review at chip.de

[nggalai]

OK, review is up (in German, once more):

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

LOTS of bugs in all tested games but Dungeon Siege . . . eek.

ta,
.rb

[Muad'Dib]

My comprehension of German is about equal to Nullset, but there si something that confuses me.

BTW, I personally could pretty much care less about fps, but this is not making sense. So the big P has lower fps in most of the 1024x768 and quite a few of the max res. graphs, but there is very little difference between the two - without the AA I'm talkin about. There seems to be about 3 - 5% (I haven't done the math, so I'm probably wrong) between the normal comparrisons at 1024 and max res. The gf4 on the other hand seems to have much great hits going from it's normals at both resolutions.

Is this probably driver related - I'm assuming yes to this?, or something holding back the big P or messing up the gf4 - any ideas?

Or am I just up way past my bedtime...

-D

[Guru]

Originally posted by thop

[nggalai]

Hi there,

Originally posted by Muad'Dib
My comprehension of German is about equal to Nullset, but there si something that confuses me.

BTW, I personally could pretty much care less about fps, but this is not making sense. So the big P has lower fps in most of the 1024x768 and quite a few of the max res. graphs, but there is very little difference between the two - without the AA I'm talkin about. There seems to be about 3 - 5% (I haven't done the math, so I'm probably wrong) between the normal comparrisons at 1024 and max res. The gf4 on the other hand seems to have much great hits going from it's normals at both resolutions.

Is this probably driver related - I'm assuming yes to this?, or something holding back the big P or messing up the gf4 - any ideas?

Or am I just up way past my bedtime...

-D

We're probably translating the review into English as soon as a) Matrox provides drivers that have anisotropic filtering working properly and the review can be extended in that respect and b) most of the mistakes have been corrected.

For now, I think one of the most interesting bits is this here:

Leider hat der aktuelle Treiber 1.0.1.225 noch einen Bug beim anisotropen Filtern - er erreicht maximal den Level 2 (kann man gut mit Serious Sam 2 mittels der Option "gap_iTextureAnisotropy" nachprüfen), obwohl der Parhelia-Chip eigentlich auf Level 8 wie die GeForce4 Ti ausgelegt ist. An diesem Punkt sind dann auch alle geplanten Benchmarks mit anisotropen Filter flachgefallen, weil dies auf Level 2 (dem niedrigsten, Level 1 ist der isotrope Filter) mehr oder weniger überhaupt keinen Sinn macht. Wir müssen diese Benchmarks schuldig bleiben, bis Matrox diesen Treiberbug ausräumt - wenn unser Testsample dann überhaupt noch bei uns ist.

"Unfortunatelly, there's a bug regarding anisotropic filtering with drivers 1.0.1.225 -- only 2x AF is supported (tested with Serious Sam: Second Encounter's gap_iTextureAnisotropy option), even though the Parhelia supports up to 8x, just like the GF4 Ti. At this point, I decided to skip benchmarks with AF enabled, simply because 2xAF doesn't make any sense. We'll provide AF benches as soon as the bug has been fixed by Matrox--if we still have our test sample by then."

Regarding your question: the Parhelia seems to be very little bandwidth-limited (surprised, anybody? ), hence higher resolutions won't tax it as much as the GF4 Ti4600. Bummer is that as of now, those higher resolutions with FAA are too slow to be interesting for most gamers.

Also, the current drivers seem to be VERY buggy. Lots of system crashes and render artefacts indeed. I personally don't care about maximum framerates, I want to know how well a VGA fares regarding minimal framerate. I rather have a board that delivers min-max of 35-45 than one that does 30-80fps in a game. Stability is king.

I am really looking forward to tests with better drivers. Here's waiting.

ta,
.rb

[tt22]

I am very dissapointed over these reviews, i ordered a parhelia but i am gonna cancel and buy me an x-box. Yes i play games and i had great expectations to the parhelia, on paper it looks like it would chrush the geforce4.
I would say it is the other way around. The only thing i would say parhelia has for itself is the FAA, but what if R300/NV30 comes with a similar technique? Comparing FAA to FSAA does not give a good picture.

Sourround gaming is more or less as important as head-casting
Unless u have 10.000 USD to spend on the new triplehead LCD screens. 3 regular monitors just isn't very good in gaming.

And FPS is important, while 20 fps in future games just isn't good enough. Even if each picture looks very nice, it doesn't help if you are looking at a slide-show.

With all this power on the paper, i would say something is seriously wrong with either the card or the drivers.

I like Matrox very much, even owned a 400 MAX, but the new parhelia don't give the expected competiton to the videocard marked, looks like SIS are doing a much better job there.
I hope future drivers will help the new card, i would like at least 50% improvement before even considering buying this card.

[Drizzt]

Originally posted by nggalai
Hi there,

[cut]

Also, the current drivers seem to be VERY buggy. Lots of system crashes and render artefacts indeed.

Hi nggalai.
This really a strange thing. We all know that BBz have their parhelia in their primary system since a bit, and I don't think that if the drivers were so buggy they were as happy as they are...

What the hell are you reviewing?

[Reckless]

I'm a little dissapointed considering the price of the unit. Matrox may well 'win' with FAA enabled but some of the scores drop down to almost bad levels. Let's face it, if you've got a GF4 you're unlikely to run with AA enabled as it's too resource hungry.

[thop]

the review (the whole site) is down. seems puretec kicked them for the traffic

[tt22]

Originally posted by Reckless
I'm a little dissapointed considering the price of the unit. Matrox may well 'win' with FAA enabled but some of the scores drop down to almost bad levels. Let's face it, if you've got a GF4 you're unlikely to run with AA enabled as it's too resource hungry.

The ti4600 4xfsaa even beats the Matrox faa in Serious Sam i think. I don't quite understand how this is possible?!?!

With this price Parhelia will dissapear when R300 arrives soon.
I think Matrox are in trouble, the good old man is getting too old

[Reckless]

Now I'm at home I thought I look at the chip.de graphs in greater detail. The beta testers here have sure given the correct message but in such a way that needs further examination.

Yes Parhelia can run with just about playable frame rates at 1024*768 with FAA BUT
- that figure will plummet if triple head is used (the much touted and admired 'but ultimately too expensive for the masses' feature)
- more importantly - a GF4 can match it for speed at 1600*1200 therefore removing most of the need for AA in the first place.

The question on IQ of GF4 vs Parhelia hasn't been touch on much and it's probably Matrox have the edge but whether that alone is enough for a gamer to swallow an extra $200 (compared to GF4/4200) is another matter.

Yes, the Matrox does have good features and reasonable speed but after reading all reviews the fact remains that the Parhelia is overpriced - I reckon by at least $100. Obviously if I was given a card (in the form of beta testing...) I think I'd think it was damn wonderful as I hadn't spent any of my cash on it. It would then for certain be an A-list card.

When (and if) improvements in speed from drivers together with titles supporting displacement mapping the card may indeed gain value. I have decided not to buy one which will be a MASSIVE surprise for my m8's in the UT clan - they know what a Matrox supporter I've been. I'll re-visit the subject when UT2K3 is released, just maybe drivers/improvements/price may have been looked at?

[isochar]

The anandtech review has UT2k3 benchmarks.

[Reckless]

Thx for that and FFS

But rest assured that Matrox's Fragment Anti-Aliasing looked nothing short of amazing. We understand that the performance of the Parhelia to this point isn't representative of a $300 - $400 card but if anything is a redeeming quality it is the GPU's AA quality.

Looks nice runs fairly poor Bah, and I was hoping for some light at the end of the tunnel

Edit: I should have explained the reason for being so upset:

Currently FAA will not work with any use of stencil in a game, which is one thing Epic had to disable in the UT2003 demo in order for FAA to properly work. Although the demand isn't necessarily great for UT2003 to support stencil right now, eventual support is necessary and if FAA doesn't properly work with it enabled then Parhelia will be forced into 4X supersampling mode which is no better than what the Radeon offers.

[Greebe]

Reckless assumption on your part.

To get an idea of how SH vs TH mode works in relation to fps compare the same amount of pixels in the different modes and the frame rate is directly proprotional to the total displayed pixels.

Example: (for comparative purposes only! Not indicative of real Parhelia speed)

SH @ 1600x1200 res = 1.92 million pixels = ~100fps

TH @ 3072x768 res = 2.36 million pixels = ~77.1fps

TH @ 2400x600 res = 1.44 million pixels = ~125fps

[isochar]

I was going to get 2 more 19" Sony's to go with the Parhelia, but it can't come close to playable fps at the monitors' intended resolution. (3840x1024 = 3.93 million pixels)

[Dr Mordrid]

--------------------- Q3Bench RESULTS ---------------------

FILE: q3b_062402_213115.txt
START TIME: Monday, June 24, 2002 09:31:15 PM
COMMENT: Parhelia-512 with "FAA-16x Always" (clocked at 275, 220, 220)

BENCHMARK SETUP:
----------------
DEMO: FOUR
CFG FILES RUN: Maximum
RESOLUTIONS RUN: 640x480 800x600 1024x768 1280x1024 1600x1200

NUMBER OF RUNS: 1 times.


MAX Configuration File Settings:
(640x480) 150.0
(800x600) 144.6
(1024x768) 116.1
(1280x1024) 76.5
(1600x1200) 55.6

NOTE: With FAA-16x on 1024x768 looks as good as FSAA 4x at a higher resolution in those "other cards"

--------------------- GL Excess ---------------------

Score: 4344

FPS (High/Low)

Scene 1: 208/125
Scene 2: 111/74
Scene 3: 76/59
Scene 4: 100/55
Scene 5: 166/100
Scene 6: 96/68
Scene 7: 62/46
Scene 8: 90/82
Scene 9: 142/108
Scene 10: 125/54
Scene 11: 168/110
Scene 12: 200/141

Dr. Mordrid