discussion whenever you like.

 

[antihelten]

This is what happens when one user tries to have a discussion, and another tries to have an argument. Insults, and everything has to be about meeting your criteria (benchmarks).

It's quite cute how you are complaining about me insulting you, when you were the one who started by accusing me of being a troll, but hey a little hypocrisy never hurt anyone right.

And you're right this isn't much of a discussion, but it isn't really an argument either. It's simply a case of someone (you) making a claim and then refusing to back it up with evidence, when asked.

Of course if you think that the criteria of actually having to provide evidence to support one's claims is too much, then you probably shouldn't be on a technical discussion forum in the first place.

It's not about 2.6%, it's about additional settings that have to be lowered simply due to one unique bottleneck, the VRAM.

Great then please provide some evidence to back up this claim. In other words just provide some evidence to back up that the 460 768MB has to lower settings compared to the 460 1GB to compensate for the lower VRAM, or is this another one of those claims where the evidence is impossible to find?

I'm discussing the uniqueness of VRAM penalties, and you're off in insult land trying to fit me into your goal post of benchmarks that don't even illustrate the issue.

It's quite astonishing how you still don't understand what the issue is. You used a 768MB card (the 460 768MB) as an example of a card that was worse of compared to a number of 1GB cards due to the lower amount of VRAM. Obviously the only way to illustrate this is to compare them, not to look at them in isolation. So taking the 460 768MB in isolation and showing that it can suffer from VRAM limitations obviously doesn't prove anything (since the 1GB may also suffer from equal limitations in the same situation), on the other taking the 460 768MB and showing that it suffers from VRAM limitations in situations where the 1GB does not (or does so to a lesser degree) would prove the issue.

Anyway this has gone quite off topic, and it's become quite evident that you are not willing to actually back up your claims with evidence, so unless you are actually willing to start doing so, I'm not going to bother responding to you anymore.

 

[crisium]

GTX 460 768MB usage in reviews is almost non existent past 2010. The 3 games I cited with my card were 2013 releases, but it is possible to find an example at launch:

http://techreport.com/review/19242/nvidia-geforce-gtx-460-graphics-processor/8

http://www.hardwarecanucks.com/foru...force-gtx-460-cyclone-768mb-oc-review-15.html

Naturally, VRAM usage increases over time. While overall game performance demands also increase, the common sense approach becomes to cross mix settings and VRAM causes limitations with this, which I was lucky enough to escape in 2013+ with my essentially GTX 460 2GB (kept Ultra textures, not so much lighting).

Having less VRAM than both contemporary competitors and older, slower cards is symptomatic. And seeing performance like this at launch is worrisome enough, and a warning to those looking to keep the card for years.

 

[antihelten]

GTX 460 768MB usage in reviews is almost non existent past 2010. The 3 games I cited with my card were 2013 releases, but it is possible to find an example at launch:

http://techreport.com/review/19242/nvidia-geforce-gtx-460-graphics-processor/8

Naturally, VRAM usage increases over time. While overall game performance demands also increase, the common sense approach becomes to cross mix settings and VRAM causes limitations with this, which I was lucky enough to escape in 2013+ with my essentially GTX 460 2GB (kept Ultra textures, not so much lighting).

Having less VRAM than both contemporary competitors and older, slower cards is symptomatic. And seeing performance like this at launch is worrisome enough, and a warning to those looking to keep the card for years.

Finally, why didn't you just post this when I first asked you, clearly it wasn't so impossible to find the evidence after all.

This quite clearly shows that the 460 768MB takes a huge performance hit when running DX11 at 1600P and 8xAA, and given that it does just fine at a lower resolution (1080P), or with lower AA (4xAA) this would appear to be a case of VRAM limitation. It's of course notable that the 460 768MB does just fine at 1600P when using DX9, so it would appear that the culprit is the added effects from DX11 (which include tessellation, HDAO and some filters).

Of course one might argue that 1600P with 8xAA isn't something the 460 768MB was ever designed for, and I would suspect that very few people who bought this card would even own such monitors, but it does at least illustrate that under extreme circumstances, having only 75% of the available VRAM can lead to a severe performance hit (about 50% relative to the normal gap between the 460 768MB and 460 1GB), necessitating lowering of settings (either resolution from 1600P to 1080P, AA from 8x to 4x, or API from DX11 to DX9).

 

[crisium]

Don't forget, yesterday's 1600p and 8x AA is tomorrow's 1080p. Over time, the hardware demands of what seems hyperbolic settings becomes routine.

I think you're right about DX11 upping the ante. Since DX11 really came into mainstream not long after the 460 launched, this could be a case of a mistimed market product for someone who wanted to keep the card for a long time (or perfectly timed for a stopgap upgrade user).

Unfortunately, the GTX 460 768mb rapidly dissipated from the reviewer landscape as quickly as it entered it so there's a dearth of ways to illustrate this without finding an end user, but I take the reasonable approximation of my own experience in 2013 as the closest I can find for now.

 

[Dave2150]

And yet this 480 is shaping up to be faster than any 1060 6gb released to date. At least, that's what the benchmarks are showing. THis thing is slightly edging out 390X too, I believe.

I don't know why, I'm not sure if anyone really knows why. I'm not saying this one model proves that 480 is just overall better than any 1060, but XFX has managed to make this thing outperform anything in its class.

Also, a week ago, you could get it on sale for about 250 and with various codes and rebates, people were grabbing it for ~$235. Probably won't be the last time it will be found at that price.

Like you, I'm also "waiting for Vega," but this XFX GTR 480 at that $235 price and easily available was very tempting, but I still want to spend that much to step up from my 280X, which is still managing to run the things I throw at it just fine.

I still don't get the 1060 will have "greater longevity" argument, considering the 2 successive generations of history we have with nVidia and their lack of support for their cards. Sure, one can say "you can't possibly know for a fact that 1060 will fall off a cliff compared to the 480!" Sure, of course you can't--but it would be foolish to ignore what we actually do know about this situation and what nVidia/AMD have done, repeatedly, with their products. On top of this, Pascal, while very good, seems to be following the Maxwell trend and DX12 really does underperform here compared to AMD. Those are great cards and I think it's fair to see a real improvement over Maxwell, but I don't think Pascal is going to be the "future proof" option that Volta might be. But then again, nVidia has essentially zero history of being concerned about future proofing. It just isn't their strategy (much to their investor's delight)

Please show us all the benchmarks where this RX480 is significantly faster than a 390X and other RX480's.

 

[antihelten]

Don't forget, yesterday's 1600p and 8x AA is tomorrow's 1080p. Over time, the hardware demands of what seems hyperbolic settings becomes routine.

I think you're right about DX11 upping the ante. Since DX11 really came into mainstream not long after the 460 launched, this could be a case of a mistimed market product for someone who wanted to keep the card for a long time (or perfectly timed for a stopgap upgrade user).

Unfortunately, the GTX 460 768mb rapidly dissipated from the reviewer landscape as quickly as it entered it so there's a dearth of ways to illustrate this without finding an end user, but I take the reasonable approximation of my own experience in 2013 as the closest I can find for now.

As far as I can tell in the context of this thread, yesterdays 1600P and 8xAA is also todays 1600P and 8xAA (or 1440P given the death of the 16:10 format). The 460 768MB was targeted at 1080P (possibly with a splash of AA) and the 1060 3GB is also targeted at 1080P (possibly with a splash of AA). When the 460 768MB was launched market share of 1600P/1440P monitors were somewhere in the 1-2% range and today they are still in this range . So really nothing appears to have changed in this regard, and likely won't be changing anytime soon either.

Probably the only significant change that has happened resolution wise is the adoption of 4K monitors. They still remain quite rare, but they will likely see accelerating adoption in the future, especially with the new consoles targeting 4K. Of course that's rather irrelevant in this context, since no one buying a 1060 3GB (or any of the other cards mentioned in this thread for that matter) would ever reasonably expect to use that to drive a 4K display.