How Much RAM Does Your Graphics Card Really Need?

[terr281]

I've had to tell numerous friends about this exact issue in the past, but they still don't seem to listen. ("Hey, this older GPU card has twice the ram, that makes it better than the second to the newest one with half the ram, right?")

As another poster commented on, however, it would be interesting to see a basic multiple monitor configuration that includes SLI'ed/Crossfire'd cards. (I don't mean game resolution on multiple monitors, I mean a webpage on a second monitor.) Many MMO players play this way, where they can keep their guild webpage, active logging program, IM window, etc. open at all times as they game. (And, I have been resisting my mate's requests for such a setup until our pc "core component" rebuilds due to our hardware's age.)

 

[dreamer77dd]

I like articles like this that take the myths out of what your actually getting. Do we need a gig of gpu ram for windows, does it effect fold@home, or other things? Maybe that RAM could be used for not only graphics but physics? How does RAM effect the HD theater experiencing? Does a blu-ray disc need a lot of RAM for colors and things?

 

[ismol]

I think any vga benefits from higher amount of vram, because if it's a high end one, for sure it will have enough memory(1gb aprox.). And if it is a low end card, the core will never be able to handle more than 512 mb. 2-3 fps more or less don't justify 30 dollars. So my advise is...get always the cheapest DDR3 graphic card!!!

 

[cleeve]

A much easier way would be to simply look at the hardware monitor in riva tuner to see how much RAM game uses. No need to do benchmarks.

Unless I'm misunderstanding you, there's no way Rivatuner would show us the performance hit that would be caused by using a card without enough VRAM.

 

[precariousgray]

The animated GIFs are nice, but insanely frustrating. A better approach would be to have them change on mouse over, that way one can compare differences at their leisure, not every x seconds. I'm sure I've seen this implemented before, if not on this site, then elsewhere. GIFs tend to result in dithering and such things with full-color images, as well; go with PNG and onmouseover changes.

 

[cleeve]

I'm paying attention by the way, I'll see if I can set up a follow-up article that looks at CrossFire/SLI and the new 5870 over multi-monitor resolutions.

Really though, since CrossFire and SLI don't share memory between cards, I doubt the results will be any different. If you're using 512MB cards in Crossfire, you're going to see the same IQ limitations compared to using 1 GB cards in Crossfire...

 

[fancarolina]

Great article however I would like to see Crossfire SLI numbers at 1gb, 2gb, and 4gb Configurations mixed in for comparison.

 

[lemonade4]

This is the best article I've read on Tom's for a while. It was very informative and had excellent visuals to demonstrate your point. There were no apparent spelling mistakes either.

 

[Guest]

very interesting article i think it opened some eyes. i know it opened mine

 

[scooterlibby]

In the conclusion you state GTX260 has 768mb RAM. That was actually the 8800GTX. GTX260 has 896mb.

 

[mapesdhs]

Just curious, did any of the GTA IV tests use AA/AF? Both images
appear to have significant staricasing artifacts along various edges.

Ian.

PS. S'funny, I have an old entry-level high-end SGI with 5GB VRAM
and 1GB TRAM, but alas its fill rate and GE speed suck by modern
standards, so a bit useless for games, though it runs Quake2/Quake3 ok.
Ideal for visualising a sliced up dead dude though, or driving
eight HD displays at once. 😀

 

[kraken18]

"there's no way Rivatuner would show us the performance hit that would be caused by using a card without enough VRAM. "

It's common sense. If a game takes up 600MBs of RAM there's going to be a performance hit if you have 512MBs.

 

[phantomtrooper]

Why doesn't Tom's use Empire: Total War as a benchmark test? I don't know of any better looking RTS game. Also, more RTS's should have been used in this test since they typically use more video ram.

 

[youssef 2010]

One of the most illuminating articles I've ever read on this site.Thank you guys,you're the best.

 

[cleeve]

It's common sense. If a game takes up 600MBs of RAM there's going to be a performance hit if you have 512MBs.

I don't think you understand what I mean... it's not common sense unless you can demonstrate it.

How do you measure the amount of performance hit it will cause? If it causes a 1% performance hit it's irrelevant, if it causes a 50% performance hit it's important.

Without measuring this, common sense tells you very little, and no conclusions can be drawn.

You can't assume the performance hit will be anything in particular without measuring it.

 

[cleeve]

Just curious, did any of the GTA IV tests use AA/AF? Both images
appear to have significant staricasing artifacts along various edges.

No, we didn't need to, with GTA we were able to push the video ram usage extremely high without using AA.

 

[mapesdhs]

Ah, I see. In the last hour I've read that GTA IV doesn't support AA anyway
due to the nature of its engine - is that true? A big surprise for me if so.

Ian.

 

[Guest]

Could you guys also add some rendering programs to the mix? I would love to see the differences on 3ds max and photoshop.

 

[cangelini]

[citation][nom]mapesdhs[/nom]Ah, I see. In the last hour I've read that GTA IV doesn't support AA anywaydue to the nature of its engine - is that true? A big surprise for me if so.Ian.[/citation]

This is true, Ian. It's a console port with its own render pipeline. There's no in-game AA setting and forcing AA through the graphics driver has no effect on the displayed image. AF is supported, but it's not called AF in-game.

 

[Manos]

I havent seen worth upgrading from my 8800GTX since i still put everything on max details and with its 768MB on it has proved to me at least that on my setup makes a difference alo from the 512MB ones so Im glad i got that one when it came out. I know ill upgrade with the Fermi now if I dont get too impressed from the Ati new offerings, but so far I had other stuff to upgrade in my life than my GPU worth the money. The new Nvidia ever since 8800GTX had nothing impressive new to offer since i still work everything on max without issues except SOME frame drops with few titles. New GPUs do seem an sound promising and more than worth moving on to.

 

[aford10]

Good article..This is why it's important to know what people plan to use the PC for when putting together a system. A large part of the budget can be unnecessarily taken up by going overboard on graphics.

 

[nachowarrior]

I'd think it would be pertinent to include some hd 58xx series cards in here at the HIGHER rez specs. I understand, blah blah, current gpu, blah blah, different memory configs. Got that, but throwing the 58xx series in there allows you to push TONS more pixels, about 4-10 million more pixels, and would add a good "here's how adding a butt load more pixels affects performance" tangent to the mix. I guess there's other reviews where we can see that. but it would have been a nice addition. 😛 Other than that, Nice job on the comp specs within the limited scope.

 

[cleeve]

[citation][nom]scooterlibby[/nom]In the conclusion you state GTX260 has 768mb RAM. That was actually the 8800GTX. GTX260 has 896mb.[/citation]

Thx, fixed!

 

[spectrewind]

[citation][nom]nonamenow[/nom]Seeing a comparison screenshot is nice, but gif...seriously? Seeing the game in 256 colors and low resolution tells very little about the actual differences. If you need animation, please use e.g. apng. Just two jpg images sould do the trick, though.[/citation]

It's actually worse than that. The Compuserve .GIF format supports a 216 color gamut. 256 would be an improvement. Ideally, side-by-side PNG (Lossless RGB-alpha) files would have been better for this article.

.GIF considerations: [http://www.alphabetsystems.com/color/image216.htm]

 

[spectrewind]

Very well written article. This is the quality I miss from THG that used to be around in the later half of the 1990's.

I want to see more well-thought-out articles like this, by Don Woligroski. I have wasted too many minutes on the useless, often wrong, and journalistally appalling crap paragraphs that Kevin Perrish produces.