[Geeforcer]

Originally Posted by NewbAnhilator

Do you have any Idea what it takes to get to 30K+?? I'll give you a hint, its not grandma's Dell

No, I am entirely incapable of reading project description, for I am blind. Many thanks for your helpful hint.

[Doomtrooper]

Originally Posted by Pete

1. Isn't it possible Vegetto's tester was a dev (dual Opteron)?

2. Should I read into the fact that none of the "big" sites I frequent have linked to this thread (the only one with numbers, real or not)? Same with the screenshots. Same with Dave and Rev staying out of this thread completely (I think--the last 14 pp. have been kind of a blur ).

As stated before, 3Dmark 2001 and UT 2004 is single threaded...having a second processor mean nothing Pete.

[Pete]

DT, I was trying to mitigate duncan's doubt, not dispute your dispute. I don't know what to think of the 3DM2K1 score, but I think we've established that those UT2K4 numbers are in the realm of possibility, no? Or are there other UT2K4 botmatch numbers that show CPU limitation (which Xbit's don't)?

NewbAnhilator: Do you have any Idea what it takes to get to 30K+?? I'll give you a hint, its not grandma's Dell
Geeforcer: No, I am entirely incapable of reading project description, for I am blind. Many thanks for your helpful hint.

Oooooh, burned. He annihilated you, nub.

PS - It's "annihilator." Just call me the SpellingErrorAnnihilator. SpellChecker, for short.

[Geeforcer]

Originally Posted by Doomtrooper

As stated before, 3Dmark 2001 and UT 2004 is single threaded...having a second processor mean nothing Pete.

True, but what makes you think a single CPU would not allow you to reach at least the UT numbers?

[Rugor]

Thanks for sharing Vegetto.

Anything is possible, so I'm going to run down the logical possibilities of this.

I see four basic possibilities:

1) He's telling the truth as he knows it and the numbers are accurate.
2) He's telling the truth as he knows it and the numbers aren't accurate.
3) He's not telling the truth as he knows it but the numbers are accurate.
4) He's not telling the truth as he knows it and the numbers aren't accurate.

These would actually hold true for each set of numbers individually, but it's easiest to treat them as a whole. Of the four possible options, number three is by far the least likely and should probably be discounted; this leaves three remaining possibilities, in two of which he is telling the truth, and in two of which the numbers would not be accurate.

So without anything else it looks to me that it's more likely that Vegetto is telling the truth as he knows it, and didn't deserve the attacks.

I do think it's interesting that Dave and Rev have avoided this thread, and that nobody's linked to it either.

Ahhhhhhh the possibilities.

[Doomtrooper]

Doubling current High end cards overclocked to 600 + core with volt modding and vapo chilling on a CPU limited benchmark (Botmatch) is dreaming...I linked to the X-bit article..9800 XT 540/840...35 fps.

[JohnH]

Originally Posted by zeckensack

That should read "deferred renderer", not "tiler"

*runs back into the woods, giggling*

Nope, the test concerned uses multiple translucent layers, deferred rendering by itself wouldn't help.

[PaulS]

Originally Posted by Pete

2. Should I read into the fact that none of the "big" sites I frequent have linked to this thread (the only one with numbers, real or not)? Same with the screenshots. Same with Dave and Rev staying out of this thread completely (I think--the last 14 pp. have been kind of a blur ).

Well, firstly it's a weekend so a couple of the big sites drop off the radar. Secondly, how many sites link directly to anything on B3D forums, let alone (what amounts to) rumours? I've got no problem believing these figures are true, but anything that isn't a preview or review is always going to be labelled as "rumour", and so many sites won't touch it with a barge poll. Having said that, both NVNews (no surprise) and EliteBastards have linked.

Originally Posted by Rugor

I do think it's interesting that Dave and Rev have avoided this thread, and that nobody's linked to it either.

Dave isn't (or at least, wasn't - may be back now) even in the country.

[Geeforcer]

Originally Posted by Doomtrooper

Doubling current High end cards overclocked to 600 + core with volt modding and vapo chilling on a CPU limited benchmark (Botmatch) is dreaming...I linked to the X-bit article..9800 XT 540/840...35 fps.

Correct me if I am wrong, but...

Xbit is using a custom demo
Xbit is using UT2003
Xbit's results show that it is not CPU limited at 37 FPS.

[Doomtrooper]

FYI... the top dog for 3Dmark 2001 scores 31353...

AMD FX51@ 256x12= 3074MHz,,, ASUS SK8V,,, Corsair RE 3200x2,, 256MHz,,2-2-5Cas2,, 3.25V,,, ATI Radeon 9800XT 654/466,,, CPU/GPU cooling provided by "CaptainCascade"@Los Angeles Cascade,,,Corsair provided by *DDTUNG*

That processor is overclocked almost 1000 mhz over stock as a Fx51 comes in at 2.2 ghz. So to say the person in question that had his hands on this card just happend to have a liquid nitrogen setup..I'll eat my shoe.

[JohnH]

Originally Posted by Chalnoth

This wouldn't a single-texturing test with no z-buffer usage.

Edit:
By the way, you don't need to resort to those things you explained to get 7Gpix/sec out of 38GB/sec of memory bandwidth.

Here's the scenario:

1. Scene has z-buffer disabled. Each pixel requires one color write.
2. Scene uses a low-resolution texture that requires very little texture memory bandwidth.
3. V-sync is disabled, such that the contents of the framebuffer are output to the RAMDAC once every screen refresh, but the framerate is much higher.

Or, another possibility:

1. Scene has z-buffer disabled, but overdraw is large. Each pixel requires one color write.
2. Since scene has large overdraw, there are many pixels written for each one output to the monitor.

Anyway, if one takes either case to the extreme limit, it may be possible to have single-textured fillrate approach 4 bytes per pixel required bandwidth, which would be only 28GB/sec of bandwidth required for 7GPix/sec.

I beleive the figure was quoted from the 3DMark2K1 fillrate test which consistes of multiple translucent layers, so would not be explained by any of the above.

John.

[Doomtrooper]

Xbit is using a custom demo

YES

Xbit is using UT2003

There has been no engine changes according to interviews with epic

Xbit's results show that it is not CPU limited at 37 FPS.

@ 1600 x 1200...YES you are correct

[ninelven]

Just for clarification: are AA & AF CPU intensive?

EDIT: Because if not, an o/ced 5900 Ultra is getting 60fps @ 1600x1200 and this is on a P4 3000 ... which gets abused by the FX53 which is basically the equivalent of what the guy claims to have used.

[Bry]

Originally Posted by PaulS

Well, firstly it's a weekend so a couple of the big sites drop off the radar. Secondly, how many sites link directly to anything on B3D forums, let alone (what amounts to) rumours? I've got no problem believing these figures are true, but anything that isn't a preview or review is always going to be labelled as "rumour", and so many sites won't touch it with a barge poll. Having said that, both NVNews (no surprise) and EliteBastards have linked.

Dave isn't (or at least, wasn't - may be back now) even in the country.

This is actually mentioned in Futuremarks forums
http://discuss.futuremark.com/forum/...=5&o=0&fpart=1
Seem to be having the same discussion about validity..
I wonder why they used basically a dx-7 game to test the card(seems like it uses a ps1.4 shader or two but thats it) ..Does not answer any questions about ps2.0 and if they fixed the problems.
Also they did not really say what settting were used..vsync etc. We have no real system specs so have no idea to compair this with current cards.
As for dave and the others..I am sure their is still an NDA out..I would not want to speculate if I was under NDA.

[Geeforcer]

Originally Posted by Doomtrooper

Xbit is using a custom demo

YES

Xbit is using UT2003

There has been no engine changes according to interviews with epic

Xbit's results show that it is not CPU limited at 37 FPS.

@ 1600 x 1200...YES you are correct

Yey, I win...where is my $1.57?

The question regarding 03 VS 04 was getting at the fact that demos (if they were stock in the first place) would probably be different. Which brings up another question, what map are the alleged NV40 scores from?

[sonix666]

Some people are saying that Botmatches in UT2k3/UT2k4 are CPU limited, however the different official benchmarks in this thread seem to indicate that Botmatches are both speeding up evenly with a faster CPU or a faster GPU.

[Geeforcer]

Regarding 3dmark scores, there is a fairly easy way to settle this. We can find the same CPU (Operton) in the database. Even if NV40 is 100000x faster then any of the previous generation GPUs, it will run into memory bandwidth bottleneck sooner or later. Since the formula for score cautions is know, it is possible to see wheatear it is physically possible for NV40 to archive such scores with a given CPU before running into bandwidth wall, since we know the raw bandwidth.

[Evildeus]

Originally Posted by Pete

Same with Dave and Rev staying out of this thread completely (I think--the last 14 pp. have been kind of a blur ).

I don't know if these bench are begnin or not (thanks to the poster anyway), but it's the non-presence of DB and Rev in this thread that i find the most disturbing

[Doomtrooper]

Originally Posted by sonix666

Some people are saying that Botmatches in UT2k3/UT2k4 are CPU limited, however the different official benchmarks in this thread seem to indicate that Botmatches are both speeding up evenly with a faster CPU or a faster GPU.

It depends on the resolution, @ 1024 x 768 a faster CPU raises the FPS linear. As you increase the resolution the graphic card becomes the bottleneck. Now I've played with UT since its inception have spent lots of $$ to ensure the game plays well. In my experience a CPU/Motherboard has always impacted frames vs. a grapic card.

I could almost gurantee if we could scale a 4ghz FX 51 @ 1600 x 1200 the frames would also scale.

[Geeforcer]

The question is, how would CPU load increase with resolution increase (say 800x600 -> 1600x1200). Under ideal circumstances, the load increase should be minimal.

[ninelven]

I could almost gurantee if we could scale a 4ghz FX 51 @ 1600 x 1200 the frames would also scale.

This is actually fairly doubtful. Looking at the benchmarks of X-bit Labs (w/ an A64 3400+), it is only a rare occassion when the system encounters a CPU upper bound on performance before limititations of the graphics cards being used, and these are only at 1024x768 with no AA/AF for all but the most elite cards. Even for the 9800xt and 5950Ultra, they become far more significant limiting factor at resolutions slightly above 1280x1024, and they only make it to that resolution 2 of 5I believe (where the CPU upper bound is significant to measure this - in 1 example it is not). Thus whichever upper bound the system encounters first (video card or processor), upgrading the opposite piece of equipment will probably have minimal effect, and certainly less of an effect than upgrading the problem piece.

From what I can tell, the increase in resolution has prettymuch zero impact on the CPU load.

[duncan36]

Originally Posted by Evildeus

I don't know if these bench are begnin or not (thanks to the poster anyway), but it's the non-presence of DB and Rev in this thread that i find the most disturbing

Dont worry I would imagine they're having a good laugh over it.

[volt]

Originally Posted by Doomtrooper

There has been no engine changes according to interviews with epic

Correct, larger textures though?

[volt]

Originally Posted by duncan36

So no developers have NV40's yet, but friend of Vegetto is busy benchmarking away on it? Count me a little skeptical.

That statement is incorrect AFAIK.

[Chalnoth]

Originally Posted by duncan36

Dont worry I would imagine they're having a good laugh over it.

More likely they're under NDA and are actually adhering to it.