[Chalnoth]

Hrm, after your post, Baron, I went back and looked at that Inquirer story. They seem to state that it's 8 pixels per pipeline when doing pixel and vertex work, 16 pixels per pipeline when doing just pixel work.

That doesn't make a whole lot of sense to me, as what company would design a part that is so overbalanced in terms of vertex processing power?

[DoS]

This is getting ludicrous...Inq said this Inq said that, guys here "in the know" say otherwise, i am not falling into that any more :?
Let's wait 3 weeks and see

[Unknown Soldier]

Originally Posted by Joe DeFuria

I find this particularly insulting:

That's until ATI release the R420 and beat the living crap out of the NV40 much like the R300 did vs the NV30 and Nvidia turn around and say the 3DMark2001 benchmark is crap.

US

[Joe DeFuria]

Originally Posted by DoS

Let's wait 3 weeks and see

Why do I get the feeling that in 3 weeks, we still won't really know?

[Evildeus]

Well, at least we will be discussing on official (PR? ) informations, and not on "one time it's that, oh sorry it's now that, and no, finally it's the contrary, oups...."

I'm sure, with all their stories, TheInq must have one time be right at least

[Skinner]

I get a deja-vu feeling about this

The NV hype machine against the quiet but strong Canadian "enemy"

Why do I have still the idea that ATI's R420 will be the best card again

[Geo]

From Inq's "Hands on" Article http://www.theinquirer.net/?article=14827 :

"Eight 32MB video memory modules are installed – four on each side of the PCB and as we first reported from Computex in Taiwan – these are of the new GDDR3 type."

Umm, that pic shows 8 on one side, so Inq spraying all over the place and being right once in awhile just on the odds? Again?

[akira888]

One thing can be confirmed at this point - 3Dfx has been thoroughly assimiliated into NVidia.

"See this 3 year old benchmark - our card kicks ass in it!"

[Fodder]

Originally Posted by geo

Umm, that pic shows 8 on one side, so Inq spraying all over the place and being right once in awhile just on the odds? Again?

Inq's card is A2, the pictured one is A0, that may make a difference to board layout.

[digitalwanderer]

Originally Posted by Skinner

Why do I have still the idea that ATI's R420 will be the best card again

Mebbe because ATi hasn't felt the need to be spreading misinformation and FUD a couple of weeks before their launch?

[martrox]

Originally Posted by Evildeus

Well, at least we will be discussing on official (PR? ) information

Yes....official........hmmm......didn't "official" info on the NV30 infer that it was 8 pipes? Guess someone was using those 4 "implied but not there" pipes to smoke something ....um...you know!

[Bouncing Zabaglione Bros.]

Ugh! Sacrificing IQ for frames designed into the hardware? What's the point in having 32bit colour, AA & AF if you are going to hack every second texture?

I suppose it's so you can claim 16x1 when in actual fact no one will use it that way. Looks like Nvidia hasn't changed at all in the last 18 months.

[Tim Murray]

Uh, BZB, that assumes it's true, which it's almost certainly not?

[tombman]

Ok guys, maybe this can help you a bit:

Numbers are from my machine:

FX51 @ 2.5ghz, 9800XT @ default

4xAA/16AF, CTF BOTMATCH: 55fps (NV40 >70fps)

FX51 @ 2.75ghz, 9800XT @ 445/380

4xAA/16AF, CTF BOTMATCH: 58fps
(but Minimum was much higher than with non overclocked machine )

So you see: the posted NV40 results WERE CORRECT !!!!

COnsidering new drivers and beta A2 hardware the NV40 just behaves
like any next generation card @ start.

have a nice day

[Tim Murray]

Careful, tombman! You tried to be helpful! They'll lynch you for sure now!

[Joe DeFuria]

Originally Posted by tombman

Ok guys, maybe this can help you a bit:

Well, you changed BOTH the CPU speed and the GPU speed at the same time. Doesn't help us at all.

BTW...which specific botmatch benchmark did you run, and is this the retail UT2004, or the Demo? If it's the retail...how did you run the benchmark?

Oh...and at what resolution?

[digitalwanderer]

Originally Posted by The Baron

Uh, BZB, that assumes it's true, which it's almost certainly not?

Dod gamnit Baron, you are REALLY getting me confuzled along with all of the rest of this speculation!

Time to cheat, I gotta get into the "smoking something hallucinagenic" mindset and see if it makes anymore sense that way...

[tombman]

But the difference is only 3 fps... so it doesn´t matter.

It is the demo i benched, 16 bots.

At 1600x1200 4aa/16af ut2004 is CLEARLY GPU limited, NOT cpu limited.

So if nv40 can push > 70fps it sounds perfectly reasonable to me.

Hint for all to find out where the limitation is:

change resolution from low to high and fps stays the same --> cpu limited
change resolution from low to high and fps drops ---> gpu and/or gpu ram limited
change number of bots from low to high and fps drops ---> cpu limited. (maybe main ram too)

[Tim Murray]

Dig, find me on AIM when I get home (and make sure you're not high) and I'll explain it all. It's rather simple.

[digitalwanderer]

Originally Posted by The Baron

Dig, find me on AIM when I get home (and make sure you're not high) and I'll explain it all. It's rather simple.

How long until you're home?

[Tim Murray]

Originally Posted by digitalwanderer

How long until you're home?

Four hours. Clean the carpets or something.

[Joe DeFuria]

Originally Posted by tombman

But the difference is only 3 fps... so it doesn´t matter.

Of course it matters. We don't know what attributed to that 3 FPs gain...GPU, CPU, or both.

So if nv40 can push > 70fps it sounds perfectly reasonable to me.

Presonally, I don't know if it's reasonable or not...because we have no idea what map and demo was benched, so we have no baseline to compare it to.

Hint for all to find out where the limitation is:

change resolution from low to high and fps stays the same --> cpu limited
change resolution from low to high and fps drops ---> gpu and/or gpu ram limited
change number of bots from low to high and fps drops ---> cpu limited. (maybe main ram too)

Why doesn't your own testing methodology follow your own hints?

[tombman]

Ok, here is proof that ut2004 is gpu limited at 1600 4aa/16af

1600x1200 4xAA/16xAF : 58 fps avg
640x480 4xAA/16xAF : 143 fps avg

---> GPU LIMITED! (and gpu ram bandwith of course)

(cpu @ 2750 FX51 on both runs)

I can give the .bat and .ini files i benched with so you can do some benches yourselves.

[Bouncing Zabaglione Bros.]

Originally Posted by The Baron

Uh, BZB, that assumes it's true, which it's almost certainly not?

Well I'm obviously making that assumption for the purposes of this discussion. Couldn't you tell from what I wrote that I would consider it somewhat bizarre, yet not altogether out of sorts from what we have seen from Nvidia in the last couple of years? Maybe we should get DB to just make sure that every reply in these threads has "If this is true" prepended to it?

Anyway, do you *know* for sure what NV40 is going to be? I'm not asking you to spill the beans, just to say whether you do *actually* know, or if you are just surmising from rumour and innuendo like the rest of us.

[Tim Murray]

Do I know it's true? Of course not. I don't have a board in my hand and won't for at least some time. However, there are very specific things that I would say that I know that imply that Fudo's stuff is totally wrong.