[Eurogamer] NV GeForce GTX 960 2GB vs 4GB review [How much VRAM do you really need?]

[Mondozei]

Interesting review.

The R9-380 also does very well, even after overclocking the GTX 960.

It makes the upcoming full fat Tonga, a.k.a 380X seem like an increasingly interesting proposition this autumn for those who can't quite afford a 970/390 but want something a bit better than an entry-level $200 dollar card.

 

[Techhog]

Oh boy. A certain someone is going to be in here to make this thread annoying very soon...

They didn't even include the clock speeds here. Weird. Then again, that site really is more console-focused.

 

[Silverforce11]

Then again, that site really is more console-focused.

They run Digital Foundry on Youtube, one of the most comprehensive benchmark site, frame time charts in real-time, side-by-side comparisons. Pretty much beyond what most tech sites do.

The article stated the clock rates: "we were still regularly boosting to an impressive 1475MHz"

 

[Techhog]

They run Digital Foundry on Youtube, one of the most comprehensive benchmark site, frame time charts in real-time, side-by-side comparisons. Pretty much beyond what most tech sites do.

And those videos are usually something like 80% console talk and 20% PC talk, if I'm being generous to PC.

 

[happy medium]

"Now, the R9 380 doesn't have much in the way of overclocking headroom available (1-3fps extra on seven of our eight test games at 1080p, with Tomb Raider doing best, rising by 5fps), so we were interested in revisiting GTX 960 performance on two counts. Firstly to see if the easily overclockable Nvidia card can catch up or exceed its competitor once the OC was in place."

" the GTX 960 either draws level with R9 380, or mildly exceeds it, but there's no knockout blow against the AMD "

" Nvidia's memory management is typically more robust than its AMD counterpart, meaning that its 2GB tends to go further (look at Far Cry 4 for the best example "

" Buying a 4GB card at this point remains an insurance policy of sorts"

1920x1080 (1080p)

R9 380 2GB,.R9 380 4GB,.GTX 960 2GB,.GTX 960 4GB,.GTX 960 4GB OC

Assassin's Creed Unity, Ultra High, FXAA
32.8 37.1 26.7 34.5 39.0

Battlefield 4, Ultra, 4x MSAA
47.9 47.8 49.4 47.6 53.6

Crysis 3, Very High, SMAA T2x
50.3 49.8 47.2 46.2 51.6

Far Cry 4, Ultra, SMAA
50.7 53.8 48.1 47.2 53.1

Shadow of Mordor, Ultra, Medium Textures, FXAA
62.7 64.1 52.8 52.2 56.9

Ryse, High, SMAA
47.8 49.5 42.4 41.1 46.5

Tomb Raider, Ultimate, FXAA
69.5 70.0 66.9 65.2 72.5

The Witcher 3, High, No HairWorks, Custom AA
49.7 49.2 47.3 47.2 53.0

 

[Mondozei]

And those videos are usually something like 80% console talk and 20% PC talk, if I'm being generous to PC.

If you have seen their benchmarking videos, you'd know that they are thorough in their testing. They also did some of the best Skylake CPU reviews, so there are no issues with their PC reviewing which you seem to suggest.

So I'm not sure where your attacks are coming from, or why you have such a deep desire to discredit them, but why don't we stay on topic for once instead of pushing an agenda?

 

[Silverforce11]

Here's the videos if you want to see it in detail:

1080p
https://www.youtube.com/watch?v=W0HzWW6mDno

1440p
https://www.youtube.com/watch?v=ypr03mNrM_Y

Tonga OC just fine, more than 1-3 fps.
http://www.guru3d.com/articles_pages/asus_radeon_r9_380_strix_review,22.html

 

[boozzer]

Oh boy. A certain someone is going to be in here to make this thread annoying very soon...

They didn't even include the clock speeds here. Weird. Then again, that site really is more console-focused.

that is some lightning fast thread crapping skills. :twisted:

 

[Spanners]

snip

Your post is entirely composed of quotes and numbers from the article.

I don't think your selective quoting gives a very fair representation of the article either. Here's the full sentence that you cut in half, it has a slightly different meaning in complete form.

"Buying a 4GB card at this point remains an insurance policy of sorts - something that pays off spectacularly occasionally, especially in sub-optimal PC ports (Batman: Arkham Knight, we're looking at you)."

 

[Erenhardt]

Why would they test 2GB vs 4GB and lower texture resolution? Max it out man!
I find that increasing texture quality has almost 0 impact on performance untill you reach BW or capacity limit of your card.

 

[arandomguy]

My problem with these tests is that they often don't seem to actually examine the cases, even if they are edge cases, where the difference may be most felt. In some cases it seems like they go out of their way to avoid such a scenario.

Notably here is the lack of forcing textures to the higher values, with Shadows of Mordor -

Shadow of Mordor, Ultra, Medium Textures, FXAA

It seems at the very least they should test for both cases, the lower texture case and higher texture case, and then at the very least do an image quality comparison between the two.

I also do no seem the point of testing games like Crysis 3, Tomb Raider or even BF4. The concern over VRAM is directed at current generations console ports as the new platforms have much higher memory.

The other issue is memory deficiency issues often do not show themselves in more limited benchmarking runs. It may take long playthroughs or even manifest only in specific segments and actions in games.

 

[Erenhardt]

My problem with these tests is that they often don't seem to actually examine the cases, even if they are edge cases, where the difference may be most felt. In some cases it seems like they go out of their way to avoid such a scenario.

Notably here is the lack of forcing textures to the higher values, with Shadows of Mordor -
It seems at the very least they should test for both cases, the lower texture case and higher texture case, and then at the very least do an image quality comparison between the two.

I also do no seem the point of testing games like Crysis 3, Tomb Raider or even BF4. The concern over VRAM is directed at current generations console ports as the new platforms have much higher memory.

The other issue is memory deficiency issues often do not show themselves in more limited benchmarking runs. It may take long playthroughs or even manifest only in specific segments and actions in games.

Good points.
Also looking at raw average fps over a period of time doesn't actually tell us anything about the VRAM performance. Sometimes there difference in average fps is very little between two cards that give very different experience due to stuttering or the lack of it.

 

[Face2Face]

Maybe slightly off topic, but Shadow of Mordor VRAM usage is very odd to me. I've installed the Ultra texture pack and ran the game - Max settings at 1080p, and my usage/caching via Afterburner is 4.8-5GB... My card is only a 3GB card, and the performance is great; no stuttering/hitching at all..odd

 

[tg2708]

Maybe slightly off topic, but Shadow of Mordor VRAM usage is very odd to me. I've installed the Ultra texture pack and ran the game - Max settings at 1080p, and my usage/caching via Afterburner is 4.8-5GB... My card is only a 3GB card, and the performance is great; no stuttering/hitching at all..odd

If I'm not mistaken is does not actually enable ultra textures if the vram is under 6gb even though its selected. I read that in the settings of the game.

 

[Face2Face]

If I'm not mistaken is does not actually enable ultra textures if the vram is under 6gb even though its selected. I read that in the settings of the game.

Do you mean this?

 

[arandomguy]

Maybe slightly off topic, but Shadow of Mordor VRAM usage is very odd to me. I've installed the Ultra texture pack and ran the game - Max settings at 1080p, and my usage/caching via Afterburner is 4.8-5GB... My card is only a 3GB card, and the performance is great; no stuttering/hitching at all..odd

If you are on Windows 10 you need a different program to read usage. Currently afterburner will combine both dedicated and shared memory pools into one reading.

Try something like process hacker.

 

[tg2708]

Do you mean this?

yes that's it, I was wrong though.

 

[Face2Face]

If you are on Windows 10 you need a different program to read usage. Currently afterburner will combine both dedicated and shared memory pools into one reading.

Try something like process hacker.

Yeah I'm using Windows 10. I take a look at Process Hacker, maybe even try looking at HWInfo and GPU-Z. Thanks! Even so, I would think performance would have equaled a stuttery mess, but it's smooth and fast - Here's a vid of the performance if anyone's curious.

1080p - https://www.youtube.com/watch?v=WZsC3LTSQ0A

Funny thing is, with the same settings @ 1440p - My VRAM usage is lower than 1080p? - https://www.youtube.com/watch?v=5t2zOxhc6qc

 

[arandomguy]

Yeah I'm using Windows 10. I take a look at Process Hacker, maybe even try looking at HWInfo and GPU-Z. Thanks! Even so, I would think performance would have equaled a stuttery mess, but it's smooth and fast - Here's a vid of the performance if anyone's curious.

1080p - https://www.youtube.com/watch?v=WZsC3LTSQ0A

Funny thing is, with the same settings @ 1440p - My VRAM usage is lower than 1080p? - https://www.youtube.com/watch?v=5t2zOxhc6qc

Form what I remembered hearing about Mordor there may be certain rare segments that stutter with the GTX 970 that would not with a 6GB card when using Ultra textures. I think there were some comments that 16gb of ram helped as well along with a ssd. From launch impressions reports some people even reported it playing fine with 3GB cards. Which isn't surprising as the cut off point for individuals may vary.

How much VRAM actually gets allocated involves a lot of different factors. How much that does end up getting allocated may not be the minimum necessary. I wouldn't rely on the allocation data by itself to determine whether or not you'd have a practical benefit from more VRAM.

It's on the other end of the spectrum of the problem described here -

Good points.
Also looking at raw average fps over a period of time doesn't actually tell us anything about the VRAM performance. Sometimes there difference in average fps is very little between two cards that give very different experience due to stuttering or the lack of it.

But for some reason there seems to a lot of usage of avg fps or memory allocation data and the reliance of those. While what we actually need is other data points for this particular issue.

 

[tg2708]

My fps is is significantly lower than F2F with an r9 290 at 1440p.

 

[Face2Face]

My fps is is significantly lower than F2F with an r9 290 at 1440p.

Same settings? Also, keep in mind my GTX 780 have a pretty decent OC. Using Guru3D's recent benchmarks @ Very High, it's very close to a stock GTX 980.

 

[xorbe]

"Buying a 4GB card at this point remains an insurance policy of sorts"

I like that quote, wrt the 960. It probably doesn't have the oomph to really dance with a full 4GB working set, but protects against peaking over 2GB for the next 18-24 months of ownership. It was on sale for the same price as 2GB cards, so of course I got the 4GB model (for one of many machines ...)

 

[tg2708]

Same settings? Also, keep in mind my GTX 780 have a pretty decent OC. Using Guru3D's recent benchmarks @ Very High, it's very close to a stock GTX 980.

Yes I'm getting an average of 40fps with dips in the teens.

 

[Face2Face]

Yes I'm getting an average of 40fps with dips in the teens.

Somethings not right. I'd make sure you're on the latest drivers.

 

[nurturedhate]

I had the same issues with Mordor and my 780ti, enabling ultra textures was a stuttering mess. Fps would be fine for a little and then tank. Just for reference this was around when the game first came out.