[Eurogamer] GTX 1060: 3 GB vs 6 GB

Page 9 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Status
Not open for further replies.

Keysplayr

Elite Member
Jan 16, 2003
21,211
50
91
Wow, NVDIA has amazing memory compression and their 3 gb can be better than actually having 1 gb more physical memory on the competitor's card. Why have more physical memory, let's just compress it the NVIDIA way. Think of all the savings from not having to put more physical memory on the cards. /s
And management.
 

railven

Diamond Member
Mar 25, 2010
6,604
561
126
Welps, as I'm finding from a few friends/family - price is gonna really dictate the success of these cards. Recommended a few people I know asking to up their budget, 2/3 still went with the GTX 1060 3GB. Welps, for their needs I'm sure it will be more than enough.

At this point I should just stop recommending AMD cards. Even when I was hardcore AMD, no one would take my recommendation for them haha. That NV brand presence. Woof.
 

antihelten

Golden Member
Feb 2, 2012
1,764
274
126
Techpowerup has their review up.

The review seems to confirm that Hitman and Tomb Raider prefers more VRAM at 1080P (3GB is 16% and 24% slower than the 6GB). Although both still achieve playable framerates (62.7 FPS for Hitman and 41.2 FPS for Tomb Raider

Weirdly enough though the situation versus the RX 470 seems to be the opposite of what we have seen in most of the other reviews with the 1060 3GB being faster in Hitman (by a tiny amount) and slower in Tomb Raider (by 22%), This is basically the exact opposite of what Eurogamer, ABT, Techspot and Hardware Unboxed saw (the later two only tested Tomb Raider). They all had the 1060 3GB as 15-25% slower in Hitman and a 5-10% faster in Tomb Raider (with very high textures). This is only made weirder by the fact that TPU has the 1060 catching up to the 470 in Tomb Raider when moving up to 1440P

Looking at the exact FPS numbers, it seems that TPU's Hitman numbers for are about 10% higher than Eurogamer and ABT for the 1060 and 10% lower for the RX 470 (a 22% relative performance swing in favor of the 1060), but it's their Tomb Raider numbers that are really crazy, 37% lower for the 1060 and 13% lower for the RX 470 than the other 4 sites (a 38% relative performance swing in favor of the 470).

Other than that it is 6% slower than the 6GB on average (pretty much confirming the 5% claim by Nvidia), and 20% faster than the RX 470.
 
Last edited:

96Firebird

Diamond Member
Nov 8, 2010
5,737
334
126
TPU's ROTR results are odd... At 1080p, 1060 3GB is much slower than 470 4GB. At 1440p, 1060 3GB is about even with 470 4GB. At 2160p, 1060 3GB is again much slower than 470 4GB. I'm not sure if VRAM is the reason for 1060 3GB being much slower at 1080p, when you look at the other results.
 

DamZe

Member
May 18, 2016
188
84
101
I dont see how they could have been as I am open about such things. Unlike many here.

I don't question your preferences, but I am forced to question your honesty if you think 3GB of vRAM will be enough for comfortable 1080p AAA title gaming in the forseeable future. nVIDIA focus group or not, you cannot tell me with a straight face that 3GB isn't DOA in late 2016.


Insulting other members is not allowed
Markfw900
 
Last edited by a moderator:

Keysplayr

Elite Member
Jan 16, 2003
21,211
50
91
I don't question your preferences, but I am forced to question your honesty if you think 3GB of vRAM will be enough for comfortable 1080p AAA title gaming in the forseeable future. nVIDIA focus group or not, you cannot tell me with a straight face that 3GB isn't DOA in late 2016.
It'll do just fine with reasonable settings expected from a mainstream device.
Face is straight.
 

antihelten

Golden Member
Feb 2, 2012
1,764
274
126
you cannot tell me with a straight face that 3GB isn't DOA in late 2016.

There is currently nothing that indicates that this will happen. Even in the games where the 1060 3GB shows VRAM limitations it is still managing 40-50 FPS.

So given the available evidence, the question is more how you could possibly claim that the 1060 3GB will be DOA by late 2016. I would love to here your reasoning for this.

It's perfectly possible that the 1060 3GB will start having severe VRAM issues 2, 3 or 4 years from now, but by late 2016? not a chance.
 

boozzer

Golden Member
Jan 12, 2012
1,549
18
81
Welps, as I'm finding from a few friends/family - price is gonna really dictate the success of these cards. Recommended a few people I know asking to up their budget, 2/3 still went with the GTX 1060 3GB. Welps, for their needs I'm sure it will be more than enough.

At this point I should just stop recommending AMD cards. Even when I was hardcore AMD, no one would take my recommendation for them haha. That NV brand presence. Woof.
that is just the superb, master level marketing at work :) it is still kinda weird how casual friends/family of yours have a brand preference, hell, even know the difference between amd or nv :)
 

Bacon1

Diamond Member
Feb 14, 2016
3,430
1,018
91
It could very well be that Nvidias memory compression technology is much further advanced than AMD can ever hope to offer. 3GB Nvidia may actually be better than 4GB AMD.

Thats not how compression works...

Compression helps with bandwidth not usable space.
Techpowerup has their review up.

After I found out that TPU doesn't retest all of the cards in their suite, even when newer drivers or even game patches occur, I gave up faith in how accurate their tests could be. Many cards are months out of date which is why their results can be completely wrong.
 

Keysplayr

Elite Member
Jan 16, 2003
21,211
50
91
that is just the superb, master level marketing at work :) it is still kinda weird how casual friends/family of yours have a brand preference, hell, even know the difference between amd or nv :)
Thats not how compression works...

Compression helps with bandwidth not usable space.


After I found out that TPU doesn't retest all of the cards in their suite, even when newer drivers or even game patches occur, I gave up faith in how accurate their tests could be. Many cards are months out of date which is why their results can be completely wrong.
And management.
And enlighten me. How is it that compression of anything doesnt take up less space?
 

railven

Diamond Member
Mar 25, 2010
6,604
561
126
that is just the superb, master level marketing at work :) it is still kinda weird how casual friends/family of yours have a brand preference, hell, even know the difference between amd or nv :)

Well, my more "online savvy" friends are most likely aware of the two companies. When the NV logo is slapped across multiple game releases, they might start to assume it's better.

Most family members are not, and the few times I've been with them shopping for parts, sales reps definitely push things on them. Had my poor mother almost forced into an Intel laptop for $500 when the the one I knew would cover her needs (an AMD model) was on sale for $200. Pretty sure if I weren't there to tell the rep we already had one in mind she might have walked out with the $500 Intel.

I've lived in a see of green for years. But thanks for asking about my friends/families preferences :D
 

96Firebird

Diamond Member
Nov 8, 2010
5,737
334
126
And management.
And enlighten me. How is it that compression of anything doesnt take up less space?

Delta color compression still needs the full texture in VRAM before it can compress it. It gets "compressed" (really it just throws out anything that hasn't changed) on a frame-by-frame basis.

http://www.anandtech.com/show/10325/the-nvidia-geforce-gtx-1080-and-1070-founders-edition-review/8

The second paragraph explains it quite well.

Edit - Memory management is another story, I'm sure that can alleviate the need for mass amounts of VRAM required.
 

Headfoot

Diamond Member
Feb 28, 2008
4,444
641
126
Yeah - at-rest compression seems like an avenue for future improvement on the low end cards
 

Keysplayr

Elite Member
Jan 16, 2003
21,211
50
91
Delta color compression still needs the full texture in VRAM before it can compress it. It gets "compressed" (really it just throws out anything that hasn't changed) on a frame-by-frame basis.

http://www.anandtech.com/show/10325/the-nvidia-geforce-gtx-1080-and-1070-founders-edition-review/8

The second paragraph explains it quite well.

Edit - Memory management is another story, I'm sure that can alleviate the need for mass amounts of VRAM required.

Ok. Thanks very much for that. I am enlightened. So, memory management it must be.
 

boozzer

Golden Member
Jan 12, 2012
1,549
18
81
Well, my more "online savvy" friends are most likely aware of the two companies. When the NV logo is slapped across multiple game releases, they might start to assume it's better.

Most family members are not, and the few times I've been with them shopping for parts, sales reps definitely push things on them. Had my poor mother almost forced into an Intel laptop for $500 when the the one I knew would cover her needs (an AMD model) was on sale for $200. Pretty sure if I weren't there to tell the rep we already had one in mind she might have walked out with the $500 Intel.

I've lived in a see of green for years. But thanks for asking about my friends/families preferences :D
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 :)
 

Bacon1

Diamond Member
Feb 14, 2016
3,430
1,018
91
So, memory management it must be.

Source?

It seems that Nvidia uses more memory than AMD in the tests I've seen:

1456505100AQVrwEp3px_13_1.gif


http://www.hardocp.com/article/2016/02/29/rise_tomb_raider_graphics_features_performance/13
 

Bacon1

Diamond Member
Feb 14, 2016
3,430
1,018
91
Does your chart show Nvidia trying to use more than it has?

No, they stopped testing the Nvidia cards once they "maxed" out their memory. Notice that the 980 TI wasn't tested 4k with 2x or 4x SSAA. Meanwhile the 8GB 390 is using less memory (more efficient?) than the 980 Ti. Opposite of what you are claiming.

You said that Nvidia has better memory management. Where is your source for that?
 

Bacon1

Diamond Member
Feb 14, 2016
3,430
1,018
91
Memory allocation and actual memory requirements are not the same things. As such you have to be careful with drawing too many conclusions from tests such as these.

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.
 

Sweepr

Diamond Member
May 12, 2006
5,148
1,143
136
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 feel free. Let's try something a bit more recent than the old test you cherry picked:

bf1_3840_11.png


bf1_vram.png
 

Keysplayr

Elite Member
Jan 16, 2003
21,211
50
91
No, they stopped testing the Nvidia cards once they "maxed" out their memory. Notice that the 980 TI wasn't tested 4k with 2x or 4x SSAA. Meanwhile the 8GB 390 is using less memory (more efficient?) than the 980 Ti. Opposite of what you are claiming.

You said that Nvidia has better memory management. Where is your source for that?
Is it "maxxed" out, as you put it? Or most efficiently utilized? Whos to say...
And I need to agree with antihelten here and echo that you draw too large of a conclusion from those memory usage tests. Memory utilization is not the same as memory requirements.
 
  • Like
Reactions: Sweepr

USER8000

Golden Member
Jun 23, 2012
1,542
780
136
Is it "maxxed" out, as you put it? Or most efficiently utilized? Whos to say...

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??
 
Status
Not open for further replies.