Comparative image quality

[MWM]

Could anyone offer a comparision of 2D image quality between the Parhelia and Radeon 8500 cards at high resolutions (IE 1600 x 1200/2048 x 1536) please.

My decrepit video card's time is at hand and the candidates are Radeon 9700 and Parhelia. I would expect an ATI manufactured 9700's signal to at least equal if not surpass the 8500.

Any feedback would be welcome.

Thanks.

-------------------------------------------

Thanks for responses. (Esp Xortam)

[103er-Fan]

Well, all I can say is that my V5 5500 was superior in terms of 2d quality to the Radeon 8500, this was at 1280x960. The Parhelia is way superior to the Voodoo, at all resolutions I tried yet, so I can say that the Radeon doesn't come close.

[xortam]

You can read the Matrox White Paper, "UltraSharp Display Output Technology", which compares the P output with an ATI 8500 and a PNY Verto GeForce4 Ti 4400.

[Ryu Connor]

The 8500 does not produce as good an image as the Parhelia.

http://www.tech-report.com/onearticle.x/3747

It's rise and fall times are not very balanced and that can cause spatial resolution problems.

It's average swing takes it to 750mV which will cause an extra heaping of brightness and distort the colors some. Of course it becomes subjective whether or not you like the slight washed colors more or less.

[Indiana]

I'd doubt the Radeon8500 having worse 2D than a Voodoo5500 - at least my Radeon8500 has better than the Voodoo5's I've seen (at 1600x1200, that is).
But unfortunately I have to say that the 2D of the Radeon8500 is quite noticeable worse than that of my G400 and even a bit worse than my old RadeonVIVO .

[Muad'Dib]

I've never seen an ATI card working before - simply because I've never looked for one out of it's box - but the 2d of my Parhelia is vastly superior to that of my MAX imho. I'm running at 1600x1200x32 and the colors are sharper, the fonts smoother, the pictures crisper.... Beautiful card!

-Dimitri

[Chrono_Wanderer]

hello does anyone know whether the Radeon 9700 will theorically have better texturing quality than P or not? which card has higher iq and which cards will have a sharper DAC?

i m kinda confused by the reviews flowing on the web regarding the technologies found on the 9700.

[GT98]

Originally posted by Chrono_Wanderer
hello does anyone know whether the Radeon 9700 will theorically have better texturing quality than P or not? which card has higher iq and which cards will have a sharper DAC?

i m kinda confused by the reviews flowing on the web regarding the technologies found on the 9700.

Well most of the Radeon 9700 (which looks like a nice FPS Gamers card, but I'm getting sick of the 6 month replacement Video Card sechule I've been having the past 2 years, so the P will be around for at least a year or so) Reviews are Previews of the card, which means that clock speeds and the like could change between now and when the card hits mainstream manufacturing and there hasn't been any specific remarks about how they will go about fixing image Quatily like 5th order filters and the like, like Matrox was pushing as part of its roll out for the Parhelia.

[Chrono_Wanderer]

Actually when ppl were leaking specs and pics of the alpha P board (hehe... and remember to hang the NDA breakers ) i had my hopes up about Matorx being the King for this year and i was really expecting P will be a fully dx 9/GL 2.0 part w/8 pixel pipes and agp 8x (well guess 8x doesn't do a lot after all) just like the 9700. But when the specs were launched i was shocked the 512-bit thing was so "quality oreinted" and couldn't beat a GF4 ti. However, after seeing the specs for 9700 it seems like they have got every quality features P has plus much greater speed. Forget about the analog output quality, signal degeneration etc of the cards for now, if one compare the technical specs/whitepapers of the two processors, does anybody think the 9700 silicon itself can draw 3d more beautiful/complex than the Parhelia GPU?

[TiG]

Good point GT98. I've been wondering that myself : how the hell can all these sites claim they have a "9700 Review!" when they are little more than a tech preview? Basically spewing out White Paper specs and the always credible "Trust me, it's real fast. C'mon you can trust me, I'm the one who told you last month that the Parhelia was going to smack the gf4ti4600 and radeon8500 around. And it could do your tax return." Note : they don't always add that last bit.
Even the much loved benchmarks are a "percentage comparison" to a gf4ti4600. In a same spec pc, setup and supplied by ati. Presumably with an ati employee handcuffed to it. Level Playing Field(TM)

It is the interesting nature of HYPE that reviewers/users who made bold/crazy/ nay ridiculous claims about a similar product release less than a month ago, are prepared to do it all again. With even less hard data. And a big fat floppy power connector obstructing the airflow

[Indiana]

Well the texture quality will surely be better with the Radeon9700 - at least for now with the broken/insufficient anisotropic filtering of the Parhelia.
Anisotropic was always the point where the newer ATI cards shined: even a Radeon8500LE could outperform a GF4 Ti4600 there (anyone want to take a guess why THG never benchmarked this? ).
The 9700 fixes some of the remainig 8500 problems with this and thus will be even better.

2D-quality, however might still be better with the Parhelia - until there are real cards out there, noone can tell for sure, but I'd be really surprised if the R9700 delivered superior 2D Quality when apparently noone out there seems to care

[Ryu Connor]

Anisotropic was always the point where the newer ATI cards shined: even a Radeon8500LE could outperform a GF4 Ti4600 there (anyone want to take a guess why THG never benchmarked this? ).

It's not an apples to apples comparison. The RADEON series can only do bilinear filtering plus anisotropic filtering. The end result is that your samples are halved.

Bilinear + 16-tap aniso = 64 samples.

The GF3 and 4 being able to trilinear + 8-tap aniso actually allowed it to match the number of samples of the RADEON, despite the availability of a higher tap on the RADEON.

Trilinear + 8-Tap aniso = 64 samples.

The RADEON also used a random sampling pattern for the aniso meaning that it was not applied evenly. Certain angles and areas were missed, even if they really needed to be filtered. I affectionately called this algorithm Cheat-Mapping. As between being forced to use bilinear and not having all the samples applied across the screen the RADEON got a huge performance advantage. The cost of course was that certain applications looked worse. In some cases it was hard to tell the difference.

Things changed recently for NVIDIA cards. The 29.42 drivers had optimizations for aniso filtering, which has lowered the performance impact rather significantly.

The 9700 fixes some of the remainig 8500 problems with this and thus will be even better.

The 9700 can actually use Trilinear + Ansio and it retained its 16-tap option meaning that the 9700 can generate 128 samples. ATI also finally kicked Cheat-Mapping to the curb and the samples are applied across the screen correctly. It is the best filtering option yet presented by the 3D Titans.

[Chrono_Wanderer]

Even if matrox has successfully corrected the anisotropic filter for the P drivers, R9700 can still filter than P right? According to the P whitepapers, P can do "8 sample anisotropic and trilinear filtering on 4 dual-textured pixels/clock" or "16 sample anisotropic filtering on 4 SINGLE-TEXTURED pixel/clock" soooo does it mean P can do both:

Bilinear + 16-tap aniso = 64 samples.
AND
Trilinear + 8-Tap aniso = 64 samples.

Just wondering

[Frost]

For a simple 'sharper DAC' test, fill out a huge bitmap at the size of your monitor setup (single, dual, triple, etc.) and fill it with a 1-pixel checkerboard pattern of choice (white vs black being my favorite). You will clearly see the superiority of the RamDAC(s) (close to having no moire patterns) compared to other cards. Of course, the results only shows through the visual presentation on the CRT(s), and grabbing screenshots of the differences won't work. Just make sure to save the bitmap to GIF or other lossless format.

[JaG]

So much was spoken about anisotropic filtering and more 3D details here! That's interesting, but returning to the beginning question, I would agree with Indiana, that 2D quality might stay better with Parhelia, although it's difficult to talk about the Radeon 9700 before you have been able to test it.

It's ATi tradition to do all necessary to get 'sufficient' visible quality. But it's not ATi tradition to be the best in this discipline.

Parhelias 2D quality is incredible. As well in those high resolutions! It's much better than I expected when reading the first reviews while the pre-order was still proceding. In my eyes it's a nice step ahead in Matrox 2D quality, again. And that although I generally would expect it to be more complicated to increase visible 2D quality the more you develop chips with higher and higer thermal power (which is an other reason to expect weaker 2D from R300).

Well, but look to the markets these chips are designed for. The Radeon 9700 will be ATi's ultimate gaming card, I expect.
Not less - not more!

For office and professional workstations, ATi will produce other products again, probably based on a shrinked (not as hot) 0,13µ R300 - but later.

Well, and in the gaming market segment is that counts maximum, producing the chip with the highest 3D benchmark results.

ATi's primary goal here surely was to beat nvidia, 'cause these guys costed them too much clients, too much sells in the last months. And they seem to win this contest for a while, now.

The smaller, privatly owned, Matrox mostly targets more than only one market segment with their first product of a new line. So a compromise is obvious. They needed a chip, which is on the one hand capable to satisfy gamers and the other hand enough for modern, professional OpenGL CAD business applications, in order to prevent to loose more and more gamers and professionals to these two competitors.

Well, home users are mostly not too difficult to satisfy in 2D quality. If they'll get their good benchmark results, most of them buy the card.

Business clients look at both. 2D quality and sufficient 3D performance (with reserves for the following revisions of the used software).

So Parhelia has to do both and it does!! Although the OpenGL performance still could be better (and probably will get better in some driver releases), it's as well a nice card for professionals. By the way, here it's as well nice to have two similar TMDS like Parhelia has and R300 will not have, because DualHead is becoming more and more popular in business, digital TFTs are standard here most of the time and all that counts is total cost of ownership. TCoO is the second thing, that the business market leads to products which live longer than 6 months. Products which run rock solid and needn't to be exchanged until the designated successor's coming. So there should be good support, which Matrox has.

So you should design a chip with hidden reserves (which Parhelia obviously has) to convince the owners while using their product for a prolonged time, that they set on the right horse, by increasing driver performance more and more.

So, with Parhelia we are at the beginning again, now.
Not less - not more. That's at least my impression.

Finally, it's obvious that Matox has more interests to realise a higer 2D quality for Parhelia than ATi for Radeon 9700. That's why I still would prefer Matrox Parhelia, at your place, if 2D image quality is important for you, MWM.