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

[golem]

Why does this matter again?

 

[dug777]

Originally posted by: golem
Why does this matter again?

it's called the video forum, and i'm commenting on a new video card.

 

[golem]

But what point are you trying to make?
That nvidia should have made it a 256mb product?
That it is poorly designed and can't take advantage of 512mb?
or
That it's so well designed it could have beat all competitors even if it only had 256mb?

 

[dug777]

Originally posted by: golem
But what point are you trying to make?
That nvidia should have made it a 256mb product?
That it is poorly designed and can't take advantage of 512mb?
or
That it's so well designed it could have beat all competitors even if it only had 256mb?

none of that. stop trolling.

I'm simply commenting on the fact that at least from the AT review, the difference between the 256 and 512 GTXs is due to the latter's far higher core/mem speeds..which last time i checked didn't involved any of those options you suggested Your last point is closest tho, with 256 at the 512 GTXs speeds i'd imagine it would perform almost identically...

 

[golem]

Sorry if I sounded trollish. Actually looking at the xl800 results... do they seem a little weird? There seems to be big gains at 800x600, where you wouldn't expect more memory to matter. Or am I looking at this wrong?

 

[dug777]

Originally posted by: golem
Sorry if I sounded trollish. Actually looking at the xl800 results... do they seem a little weird? There seems to be big gains at 800x600, where you wouldn't expect more memory to matter. Or am I looking at this wrong?

no worries mate :beer:

yeah i don'ty quite understand that, maybe the core isn't bogged down with data at that stage and can utilise the memory more effectively?

 

[IntelHydralisk]

I think it was quite obvious from the start that the performance of the 7800 gtx 512 was due to the clocks and not the 512 mb of RAM ... right now anyway. This is evidenced by the previous 512 cards... all of which had practically no gains in today's games except for COD 2.

 

[CaiNaM]

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.

QFT. Been mentioned before. As a current example Q4 has hitches/pauses often when running in ultra mode.

Originally posted by: BFG10K
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".

how do you reach the conclusion AT uses 3-4 second benchmarks?

 

[dug777]

Originally posted by: CaiNaM

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.

QFT. Been mentioned before. As a current example Q4 has hitches/pauses often when running in ultra mode.

as correct as this is, in this situation, without hard evidence to back this theory up, you're just speculating that the 512mb GTX offers much smoother gameplay...I think AT would have noticed and commented on this fact if it was actually a noticeable difference. Did hardocp, who profess to benchmark games in a true 'gaming' fashion make comment or notice anything like this? (i don't know the answer to that question, anyone care to enlighten me who did read their review?).

EDIT: just for interest's sake, i remember my 9800 pro (128mb) used to hitch in d3 on high 1024 when it first came out, but when i lasted played d3 on it, it didn't hitch at all on high, and the amount of VRAM i had certainly hadn't changed, all that had changed was i had another 512mb of RAM and the latest ATI drivers....now i blamed the hitching on only having 128mb of VRAM back when i first got the game, but more system RAM and new drivers fixed the hitching

 

[CaiNaM]

first of all, it's only logical that if the texture size exceeds that of the frame buffer, swapping is going to occur.

secondly, it's not really speculation as we did this on an x800 512 vs 256 (apples to apples). while it's not a gtx, were comparing frame buffer size, not architecture. the 256mb played fine once the textures loaded, but would again stutter intermittently here and there (and farily often).

the flipside tho is that there is no discernable visual difference between "high quality" and "ultra quality", as whatever losses attributed to minor compression is not noticeable visually.

so yea, there is an advantage to 512mb, but it's easily argued that advantage is not of much/any use (from an iq standpoint). this may change however if/when texture size increases.

 

[BFG10K]

i've interpreted them fine mate

Nonsense.

not exactly sure what you are trying to prove, or say here

Let me say it as simply as possible:

A few seconds of gameplay from a demo cannot be used to prove that there is no difference between 256 and 512. ALL you can prove is that FOR THAT DEMO ALONE there is no difference.

I really can't explain it any simpler than that I'm afraid.

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

You cannot prove this sort of thing by running a demo for a few seconds. Do you understand why this is the case? I get the feeling you still don't get it.

I'm simply commenting on the fact that at least from the AT review, the difference between the 256 and 512 GTXs is due to the latter's far higher core/mem speeds.

The evidence you have cannot prove that it was entirely a faster core/memory that made the difference. What you can say is "in the benchmarks AT ran there was no difference". That's it. You can't then claim there is no difference period.

as correct as this is, in this situation, without hard evidence to back this theory up, you're just speculating that the 512mb GTX offers much smoother gameplay..

Likewise, you're simply speculating there isn't smoother gameplay since you have no evidence to back that theory up.

And actually yes, it is provable since it's very easy to create scenarios now that texture swap on 256 MB boards. I know this for a fact based on developer comments (e.g. Doom 3/Quake 4 Ultra quality), benchmarks that show differences (e.g. Call of Duty 2) and my own testing. All you really need is a basic knowledge of 3D rendering and a bit of gaming experience and it's quite easy to prove.

 

[BFG10K]

how do you reach the conclusion AT uses 3-4 second benchmarks?

I might ask you the same question because that's certainly not a conclusion I ever reached.

 

[dug777]

Originally posted by: CaiNaM
first of all, it's only logical that if the texture size exceeds that of the frame buffer, swapping is going to occur.

secondly, it's not really speculation as we did this on an x800 512 vs 256 (apples to apples). while it's not a gtx, were comparing frame buffer size, not architecture. the 256mb played fine once the textures loaded, but would again stutter intermittently here and there (and farily often).

the flipside tho is that there is no discernable visual difference between "high quality" and "ultra quality", as whatever losses attributed to minor compression is not noticeable visually.

so yea, there is an advantage to 512mb, but it's easily argued that advantage is not of much/any use (from an iq standpoint). this may change however if/when texture size increases.

do you have a link to this review or did you do it yourself/privately? That would be fascinating reading, since i don't think i've ever read a review that looked at that. It makes sense, but in the multitude of 256/512mb x800xl reviews i read they all concluded the extra RAM was a gimmick, and none mentioned that gameplay was actually noticeably smoother, that i can remember anyway.

@BFG, i was simply basing my comments on the data provided by AT (i have to assume the tests lasted longer than 3-4 seconds, and that if the 512mb GTX was noticeably smoother they would have at least commented on it at least once in their look at the difference the extra RAM made), and i never suggested it was based on anything else.

 

[BenSkywalker]

B3D has some numbers that show some difference. The impact the additional RAM has can be staggering- but you have to run into one of those instances. Dave did.

 

[BFG10K]

@BFG, i was simply basing my comments on the data provided by AT (i have to assume the tests lasted longer than 3-4 seconds,

I would have to assume this as well and I'm not sure where this 3-4 second nonsense is coming from.

and that if the 512mb GTX was noticeably smoother they would have at least commented on it at least once

Have you ever ran a benchmark? If it blazes past in a few seconds there is no possible way you can tell this sort of thing. Like I said before, they'd need to physically play through substantial portions of the game to see any benefit.

In any case here are a few benchmarks:

HL2.
Call of Duty 2.

 

[Dribble]

The 512mb of memory is mostly future proofing at the moment.

In the past some cards have never really used the extra memory (e.g. 9800 pro 256mb was never really any better then the 128mb card) so perhaps it's a waste. However I remember playing COD on a geforce 4 64mb and wishing I had a 128mb version cause I had to turn down the texture details quite a bit to stop hitching despite the card itself managing a very high fps.

Bottom line if I spent that much money on a card I'd expect to be future proofed.

 

[ScrewFace]

You better watch your mouth, BFGoof10K. Calling people Trolls and Idiots is the work of a 12-year old. I guess you want to get banned.:disgust::|

 

[CaiNaM]

Originally posted by: BFG10K

how do you reach the conclusion AT uses 3-4 second benchmarks?

I might ask you the same question because that's certainly not a conclusion I ever reached.

funny.. you keep repeating it... or should i say you've repeated it a few times (which obviously would not mean you've repeated it several hundred, or several thousand...):

few seconds, ie couple, several, etc. -- more than one but indefinately small in number.

as a matter of fact, i'm not sure i saw anywhere what they were using to measure fps, or the duration of the segment they used to test it. i only ask because you seem adamant to hammer the point.

 

[jiffylube1024]

Originally posted by: BFG10K

I'm simply commenting on the fact that at least from the AT review, the difference between the 256 and 512 GTXs is due to the latter's far higher core/mem speeds.

The evidence you have cannot prove that it was entirely a faster core/memory that made the difference. What you can say is "in the benchmarks AT ran there was no difference". That's it. You can't then claim there is no difference period.

I'm saying that the difference isn't equal to zero, I'm saying it's negligible.

Regarding 14-18% being within experimental error, in this case it isn't but if you're talking about 10 fps vs 11-12 fps, then it very well could be. Heck even if you're benchmarking in the 20's, a hard drive hitch could be responsible for the difference. But I do agree, in the Anandtech review, those 14%-18% improvements are from the extra video memory. Probably an intangible gain from a "look and feel" perspective, especially considering those were the largest increases in the whole review, and it took 20xx by 15xx with 4X AA, at around 30fps to do it.

as correct as this is, in this situation, without hard evidence to back this theory up, you're just speculating that the 512mb GTX offers much smoother gameplay..

Likewise, you're simply speculating there isn't smoother gameplay since you have no evidence to back that theory up.

And actually yes, it is provable since it's very easy to create scenarios now that texture swap on 256 MB boards. I know this for a fact based on developer comments (e.g. Doom 3/Quake 4 Ultra quality), benchmarks that show differences (e.g. Call of Duty 2) and my own testing. All you really need is a basic knowledge of 3D rendering and a bit of gaming experience and it's quite easy to prove.[/quote]

Yes but if Q4 really does hitch or stutter on Ultra quality on a 256MB board, then just run it on high. The quality difference is negligable, and if your manhood can't take not running the absolute highest setting on a new video card, then fine, be my guest and buy a 512MB card!

COD2 didn't seem to show any gains in the Anandtech review. Does it stutter for you on a 256MB card, BFG? And if so, are you sure it's video related, and not hard drive/RAM related?

 

[IeraseU]

Bottom line is I think there's a market for a 256mb 7800GTX with the same clockspeeds currently offered in the 512mb version. IMO by the time it makes a significant difference to have 512mb of ram, the 7800 will be woefully dated......therefore right now higher memory and core clockspeeds without the added expense of memory would be preforable.

 

[dguy6789]

This also proves that ID was completely full of it when they said Doom 3 uses 500MB of texture memory. There is no difference between Doom 3 with a 256 or 512 GTX at the same clocks.

 

[Dethfrumbelo]

Originally posted by: IeraseU
Bottom line is I think there's a market for a 256mb 7800GTX with the same clockspeeds currently offered in the 512mb version. IMO by the time it makes a significant difference to have 512mb of ram, the 7800 will be woefully dated......therefore right now higher memory and core clockspeeds without the added expense of memory would be preforable.

Agreed. I'd like to see a 256MB version of this card for $150 less, but it probably won't happen.

What would they call it anyway?

 

[dguy6789]

Originally posted by: IeraseU
Bottom line is I think there's a market for a 256mb 7800GTX with the same clockspeeds currently offered in the 512mb version. IMO by the time it makes a significant difference to have 512mb of ram, the 7800 will be woefully dated......therefore right now higher memory and core clockspeeds without the added expense of memory would be preforable.

This whole "outdated before the card can use the memory" idea I believe is wrong.

The 9800 Pro 256MB can run fear, and so can the 128MB 9800 Pro. The 256MB can run at higher settings than the 128MB can. The same can be said for Serious Sam 2.

While the difference between the memory amounts do not matter much now, they will matter near the end of the card's life.

 

[jiffylube1024]

Originally posted by: Dethfrumbelo

Originally posted by: IeraseU
Bottom line is I think there's a market for a 256mb 7800GTX with the same clockspeeds currently offered in the 512mb version. IMO by the time it makes a significant difference to have 512mb of ram, the 7800 will be woefully dated......therefore right now higher memory and core clockspeeds without the added expense of memory would be preforable.

Agreed. I'd like to see a 256MB version of this card for $150 less, but it probably won't happen.

What would they call it anyway?

No kidding it won't happen, $150 less than the 7800GTX is less than the 7800GT costs! It would be more like a decrease of $50 or so if it ever happened...

Originally posted by: dguy6789

Originally posted by: IeraseU
Bottom line is I think there's a market for a 256mb 7800GTX with the same clockspeeds currently offered in the 512mb version. IMO by the time it makes a significant difference to have 512mb of ram, the 7800 will be woefully dated......therefore right now higher memory and core clockspeeds without the added expense of memory would be preforable.

This whole "outdated before the card can use the memory" idea I believe is wrong.

The 9800 Pro 256MB can run fear, and so can the 128MB 9800 Pro. The 256MB can run at higher settings than the 128MB can. The same can be said for Serious Sam 2.

While the difference between the memory amounts do not matter much now, they will matter near the end of the card's life.

At what resolution is FEAR playable on any 9800 Pro? Personally I found FEAR borderline unplayable on an X800XT PE, although I was running at my LCD's native resolution of 1680X1050.

I can see FEAR being playable at maybe 800X600 on a 9800 Pro, which I wouldn't figure would use any more than 128MB of video memory.

 

[kylebisme]

Originally posted by: Dribble
The 512mb of memory is mostly future proofing at the moment.

In the past some cards have never really used the extra memory (e.g. 9800 pro 256mb was never really any better then the 128mb card) so perhaps it's a waste. However I remember playing COD on a geforce 4 64mb and wishing I had a 128mb version cause I had to turn down the texture details quite a bit to stop hitching despite the card itself managing a very high fps.

Bottom line if I spent that much money on a card I'd expect to be future proofed.

Funny you bring up the orginal CoD as a game that used lots of texture ram right after you say the 256mb 9800pros were never any better as CoD was one of the first games to make use of over 128mb of video ram and the 256mb 9800pros were basicly the best cards out for that game when it came out. Granted you didn't see the difference much in benchmarks, again because short benchmarks reporting average framerates simply can't show which card is choking on too much textures for the memory on the board and which card loads it all smoothly.