Matrox and Parhelia questions?

[Greebe]

Well for someone searching for answers you sure know how to bait a responce.

If you knew what the comparsion was and why then you should have stated it in the first place instead of having me attempt to read your mind ie guess.

Wasting my time does not make a happy Greebe, especially after a 19 hour day.

[isochar]

I'm curious myself. Bandwidth delivers here, and the Parhelia is dead last.

http://www.anandtech.com/video/showdoc.html?i=1647&p=5

But the main reason we pursued this test was to see how the Parhelia performed; in this case, the 27.6 fps is an indication that its 256-bit memory bus isn't doing much for the Parhelia at all.

[Silencer]

ah.. well. if I wasting your time, why you answer my posts?

but maybe you can give me some answers about the low parhelia performance, or don´t you search for answers at this point?

[Greebe]

Now your just behaving like a flippant azzhole. READ!

[az]

Hey Greebe, don't give him such a hard time. It's obvious his english isn't the best, so he likely can't express himself the way he wants to - I'm sure he didn't mean to provoke you. And his searching for answers is quite understandable, Parhelia should come out way ahead in bandwidth tests, but instead it's last.

I certainly don't know the reason - might be due to UT2K3 (and most, if not all other software) only using two of P's 4 texture units, because it doesn't know the card, might be a driver problem, which'd be understandeable given the just recent introduction of this new architecture (but then, this really wasn't the right time for benchmarks), or a problem with the chip not being able to make good use of the bandwidth provided, or a combination of some or all of these, and probably even others I can't think of at the moment.

You might know more, and you might or might not be allowed to share your knowledge, but if you're allowed, please do so - or take an educated guess, since you probably know more than we do.

AZ

[Silencer]

well, you´ve right. I´m sorry.

sometimes I exaggerate... I want a clear discussion about the parhelia and its performance.. but up to now, I´m shocked.
I´ve hoped months for the Parhelia and its great performance. Now I stand before a big shattered remains.
forgive me my reaction... but sometimes it is so hopeless to believe in a great thing..... sigh

[Marshmallowman]

the ut benchmark was a bandwidth based, but there were some strange discrepancies with the 8500 as well, for the nvidia card it did seem to scale...but I would not attache much importance to it.

There has been at least one or two benchmarks showing parhelia with a good synthetic bandwidth...eg there are to many factors clouding the whole area at this point.

But Think about this, currently parhelia has bandwidth it is not utilising fully, driver optimisation will take up some slack....but with the double sized memory interface than any other current card it will quite obviously scale well as the core speed goes(leaving memory as is) as well as other tweaks on the way.

Current competing cards have real peak fill rate/bandwidth that are pushing the limitations of there memory interface(hence all there bandwidth saving stuff).
In order to progress past the next few generation they will have to make the transition to a wider interface.(or use VERY expensive memory).

But matrox has already made the transition, parhelia will improve a lot and the next iterations(.13..etc) will surely build on this next generation base.

Parhelia is the first of the next generation!

[Greebe]

I most certainly can tell the difference between a language barrier and an accusation like...

or don´t you search for answers at this point?

Dispite whatever you may wish to think AZ he has baited responces and flamed. Another (mikeul) was warned already today for some of the same activity and it goes a long way in having a little respect toward those you get replies from... especially beta testers.

Actions like this simply do but only one thing, force us not to answer ANYTHING! We are under NDA and thus cannot comment or speculate on what "we" think per se. Yes there are times when our NDA's do indeed get in the way regretfuly, but bashing us because we don't answer after being baited like has happened here ... well all I can say is he has some gall!

[Greebe]

Silencer I can forgive and forget. Thus seeing what you stated above does indeed justify that

There are some things that I was trying to put into words but couldn't in such a manner inwhich I knew wouldn't violate my NDA and thus will refrain until I have authorization todo so... say later today possibly. Everyone on this side of the world that matters is in bed, as I should be.

[Fubar|wtf|]

Seriously though. The P has every advantage on menmory bandwidth, and should annihilate the competition in those tests. It doesn't help the bandwidth credability with comments like these floating around either:

"The performance was really disappointing for the first 256 bit DDR card. I tried to set up a "poster child" case that would stress the memory subsystem above and beyond any driver or triangle level inefficiencies, but I was unable to get it to ever approach the performance of a GF4." (http://www.shacknews.com/finger/?fid...idsoftware.com)

I am a Matrox fan, and have owned quite a few of them through the times, but either there are some _serious_ optimizations requiered in the drivers, or the hardware is "flawed". Either way I will wait to see what new driver revisions and the R300 brings to the table.

[Jon P. Inghram]

It's a design flaw, instead of having a 256 bit wide bus, they make a slight error and put 256 1 bit wide busses in.

[Reckless]

Originally posted by Greebe
I....goes a long way in having a little respect toward those you get replies from... especially beta testers.

Beta testers deserve the same respect as ALL posters. Just 'cause you have a Parhelia and access to information no other can hope to obtain, doesn't (automatically) gain you more respect.

You were hard on the guy but agreed he didn't phrase his statements well either. You've slightly redeemed yourself with the later post but it doesn't really excuse the fact that you maybe consider everyone else 'less' than yourself....

Am I right in thinking (by your aggravated link posting) that no-one is able to answer the "why low bandwidth" question? If that is the case then Matrox will no doubt have a hard time shifting the current generation of Parhelia boards. However, going by your last sentance, maybe there's a little light at the end of the information tunnel?

[AdmiralThrawn]

Parhelia's Memory Interface!

At first I want to say hello to you all, since I'm new here.

Oh, since I'm an austrian and therefore a native german speaker, don't expect high-quality english from me too!

Let's get to the point:

The Parhelia's extremely Low Memory Throughput in practical tests seems to be quite confusing. Until now, i think it's not clear if a full utilization of Parhelia's Texture Units will truly improve the performance by a reasonable amount. It MAY be that the memory subsystem just isn't getting fully used and the core is the true bottleneck as long as it's texture units are not used properly.

But it may also be that something is really, really WRONG with Parhelia's Memory Interface, i just don't know. But there has to be some limiting factor, either core or memory, because i cannot believe the drivers to be the cause of such a decrease in performance.

I've seen D3D / OGL Drivers since the Voodoo1 Cards, and - as always - driver development did improve performance, but NOT by more than 25%... If that should be the case with Parhelia's soon-to-come drivers, I'll be one of the first to party for sure! But i don't really asume, that the current drivers are that bad, considering the duration of the driver development.

I guess, we will not see any truly clear answers very soon. But some of us might be really, i mean REALLY BADLY dissapointed, because of course there has been the great hope of a Ti4600 Killer (as we aaall know) with much better image quality... Well, not as soon as we hoped I guess...

[Reckless]

Re: Parhelia's Memory Interface!

Originally posted by AdmiralThrawn
But some of us might be really, i mean REALLY BADLY dissapointed, because of course there has been the great hope of a Ti4600 Killer (as we aaall know) with much better image quality... Well, not as soon as we hoped I guess...

I think it's fair to say that not many on these boards (well, at least me!!) expected a GF4 killer. However, a card that rivals a GF4/4200 would have been a reasonable expectation giving the pricing point ... I do agree that with driver improvements a huge speed increase is unlikely (almost impossible) unless the driver team simply haven't used areas of the GPU (for stability/time/other reasons!)

[AdmiralThrawn]

In my opinion, the big problem is that the quality features of the Parhelia are not very useful as long as the base performance is that low. Who really want's to use FAA or even Triple-Head when you're not getting enough fps for a smooth gameplay? But i guess that has already been discussed elsewhere in this forum!

If the Parhelia had been somewhat equal to the Ti4200 I might have considered buying one, but not that way.