Was buying this 4GB 770 for playing at 1080p a mistake?

Interesting that there are no FCAT tests to demonstrate how identical cards with different vram may perform. Stutter is not strictly related to vram capacity, as AMDs 79xx cards with more vram vs Kepler (680) had shown the latter to perform better in FCAT tests a year or two ago. A problem AMD acknowledged and addressed later on. Granted that the problem was not severe enough for AMD owners to notice, I do wonder how a 'stutterfest' (which would be noticeable) can occur without affecting FPS, especially between 2 identical cards with different vram. FCAT benches would probably be off the charts in such cases.

 

I'll just leave this here (Bioshock Infinte has been known to use more than 2GB at 1080 this is a 770 2GB at 2560x1440)



Only spike above 40ms (perceivable stutter is the 7970 which has more VRAM)

 

Nah the 4gb and 2gb card will perform nearly identical. Thats why I had no problem going from 4gb to 3gb.

 

Naturally stutter won't be perceivable when you're only chugging along at 30fps max to start with, but I'm glad you posted that. Mind if I leave this here?




You'll notice the 770 4GB at the same resolution, same game and same settings, getting nearly double the fps as the 2GB 770 in your post.

Personally I don't have much faith in the Alienbabeltech graphs because SSAA was only used in 1 or 2 games.....the rest had FXAA turned on.

 

^ What are you smoking Pill? There is no 4GB 770 in that bench. The 760 mars is a 4GB dual GPU card with 2GB per GPU? The only 770 in that bench is a reference card with 2GB.

 

Pretty much exactly what I was going to say.

My 3GB 7970 is still going strong as it's a good overclocker but playing TITANFALL with Insane textures (which still look like ****..) and I had a max vram usage of 3062 of a possible 3072 and that was only using 4xMSAA (which made no difference in IQ...). So going for a 4GB vram card now is 100% NOT a bad idea at all. I might keep my 7970 for another year and just turned down texture resolution and AA in some of the newer games until I can afford a 4GB 290 with custom cooler but by then Maxwell will be out and AMD might have a refresh of the R9 series so I might either get a great deal or the newest card.

 

Irrelevant. You don't microstutter like a multi-GPU setup when you run out of VRAM. When you run out of VRAM your FPS tanks, plain and simple.

You don't see something getting smoothed out because you have more VRAM, doesn't work that way. Never did, never will.

Not saying 2GB is enough or not. But all evidence points to the fact that in the majority of scenarios, it's more than enough. Would I be comfortable with a 2GB card at my resolution? No, I'd like a bit of headroom, but 2GB isn't bad by any means.

 

What is actually used vs cached vs what is needed is a big question mark. So far very little documented data has shown measurable performance differences between 2gb vs 4gb other than with ultra high res or multi-monitors (7680x1080) or silly levels of AA. By the time you need the 4gb, a 770 will sputter out of GPU power in and of itself.

 

There have been a few reviews of the 680 4gb vs 2gb (close to the 770), including here at guru3d.

http://www.guru3d.com/articles_pages/palit_geforce_gtx_680_4gb_jetstream_review,1.html

Hilberts findings:

Multi-monitor performance, 4gb vs 2gb. It does make a little difference... at 7680x1080:

http://www.legionhardware.com/articles_pages/gigabyte_geforce_gtx_680_4gb,4.html

http://www.pugetsystems.com/labs/articles/Video-Card-Performance-2GB-vs-4GB-Memory-154/

Now there may be newer games out now or on the horizon that may actually use that vram (not just cache it), but I'll be waiting to see actual differences between two identical cards having 2 vs 4gb vram. There is no doubt that there are certain situations which may indeed use 3gb or more of vram, but again it may require more GPU power than what single cards can deliver. And still probably at very high res, silly AA levels. Until we see documented data on this, its mostly just talk.

 

True. I remember having a 7800GTX with only 256MB back in the day. S.T.A.L.K.E.R. came out, I could get decent framerates at max detail but every now and then I would dip to 5 fps or worse and after a couple of seconds I was back to my usual 40+ fps...

My mate with his 7900GTX with 512MB Vram had no such issues.

No idea if it's necessary atm, but for future games, you never know if the 4GB VRAM could be usefull.

 

The numbers posted don't show such fps variation, fps wise 2gb and 4gb are equal, your frame drops are a completely separate issue.

 

The main difference between 2GB and 4GB is between a completely freezing and a not freezing game when the card exceeds 2GB with all the load forced to VRAM instead of a shared virtual memory pool.
Using shared system ram adds stuttering and fps drops when single rendered frames exceeds 2GB pr. frame on a card with 2GB VRAM.
Even with lower bandwidth on one card compared to a card with more bandwidth, that is still much faster than using system ram, GPU speed also has a saying thou.

 

^evidence please. Documented data rather than anecdotal or theory.

 

Not really a difference between 2gb 680s and 4gb 680s.
You would be suprised what a 2 gb 680 770 is capable of.
I only payed 50 bucks extra for my 4gb 680s so I don't feel too bad. Multi monitor I would go wwith the 4gb cards. Gk104 will long run out of gpu power than vram.

 

I concur, while I can make a situation where I use more than 2GB of VRAM those situations would have made 4GB versions of the sames cards buckle from lack of power. Even on multi monitor if your smart with your setting 2GB is very usable. Trust me.

 

Like I said, that VRam poohaa is overrated and'll make you spend more than you need to. And yes, I'm always right. And no, ATi or nVidia don't ever release single card solutions that are bottlenecked by its vram.

 

Well then the 4GB vs 2GB debate is merely about how long he wishes to use this card. Eventually games will need more (even if now wouldn't be the case) and then it might become an issue.

The issue I had in Stalker was VRAM, I checked it over and over back then and it's really frustrating being able to "handle" a game but running out of VRAM and then suddenly dropping to 5 fps for several seconds on end.

 

When you quoted Pill he was talking about a micro stutter type situation that would not show up in a average based or minimum frame rate bench. He was claiming that the lack of VRAM could not effect frame rates but still cause stutter which simply is not true.

 

I would get constant CTD's with a 2GB card playing Skyrim with the mods I use, been there done that.

 

Modded skyrim is the exception for 99% of games there is little difference.