Quote:


A 4500 will show a difference in benchmarks, a 5500 might show an even greater amount since it can do more work them a 4500.

Since FSAA takes away a decent amount of Video RAM, I'd say the performance difference might be enourmous if high Resolution + FSAA are used.

Otherwise, more Video RAM still means more Texture RAM available and therefor more Texture Detail (which IMHO still has the biggest impact on Quality)

Quote:


False logic- Even if the test is biased against 3dfx hardware *it shows a difference between two 3dfx boards*

Were not talking 3dfx vs NV but a 4500 vs a 4800, the only difference between the two boards is the ram amount.

A board running high res and AA will stutter if more textures have to be paged out of main memory, much less so with a video board with more ram. There are no shortage of Q3 levels with loads of textures a 4500 can run with good speed but will be hobbled by the limted ram.

true there but 3D mark is too wrong to use , Using a nutral benchmark utillity is a far better way for testing the cards and the games that they were built for.

3D mark has a very bad rep here in the netherlands just because it's so Pro NV based.

But yes Gary I do agree that comparing the Voodoo4 4500 and 4800 in 3D mark does give a difference but it if you'd test it in game built for 3dfx cards or a benchmark that is more suitable for them you might see a bigger difference

The most dutch people use the games the cards are going to run with, so do'n test's in Unreal Ver 2.26 for instance like Rolo01 did with his Voodoo 5 5000 AGP 32MB versus Voodoo 5 5500 AGP 64MB, Voodoo 5 6000 AGP 128MB, Voodoo4-2 4800 AGP 32MB DDR and Voodoo4 4500 AGP 32MB.

Do remember that a 5000 AGP with 32 MB is actually a castrated 5500 AGP 64MB

His test's show totally no performance hit between the 5000 AGP as 5500 AGP using the same driver, now try to explain that?

The 5000 AGP has 2x 16MB SGRAM while the 5500 AGP has 2x 32MB SDRAM while the differences in Unreal Ver 2.26 were nihil!
This test gave me the idea that even a 6K with 4x 64MB would be done the difference would be minimal.

Otherwise ask Rolo01 he's the one that did this test quite a long time ago.

ah here it is
http://www.3dfxzone.it/enboard/topic.asp?TOPIC_ID=784&SearchTerms=5000


well look at that the 5000 AGP is just in most cases faster than the 64 meg 5500 heh, but when FSAA x4 is uitlised the 64 meg card has no performance increase what so ever the 32MB card as the 64MB have exactly the same speed.

But if you want to use the cards in newr applications then I would agree that extra ram can be usefull especially apps like HL2, Call of Duty 1.51 things like that would love the extra ram, but games where the cards were built for like Unreal 2.26 as you see the change is very minimal and actually reversed.

that Voodoo4-2 4800 AGP is actually the Daytona AGP + TV-Out it runs core/mem 143Mhz but it has been OC'ed to 166Mhz core/mem, it has 64Bit DDR that's why it's somewhat slower in FSA x2 than the 4500 AGP, the 128 Bit SDRAM seem's to give the 4500 a little more bandwidth heh.

But yes a 4 x 64MB V6K would have a nice impact on newer games on higher reso's sure there

hehe, indeed.

This test was meaningless, it becomes interesting when using Games from 2001 and later (32MB Textures +)

It requires no testing to see that Unreal I does not benefit.
UT2003 would be the first to see a major improvement going from 32 to 64MB.

UT2003 should work in Lower Detail in 800x600 16bit, maybe still playable in 1024x768 16bit (depends on level loaded though).

1600x1200 32bit will work, but only give a nice Slideshow most of the time

Quote:


Not an apples to apples test- A prototype board with unknown bios vs a production board?

I'll repeat since I'm not getting through- In a 4500 vs 4800 test (boards with same bios aside from extra bank being disabled) I have observed a difference in FPS between the two boards.

Ergo the extra ram made a difference and in an even more demanding application will have a larger impact as bigger texture come on line causing more and more page hits to main memory.

Try playing Quake3 with an 8MB Voodoo II vs a 12 MB board.

well the 6000 should run CoD 1.51 very fluent it has 4 GPU's with each 32MB it should crush the 4800 with ease!, My V6K ran CoD 1.51 flawless @ max settings @ 1600 x 1200 x32 with Mesa FX 6.2.0.2 & AM 3.1 R11 under Win2K Pro:


no glitches no graphical errors nothing but fluid gameplay, was at a lan party everyone was very supprised that a the 6K did so darn well @ max settings @ 1600 x 1200 x32 no AA though.

The Voodoo 5 6000 is a star in that game, if it's meant for heavy duty the 6K will always be the best at it for a 3dfx card, but yes with the 4x 64MB upgrade it might be able to use the FSAA x2 setting alot easier at that nice setting.

the setup I used for that test was:

AthlonXP 2700+ @ 2167Mhz default
1x 512MB PC 2700CL2.5 from Samsung Original
EPoX EP-8K3A+ Ver 1.1
3dfx Voodoo 5 6000 AGP 128MB Rev.A 3700 + Internal Hookup & PCI-Rework by Hank
SB Live 5.1
Win2K Pro + SP4 & Rollup 1 for SP4
DX 9.0c
AmigaMerlin 3.1 R11 & MesaFX 6.2.0.2 opengl32.dll

And Thanks for your help Gary & Hank for that 6K my thanx are still greatfull for you two blokes