quadro fx 1300 or 3400?

[hydrophobic]

There is an article in the November 2004 issue of Digit called "3d Graphics Cards", and they compare some of the best cards. They compare the Quadro FX 1300 and the 3400. They used Cinebench, Maya, and 3ds Max to test the cards performance. They gave the 1300 4.5 stars and rated it a best buy and gave the 3400 3 stars. They said the 3400 was priced too high for its performance.

Cinebench: tests both OpenGL and raytrace. Higher scores are better.
1300: 2450 (6.5x)
3400: 1240 (5.97x)

Maya: dense polygonal scene of 100 spheres and large texture animated over 60 frames. Lower scores are better.
1300: 16.0s
3400: 18.1s

SpecView(3ds Max): runs through a series of scens. Higher numbers are better.
1300: 25.13
3400: 23.5

However, they said that the 3Dlabs Realizm 100 was an excellent card. They gave it 4.5 stars.

[greyface]

What the hell? How can the 1300 perform better than the 3400?

[hydrophobic]

I'm not sure why the 1300 beat out the 3400. The machine that they ran the test on was a Dual Pentium Xeon 3.4GHZ PCI Express machine. Also, price had something to do with it. The 3400 is $400 more than the 1300.

[CiaránMurphy]

Well I have a 3400 (Precision 670 3Ghz Xeon) - If I were Digit I'd take a look at their figures again!

Both Tom'sHardware and SPEC itself seem to suggest otherwise - both their results
show the 3400 about 25% faster than the 1300 which tallies with Nividias on tri/sec values.

http://graphics.tomshardware.com/graphic/20040816/opengl-pcie-16.html
http://www.spec.org/gpc/opc.data/vp8/3dsmax-perf.html

The range for the 3400 was from about 31 to 34 (depending on CPU in test machine) - that's a lot faster than Digit's 23.5 (Possibly a typo... 2 & 3 swapped?)

The official hardware figures from NVidia in Triangles per sec are
3400 - 117 Million/Sec
1300 - 88 Million/Sec
700 - 69 Million/Sec

And subjectively the FX3400 is Fast - I can spin a heavily modelled appartment block around far faster than I can on my FX700 - yes probably close to twice as fast.

Also I've never really liked Cinebench - the data sets are small, the textures are small, so the benefit of higher bandwidth card like the 3400 over a 1300 will not be apparent. Try load a 50MB file in that loads lots of textures and you see that your 3400 will be worth it.

I do Arch Viz - so lots of high poly counts and lots of high res textures. Cinebench is a fair indicative benchmark, but only if you are running Cinema4D, that's all it's meant to indicate. It should not be construed as a card benchmark - its a platform system test.

In any case if you are thinking of spending close to a grand on a Quadro card I imagine it is for Max (which the Quadros specifically accelerate through Maxtreme) So don't pay any attention to Cinema 4D. These benchmarks always test OpenGL or DirectX. But who uses those? I never use either of them because they are far slower than Maxtreme.

But yes the 3400 is expensive (more than 25% extra) but if you are getting it bundled with a Dell you can get it for about E800 extra. I'm sure the 3dLabs card is good too - but again I'd make sure it has optimized application specific drivers.

Edit (I'm an IDIOT :-)) You're obviously concerned with XSI!!! But in any case the same holds true - higher bandwidth cards better for bigger data sets and don't presume that performance figures for Cinema 4D or 3dsMax will translate directly in XSI.

[CGTalk Moderation]

This thread has been automatically closed as it remained inactive for 12 months. If you wish to continue the discussion, please create a new thread in the appropriate forum.