So it appears the extra 256mb of RAM on the GTX does almost nothing (in current games)...

[dug777]

http://www.anandtech.com/video/showdoc.aspx?i=2607&p=10

But then we've seen a big boost for the x800xl 512mb vs 256mb haven't we? (i can't remember where, but i think it was a quake 4 review... i know it was recent). What's going on? Has anyone clocked the x1800xt down to x1800xl speeds and compared them?

EDIT: i can't find that review that showed the 512mb x800xl taking a lead over the 256mb one..maybe i was dreaming

EDIT2: found it http://www.firingsquad.com/hardware/call_of_duty_2_performance_ati_nvidia/page2.sp

 

[Schadenfroh]

not surprised

 

[dug777]

haha i found it:http://www.firingsquad.com/hardware/cal...uty_2_performance_ati_nvidia/page2.asp

 

[dug777]

we need a COD2 test for the 512mb GTX

 

[SynthDude2001]

Originally posted by: dug777
we need a COD2 test for the 512mb GTX

Indeed we do. I'm not sure why AT omitted that game. Wonder if it's really the only one that shows a difference from 256MB-512MB right now?

(Yeah, they said they had an issue running it without sound, whatever...)

 

[kylebisme]

You won't see the difference in short average framerate benchmarks. The difference comes in when a game wants to use more ram than the card has and then you get some hitching for a moment when things swap.

 

[BFG10K]

http://www.anandtech.com/video/showdoc.aspx?i=2607&p=10

Uh, did you look at the bottom graph?

 

[dug777]

Originally posted by: BFG10K

http://www.anandtech.com/video/showdoc.aspx?i=2607&p=10

Uh, did you look at the bottom graph?

huh?

same story...

 

[BFG10K]

huh? same story...

What are you talking about? Black & White 2, Day of Defeat and Battlefield 2 show significant gains from 512 MB.

And why do you feel a single benchmark is the end-all indicator of total game performance? As an example even the original Call of Duty will take advantage of 512 MB cards when you run high settings like 1856x1356 with 6xAA and that game is more than two years old.

 

[dug777]

Originally posted by: BFG10K

huh? same story...

What are you talking about? Black & White 2, Day of Defeat and Battlefield 2 show significant gains from 512 MB.

And why do you feel a single benchmark is the end-all indicator of total game performance? As an example even the original Call of Duty will take advantage of 512 MB cards when you run high settings like 1856x1356 with 6xAA and that game is more than two years old.

AT came to the same conclusion i did...

BF2 is insignificant in that graph, in that case look at the 256 beating the 512mb in d3 ...and the others are only two games, and only 3-4 frames at fps too low to matter...

So basically the only real, noticeable and arguably important difference is for B&W2...what did you say to me again about a single game?

I don't quite understand you at all here i'm afraid

 

[BFG10K]

AT came to the same conclusion i did...

Again I'll ask why you feel a single demo running for a few seconds is an indication of the entire game?

and the others are only two games...

Right, so there is a difference then. What is the purpose of this thread?

 

[malG]

Originally posted by: BFG10K

huh? same story...

What are you talking about? Black & White 2, Day of Defeat and Battlefield 2 show significant gains from 512 MB.

QFT. Hence 512MB would benefit more if they bench with 8X AA and 16X AF enabled.

 

[Matt2]

regardless of the advantages of 512 vs 256, the 7800 512 owns.

 

[dug777]

Originally posted by: BFG10K

AT came to the same conclusion i did...

Again I'll ask why you feel a single demo running for a few seconds is an indication of the entire game?

and the others are only two games...

Right, so there is a difference then. What is the purpose of this thread?

read my edit

i said in my title 'almost nothing' if we are going to nit pick

 

[crazydingo]

Originally posted by: dug777
we need a COD2 test for the 512mb GTX

http://img.hexus.net/v2/graphics_cards/nvidia/g70gtx512/images/bench-cod2.png

 

[dug777]

Originally posted by: malG

Originally posted by: BFG10K

huh? same story...

What are you talking about? Black & White 2, Day of Defeat and Battlefield 2 show significant gains from 512 MB.

QFT. Hence 512MB would benefit more if they bench with 8X AA and 16X AF enabled.

you are an idiot mate look at the fps at the current settings, sure you might be right but what use would a faster card be at those settings? None, unless you like powerpoint-like fps...

added to which there's NOTHING significant about the BF2 difference, even % wise

 

[dug777]

If you have a problem with the way Anand tests, address complaints to anand@anandtech.com, not me

i'm quite happy with the way he tests personally.

 

[malG]

Originally posted by: dug777

you are an idiot mate

At least I'm not a poor bastard like you with a lowly 6600GT.

 

[crazydingo]

Originally posted by: malG

Originally posted by: dug777

you are an idiot mate

At least I'm not a poor bastard like you with a lowly 6600GT.

worst.comeback.ever

 

[BFG10K]

i said in my title 'almost nothing' if we are going to nit pick

Actually if you want to nitpick it would be "almost nothing if you choose to accept the flawed premise that a few seconds of benchmark data is all you need to show how the entire game will work".

If you have a problem with the way Anand tests, address complaints to anand@anandtech.com, not me

The problem isn't Anand's tests, the problem is how people like yourself wrongfully interpret them.

 

[airborne82nd]

Originally posted by: malG



At least I'm not a poor bastard like you with a lowly 6600GT.

Why do you have to bust on the mans card? Must not be intelligent enough to debate in an open forum.

I had a 6600GT and it kicked ass.

 

[jiffylube1024]

Originally posted by: BFG10K

huh? same story...

What are you talking about? Black & White 2, Day of Defeat and Battlefield 2 show significant gains from 512 MB.

And why do you feel a single benchmark is the end-all indicator of total game performance? As an example even the original Call of Duty will take advantage of 512 MB cards when you run high settings like 1856x1356 with 6xAA and that game is more than two years old.

Maybe your definition of significant differs from mine. Here's the 'big' gains:

Black & White 2: 18% (14 to 17fps; a gain of 3fps)
DOD: Source 14% (27.7 to 31.7 fps; a gain of 4fps)
BF2: 3% (28.6 to 29.4 fps, a gain of 0.8fps)

Those are not significan gains, those are almost within margin of error! The impact of memory size is within 3% at 2048X1536 with no AA, btw.

8X AA is a silly suggestion since these games are already averaging 15-35fps with 4X AA. How unplayable do you want it to be? Will a gain of 8fps to 12fps (50%) be amazing to anyone?


Anandtech is spot on, the memory size does almost nothing for any of these games. The overwhelming majority of the performance increase is from GPU clockspeed and RAM clockspeed.

----------------

I don't understand your argument about Anandtech's testing methodology being flawed. Which position are you taking: that the extra memory helps in the 'look and feel' of the game (ie. less stuttering), or that Anand didn't run enough tests, and that more extensive tests would have showed more significant gains?


IMO if 2048X1536 with 4X AA can't show 25%+ type of gains right off the bat in demos on seven different games, and every lower resolution shows even smaller margins, then the extra RAM does nothing.


Although I would like to see some 8X AA results at playable resolutions, namely 1280X960 and perhaps 16X12 as well.

 

[kylebisme]

Originally posted by: TheSnowman
You won't see the difference in short average framerate benchmarks. The difference comes in when a game wants to use more ram than the card has and then you get some hitching for a moment when things swap.

 

[BFG10K]

Those are not significan gains, those are almost within margin of error!

14% and 18% within margin of error? I don't think so. You may have a point with 3% but certainly not with the former two.

Anandtech is spot on, the memory size does almost nothing for any of these games

No, Anandtech's few second benchmarks show almost no gain. When it comes to this sort of thing you can't extrapolate an entire game based on a few seconds of gameplay.

I don't understand your argument about Anandtech's testing methodology being flawed.

Unless you get very lucky and happen to get a demo that shows a difference right off the bat (e.g. CoD2) your only real way of testing this sort of thing is to play the game from start to finish and test for stuttering/hitching/pausing during actual gameplay. Either that or specifically look for problem areas in the game and build a benchmark in these areas.

In general arbitrary few second demos are next to useless for showing VRAM differences, much like they are for showing system RAM differences.

 

[dug777]

Originally posted by: BFG10K

i said in my title 'almost nothing' if we are going to nit pick

Actually if you want to nitpick it would be "almost nothing if you choose to accept the flawed premise that a few seconds of benchmark data is all you need to show how the entire game will work".

If you have a problem with the way Anand tests, address complaints to anand@anandtech.com, not me

The problem isn't Anand's tests, the problem is how people like yourself wrongfully interpret them.

i've interpreted them fine mate

not exactly sure what you are trying to prove, or say here...only one game showed any meaningful difference, and thats arguable given how low the fps were. I came to EXACTLY the same conclusion that AT did, so you are saying they're wrong as well