Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

512MB vram Vs. 256MB vram

makowb said:
http://www.macworld.com/article/132407/2008/03/macbookprocore2duo.html?t=201

Look here. Check out the FPS between the 2.4 and 2.5 on UT2004.

2.4 penryn 73.4 fps
2.5 penryn 89.4 fps

Tell me that's not a huge difference...yes the processors are 3mb vs 6mb cache, but 16 fps is huge.

Edited, macworld uses a botmatch not a flyby.
Click to expand...

That's definitely thanks in part to the processor, if you'll notice it also performed better on most of the other tests by a good margin. If you show me a benchmark of two identical systems running the different versions of the card, I'll fully admit I'm wrong (And therefore my sources are wrong, which will lead to a whole bunch of fist shaking. :D) and move on. But believe me, I've spent the better part of two hours trying to find a reliable site, and it isn't easy.

For now, I'll stick to my guns still someone else makes me look bad. :p

EDIT:
Erasmus made a great post in the other parallel thread that explains things much better then my admittedly poor attempts:

Erasmus said:
The argument over 256 versus 512 has been going for a while now. I'd like to throw a few extra facets out there, on both sides.

Firstly, now that the MBP uses DDR3 RAM at 1066MHz, there would be minimal difference between the data transfer speeds between the GPU's own RAM, and system RAM. Yes, there will be a large difference in latency, but if the texture files that the GPU are big enough, I doubt whether this would be an issue. Seeing as the 9600 has Turbocache, I'm sure this seriously reduces the benefits of large amounts of VRAM.

Yes, games will not benefit that much from more VRAM, because I imagine they involve rapidly swapping in and out different data about the virtual world as we proceed through the game. Therefore games would be bottlenecked by bandwidth, and how much texture and polygon information can be pumped to and from the VRAM, main system RAM, hard disk, CPU and GPU core. What about CAD programs though? These ultimately do not require rapidly changing texture maps, vertices and shaders, and therefore I would imagine would not require overly large bandwidths. Therefore, they are more likely to be limited by the amount of data about the object that you can fit in the VRAM.

Therefore, if you're a gamer, I don't think you need to worry about the VRAM too much. 256MB is enough for the moment. If you're into some seriously heavy CAD, maybe you want to go for the 512MB version.
Click to expand...

In light of the above information, I would add that doing such heavy lifting on a notebook is a little misguided, desktops can power through that stuff much faster. So while you would be better off with the extra ram for heavy CAD applications, you'd be even better off spending that money on a desktop in the first place! :p
 
MichaelThomas said:
So the 512MB will only utilise 256MB REGARDLESS of what you through at it?
Will this change in snow leopard?
Click to expand...

NO. It DOES use all 512MB.

The 9600GT is just another 8600GT part. The issue is with a low end GPU, adding a bunch of RAM to it dosen't magically make it faster. That said, it could still help with multiple displays, could still help for programs that actually do do some work on the GPU (as more are going to start doing), etc.

Personally I think 256MB is kind of skimpy, but I doubt right now you'd notice a difference except maybe in games (but then the biggest limitation is still the low end GPU).
 
Wolfpup said:
NO. It DOES use all 512MB.

The 9600GT is just another 8600GT part. The issue is with a low end GPU, adding a bunch of RAM to it dosen't magically make it faster. That said, it could still help with multiple displays, could still help for programs that actually do do some work on the GPU (as more are going to start doing), etc.

Personally I think 256MB is kind of skimpy, but I doubt right now you'd notice a difference except maybe in games (but then the biggest limitation is still the low end GPU).
Click to expand...

I disagree...from what I've read, a bus width of 128bits which this card has will limit the memory usage. You are correct in saying that using an external display the extra memory will help (if were talking a 30inch cinema) but for gaming, the card is still limited by that 128bit bus. Apple dropped the ball in not offering one with a 256bit bus width..

someone correct me if im wrong
 
How does these arguments translate if you are using Apple's Motion 3 application in Final Cut Studio. Motion is suppose to be optimized by tapping the GPU versus the CPU... Your thoughts...

Thanks in advance,
Lonnie
 
thirtytwopaths said:
I disagree...from what I've read, a bus width of 128bits which this card has will limit the memory usage. You are correct in saying that using an external display the extra memory will help (if were talking a 30inch cinema) but for gaming, the card is still limited by that 128bit bus. Apple dropped the ball in not offering one with a 256bit bus width..

someone correct me if im wrong
Click to expand...

I wish I could find some benchmarks comparing this kind of thing for low end GPUs. The issue is kind of whether the GPU will run out of video RAM before it runs out of processing power. I know on like the 64 processor parts and better, 512MB really does matter. And I know at least on the games of the time, Apple's MBP with a 256MB part with faster RAM was beating out 8600GT parts with 512MB of slower RAM. I don't know, I think for games it's better to error on getting 512MB, as that does seem to be the standard modern games are benefiting from.

It can definitely SEE and address the RAM though.

mamasboy said:
How does these arguments translate if you are using Apple's Motion 3 application in Final Cut Studio. Motion is suppose to be optimized by tapping the GPU versus the CPU... Your thoughts...

Thanks in advance,
Lonnie
Click to expand...

Honestly I know nothing about how that program uses the GPU, and how much RAM it uses. I'd guess you'd be future proofing it a bit, as future versions of it (and the other programs that are going to be using the GPU) will probably use more and more.

I don't know...
 
i'm confused with why more ram is good for multiple desktops but not very effective for running certain programs.

would it suffice to say that for your everyday consumer, 256 is sufficient?
 
nanvinnie said:
i'm confused with why more ram is good for multiple desktops but not very effective for running certain programs.

would it suffice to say that for your everyday consumer, 256 is sufficient?
Click to expand...

512=256 if you are only 128 bit. Apple dropped the ball.
 
I can't hardly believe what I read. Why the hell would Apple put 512 VRAM inside if it has NO EFFECT !? I wouldn't expect Apple to cheat costumers like that.

By the way, is it the same for the early 2008 MPB ?
 
thirtytwopaths said:
I disagree...from what I've read, a bus width of 128bits which this card has will limit the memory usage. You are correct in saying that using an external display the extra memory will help (if were talking a 30inch cinema) but for gaming, the card is still limited by that 128bit bus. Apple dropped the ball in not offering one with a 256bit bus width.
Click to expand...

Bus width has nothing to do with addressing memory.

someone correct me if im wrong
Click to expand...

You are wrong.

The bandwidth of a bus between a GPU and its onboard memory is not as important as access times. If the bus is clocked higher, for instance, the card can still pull data off a narrow bus from the on-board memory before the GPU will have exhausted its pipeline (which is the primary goal in card design.)

The idea is to make sure that the card isn't starved for data to process, which is why you cache commonly used textures (or other data) in the card memory. But if the card isn't fast enough to be starved by a slower bus, then there's no sense in putting in a faster bus that can't be used but which has thermal and power consequences in the form of wasted refresh cycles.

Having more local memory to cache data like textures will mean better performance at higher resolutions in things like games which have textured backgrounds that need to be constantly processed as the POV changes, but not so much in applications where the transforms are on large sets of data that can't be cached. Small amounts of data that are repetitively transformed by really fast cards are why you need high bandwidth access to large local caches.
 
bbplayer5 said:
512=256 if you are only 128 bit. Apple dropped the ball.
Click to expand...

Nonsense. The 128 bit bus does not mean you can only access 256mb RAM. It means you can only put 128 bits or 16 bytes on the bus at one time. Even if your assumption was correct (that the addressable space is limited by the bus width) 128 bits would give you an addressable space of several Petabytes of data, not just a measly 256 Megabytes.
 
nanvinnie said:
i'm confused with why more ram is good for multiple desktops but not very effective for running certain programs.

would it suffice to say that for your everyday consumer, 256 is sufficient?
Click to expand...

If you're browsing the web, using Office, stuff like that, it's fine. If you play games, or want to be able to take advantage of programs that use the GPU down the line (Apple's going to be pushing this it sounds like in the next OS), then 512MB is the way to go.

Though if you can't afford the pricier system, the cheaper MBP is still a much better deal than the Macbook if you do things that use the GPU!

bbplayer5 said:
512=256 if you are only 128 bit. Apple dropped the ball.
Click to expand...

Glad a few other people are saying it too. Of COURSE it can address the whole thing. You could do that through a 2-bit path, that has nothing to do with anything.

I think what's confusing people is articles that talk about more RAM not being USEFUL. That dosen't mean it isn't being addressed. A common trick on low end video cards is to throw a lot of RAM on them. People who don't know better, think their low end card with more RAM on it than even a high end card might need is somehow better, but it's not.

And low end cards don't need as much RAM anyway, because they can't use as high of resolutions. But...at least for higher end cards, not only are we seeing a benefit from 512MB (a real benefit), even the 1GB cards are doing better than the same card with 512MB. But of course that wouldn't be the case for the Geforce 8600GT these things use...but still they might hit situations where they need 512MB.
 
NuVector said:
Nonsense. The 128 bit bus does not mean you can only access 256mb RAM. It means you can only put 128 bits or 16 bytes on the bus at one time. Even if your assumption was correct (that the addressable space is limited by the bus width) 128 bits would give you an addressable space of several Petabytes of data, not just a measly 256 Megabytes.
Click to expand...

Thank you.
 
bbplayer5 said:
Ok so its "ok" to get the more memory version. How about the L2 cache?
Click to expand...

Darn, I'm fairly unclear on that too. It's going to really vary depending on what you're running. If I'm remembering right, for most things on Windows, there was a bigger difference between the 2 and 4MB Core 2s than between the 3 and 6MB ones. But OS X might be different, and of course programs are always getting bigger.

Personally I'm getting a 6MB version when I get a laptop. I mean the price difference between those chips isn't huge, and you do get the faster clock speed-though if you can't afford it, don't worry about it. The 2.4GHz model is still plenty fast!
 
You must log in or register to reply here.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back