[Chalnoth]

Originally Posted by jvd

No i've seen you defend nvidia on worse than this . I figured you'd defend the unreal numbers or say something about then yet you had not .

Those numbers just don't mean anything to me. Besides, I don't care all that much about performance anymore. For me it's all about the image quality and reliability.

This is why I continue to complain about ATI's texture aliasing, and why I am rather excited about the pictures. That FSAA looks very nice....I just hope it's what we'll get.

Oh, and I do expect the NV40 to outperform the R420 (something that becomes somewhat less likely if ATI went for a sub-PS/VS 3.0 card), given that they have a much longer history of higher-performing parts.

[Geeforcer]

So, to summarize:

UT2004 numbers are ambiguous.
3dmark score is extremely unlikely, at least on the described PC.
3dmark filtrate is impossible
Stated AA mode does not match the on-screen result.

Nothing against Vegetto, who himself is apparently only a secondary source, but the whole things smells.

[Vegetto-eX]

Originally Posted by Chalnoth

Those numbers just don't mean anything to me. Besides, I don't care all that much about performance anymore. For me it's all about the image quality and reliability.

This is why I continue to complain about ATI's texture aliasing, and why I am rather excited about the pictures. That FSAA looks very nice....I just hope it's what we'll get.

Oh, and I do expect the NV40 to outperform the R420 (something that becomes somewhat less likely if ATI went for a sub-PS/VS 3.0 card), given that they have a much longer history of higher-performing parts.

I have the same feeling

[Doomtrooper]

I remember Chalnoth ranting about Nvidias superior engineering with the Nv30..blah blah...I also told him he was wrong.

He is wrong again....I don't cheer for companies that intenionally lie and cheat to their customers. People that are not offended by such actions since they are the victims being played need to get their head examined at their earliest conveniance.

[Dave Baumann]

One of the biggest levelling factors is likely to be bandwidth this time around. Unless something funky goes on with the bus width, which is unlikely seeing as how long it took for the high end to transition from 128-bit to 256-bit, then that bandwidth is going to be related available memory speeds at production quantities, which in turn will probably vary dependant on your time to market.

[Geeforcer]

I expect 420 to outperform NV40 in several cases, maybe in the majority of situations. What am I basing this at? Nothing, but that doesn’t seem to stop anyone these days. Considering that of the two major companies Nvidia has a dubious honor of being the only one that in the last several years has released a high-end part that failed to achieve parity with competitors’ counterpart already on the market, the burden of proof is on them.

[DemoCoder]

Originally Posted by Doomtrooper

I remember Chalnoth ranting about Nvidias superior engineering with the Nv30..blah blah...I also told him he was wrong.

He is wrong again....I don't cheer for companies that intenionally lie and cheat to their customers. People that are not offended by such actions since they are the victims being played need to get their head examined at their earliest conveniance.

Cheering for any corporation is dumb. The motivation of corporations is to make a profit. You're already a victim, you just don't know it. You spend hours and hours of your time promoting a corporations products, effectively being a sales agent for them, on all of these message boards, and you're not being paid for it, if at all.

Stooge and shill.

[Joe DeFuria]

Originally Posted by DaveBaumann

One of the biggest levelling factors is likely to be bandwidth this time around. Unless something funky goes on with the bus width, which is unlikely seeing as how long it took for the high end to transition from 128-bit to 256-bit, then that bandwidth is going to be related available memory speeds at production quantities, which in turn will probably vary dependant on your time to market.

That's my gut feeling as well.

16 pipes or 12 pipes....both of these should be pretty starved for bandwidth at high-res and AA...given the clock-speeds that are being rumored about. (approx 500 Mhz for the cores, and 500-600 Mhz, 256 bit DDR for the memory.)

That being said, the picture might change with pixel shading operations.

[DemoCoder]

Originally Posted by DaveBaumann

One of the biggest levelling factors is likely to be bandwidth this time around. Unless something funky goes on with the bus width, which is unlikely seeing as how long it took for the high end to transition from 128-bit to 256-bit, then that bandwidth is going to be related available memory speeds at production quantities, which in turn will probably vary dependant on your time to market.

It will level maximum fillrate on diffuse shaders, so no card is going to do more than 4-5 gigapixels/s as max, but when executing PS2.0/3.0 shaders, there is alot more variation. We saw this on the NV35 vs R350. In old games, NV35 and R350 both hit a fillrate wall and maxed out near similar levels. On PS2.0, NV35 took huge hits.

The question is, what's going to level computational throughput? Clock speed and transistor budgets, and of course, smart algorithms.

[John Reynolds]

Originally Posted by DemoCoder

Cheering for any corporation is dumb. The motivation of corporations is to make a profit. You're already a victim, you just don't know it. You spend hours and hours of your time promoting a corporations products, effectively being a sales agent for them, on all of these message boards, and you're not being paid for it, if at all.

Stooge and shill.

Phantom marketing online has been on the upswing in recent years.

[ChrisRay]

Originally Posted by DaveBaumann

One of the biggest levelling factors is likely to be bandwidth this time around. Unless something funky goes on with the bus width, which is unlikely seeing as how long it took for the high end to transition from 128-bit to 256-bit, then that bandwidth is going to be related available memory speeds at production quantities, which in turn will probably vary dependant on your time to market.

I was sitting there thinking the same thing. 2 years ago I remember aplications and cards (Geforce 3 4, 2, Radeon 8500 ect) All being starved for bandwith.

And now the situation has been reversed, These cards plenty of bandwith and the bandwith is only used for Anti Aliasing anymore (For the most part) And we've been in a situation where the fill rate has been trying to catch up with bandwith.

Of Course this only applies to the 5900/9700/9800 line of cards, And most mid range users arent having this problem 8)

[Dave Baumann]

Originally Posted by DemoCoder

The question is, what's going to level computational throughput? Clock speed and transistor budgets, and of course, smart algorithms.

You can still be quite bandwidth bound if your computational power exceeds that of current titles, so, again, you are playing the bandwidth game.

However, right after release / preview, for the cases where you may be computationally limited, I'd say the driver shader compilers will be the one of the biggest influences as far as computational processing performance goes.

[jvd]

Hey i cheer for whatever company gives me the fastest card with the best image quality. But that changes each time i buy a card .


I don't like being lied too .

[jpaana]

Originally Posted by Doomtrooper

3Dmark 2001 is single threaded

I did some digging (which everyone can repeat if necessary) and xp's task manager shows 8 threads for 3DMark 2001SE in the menu window and all other tests except the game test 1 which shows 9. Some of the threads are probably created by D3D etc. but anyway.

Playing single player UT2K4 shows 10 threads btw.

Obviously only a single thread might be doing all the work, but for 3DMark I could imagine the physics simulation threaded (calculating next frame physics while rendering the previous one) and UT has lots of uses for threads, sounds, bot AI and server functions being obvious ones.

(and yes, there are single threaded programs like cmd.exe too )

[PaulS]

Interesting if they ship with 600/1200 DDR2 on the NV40. You'd think they'd go with 800/1600 (perhaps downclocked a little), which has been the expectation for some months now. It would certainly shift their bandwidth to ~45/50gb/s, which you'd imagine would come in handy with the high res + AA/AF situations these cards will be competing in.

Maybe the samples were too expensive? Can't see quantity being a problem.

[Mike89]

If ATI can manage to beat a card like that, I will have Dave Orton's babies.

Talk about putting all your cards on the table before the game has even begun! Jeez.

You might want to get some birth control pills cause a baby shower may be just around the corner.

[Chalnoth]

Bah, you could at least give the name of the person you're quoting.

[MuFu]

If ATI can manage to beat a card like that, I will have Dave Orton's babies.

Jen-Hsun maybe. Imagine that.

[PaulS]

Maybe the babies are hallucinogens.

[BetrayerX]

Originally Posted by DaveBaumann

However, right after release / preview, for the cases where you may be computationally limited, I'd say the driver shader compilers will be the one of the biggest influences as far as computational processing performance goes.

Do you expect more of that nice NV magic driver pixie dust that we all love?

[StealthHawk]

Originally Posted by Chalnoth

Oh, and I do expect the NV40 to outperform the R420 (something that becomes somewhat less likely if ATI went for a sub-PS/VS 3.0 card), given that they have a much longer history of higher-performing parts.

But that was the same reasoning used when people though NV30 would be so much better than R300, that NV30 would use 3dfx quality FSAA, etc, etc. Didn't pan out. I don't think we can make assumptions based on the past anymore. ATI's drivers used to be horrible too, and they always played second fiddle to NVIDIA.

Things change.

[Chalnoth]

And, if you remember, there were lots of rumors at the time that nVidia was having some major process problems. Some even said that the NV30 was meant to be released (originally) around the Spring of that year.

Anyway, the NV3x doesn't change anything. One sample is not of statistical significance. And this time, we haven't been hearing any "process trouble" rumors. Additionally, nVidia doesn't yet appear to be pre-hyping the launch, which would seem to indicate that they thing it's a better product than the NV3x was.

[SsP45]

Originally Posted by Mike89

Talk about putting all your cards on the table before the game has even begun! Jeez.

You might want to get some birth control pills cause a baby shower may be just around the corner.

I was trying to be funny. I guess I needed more of these to get my point across:
________
Ecigarette

[digitalwanderer]

Originally Posted by Chalnoth

And, if you remember, there were lots of rumors at the time that nVidia was having some major process problems. Some even said that the NV30 was meant to be released (originally) around the Spring of that year.

Yeah, just like the nV40 is slated to be "released" around the "Spring" of this year.... (LOTS of quoteys there, just to be annoying. :P )

[Tagrineth]

Originally Posted by Chalnoth

Additionally, nVidia doesn't yet appear to be pre-hyping the launch, which would seem to indicate that they thing it's a better product than the NV3x was.

More like, they don't want to say anything lest it all backfire like it did last time.