If you doubt the requirement for >64MB, take a gander at all the professional graphics cards with at least 128MB, the stuff that's intended to be used at very high rez/refresh rate for 3D modelling, f'rinstance, and costs more than the rest of the computer... high-end Oxygen or FireGL, or better yet a Wildcat.

Not something a gamer would appreciate at 10x7, for sure -- but just wait, you'll get there. ;)

Not to mention the increased memory requirements of dualhead (3200x1200 or 1600x2400) and quad-buffered stereo (very cool stuffs). Besides that, a lot of us are working at 1600x1200 or above now, and would greatly desire to do our 3D at that res as well.

3D requires a LOT more memory than 2D, but that wasn't really the question.

3D rendering requires a bit more memory than 2D. I believe you have to factor a Z-buffer into your 7.4MB calculations. A 24-bit Z will kick you up to a ~22MBs starting point for 1600*1200*32, like R2P2 said. This is before texturey goodness.

There's a reason why Voodoo2 cards were limited to 800*600 with a 4MB frame buffer. Using a simple X * Y * bit-depth calculation, it should've been able to hit 1600*1200 as well.

Originally Posted by nuclear
doesn\'t equal 7.5 megs, it equal to 7.32 if we follow your calculation :D

32 bits *1600 * 1200
________________
8bits/byte * 1024

=7500 kilobytes = 7.5MB

Originally Posted by nuclear
a card with 8 megs like the old matrox millenium (before the g nominations) were able to drive a monitor to 1600 X 1200 X 32 bit in 2d. The only limiting factor is the refresh rate which is now derived from the ramdac.
For 3d, that\'s another thing, because of all the rendering.
128 megs might also help for higher quality textures and bigers textures without having to resort to texture compression.

Damage -- I used: (1600 * 1200) pixels * (4 bytes x + 4 bytes y + 4 bytes color = 12 bytes) per pixel = 23040000 bytes. 2304000 / 1024^2 = 21.97MB

Crap. Lemme guess -- they don't store x and y values, they just use an array? Don't mind me...

I think 64MB extra memory will make about 15% Differences
in today game performance.
256bit GPU on Radeon 8500 using 128MB memory. Is not
very good idea.
Because I think you would need 512bit GPU to handle 128MB
memory on the card in order to get true performance increase.

If you people don't aggree with me just say so!

Hmm. I get 7.5MB for 1600x1200x32. You sure about that, R2P2?

Oh, and don't forget that you want to use double-buffering, if not triple-buffering for 3D rendering, so you're looking at 44 or 66MB before you've even gotten to the z-buffer and textures.Now that I think about it, 44MB for buffering at 1600x1200 seems really high; I'd think most newer games would stutter a lot with only 20MB of texture memory at 1600x1200x32 double-buffered.

#22 -- Actually, 1600x1200x32 requires about 22MB, using 32-bits each for the x-coodinate, y-coordinate, and color of each pixel. That's 12 bytes per pixel, and 1600x1200 is a *lot* of pixels.

Originally Posted by jbirney
Neo69,

as it stands right now (ie before the press release) the Ti200 and the LE are at the same price point of a MSRP of 199.
So you have to look at the other factors, like features, perfromance, ect... yes its a close call

AG9

\"will the extra 64 mb allow higher resolution/color depht/refresh rate in 2D? \"

Not really. I thought only 6 mb was required to run at 16x12 at 32 bit. So the card can support higher if memory was the only factor. But its not. The RAMDACs help to set what the refreshrate is. Extra Memory really does not help in this case.

Now I am wondering if it will help with FSAA scores. You might be able to run at a higher res (although at a huge perf hit).

the 8500LE looks VERY good though, at 199 dollars...

this ati announcement is good n well, but we still haven't got 8500s here in Australia. Maybe this means the 8500 and 8500DV will finally reach us....*sigh*

BTW, those prices don't look cheap to me. 'Aggressive pricing moves?'

Biff - All the new Ti's support dualhead, and the hardware to do it is present on GF3 Ti500 (and likely Ti200/classic as well). The Ti4400 might be halfway affordable...

mmm.

I smell a cheap CAD card.

100 dollars for 25 mhz in the core & memory? LOL
(and i dont really care about DVI)

The MX has different technology that still requires it's market presence, primarily dual-head ...mmm, dual-head... :) which is probably why it still exists. It's 'market' doesn't have the same performance requirements as the non-MX and thus the existance of the Ti200 and the like.

Of course, if nVidiots would put dual-heads ...mmm, dual-heads... :) on their main line stuff that don't start at $399, that would be all the better.

yeah, well, i think that the gf3 ti200 is a far more attractive offer because of the vertex and pixel shaders and cheap price. the extra 64MB are such a fucking waste (proven at digit-life's 128mb shootout) so....bring on the gf4 ti 4600 :-)

AG3: You hit the nail right on the head with your post here.

AG4: Not likely at all. This is something that would create something of a significant public backlash against nVidia, and no company wants that, monopolist or not..

sort of off topic...but what does everyone think of the Appian Graphics cards...are these totaly worthless for anything but high-end cad/cam?

Uh, Forge, not to be immodest, but look here: http://www.tech-report.com/news_reply.x/3367/

Spune -- At least ATI will probably give TR review cards, just to look like they're not mad about the LE thing ;)

http://www.tech-report.com/forums/viewtopic.php?topic=230&forum=3&5

Damage- "I'm not under NDA of any kind since NVIDIA's PR team has decided TR doesn't "deserve" a GF4 review on release"

No Geforce4 review for us :( not yet anyway

Does the 8500 core have the ability to use the use the extra memory bandwidth? It's not like any reviewers stated that "This card is crippled by having only 64MB of memory." And, putting 128MB on an 8500 LE sorta smells like the GF MX 400, which was pretty much a marketing ploy to get rid of plentiful, slow memory chips that couldn't be used (well, 90% of the time) by the GPU.

will the extra 64 mb allow higher resolution/color depht/refresh rate in 2D?

128MB haven’t done a whole lot of good for the GF3 Ti200 line and I don’t think it will for the 8500. 128MB Ti200s are slower than the 64MB versions.

AG #3 - As far as the performance points, I have no clue WTF Nvidia is thinking. As far as naming goes, I think it was LocalYokel who made it all very clear and simple for me: Ti200 = downclocked Ti500. GF2MX or GF4MX = crippled product.

See:

GF2 MX @ GF2 GTS speeds != GF2 GTS. There are 4x2 pipes on GTS and 2x2 pipes on MX.

GF3 Ti200 @ GF3 Ti500 speeds ~= GF3 Ti500. Pipes, shaders, core exactly the same. Even the ram is almost exactly the same.

So basically, the Ti200 wasn't crippled enough to warrant a GF3 MX name. Make sense?

AG #4 - You betcha. :) I actually heard Nv has stopped GF3* production entirely, as of two or three weeks ago.

AG #5 - Not unless Damage's poker face has increased in effectiveness about 500% in the last few weeks.

are the 128MB cards ATI products or third party vendors? I have heard some third party manufacturers are gonna make 8500 cards with 128MB, but not aware ATI was releasing any. also... no mention of them on the ATI site.

-danny e.

[quote]
Once you've read all that, you'll be ready for what's next.
[/quote]

Hmmm. Could [b]What's Next[/b] be a review of a certain new Nvidia-based video card currently being benchmarked in the bowels of Damage Labs and protected by NDA?

Mum's the word...