Printable View
http://www.xbitlabs.com/articles/vid...0-roundup.html
even at 2560x1600 there is no diference between the 2gb sapphire model and then 1gb msi card, which both clocked about the same... the sapphire card is clocked only 10mhz lower, but that was enough to always pull ahead of the 2gb sapphire card... so 2gb seems to make no sense at all so far?
Makes perfect sense. Let's see again after 12 months.
Shame they did not test GTA IV, the 2GB card would rock the socks off in that game :)
I have found that 2GB cards have a much higher minimum frame rate at resolutions equal to or greater than 1920x1200, for example in Crysis you can get a 3fps minimum frame rate boost! (on a 2GB GTX 285), no difference on Maximum Framerate tho.
John
I was under the impression that GTAIV was one of the main games that benefited from 2gb over 1gb, surprised they didn't bench it.
Exactly, I have read somewhere that GTA IV requires 1.6GB of Video RAM for Maximum Details/Draw Distance with highest texture qualityQuote:
Originally Posted by acidpython
John
GTA4 is just a horribly coded game.
And in fact, with tessellation and new high-res texture compression methods, I doubt we will see the demand for more than 1Gb any time soon.
DirectX 11 should be very demanding GPU performance wise, though.
would be interesting to see how that changed once you add 1gb or so of texture mods to fallout 3 and increase the ugrid rather than running it at stock settings since its only made to use ~350m as thats around the xbox360 limit i think
more memory on gpus is really needed as games look so much better when there not just a blur of fugly textures although you can get around this by having frequent loads which isnt great
Also DASA correct me if I am wrong, but isnt there an official HD texture pack for Fallout3?
Also Fakefactory's Cinematic mod for Half-Life 2 would benefit from more VRAM as each incarnation of this fantastic mod gets more and more texture lovin!
John
Quote:
Originally Posted by JohnZS
yes that mod was the beginnings of texture mods for fallout3 its good but fairly low quality in comparison to some
http://forums.guru3d.com/showthread.php?t=287764
there is some nice 2048 textures to replace the old 512 ones
good point with the cinematic mod to its a great mod for hl2 which makes a huge difference
its a shame vista stopped gpu memory usage monitors from working it would be handy to know just how much extra games use with these mods and aa\af at high res
where is gta4? where is cinematic mod? where is fallout 3 with hd textures pack? :)
HD4890 1GB in Crossfire vs. HD4890 2GB in Crossfire.
Using a single HD4890 2GB in Crossfire doesn't give it justice... use 2 of them in 2560x1600 and compare it to 1GB cards...
put some 16x AA and watch what happens.
But can you even see any difference in quality between 16x and 4x at 2560x1600 or even 1920x1200? :p:Quote:
Originally Posted by Chumbucket843
Just saying normally you wouldn't use 16x anyway cuz you won't see any difference so there's no reason to use it and then the 2GB is still unneccessary if showed a performance difference, so it doesn't change the point at all. 4x AA is more useful to test IMO.
yesQuote:
Originally Posted by RPGWiZaRD
Seriously?
4x more than enough when you get to 1920x1200 +.
I can live with 4x.Quote:
Originally Posted by ColonelCain
When i get Pixar quality edges then it will be enough for me.
Video memory serves a very simple function. It stores texture data, vertex data, a frame buffer, and a number of other things. It isn't something that is going to increase performance unless there wasn't enough memory to fit it all before. So, it isn't fair to look at ROFLBENCH99 which uses 200MB of VRAM and conclude "The 2GB model is pointless as it doesn't perform any better." That comparison compares two things which don't test the variable that actually changes.
Some people here pointed out that GTA4 would be a good game to test since it is known to be able to use over 1GB of VRAM. This is exactly the kind of test which would matter--seeing the increased performance in a scenario where on one card it has to lean on system memory and on the other it does not. It's quite clear which will be faster.
you'll be waiting a while then, as in 2 or 3 years.Quote:
Originally Posted by Final8ty
GPGPU computing is the major candidate to benefit from large amounts of video memory, but only for professional app's as of now.
you guys are game modders and such and still dont know its a gimmick like every thing else in this industry.
enev if the textures were high enuff to need 2g it woulg give you 2-3 frames xtra woopy doo
I thought 1gb vs 2gb was common knowledge after the nV GT260 cards, even was a review on it.
Review had GT260>X280 and 4870 512MB&1024MB variants. :cool:
That soon :)Quote:
Originally Posted by Helloworld_98
All games can be modded to use full system resources so this is pointless,it only applies to those who want to utilize it as you can see they didn't MOD a single game.The point of having more resource's is for more Eyecandy,once you have acceptable performance already.:yepp:
So this review os useless. Next!
What surprises me is that XBit labs usually do quite indepth reviews, the omission of GTA IV is very sloppy and I am sure that if they actually tested the "proper" Crysis @ 1920+ with Very High Details you will see the 2GB card having higher MINIMUM frame rates (albeit both cards producing woefully unplayable frame rates...)
It shocks me they did not chose games which we know use large amounts of VRAM
John
It may only show up as a few FPS lower on average, but thrashing can be very visually distracting. It also has a very negative effect on gameplay because it often happens when things are moving the fastest and there is the most action.Quote:
Originally Posted by cowie
So what's your threshold for minimum vram size?Quote:
Originally Posted by cowie
I like how people still want to claim 1GB of video memory is enough. If you are playing 2560x1600 with 4xAA it is not enough. The thing I'm hating the most about all the game reviews is you don't know what section of a map/game they used. The game benchs are worthless without min/max frame rates included because no matter what there's always a spot in a game that is going to slow way down and that's where more vram is going to matter.
Didnt use a single game which uses more than 1gb of video memory.... go figure.
hmmm so you think nv gpus need more mem than ati gpus?Quote:
Originally Posted by JohnZS
could be... depends how they handle stuff in their drivers i guess
good point! pcgh tested those mods in the past, but i dont remember if they tested 1gb vs 2gb cards...Quote:
Originally Posted by Spectrobozo
oh yeah they did!
for some reason they set the view distance to 50% though... :doh:
and as a result, the 2gb card was slightly SLOWER than the 1gb card, LOL :D
they also tested hl2 cm and there is 0 diference between the 1gb and 2gb gtx285, nada... even at 2560x1600 hmmmm
interestingly far cry2 shows a 2-3fps gain for the 2gb card in 1920x1200, and in 2560x1600 the boost is huge, BUT, 22min 30av fps arent really playable... well you can argue about that... so with 2gb memory you can arguably play far cry2 at 2560x1600 with a single gtx285 while with 1gb you def can not (13min 23.8av)
http://www.pcgameshardware.de/aid,69...e/Test/?page=8
the largest gains were in the fallout3 mod, but again, only at 2560x1600 and again its arguable how playable it is with 26min 32.8av... but its notably more than 1gb (6min 11.1av)
good point, id like to see that as well... seems nobody is testing this :confused:Quote:
Originally Posted by Smartidiot89
Exactly. It's unfortunate that this forum is allergic to one of the few hardware sites that actually shows their data however. It's too bad really.Quote:
Originally Posted by safan80
i dont think its that hw sites want to hide that 2gb makes sense or try to prove that 1gb is enough... i think they just dont try hard enough to really make use of it and then do a reality check if its worth it and how big the diference is... thats a shame...
He made no mention of ATi/nV at all?Quote:
Originally Posted by saaya
Quote:
good point! pcgh tested those mods in the past, but i dont remember if they tested 1gb vs 2gb cards...
oh yeah they did!
for some reason they set the view distance to 50% though... :doh:
and as a result, the 2gb card was slightly SLOWER than the 1gb card, LOL :D
they also tested hl2 cm and there is 0 diference between the 1gb and 2gb gtx285, nada... even at 2560x1600 hmmmm
interestingly far cry2 shows a 2-3fps gain for the 2gb card in 1920x1200, and in 2560x1600 the boost is huge, BUT, 22min 30av fps arent really playable... well you can argue about that... so with 2gb memory you can arguably play far cry2 at 2560x1600 with a single gtx285 while with 1gb you def can not (13min 23.8av)
http://www.pcgameshardware.de/aid,69...e/Test/?page=8
the largest gains were in the fallout3 mod, but again, only at 2560x1600 and again its arguable how playable it is with 26min 32.8av... but its notably more than 1gb (6min 11.1av)
good point, id like to see that as well... seems nobody is testing this :confused:
With HL2 you may need to set mat_min_heapsize and mat_max_heapsize larger (value measured in MB btw, I run 1024 and 2048 respectively) to "fill" the video memory (yes I realize it is more related to system memory).
There may also be a console variable that automatically sets the preferred video memory amount... all I have is my 4870X2 so I cant test really.