Performance of Parhelia . . .

[frankymail]

Hey, I've just read that 3dMark 2001 /SE has support for only 2 texture mapping units per pipeline. That means that 8 out of Parhelia's 16 textyre mapping units aren'T used at all . . . and it still gets a G4Ti-ish score (and looks sooo much better... ); ATI's R300 with its 8-pipelines/2-tex-per-pipes gets full usage of all it's texe-map units, has a 256-bits mem bus but still gets -only- 14000? Man, I was a little scared about the 3D performance of my preordered Parhelia (for hum-occasional-hum gaming ) but I'm now very confident about my purchase... I'd like to see some Doom 3 scores... that should put a sock in some pie-holes (well, hat-with-ranch-dressing-holes)

And man, I already cannot live without DualHead... can't wait for some TripleHead luvin' (yuk, that sounded quite geeky didn't it? Well that suits me just fine ) The 3rd monitor is already on my desk, waiting to exhale (C)

BTW Ant, u still got those "Beginner's guide to hat cooking", "1001 easy hat recipes" and "Hats may still be hard to swallow, but they're now easy to cook" guides? I guess we'll need them on tuesday: ther'll be a o-lot o' cookin' to do...

[borat]

where do you get those benchmarks from considering that parhelia benchmarks ar under nda untill tomorrow and the r300 is only in beta silicon and drivers and no such figures would be released to the public in that state. aside from that nvidia will always have the edge in 3dmark anyway as their programmers spend time trying to optimise drivers for the benchmark to make their card seem faster than they actually are to potential buyers. anyone got that link to the way they render the intro screens on 3d mark which can boost your 3dmark score by 500 or more mark.

[knirfie]

i read it somewhere too, that the beta silicon from the r300 is getting a score of 14000

[Dr Mordrid]

When you read those 10000+ 3DMark scores take 'em with a whold shaker of salt.

1. most times its a very overclocked board

2. if the cards drivers are optimised for 3DMark (and both Nvidia and ATI do this) the results mean nothing in *real* software.

3. Nvidia for one has been cheating by having drivers that would record frame rates based on the still frames between modules. Granted they "fixed the bug" in their drivers, but only after they were called on it in the computer media

Dr. Mordrid

[Michel]

hmmz.. talking about +10000 3dmark scores:



found it here

[Dogbert]

WTF ? I guess he did some photo editing there...

[Drizzt]

If that is true, I'm gonna buy one!

[GT98]

yeah its been edited...you can see ever so slight differences around the benchmark numbers and where it says Radeon9500 MAX. Boy I love my PCI G450 at work LOL

[frankymail]

Well, even though it hurts my pride a lot to say this, this autumn, Parhelia is nto THE card to get if you want the best in 3D performance… I'm eagerly waiting for my Parhelia, but I don't feel that excited about it anymore… I know it will be fast enough are look very pretty, but usually when you change video card, you somewhat look forward having better performance, don't you? Well, I already know I won't I just hope that there will be a lot of driver improvement, because the currently, even the GeForce 2 Ti beats parhelia in some tests... (nVIDIA's G2GTS/Ti, G3 and especially G4, like ATI's Radeon 8500, all have very advanced bandwidth-saving features that are very effective... which kinda nullify Parhelia bandwidth advantage... However, I look forward overclocking the hell out of my board (well, the GPU mostly, and the mem a little). If I can get the chip to run stable at 275-300 MHZ (which if probably not farfetched using the setup i'll use), i can probably get much better performance Now if they could hurry up with the shipping a little...

[DentyCracker]

There is something strange happening here. I have a slight difficulty reconciling the statements that the Parhelia produces unplayable framerates at certain resolutions in certain games and those that the BBz have made. I'm not even sure where I should post this but I didn't want to start a new thread. Maybe it means that there is a fundamental flaw in how we report benchmarks.

[borat]

yes dent cracker i was thinking the same thing here can any bbzs enlighten us on this one?i will buy a parhelia but only if frames are smooth.

[seether]

I noticed one person's comments about non-playable framerates was related to a particular bench in which the Parhelia was scoring ~90fps. What I've inferred from that and what the BB's have said is that 50, 60, 90, 150 fps are all smooth to look at, as they should be, but if you are used to a card that consistantly *reports* 100+fps, but still gives noticable stuttering at times, than it's easy to see where someone less educated (I mean the reviewer Denty, not you ) would take that as a sign that the card doesn't provide playable performance, since the game benches are looked at instead of the game play itself.

[TdB]

this my own speculation about this, from:

http://forums.murc.ws/showthread.php?s=&threadid=34007&pagenumber=5

regarding the "smoothness" on the parhelia; it actually makes sense:
when I am running games, the situations that are making the framerate drop below acceptable, is always tied to the use of alpha-textures or "cut-thrugh(sp?)"-textures(explosions, smoke, clouds, and trees, etc. extreme-stress scenarios for fillrate), these framerate-drops are almost unbearable with fsaa, because these texures are render with 4 times the precision and 4(almost) times the framerate-drop, requirering 4 times the fillrate.
However the parhelia doesn´t apply 16xfaa on those textures, they are ignored because they are not polygon-edges, and would therefore not have the same lowpeaks compared to traditional fsaa, and the betaboyz seems to have 16xfaa enabled all the time, so i can see why they are saying it runs "smoother" than other cards with fsaa enabled, even though the average framerate is the same.
Simply put: 16xfaa should not give the low-peaks in framerates as normal fsaa.
this is only an educated guess, and I would love to know if I am right, betaboyz please comment.