RSS Feed

HCW Tech Blog

For the latest info on computer hardware, tech, news, video games, software tips, and Linux, check out our new improved front page: HCW Tech Blog

Reviewed by: Bryan Pizzuti [04.29.03]
Edited by: Carl Nelson

Card Manufacturer: InnoVISION
GPU Manufacturer: NVIDIA
MSRP: $80-$100

Discuss this article in the forum!
Registration NOT Required!



Well, finally the GeForce FX series has arrived.  Well, at least part of it.  The FX5800 series seems to be the most reviewed product that no one can actually buy since the Athlon 2800+.  But that's not the entire FX line, luckily.  NVIDIA is ditching the Ti and MX ratings, and following in the footsteps of ATI by launching 3 levels of numerically designated cards.  Today we look at Inno3D's Tornado GeForceFX 5200.

So what's with the new numbers?

Well, as I said, they're sort of taking a page from ATI's book, who have a range of products designated by numbers. NVIDIA WAS doing this in a way, but they also subdivided their lines with letter designations as well. And let's not forget the TI4800 and TI4800SE confusion, right?  Anyway, ATI's upcoming lineup for the 9XXX series includes the value 9200 and 9200 Pro, mid-range 9600 and 9600 Pro, and high end 9800 and 9800 Pro.  The 9 supposedly signifies DirectX 9 compatibility, though the 9200 series will not have it.

The FX series consists of the baseline 5200 and 5200 Ultra, the midrange 5600 and 5600 Ultra, and the high end (and rare) 5800 and 5800 Ultra (complete with blow dryer).  Eerily familiar?  Yes. But it's also a model of simplicity, and at least the numbers match up enough to instantly know which is supposed to compete with what.  Instead of 9, NVIDIA uses a 5, probably to signify "GeForce Series 5" or something.  And NVIDIA likes the word "Ultra" rather than "Pro." But other than that, it's very easy to find a direct comparison.  So this can't last long. ;)

NVIDIA has been very closed-mouthed about the internal architecture, claiming only the number of pixels it can process per clock, which, on this card, is 4. Notice the word "can."  That doesn't mean it ALWAYS processes 4 pixels per clock.

Supposedly, pixel shader effects are going to make multitexturing technology unnecessary, to the general trend is moving towards one texture unit per pipeline, like the way ATI's DX9 cards are designed.  What we DON'T know is what the GeForceFX core does when it needs to apply more than one texture per pipe.  One possibility, for example is sending a 2 copies of one pixel through 2 pipelines, each receiving a different texture operation from each texture unit.  That would mean that, when multitexturing, only 2 pixels per clock are processed.  That's not necessarily bad, since the GF4MX series was able to process 2 multitextured pixels per clock.  But it could also only process 2 single-textured pixels per clock, whereas the FX5200 can do 4.

There's also the theory that somehow this particular FX card is simply a 2 pipe/2 unit architecture like its predecessors, which somehow is capable of squirting out 4 pixels per clock. 

But this is all theoretical at this point.  NVIDIA just isn't divulging that much about the FX core and its internal architecture.  But it's important to note what they HAVE told us, and what it might mean.  One other thing to keep in mind is that, when 3DMark2003 came out, NVIDIA squeaked, complaining that the first game test, which uses single texturing, wasn't a valid test of the way games are written and played.  It may be a safe bet that if NVIDIA doesn't like seeing something that uses single texturing, then their architecture is somehow more effective when doing multitexturing, which suggests multiple texture units per pipe.  This is, of course, just fine if you like to play games that use multitexturing. :)

But anyway, we don't REALLY know what's going on in the core, or how it does it.  And they seem to be quite happy about it.  Maybe their PR section is being run by Donald Rumsfeld?

The Card

The Tornado FX5200 is passively cooled.  This passive heatsink is a bit of a monster though, as you can see.

This makes the card comparatively heavy.  Also, if you look closely, you can see the plug for an on-board fan, so it seems like the 5200 wasn't originally intended to be passively cooled, or it may share the same exact PCB design as a higher end (and hotter) GFFX card (maybe the 5200 Pro?).  While it's a big, reasonably heavy heatsink, it's not big enough to interfere with the first PCI slot, unlike some heatsinks and video cards that we won't mention but are also made by NVIDIA. ;)

We've seen other GFFX 5200 cards using smaller passive heatsinks, so we consider this monster to be a nice extra from Inno3D.

This shot is hard to make out because it's so close-in but you can see the heatsink, and the card in the first PCI slot.  As you can see, there's plenty of room.

Next Page: (2)