Video cards (2)

[1000101-]

On a side note the geforce3 does do more work per clock than a geforce2 Ultra does; kinda like a 500mhz MIPS CPU runs circles around a 500mhz intel CPU. But yeah I'd say go with the ti200. my friend bought the visionTek ti200 and has it overclocked to a 250mhz core and about 510mhz memory; which is faster than the ti500; and it shows to.

That Ole Russell

[jtk77]

In reply to:
On a side note the geforce3 does do more work per clock than a geforce2 Ultra does; kinda like a 500mhz MIPS CPU runs circles around a 500mhz intel CPU. But yeah I'd say go with the ti200. my friend bought the visionTek ti200 and has it overclocked to a 250mhz core and about 510mhz memory; which is faster than the ti500; and it shows to.

That Ole Russell

same reason a AMD 1.2 gig chip is faster than the intel 1.4 gig p4..it deos mroe operations per clokc ...thats a very important point when looking at these numbers.
Unfortunitly,. I don;t think there is away to rate this yet...
on the MIPS thing ,, i think thats a little different , aren;t those ( in the SGI anyway) based in 64 bit instead of 32 bit? ?

Anyway , thats somthing that should be looked at..its a shame when consumers think the mac is slow becosue they see a 550 mhz chip and then a 1.4 gig chip on a PS next to it and have no idea why the mac is more expensivebecouse the public is not educated in the matter.
to be honest with you...its confusing as hell

................

[archone]

That Ole Russell,

I agree with your observation about the Geforce 3 chip doing more operations per clock cycle.

The results of your friend's overclocking his visionTek ti200 to 250Mhz core and 510mhz memory is great! Since the normal Ti 200 specs are 175MHz core and 400 MHz memory.

If you check out Tom's Hardware, on his Geforce 3 card round up, he stated that he had an Asus V8200 T2 Deluxe Ti200 get up to 260Mhz core and 554Mhz memory!!!

Gee, I can see the writing on the wall!!

Get a Geforce 3 Ti 200, overclock it, and SoftQuadro it. A cheap and powerful Maya 4 video card solution!

Thanks That Ole Russell!!!
Thanks everyone!!!
--Sir Gareth

In reply to:
On a side note the geforce3 does do more work per clock than a geforce2 Ultra does; kinda like a 500mhz MIPS CPU runs circles around a 500mhz intel CPU. But yeah I'd say go with the ti200. my friend bought the visionTek ti200 and has it overclocked to a 250mhz core and about 510mhz memory; which is faster than the ti500; and it shows to.

That Ole Russell

[jtk77]

to anyone messign with this, jsut make sure your system has plenty of room to vent ..over clocking these cards as well as say having dual AMD chips and a few 64 bit pci slots chugging away on a 450 power watt..thats some heat, be careful overclocking in a sealed case.

................

[1000101-]

Yeah that whole confusion over clockspeed to performance ratio is why my athlonXP 1500+ isn't called a Palamino 1.33ghz CPU. In some respects it even outperforms 1.6ghz p4's. I do recommend a hefty power supply; not so much for the 64 bit PCI slots (researching for a fiber channel card for my company to buy; I found they actually pull less watts and voltage than 32 bit PCI slots); I would recommend a 450 watt power supply though; especially in a dual athlon system; those CPU's pull hella juice; as will the large numbers of hard drives and fan's you'll surely install on your overclocked beast.

That Ole Russell

[senshi]

well when I wrote my post about choosing a GeForce2 Ultra, I didn't simply mention raw MegaHertz speed (which all of you put down) , but I also wrote actual data about the bandwidth of the card and the power of the card.

quote from me from my post:

"
The GeForce2 Ultra (which is faster than a GeForce 2 Ti ) has a higher raw Texels/s through put than any of other nvidia cards (even the geforce3 Ti500) and a respectable memory bandwidth of 7.6 GBytes/s compared to Geforce3 Ti 500 8 GBytes/s
"

GeForce3 has been made better because they added those extra DirectX 8 features, which Maya does not use.
Also they added the LightSpeed memory architecture which, for it's performance gain, does not justify the price increase.
Also it's got 3d generation hardware T&L, which is only used on games.

Now can you tell me still that the GeForce3 would yield better performance in Maya?

P.S. Make sure you backup your statements with concrete FACTS.

[1000101-]

It does; fil rates asside; the transform and lighting unit on the geforce 3 is somehwat improved over its older brother. An example with this would be to load up a walk cycle you might be getting 5 fps on with a geforce2; on a quadro2 card; you might get 6 or 7 average; and on the geforce3/quadro DCC you'll end up with about 7-8fps; regardless of the fact that the ultra's fill rate is higher. The geometry isn't pushed by fill rates; textures are.If you notice; higher end Wildcat 5110; the fill rate is at a "paltry" 332 Million pixel/sec. Game cards depend on high fill rates because they may have to sample a frame 2 or 4 times with at least 2 layers worth of textures; hence the need for 2+gigatexels worth of memory bandwidth. Geometric performance is more or less on the actual graphics processing unit of the card. Which every generation needs an overhaul to keep it faster than the CPU.
When I dropped my geforce3 based card into my system(a then dual pIII 750); from the 2; I noticed 2fps difference; which doesn't seem like much to a game; but to me and maya; its almost a worlds difference when moving a limb. When upping the CPU. The ultra is a fine card; but a ti200 (which you can overclock with stock heat sinks to near ti500 levels getting the better fil rate if you need it) softquadro'd is DEFINATLEY a good choice for the budget workstation needs.

That Ole Russell

P.S. If you want; I'll run a series of tests, I still have my old geforce2; but just working with the card I've noticed a hella difference.

[senshi]

I'll take your word on that man, and I could be wrong of course, I'm not saying I'm 100% riht...

So waht's the transform and lighting got to do with it again now?

EDIT

Well, I'll be...

you were right on pal http://www.aceshardware.com/read.jsp?id=45000271http://www.aceshardware.com/read.jsp?id=45000271

Edited by efishta on 01/04/02 12:53:25 AM.

[1000101-]

Its all good ;> I see I did do my homework hehhe
but to define transform and lighting; its got everything to do with geometry performance. Transformation does anyway; when you adjust an IK chain for example; it transforms(moves) the vertices on the model from one point to another. T&L handles these calculations in hardware. The lighting is designed to of course; calculate how light falls; once again freeing the CPU from having to do it. Dunno if thats much help, but thats it in a nutshell.

That Ole Russell

[exodusmaximus]

Pixel fill rates are missleading!
Wildcat5110 will kill any GeForce card! - so will FireGl4 (which I consider a better card)
When I changed quadro2 pro for FireGL4 I gained over 15fps in wireframe mode (40.5 - FireGl4 from 25 - quadro), 10fps in shaded mode (21fps from 11fps) and 6fps in light+shaded mode (12fps from 6fps) while using openGL test scene.
Duall display mode too!
Great card (fireGL4)

In reply to:
It does; fil rates asside; the transform and lighting unit on the geforce 3 is somehwat improved over its older brother. An example with this would be to load up a walk cycle you might be getting 5 fps on with a geforce2; on a quadro2 card; you might get 6 or 7 average; and on the geforce3/quadro DCC you'll end up with about 7-8fps; regardless of the fact that the ultra's fill rate is higher. The geometry isn't pushed by fill rates; textures are.If you notice; higher end Wildcat 5110; the fill rate is at a "paltry" 332 Million pixel/sec. Game cards depend on high fill rates because they may have to sample a frame 2 or 4 times with at least 2 layers worth of textures; hence the need for 2+gigatexels worth of memory bandwidth. Geometric performance is more or less on the actual graphics processing unit of the card. Which every generation needs an overhaul to keep it faster than the CPU.
When I dropped my geforce3 based card into my system(a then dual pIII 750); from the 2; I noticed 2fps difference; which doesn't seem like much to a game; but to me and maya; its almost a worlds difference when moving a limb. When upping the CPU. The ultra is a fine card; but a ti200 (which you can overclock with stock heat sinks to near ti500 levels getting the better fil rate if you need it) softquadro'd is DEFINATLEY a good choice for the budget workstation needs.

That Ole Russell

P.S. If you want; I'll run a series of tests, I still have my old geforce2; but just working with the card I've noticed a hella difference.

[1000101-]

In reply to:
Pixel fill rates are missleading!
Wildcat5110 will kill any GeForce card! - so will FireGl4

Thats what I'd been saying. I will say this; the geforce cards do actually have a faster rasterizing engine than even the allmighty wildcat. This is the most confusing factor when reading benchmark data. Wildcats and fireGL's were designed to push vertex data and wireframes; both having dedicated transform engines. Geforce and Radeon cards are designed to throw as many frames on the screen in a given second as possible, with moving geometry a 2ndary function, and accuracy and image quality being 3rd on the list.

That Ole Russell.