3dCenter.de - Parhelia vs GF4 @ 220/275

[isochar]

http://www.3dcenter.org/artikel/parhelia_theoretische_benchmarks/

[Phalanx28]

That's a neat comparison.

Phalanx28

[thop]

yep good idea, however the P still looses most comparisons. what's the default speed of the GF4 again? how much was it underclocked?

[Jazzz]

I was wondering if anyone who is registered at aceshardware, if they could ask them to do a similar thing. I would be interesting to see another site do it as well coz after seeing how much variations there are in benchmarks results for the parhelia im not sure how to trust

[thop]

hmm that's a pretty nifty underclock then 80/50 Mhz. they should have done some AA benchmarks though with the underclocked GF4.
and what i'd really like to see would be a 150/225 comparison of both cards

[TdB]

but it certainly has a LOT of 3d-features!

[az]

Originally posted by Jazzz
I was wondering if anyone who is registered at aceshardware, if they could ask them to do a similar thing. I would be interesting to see another site do it as well coz after seeing how much variations there are in benchmarks results for the parhelia im not sure how to trust

You don't need to be registered in order to post there

Just enter the username you want to use, enter the text you want to post, and, if you want, your e-mail address, and click "submit"

BTW, scroll WAY down the page (or hit ctrl+f and search for "suggestions"), there already is a P review suggestions thread

AZ

[TdB]

it seems like the geforce4ti has very optimized drivers and efficient hardware, compared to the parhelia.

[Phalanx28]

az is correct, you don't need to be registered. It's a different kind of forum layout. I started a thread in the General Forum on the article - I don't think there have been any bites, though.

Perhaps this was already posted in the suggestion thread.

EDIT: Whoops - it was already posted over there.

Phalanx28

[trk2]

I think it's obvious that parhelia's opengl drivers still need alot of work. Look at the cpu scores under the opengl tests, aren't those designed to be mostly card independent? The parhelia is still 30% behind. I guess this is good news, shows that there is alot of room for improvement.

[Stringy]

Remember all those comments about the lack of "Culling" in the P??
and then with the less than stellar performance the P has ATM I've seen many alude to this lack of "culling" as a possible contributor...
Well, this little experiment lays the smack down...
Village Mark - a Heavy Overdrawn bench, used by PowerVR to show the power of the Tile-Rendering...
well the P does 80Fps and the castrated GF4Ti does 87Fps, there is your "culling" advantage right there, a mere 9%....

not bad, but I'm sure architecture and drivers play a more important part than their culling tech.


Craig

[P5ycho]

Also check GL extreme overdraw benchmarks, Parhelia might be slower but the slowest of the onderclocked geforce is nearly thesame, so in a worstcase scenario the geforce would drop to parhelia framerate anyway.

[CHHAS]

Originally posted by Stringy
Remember all those comments about the lack of "Culling" in the P??
and then with the less than stellar performance the P has ATM I've seen many alude to this lack of "culling" as a possible contributor...
Well, this little experiment lays the smack down...
Village Mark - a Heavy Overdrawn bench, used by PowerVR to show the power of the Tile-Rendering...
well the P does 80Fps and the castrated GF4Ti does 87Fps, there is your "culling" advantage right there, a mere 9%....

not bad, but I'm sure architecture and drivers play a more important part than their culling tech.


Craig

You're forgitting that P. has twice the bandwidth of a GF4 clocked at the same speed (256bit vs 128bit), want to guess at how a 128bit P. would do ?

[nggalai]

Just some clarifications regarding that 3DC article.

Lots of people in the 3DC forums were asking for "theoretical" benchmarks--3DCenter traditionally only benches with available games or demos, hence no such benches were included in the original "Parhelia first look." Hence, Leo used the remaining time he had with the review board to cook up some of those theoretical benches.

There might be a follow-up article with the 226 drivers used (if Leo still gets the time to write it). At any rate, there WILL be a follow-up as soon as a driver set provides a marked performance boost and/or enables higher-degree anisotropic filtering . . .

ta,
-Sascha.rb

[Stringy]

You're forgitting that P. has twice the bandwidth of a GF4 clocked at the same speed (256bit vs 128bit), want to guess at how a 128bit P. would do ?

True...
if you were to look here you'd think the P had a 64Bit Memory bus....


Craig