Impact Hardware flames the Parhelia!

[Yannick]

Mikeul:

"You can't blame anyone from being honest."
The fact that a new "gamer's card" is not as good as the best available doesn't make this "not a gamer's card". Most people who will buy Parhelia have a card that is old and probably worse that most products still on the gamer market, and they're still gamers.

""Matrox employees did tell us that the Parhelia wasn't a framerate killer. "
Wrong. They insisted that their super bandwith would make their card a killer."

Haig did insist on the fact that the Parhelia would not be framerate killer.



""Happens I'm not running benchmarks, but real games... "
Serious Sam 2,Jedi Knight2, Quake 3, Aquanox, etc. all are real world games..."

... when run in benchmarks, you do not choose the settings. When you really play those games, you can choose to enable the features that matches your card's best capabilities and avoid its weaknesses. You cannot do that in a benchmark, or when you can you cannot compare the cards because different cards means different strengths.


" "the development cycle of games is too long to take a particular card into that much consideration... "
But the benchmarks above are games sold in shops..."

...most based on engines developed in the time of nVidia's dominion. Unreal was developed for Glide, and had poor performance on Direct3D cards at first. It did not mean that Direct3D cards were inefficient, only that the games needed some serious tweaking to match different chips...


" "Well, at least Matrox drivers do not need antivirus-like update like nVidia"
Yah, but Matrox G450 can't even play Jedi Knight 2 properly because of the lack of driver updates whereas the TNT2 M64 can."

Then keep your TNT2 M64... UT2003 will be so fine on it...


""I'm much more interested in visual quality than peak framerate."
Image quality nothing.
Parhelia doesn't support yet anything above anisotropic 2x whereas ati and nvidia support 8x anisotropic already... "

I thought this was a driver limitation only ? Well... you might be right, I didn't learn the specs by heart.

Still, speaking of visual quality, I'm also speaking of 2D quality.
When not gaming, I'm running my desktop at 1600x1200x32bits@70Hz. Nothing extraordinary. Image quality is perfect, thanks to the RAMDAC. A friend of mine recently switched to one of the numerous GeForce cards. He says that although 3D performance is very good, desktop quality is not as good as with his old G400. And I also read an article somewhere explaining how nVidia's third-party made cards often lacked visual quality because of poor analog circuits.


" "So the first version of the brand-new Parhelia is only that good compared to the state of the art of gaming cards, the fourth generation geForce ?"
Have you seen the price of the parhelia?
So you're willing to spend so much more just because there's matrox logo on the card?
The parhelia is more expensive than the GF4 Ti 4600 with performance inferior to the Ti 4200..."

The Parhelia may be a little bit too expensive, but there's something else to it than 3D framerate. Plus, the chip is loaded with new features that will only show their power when games are using them.




" "Most Matrox fans will buy the Parhelia not because of the company's name, but because of what you get when you buy Matrox. "
Exavtly.
The Matrox logo on the card.
Blindness personified."
I like that way of putting your answer pretending I said exactly the contrary of what I said.

[Cheesekeeper]

One item that is constantly alluded to is that the original cards may have had the Parhelia chip running at a higher rate, which was then dropped at release time. All of the benchmarks seem to indicate that if the chip was running at a slightly higher clock speed, the performance would most likely impove considerably.

This is pure conjecture, but is it possible that Matrox were intending for the Parhelia to be released at a higher clock speed, but were forced to clock it down when the production chips didn't yeild the speeds they were hoping for? If this is the case, the build-up would have been justified (as they were expecting retail cards to be faster), and would explain the discrepancy between the specs and the actual performance.

I would have to agree that everyone has been deceived to a certain extent. Matrox always said that it would not be a frame-rate killer, but let's face it, most of the people here were expecting performance numbers higher than we've been seeing. It may not be a killer, but it should at least be able to hold its own in the benchmarks. The memory bandwidth is impressive, but is a waste of time if the chip isn't running fast enough to utilise it.

The only way we will know this for sure is if someone is able to overclock the card (say to 250MHz) with better cooling, and see the performance numbers again. Matrox may be able to bump up the speed within the coming months, to a level where it can compete directly. Unfortunately, by then it will be too late, as it will be competing with a new generation of cards.........

[GNEP]

I think that I will jump in here.

I was a little bored yesterday, and I am still interested in P but undecided, so I collated a load of the benchmark numbers from a number of different sites into a spreadsheet.

I then had a think about how I would like to play games. And decided that 1024x768 with max AA was probably the best mix for me.

So I took all of the results with those settings, from all the different sites, and watdya know? Parhelia performs about the same as GF4Ti4600. Pretty consistently.

Now at the same time, I had a think about clock speeds. For Bulk and Retail cards, the ratio of core clock to mem. clock is 1:1.25. Now when they were talking about the chip specs only, 20GB/s of bandwidth was bandied around (excuse the deliberate pun). This would mean a mem clock of 312.5 MHz. And hence a core clock of 250MHz (as cheesekeeper suggested).

14% or so above what the retail is at now. Add that to the purported "10%" improvement from more mature drivers, and we have around a 25% increase in speed (assuming that FPS scale with clock).

That is in line with the "20-30% faster than a GF4" figures that were also floating around before.

It seems to me that Matrox have (due to whatever reason - worse than expected initial chip quality I guess) had to clock lower than they planned. Hence what we are seeing now.

But what intrigues me is the possibility of putting a big HS and fan on there, and getting the originally planned performance.

Maybe there will be a MAX version (probably a mega-expensive 256Mb "Pro" version once they have certified OpenGL drivers), clocked at 250/312.5. Of course they will have to have all of Santa's elves working overtime to hand-pick suitable chips from the production line for this.

Interesting.

gnep

[Goc]

This is pure conjecture, but is it possible that Matrox were intending for the Parhelia to be released at a higher clock speed, but were forced to clock it down when the production chips didn't yeild the speeds they were hoping for?

I'm with you there Cheesekeeper. I think that is exactly why we are seeing somewhat smaller performance numbers then what we expected...
To be completely honest I expected performance to be exactly as it is now, and hoped for better. Therefore I'm not disappointed at all, and I'll be buying a P-card as soon as it gets in shops here...

[pgde]

Originally posted by Cheesekeeper
One item that is constantly alluded to is that the original cards may have had the Parhelia chip running at a higher rate, which was then dropped at release time. All of the benchmarks seem to indicate that if the chip was running at a slightly higher clock speed, the performance would most likely impove considerably.

This is pure conjecture, but is it possible that Matrox were intending for the Parhelia to be released at a higher clock speed, but were forced to clock it down when the production chips didn't yeild the speeds they were hoping for? If this is the case, the build-up would have been justified (as they were expecting retail cards to be faster), and would explain the discrepancy between the specs and the actual performance.

I would have to agree that everyone has been deceived to a certain extent. Matrox always said that it would not be a frame-rate killer, but let's face it, most of the people here were expecting performance numbers higher than we've been seeing. It may not be a killer, but it should at least be able to hold its own in the benchmarks. The memory bandwidth is impressive, but is a waste of time if the chip isn't running fast enough to utilise it.

The only way we will know this for sure is if someone is able to overclock the card (say to 250MHz) with better cooling, and see the performance numbers again. Matrox may be able to bump up the speed within the coming months, to a level where it can compete directly. Unfortunately, by then it will be too late, as it will be competing with a new generation of cards.........

I think you are correct. See Haig's responses (posted today, 6/27) at http://forum.matrox.com/mgaforum/For...ML/001314.html

[ALBPM]

I don't care about his opinion. I have been testing the card for a couple of weeks now and most of what I have seen so far is complete BS from fps crack whores.

I'll stand by what Father Joel said!!

Matrox has never been known for FPS killer gaming cards but these "editors"=(figurative/subjective sense of the word) are missing the most important issues.

They must truely have their heads up their asses if all they see are frame rates. Most haven't noticed or mentioned all the new features that Matrox developed for Parhelia. Again Matrox leads the Industry in developing new technology and features and you will see nVidia and ATI scrambling to cludge together their versions of Triple Head for surround gaming (copying Matrox!!)

I've been testing the "P" for several weeks now in 3 different systems and it is beautiful and smooth. I'm loving every minute of it.


Paul

[Lemmin]

The pictures of the early boards all showed rather large heatsink/fan combos.

As I am understanding the various comments floating around, they had to replace the large slot-hogging HS/F with a svelte, slimline one to fit into a standard AGP slot, which apparently meant they had to lower the clock speed to reduce the thermal output.

I am wondering whether the upcoming 256mb board will be an AGP pro board (and will therefore be 'allowed' to take up double the room), and whether this will allow for higher clock speeds. The 256mb version will almost certainly be targeted at the professional market, and a lot of pro cards seem to be AGP pro, either for the extra room, or for the power requirements. My FireGL takes up two slots worth, and my Wildcat did as well, both apparently for the extra cooling.

I might wait and see what happens if/when the 256mb version appears.

LEM

[Cheesekeeper]

Originally posted by pgde


I think you are correct. See Haig's responses (posted today, 6/27) at http://forum.matrox.com/mgaforum/For...ML/001314.html

Thanks
I've just posed a reply there (which for all I know could get me bumped straight off the forums), asking Haig (in a very gentle and indirect manner) what to expect if an irresponsible Matrox user violated his warranty by attaching a beefier cooler to the Parhelia. It's not the sort of question they would usually answer I suspect, but I'm hoping to coax out a general yes/no as to whether the chips on the shipping Parhelias could be expected to work similarly to the Alpha boards with proper cooling.......

[Ryu Connor]

He's already answered you.

[diewlei]

hey cheesekeeper. nice work.

[Cheesekeeper]

Thanks

Well this is getting interesting. If we can bet on a 20MHz increase in speed, we are looking at a situation similar to that which Gnep suggested (but with a boost of a little under 10% of clock speed). I suppose the big question is - how much of a difference in actual speed would that change give us?

If the Parhelia chip itself is the only limiting factor then the improvement could be fairly reasonable (much more so than the 2-3% improvements of Geforce cards, where there are several limiting factors). I guess at least from what Haig has told us, it is fairly safe to try.

Of course if any of us die in the process, the deaths will have to be classified as "Gnep's Syndrome"

[GNEP]

Thanks!