If V5 5500 would of used DDR from jump....

[Remedy]

I often wonder if 3dfx would of used DDR on the V5 series would it have been able to help it keep up with the competition?

 

[Soccerman]

performance wise, it would be pretty good against the GTS, because it would have a surplus of double the memory bandwidth of the GTS, but of course, the VSA-100 isn't that powerful, so you can only do so much.

I know, in higher resolutions, you would get less of a performance hit, due to the RAM. the question is, how limiting would the VSA-100's be with nearly no memory bandwidth restrictions?

I often wondered what it would have been like with DDR SDRAM as well. imagine the V5 6K with that? holy crappers, can u say 20 gig's/second bandwidth?

 

[Shazam]

Well, judging from the many times 3dfx has said that the V5 5500 is not memory bandwidth limited (therefore it is fill-rate limited), it may not make any difference.

 

[Mr.IncrediblyBored]

It would make a big difference with 32-bit color and high rez (1024x768+), but the added speed wouldn't be worth the massive price jump.

 

[Marty]

DDR memory on the V5 would result in an excess of memory bandwidth; the payoff would be relatively small.

Marty

 

[DaveB3D]

sigh...

Allow me to explain a few things.


First, I assume you are basing your performance numbers off of scores found in reviews and such on the Internet. Don't do this. You can get substanitially better performance out of the V5 with newer drivers and also by a few performance adjustments in the advanced tab.

Second, we have the exact same bandwidth as the GTS. In the higher-resolutions we get VERY similar scores. Again, you probably don't see this in the review benchmarks, but we do.


If DDR were used, basically the performance advantages would come from 32-bit color. 32-bit color requires 2x the bandwidth as 16-bit, and so should in theory cause a 50% performance hit. You don't see this hit on all boards (coughATIcough) though, because they have sucky 16-bit implementations.


Hope that helps.

 

[Volenti]

In theory the card would be just as fast in 32bit colour as it is in 16bit, but it would also cost $450+ and would have been delayed even longer than it was...

 

[xtreme2k]

V5 5500 already has a 5GB/s+ bandwidth with SDR, the same bandwidth as the G2 GTS.

i think it is enough

the core itself is the bottleneck i think

 

[Wingznut]

As Dave said, at high resolutions the V5 is already neck-and-neck with the competition. So, there really would be a minimal benefit, if any.

 

[RagingGuardian]

Whether or not the Radeon has crappy drivers it still beats the V5 in just about all color depths and resolutions. If ATI were to somehow get some decent drivers then it would put the V5 to shame, DDR or not.

 

[NFS4]

<< Whether or not the Radeon has crappy drivers it still beats the V5 in just about all color depths and resolutions. If ATI were to somehow get some decent drivers then it would put the V5 to shame, DDR or not. >>


True, if only ATi would work on their drivers some more

 

[DaveB3D]

I would probably dispute the Radeon beat the V5 at everything because you aren't seeing V5 scores with all the optimizations in place. Reviewers need to start activating those in comparisions.

 

[NFS4]

Can you give us an idea of the optimizations that you mention. Like what are the names of these features that you speak of that give the 5500 the upper hand?

 

[xtreme2k]

maybe a few 4x FSAA demos on Q3 640 16bit fastest then V5 *MIGHT* have an upper hand

 

[DaveB3D]

I talked to Anand about them yesterday (in my 30 second mssg... don't ask ).

Basically things like depth precision being set to fastest, refresh optimization enabled, etc. Those types of things. You get a substantial boost out of that stuff.

 

[Mr.IncrediblyBored]

Yes, but do those &quot;optimizations&quot; decrease image/texture quality?

I've never really played with them so I can't judge, but I wouldn't want to increase speed at the cost of a crappy looking game.

 

[DaveB3D]

Nope

 

[Mr.IncrediblyBored]

Hmm... You've peaked my interest

I'd like to see some benches with these so-called optimizations.

 

[Soccerman]

ahh DaveB3D.. I need to ask u some stuff!

first off, what DOES that new Zbuffer option do? I have it set to FAST, becuase I don't know if Fastest is worse quality (I play Homeworld ALOT, and lack of FSAA gives me alot of non-existant polygons in the distance, ie, I can see through polygons that should be solid).

so tell me, what do these new drivers do?

 

[DaveB3D]

It optimizes performance.. about all I can say. The polygon issue is likekly a Z-buffer accuracy issue. I assume you are in 16-bit color based on what you said. You likely need a 24-bit Z buffer.

 

[Soccerman]

&quot;I assume you are in 16-bit color based on what you said.&quot;

well Duh, cause I'm on a Voodoo 3, whose Z-Buffer is only 16 bit (correct?).

so would changing that value decrease accurace of the Z-Buffer at all?

what kind of speed improvement can we expect to see?

 

[DaveB3D]

You should be fine turing on the feature. I honestly haven't looked into it with V3 though.

 

[gamingbuddy]

Hi Dave,

Could you give me the exact steps on turning these so called 'performance options' on? I would like to do a bench on base options and then your performance options with my v5.

I would also like to bench against a GTS to see if it is what you say (within a few frames of each other @ high resolutions).

Again, please give me specific step by step exact details on which options to turn off and turn off / adjust.

thanks.

 

[2dfx]

I recon some sites should do some Q3 benches with S3TC off. The compression of lightmaps and sky make it look even worse than 16bit and no person would use it if they wanted good image quality. 32mb cards like the 32mb GF2 GTS at 1024-768 @ 32bit slow down dramatically without it. It would be interesting to see how the V5 and GF2 stack up against each other in this situation.

 

[BFG10K]

Basically things like depth precision being set to fastest, refresh optimization enabled, etc. Those types of things. You get a substantial boost out of that stuff.

Let's not forget to tweak the other video cards in the same way. Otherwise I'm sure that you're all aware that the results from the benchmarks will be meaningless.