Computerbase - 3gb vs 4gb vs 6gb vs 8gb GDDR5 VRAM Frametime Testing

[Bacon1]

https://www.computerbase.de/2016-09/grafikkarten-speicher-vram-test/2/

3GB falls often at even 2nd to highest settings, 4GB falls off @ highest and 6GB even falls during Mankind Divided highest.

Would be interesting to see 970 / Fury tested along with those to see how the 3.5 on 970 and HBM for Fury react to the same test scenarios.

 

[AtenRa]

Yeap, as said by many here before and its validated by this review, the GTX 1060 3GB starts its life with compromising on Texture Quality and higher frametimes. 3GB for 2016 at this level of performance its DOA and people on the know have the moral obligation to warn those that dont know about it. Next year this card will be even worse with more DX-12 games on the market.

 

[Headfoot]

Im personally more surprised at how 4gb is showing increased frametimes in certain titles too

 

[antihelten]

There's something really weird going on with these numbers.

In Rise of the Tomb Raider with high textures the 1060 6GB is substantially faster than the 3GB version (roughly 15ms vs. 19ms, or 67fps vs. 53fps), but in the actual 1060 3gb review this isn't the case at all, with both versions managing about 52fps. So somehow the 6GB version just got significantly faster all of a sudden in this test.

Another peculiar result is their RX 480 4GB numbers in mirrors edge with hyper textures. Unlike every single other card with 4GB or less the RX 480 does just fine here (performing basically the same as the 8GB). Given that every single other 4GB card out there has been shown to take a significant hit with hyper settings, I can't help but suspect that Computerbase.de forgot to turn off the memory protection feature when testing the RX 480 4GB.

Finally it's worth noting that whilst computerbase.de recommends medium textures in deus ex for the 1060 3GB and high textures for the RX 480 4GB, there is literally zero difference in quality judging from the screenshots (medium, high).

 

[nopainnogain]

Finally it's worth noting that whilst computerbase.de recommends medium textures in deus ex for the 1060 3GB and high textures for the RX 480 4GB, there is literally zero difference in quality judging from the screenshots (medium, high).

The tree branch in the left upper corner?

 

[kondziowy]

Given that every single other 4GB card out there has been shown to take a significant hit with hyper settings, I can't help but suspect that Computerbase.de forgot to turn off the memory protection feature when testing the RX 480 4GB.

I have seen sites posting results from ME:Catalyst where 4GB and 3.5GB cards had terrible fps with hyper settings, and then after a while in another test these cards were just fine suddenly. And they never even mention if memory protection was on or off. It's getting to a point where I just use youtube videos as bechmarks as it's getting more and more confusing and you never know what settings were used and if a benchmarking place is representative of the majority of the gameplay.

 

[Bacon1]

The tree branch in the left upper corner?

The statue itself is blurry in medium and better in high, but overall they are very similar.

Also here is very high vs high:

http://www.dsogaming.com/wp-content/uploads/2016/08/Very-High-2.jpg

http://www.dsogaming.com/wp-content/uploads/2016/08/High-2.jpg

http://www.dsogaming.com/pc-performance-analyses/deus-ex-mankind-divided-pc-performance-analysis/

 

[Bacon1]

there is literally zero difference in quality judging from the screenshots (medium, high).

Statue shows it in that image, but there also isn't much since its looking up into the sky.

You skipped the rest of the images though...

Medium: https://pics.computerbase.de/7/4/5/8/0/6-1080.758232269.jpg
High: https://pics.computerbase.de/7/4/5/8/0/7-1080.241377914.jpg

The path is blurry in medium as well as the left wall.

Medium: https://pics.computerbase.de/7/4/5/8/0/9-1080.782558618.jpg
High: https://pics.computerbase.de/7/4/5/8/0/10-1080.3769635422.jpg
Very High: https://pics.computerbase.de/7/4/5/8/0/11-1080.186649973.jpg

Her face is blurry in medium, along with her suit, especially noticeable are the circles on her shoulders.Adam's neck area is blurry. All increase a lot as you go higher

Medium: https://pics.computerbase.de/7/4/5/8/0/13-1080.1627248236.jpg
High: https://pics.computerbase.de/7/4/5/8/0/14-1080.432521270.jpg
Very High: https://pics.computerbase.de/7/4/5/8/0/15-1080.1196945698.jpg

The image of the person (and other two surveillance video on left) is almost completely a blur on medium, ok on high and clear on VH. The "Locked" is blurry on medium, same with the wall texture on the right.

Basically any close up textures are much more blurry on Medium, better on High and good even at distance on VH.

 

[Vaporizer]

That is against all truth! People keep telling me that 3GB is fair enough for 1080p (at least if its on a NV card). Therefore this test must be disqualified.

 

[ShintaiDK]

What exact settings was used in the test? Just picking the highest texture level with or without AA may be far from the optimal gaming experience.

 

[Pantalaimon]

3GB falls often at even 2nd to highest settings, 4GB falls off @ highest

This test is not true! NVIDIA's memory management and compression is so amazing that a 3 GB card can overcome its smaller capacity and become even better than a 4 GB card of the competitor's. People here on this very board tells me so! /s

 

[Keysplayr]

This test is not true! NVIDIA's memory management and compression is so amazing that a 3 GB card can overcome its smaller capacity and become even better than a 4 GB card of the competitor's. People here on this very board tells me so! /s

Not compression, I've learned. But management? Seems so. 3GB overcoming 4GB? I dont believe so, but through management can certainly close the gap. I understand your need for sarcasm to try and drive your point home, but the funny thing is, you were probably more correct than you realize.

 

[antihelten]

Statue shows it in that image, but there also isn't much since its looking up into the sky.

You skipped the rest of the images though...

...

I did indeed miss the remaining images (I have to admit I didn't look too closely after the first set of images), And you are indeed right that the difference show up up much more significantly here, especially with character models and object (which were missing to a large degree from the first set).

It's interesting though to see the second set of screenshots with the cobblestone road, which clearly looks better on high. It would appear though that this is not just due to higher textures, but also due to a higher level of tessellation, so it seems that the texture settings affects things other than purely texture resolution. Note I'm not trying to start some conspiracy theory here, I think increasing tessellation quality in this case makes perfect sense, just found it a bit interesting.

Btw does it say anywhere in the review whether or not Computerbase.de tested Deus Ex in DX11 or DX12 (I couldn't find any mention at a quick glance). It is possible that the 1060 3GB (and the 480 4GB to some degree) could perform significantly better with DX11 here, given the increased VRAM requirements of DX12 (plus it's my impression that the DX12 patch for Deus EX hasn't been all sunshine and rainbows in general).

 

[sm625]

Finally it's worth noting that whilst computerbase.de recommends medium textures in deus ex for the 1060 3GB and high textures for the RX 480 4GB, there is literally zero difference in quality judging from the screenshots (medium, high).

The tree branch in the left upper corner?

There is also a difference in the pavement. And a very slight change to the sidewalk to the right of the tiny car behind the sign (apparently this is supposed to be some added detail to the curbs). But it is all pretty small. Honestly if this is what an extra 3GB gets you then it is foolish to argue that you really need more than 3GB. That's ridiculous.

 

[DamZe]

No matter what data shows, the nVIDIA stock holders and defense force will be out defending these gimped offerings, because there will always be a scenario where a 3GB Geforce Gimp Edition can be shown to be viable for today and yesterday, but what about tomorrow? I won't name anyone, but if you frequent these forums enough, you probably know who these people are by now…. Suffice to say, take their "advice" with a grain of salt, and for heaven’s sake don’t pay 200+$ in 2016 for a 3GB video card.

 

[raghu78]

for heaven’s sake don’t pay 200+$ in 2016 for a 3GB video card.

very well said. I have been following the GPU industry for close to 2 decades and the fact is good cards with low VRAM always age faster and run out of VRAM sooner than they run out of GPU horsepower. 8800 GT 256 MB, HD 4870 512 MB, HD 6950 1GB, HD 7850 1GB were a few of the cards. These cards were very pretty much identical to the higher VRAM versions but faded much sooner than the versions with double the VRAM.

 

[antihelten]

There is also a difference in the pavement. And a very slight change to the sidewalk to the right of the tiny car behind the sign (apparently this is supposed to be some added detail to the curbs). But it is all pretty small. Honestly if this is what an extra 3GB gets you then it is foolish to argue that you really need more than 3GB. That's ridiculous.

Whilst the screenshots I linked definetly don't show a difference anywhere near enough to justify going from 3GB to 4GB (i.e. from medium to high), the rest that Bacon1 linked shows a more significant difference.

very well said. I have been following the GPU industry for close to 2 decades and the fact is good cards with low VRAM always age faster and run out of VRAM sooner than they run out of GPU horsepower.

This is just blatantly false, and a classical example of selection bias. I bet you that for every GPU you can mention where the low VRAM version had issues, I can mention 2-3 GPUs where the low VRAM version had zero issues.

 

[Piroko]

This is just blatantly false, and a classical example of selection bias. I bet you that for every GPU you can mention where the low VRAM version had issues, I can mention 2-3 GPUs where the low VRAM version had zero issues.

That's a queer argument.
"I bet you that for every preordered game you can mention where the game had unfixable issues, I can mention 2-3 games where it had zero issues." Am I not allowed to counter with "Preordering is still a pretty terrible idea if I have a 25% chance to get a dud and a 75% chance to get no more than what I would get without preordering..."?

I don't think 3GB in 2016 is bad, but I do think that 3GB for a card that is supposed to be two tiers above Tahiti is pretty terrible.

 

[antihelten]

That's a queer argument.
"I bet you that for every preordered game you can mention where the game had unfixable issues, I can mention 2-3 games where it had zero issues." Am I not allowed to counter with "Preordering is still a pretty terrible idea if I have a 25% chance to get a dud and a 75% chance to get no more than what I would get without preordering..."?

You might want to read Raghu78's post again. He didn't say that there was a 25% chance to get a dud, he said that there was a 100% chance to get a dud: "good cards with low VRAM always age faster". He didn't say that they sometimes age faster or that they mostly age faster, he said always. Hyperbole like that really doesn't help anyone.

 

[Piroko]

You might want to read Raghu78's post again. He didn't say that there was a 25% chance to get a dud, he said that there was a 100% chance to get a dud: "good cards with low VRAM always age faster". He didn't say that they sometimes age faster or that they mostly age faster, he said always. Hyperbole like that really doesn't help anyone.

A bit of hyperbole is fine when even winning the jackpot means you come off worse than not playing this game of vram-Roulette at all.

I'm completely aware that this isn't the case this time, because the 3GB card is noticeably cheaper. But the discussion clearly isn't baseless. Not with the 3GB 1060, 2GB 960+680, 2GB HD285/380 or 4GB Fury. They all have a weird ratio of shader power to vram size, which is worth pointing out at every recommendation.

 

[antihelten]

A bit of hyperbole is fine when even winning the jackpot means you come off worse than not playing this game of vram-Roulette at all.

I know you're not referring to the situation at hand here, but in what in the world are you referring to? To my knowledge there has never been a case where the low VRAM version of a card wasn't cheaper, so in other words if you won the "VRAM roulette" you would always come of better given the $ savings.

I'm completely aware that this isn't the case this time, because the 3GB card is noticeably cheaper. But the discussion clearly isn't baseless. Not with the 3GB 1060, 2GB 960+680, 2GB HD285/380 or 4GB Fury. They all have a weird ratio of shader power to vram size, which is worth pointing out at every recommendation.

I never said the discussion was baseless, far from it. I'm just taking issue with the amount of hyperbole and extremism that seems to have taken hold of it. It seems like people are constantly claiming that "X always happens all the time, 100%" or "Y never ever ever happens". A lot of people seem to be basing their arguments on truthiness more than facts (the facts most of the time being that there simply isn't anyway to say for sure whether or not X/Y will happen), and it really isn't conducive to a fruitful conversation, and it certainly doesn't help with getting to the bottom of this issue.

 

[raghu78]

This is just blatantly false, and a classical example of selection bias. I bet you that for every GPU you can mention where the low VRAM version had issues, I can mention 2-3 GPUs where the low VRAM version had zero issues.

not really. Try using a GPU after 2 years from launch with the latest and most demanding games. I am speaking from personal experience where a HD 4850 512 MB and HD 4850 1GB showed clear difference with the lower VRAM card running into VRAM limits often with newer games 2+ years from launch.

 

[swilli89]

Why in Heaven nVidia didn't name the 1060 6gb as a 1060ti and the 1060 3gb has just a vanilla 1060 is beyond me.

Anyways, seems like the 1060 6gb or the 470 4gb are the minimum $200+ cards to get.

 

[ShintaiDK]

Why in Heaven nVidia didn't name the 1060 6gb as a 1060ti and the 1060 3gb has just a vanilla 1060 is beyond me.

Anyways, seems like the 1060 6gb or the 470 4gb are the minimum $200+ cards to get.

The review favours 6GB as the minimum. Example that 4GB isnt enough for the performance tested.

 

[Piroko]

I know you're not referring to the situation at hand here, but in what in the world are you referring to? To my knowledge there has never been a case where the low VRAM version of a card wasn't cheaper, so in other words if you won the "VRAM roulette" you would always come of better given the $ savings.

Wasn't the 2G/4G GTX770 almost at price parity? The 960 and 380 I remember as well as two cards that turned out to be very close in price between the two versions (less than 20$ when I bought my 4G 380). I'm only making an exception for the 1060 3G version because it's almost 60$ of price difference currently in Germany (210€ vs. 260€). My opinion would be different if those cards were only 30$ apart.