[Eurogamer] GTX 1060: 3 GB vs 6 GB

[frozentundra123456]

If one reads the article itself, they specifically state that they cant explain the low memory usage of the 390x, since if one adds together the dedicated and dynamic memory usage of the FuryX it is very close to the 980Ti. Hard to find exact comparisons, but overall, it also looks like the 980Ti is also about 50% faster than the 390x and 20% faster than the FuryX.

 

[Keysplayr]

Out of interest why are you argueing so much about 3GB being enough,when you have a 12GB Titan X?? Surely you could have bought a 6GB GTX980TI and saved a ton of money??

Start a new thread. That is waaaay off topic and will derail.

 

[antihelten]

Feel free to provide any other sources. The only conclusion I've drawn so far has been the 4GB HBM isn't limiting the Fury, as it holds up just as well as the 12GB Titan X. Wonder what happens to the 980 TI @ 2/4x SSAA @ 4k? They didn't publish those results.

I don't have any other sources, as these kind of tests tend to be the only ones that can be easily performed by tech sites. I simply wanted to point out that you should be careful with these kind of tests.

As for what happens with the 980 at 2-4X SSAA at 4K, I would assume absolutely nothing, just like nothing happens when they moved from 0x to 4x SSAA at 1440P or from 1440P to 4K. The 980 Ti already has 6GB of memory allocated and thus can't go any higher. Likewise I don't expect anything special to happen to performance, seeing as nothing special happened at 1440P (both the 980 Ti and the 390X saw a 55% performance hit when moving from 0x to 4x SSAA)

 

[Sweepr]

Start a new thread. That is waaaay off topic and will derail.

Not just off topic, but an user callout as well. So you can't recommend a card with less VRAM just because you own something better?

 

[Keysplayr]

Not just off topic, but an user callout as well. So you can't recommend a card with less VRAM just because you own something better?

Yeah I know I know. I just wont be sucked in anymore.

 

[USER8000]

Start a new thread. That is waaaay off topic and will derail.

Wait so basically you are argueing for a 3GB card,when you yourself bought a 12GB card instead of a 6GB one,and now when that is highlighted you don't want to answer.

Aftermarket GTX980TI cards are faster than a Titan X.

You got a slower card since it had more VRAM.

So you are so scared to answer a simple question you get your mate to say YOU are a victim,and then try to duck the reason why you bought the card.

Some people here were insistent a 2GB GTX960 was fine and yet it appears the GTX960 4GB I got was a better purchase.

Its very sad when hardware enthusiasts would rather get people to waste their money.

Yeah I know I know. I just wont be sucked in anymore.

Then why are you ardently argueing for a 3GB GTX1060 instead of the 6GB GTX1060??

Or is everybody who questions you somebody who hates you??

Edit to post.

The only reason you want to play the victim is so you can censure people who question you.

That is why you get your mate to try and spin the same yarn - I have lurked long enough on here for 10 years to see the tactics you are trying to play here.

I am glad I ignored people like you - at least my GTX960 4GB has been proven to have more lasting power than the 2GB version.

My previous cards - GTX660TI and now a GTX960 4GB.

I am the target market for this card - NOT you.

But I guess this is the internet and memories are short.


Personal attacks are not allowed
Markfw900

 

[Bacon1]

I feel free. Let's try something a bit more recent than the old test you cherry picked:

How is it cherry picked? It shows a lot more data than yours does, and actually hits the full card limits in multiple cases. Not to mention the game isn't beta and missing most of the actual game (possibly including some graphic settings as mentioned by Digital Foundry in their testing).

Wonder how this quote fits into your graph:

s it "maxxed" out, as you put it? Or most efficiently utilized? Whos to say...

Gotta say that test would be more meaningful with 3, 4 or even 6GB cards since its not using close to 8GB @ 4k.

 

[USER8000]

How is it cherry picked? It shows a lot more data than yours does, and actually hits the full card limits in multiple cases. Not to mention the game isn't beta and missing most of the actual game (possibly including some graphic settings as mentioned by Digital Foundry in their testing).

Wonder how this quote fits into your graph:



Gotta say that test would be more meaningful with 3, 4 or even 6GB cards since its not using close to 8GB @ 4k.

Its pointless argueing - you might as well hit your head against a brick wall.

Once you find an inconsistency in their arguments,they will start "playing the victim card".

It will continue until the cards hit VRAM limits,but by then there will be no doubt a replacement card with more VRAM.

BTW,you might want to see a very famous example of VRAM limitations within a year of launch.

The good old 8800GT 256MB with 112 shaders(G92 core). Look what the same generation G94 core did with 64 shaders,a 10% clockspeed bump and 512MB of VRAM:

http://www.anandtech.com/show/2453

Plenty of us who have been enthusiasts remember what happened to the 8800GT 256MB over time.

How can anybody who has been an enthusiast long enough,forget it??

I know friends who bought that card due to forum "experts" who regretted it not spending extra on the 8800GT 512MB.

It was repeated with other cards from ATI,AMD and Nvidia too which were VRAM limited.

So,despite this people still argue for getting a GTX1060 3GB over a GTX1060 6GB.

How can anybody with a straight face say a 3GB £200 card is fine in 2016 when the 6GB version with 10% more shaders is only £40 more??

Plus who would want the GTX1060 3GB in a few years - at least the GTX1060 6GB is going to look somewhat better if you want to give it to somebody.

 

[antihelten]

How can anybody with a straight face say a 3GB £200 card is fine in 2016 when the 6GB version with 10% more shaders is only £40 more??

Because it's £40/$50 more. It's not rocket science, more is more, and in this case some people may find a price hike of 25% for a 5% increase in performance (and an unknown increase in the future) to be too much.

 

[USER8000]

Because it's £40/$50 more. It's not rocket science, more is more, and in this case some people may find a price hike of 25% for a 5% increase in performance (and an unknown increase in the future) to be too much.

Its not rocket science,VRAM limited cards have shown much less lasting ability over normal versions. Hardware enthusiasts on forums who make arguments for the following cards,all went quiet when the cards hit issues:
1.)8800GT 256MB
2.)8800GTS 320MB
3.)GTX460 768MB
4.)HD3850 256MB
5.)HD6950 1GB
6.)HD7850 1GB

Probably a few others too which I missed out.

I knew mates who went with those cards due to the "its enough VRAM" brigade and regretted they did not pony up the extra cash.

Its more expensive to buy a new card.

Those enthusiasts on forums made them waste their money.

There has been plenty of evidence,to show those cards and this card have big issues when they hit VRAM limits.

EVERYTIME these sorts of cards are released,people go on how perfectly fine they will be,and almost everytime they don't last as long. OFC,being the internet people conveniently forget cards,and then restart the same thing for the next round.

People keep cards for years - two to three years. £200 is considered the start of the "enthusiast" pricing range for cards. £200 is the sort of money which buys who a Core i5 6600 or a 6600K,and only people buying decent rigs have those sorts of CPUs.

I have not met anybody in 15 years who could spend £200 on a card,could not find £230 for a better card.

People who generally are that tight on money,tend to be people buying £100 CPUs like a Core i3 and £100 to £150 cards - this is what the GTX750TI,RX460 and GTX1050 4GB are aimed for.

"Those who cannot remember the past are condemned to repeat it."

I suspect once these cards hit their VRAM limits,if anybody points it out,I suspect they will be accused of being unfair or calling out or something else.

Its going to be so interesting to look back at this thread in another year to 18 months time.

 

[antihelten]

Its not rocket science,VRAM limited cards have shown much less lasting ability over normal versions. Hardware enthusiasts on forums who make arguments for the following cards,all went quiet when they cards hit issues:
1.)8800GT 256MB
2.)8800GTS 320MB
3.)GTX460 768MB
4.)HD3850 256MB
5.)HD6950 1GB
6.)HD7850 1GB

Probably a few others too which I missed out.

There has been plenty of evidence,to show those cards and this card have big issues when they hit VRAM limits.

EVERYTIME these sorts of cards are released,people go on how perfectly fine they will be,and almost everytime they don't last as long. OFC,being the internet people conveniently forget cards,and then restart the same thing for the next round.

People keep cards for years - two to three years. £200 is considered the start of the "enthusiast" pricing range for cards. £200 is the sort of money which buys who a Core i5 6600 or a 6600K,and only people buying decent rigs have those sorts of CPUs.

I have not met anybody in 15 years who could spend £200 on a card,could not find £230 for a better card.

People who generally are that tight on money,tend to be people buying £100 CPUs like a Core i3 and £100 to £150 cards - this is what the GTX750TI,RX460 and GTX1050 4GB are aimed for.

"Those who cannot remember the past are condemned to repeat it."

I suspect once these cards hit their VRAM limits,if anybody points it out,I suspect they will be accused of being unfair or calling out or something else.

Its going to be so interesting to look back at this thread in another year to 18 months time.

Great, so now we're back to your flawed arguments from page 1, btw you still haven't provided a link showing that the 460 768MB collapsed compared to the 460 1GB that I asked for, I would still love to see that.

In other news Guru3D has their review up. Their Hitman and Tomb Raider number seems to support the notion that TPU's numbers are outliers.

Guru3D is also interesting because they provide FCAT results, all in all there isn't much of interest to report here though. The GTX 1060 3GB doesn't really show any serious cases of stutter apart from scene changes, which is to be expected. It does a bit worse than the RX 470 in Hitman, whilst the RX 470 in turn does a bit worse in The Division. All in all nothing to write home about.

Once again the 1060 6GB is 6% faster on average, and the RX 470 is 5% slower (the 1060 3GB and the RX 470 are neck and neck across the 5 DX12/Vulkan games btw.).

 

[Keysplayr]

How is it cherry picked? It shows a lot more data than yours does, and actually hits the full card limits in multiple cases. Not to mention the game isn't beta and missing most of the actual game (possibly including some graphic settings as mentioned by Digital Foundry in their testing).

Wonder how this quote fits into your graph:



Gotta say that test would be more meaningful with 3, 4 or even 6GB cards since its not using close to 8GB @ 4k.

Well, these aren't 4K cards we are talking about so, I guess 4K wouldn't matter. 1080p and certain games at 1440 is more like it.

 

[Keysplayr]

Great, so now we're back to your flawed arguments from page 1, btw you still haven't provided a link showing that the 460 768MB collapsed compared to the 460 1GB that I asked for, I would still love to see that.

In other news Guru3D has their review up. Their Hitman and Tomb Raider number seems to support the notion that TPU's numbers are outliers.

Guru3D is also interesting because they provide FCAT results, all in all there isn't much of interest to report here though. The GTX 1060 3GB doesn't really show any serious cases of stutter apart from scene changes, which is to be expected. It does a bit worse than the RX 470 in Hitman, whilst the RX 470 in turn does a bit worse in The Division. All in all nothing to write home about.

Once again the 1060 6GB is 6% faster on average, and the RX 470 is 5% slower (the 1060 3GB and the RX 470 are neck and neck across the 5 DX12/Vulkan games btw.).

Didn't the Eurogamer article mention that even with more memory, the AMD cards seemed to stutter more than the 1060 3GB?

 

[antihelten]

Didn't the Eurogamer article mention that even with more memory, the AMD cards seemed to stutter more than the 1060 3GB?

Yes, in assassin's creed unity. I imagine that if you look hard enough you could probably find a handful of games for both the 1060 3GB and the RX 470 with stutter, but I haven't seen any cases where I would consider the stutter to be serious.

 

[Dribble]

Its not rocket science,VRAM limited cards have shown much less lasting ability over normal versions. Hardware enthusiasts on forums who make arguments for the following cards,all went quiet when the cards hit issues

What does "lasting ability" mean? Does it mean you can't play the game? I think not, all it means is you might have to drop texturing or anti-aliasing a notch, which to most people who aren't studying screen shots will look near identical. Every game you can play on a 6 Gb 1060 will play just fine on a 3 Gb one.

 

[Madpacket]

What's the big deal here?

The gimped 3GB 1060 has enough GDDR5 RAM to handle the cards fillrate in most scenarios / most games. The price delta between the 3 and 6GB could be used for a larger SSD, faster processor, or more system memory. Heck perhaps even a decent meal.

There will always be super ultra high resolution texture packs and suplerflous sucky settings which make an almost a negligible difference in image quality once the game is in motion. Last time I checked we don't play screen shots.

Game developers artificially push the limits of the VRAM due to consoles, vendors/manufacturers try and justify that extra memory which really only adds a few bucks to the BoM, in turn we end up paying a huge premium for.

HardOCP already proved with the 4GB FuryX that with a little extra development you can make games work just fine even at 4K. Beta games don't count sorry. 1060 isn't really capable of 4K gaming.

Where that extra VRAM *might* make a difference is at a pixel density that stretches past the cards capabilities anyway. 4K is where that 4GB+ of VRAM starts to come in handy or in SLI scenarios (which the 1060 has been stripped of anyway).

Tldr;

The 3GB 1060 will likely work just fine until Nvidia fires up the planned obsolescence driver machine in order to sell their next batch of cards. If you're allergic to changing a few sliders in an options panel, skip that dinner and buy the 6GB version.

 

[boozzer]

the million dollar questions.

how is 960 2gb doing now?

how well do you think 1060 3gb will do in a year? using the 960 2gb data as a precedent?

there are some evil people in this thread.


Trolling and insulting a lot of members is not allowed
Markfw900

 

[justin4pack]

the million dollar questions.

how is 960 2gb doing now?

how well do you think 1060 3gb will do in a year? using the 960 2gb data as a precedent?

there are some evil people in this thread.

The benchmarks I could find show it runs as well as the 4gb counterpart with current gen games. Can u link some otherwise that show what your refrencing.

Sent from my VS985 4G using Tapatalk

 

[frozentundra123456]

Here is the conclusion about the 960 2gb vs 4gb from the latest test I could find. "In fact, we saw very few cases were the GTX 960 4GB made sense (Edit: compared to the 2gb card with a price difference of only about 20.00), and certainly there were no instances where the 2GB frame buffer crippled performance."

source

 

[Bacon1]

Here is the conclusion about the 960 2gb vs 4gb from the latest test I could find. "In fact, we saw very few cases were the GTX 960 4GB made sense, and certainly there were no instances where the 2GB frame buffer crippled performance."

source

on January 4, 2016

Guess you guys don't have any issues saying people should buy the $100 2GB 460 then? Apparently 2GB is fine for 1080p.

Would be nice if more reviewers spent the time comparing these. If someone wants to send me the cards I'll do the hard work

 

[Keysplayr]

Guess you guys don't have any issues saying people should buy the $100 2GB 460 then? Apparently 2GB is fine for 1080p.

Would be nice if more reviewers spent the time comparing these. If someone wants to send me the cards I'll do the hard work

It seems that "you guys" as you put it, dont have a problem saying people should buy 1060 3GB. Benchmarks show it handles what is thrown at it fairly easily. So, yeah, theres that.

 

[antihelten]

Guess you guys don't have any issues saying people should buy the $100 2GB 460 then? Apparently 2GB is fine for 1080p.

Note that the 2GB 460 appears to be significantly power limited compared to the 4GB version (likely due to the lack a 6-pin connector), and cant maintain it's clockrate. So you have to be careful with drawing any conclusions about VRAM here.

On that note Computerbase.de did make an interesting observation when they played around with texture settings. Basically they found that the gap between the 2GB and 4GB 460 would diminish when texture settings were lowered clearly indicating that the 2GB was VRAM limited, the 2GB GTX 950 on the other hand didn't care, indicating that it wasn't VRAM limited. This along with the Techspot review posted above clearly supports the notion that Nvidia is better at VRAM management than AMD.

The quote from computerbase.de (roughly translated by me, if anyone speaks german fell free to provide corrections):
"The common denominator of all games: None of them go light on the graphics card memory. One or two steps lower texture details quickly shows that for the RX 460, 2GB is not sufficient for the settings. The opposite is the case for the Nvidia card: the GeForce GTX 950 with 2GB supports sometimes much higher FPS in the problematic games, up to 42 percent (F1 2015) ahead of the 2GB version of the RX 460. It can thus only be concluded that Nvidia in borderline cases has a more efficient memory management than AMD."

 

[Keysplayr]

Note that the 2GB 460 appears to be significantly power limited compared to the 4GB version (likely due to the lack a 6-pin connector), and cant maintain it's clockrate. So you have to be careful with drawing any conclusions about VRAM here.

On that note Computerbase.de did make an interesting observation when they played around with texture settings. Basically they found that the gap between the 2GB and 4GB 460 would diminish when texture settings were lowered clearly indicating that the 2GB was VRAM limited, the 2GB GTX 950 on the other hand didn't care, indicating that it wasn't VRAM limited. This along with the Techspot review posted above clearly supports the notion that Nvidia is better at VRAM management than AMD.

The quote from computerbase.de (roughly translated by me, if anyone speaks german fell free to provide corrections):
"The common denominator of all games: None of them go light on the graphics card memory. One or two steps lower texture details quickly shows that for the RX 460, 2GB is not sufficient for the settings. The opposite is the case for the Nvidia card: the GeForce GTX 950 with 2GB supports sometimes much higher FPS in the problematic games, up to 42 percent (F1 2015) ahead of the 2GB version of the RX 460. It can thus only be concluded that Nvidia in borderline cases has a more efficient memory management than AMD."

Well now. That's a source.

 

[railven]

that is why marketing works logos on AAA games gets alot of eye balls. and the salesperson pushing the 500$ laptop it is just a matter of commission

Definitely feel that commissions at boutique stores are a hindrance. Sales clerks just want to up sell. But the alternative is Best Buy like stores where the sales clerks are often just given a department regardless of their knowledge.

You can bet your rear-end buyers in both situations will be persuaded into getting the NV card

 

[Pantalaimon]

Well now. That's a source.

When they both have the same amount of RAM. You're trying to sell the idea that NVIDIA's 3 GB is actually better than having physically 1 GB more RAM.

Edited to add:
You're actually trying to use the same tack as NVDIA's PR regarding the 970 3,5 GB vs 4 GB issue. In that case they basically said 'You ingrate users should be happy we found a way to give you 4 GB on the card, instead of complaining that we did not tell you 0,5 GB of it is slower VRAM'.

Here you're trying to say 'Users should be happy NVDIA has a secret sauce to manage and compress VRAM that makes their 3 GB on their cards better than actually having physically 1 GB more on the card. So shut up about 3 GB not being enough'.