To Haig

[Amiga Blitter]

Haig, please. Revamp this site whit fresh news.


please spill the bean (clock rate, memory, features, colors) using the cryptoapi.



we will try to understand

[Haig]

http://forums.murc.ws/showthread.php?s=&threadid=40897

http://forum.matrox.com/mgaforum/Forum8/HTML/001824.html

[UtwigMU]

Haig: how about a hint regarding core of the upcoming card:

Is it going to be a new core, core revision or Parhelia core?

[Chrono_Wanderer]

me thinks it's a sundog = Parhelia w/out FAA

[mdhome]

Silly question, but what can be gained by removing FAA from the P core; a large saving in die space?

Regards MD

[VJ]

Originally posted by mdhome
Silly question, but what can be gained by removing FAA from the P core; a large saving in die space?

The fact that no-one could complain about a mal-functioning FAA ?
Just kidding!

Perhaps higher clockspeeds ? Most chips are rated at a certain temperature/mm², but sometimes one can go higher on certain parts if parts next to it are cooler. By "leaving out" a part (in this case FAA), it might be possible to achieve higher clock in nearby parts...? Just guessing though...

So the question remains....


Jörg

[Tempest]

(Mostly to mdhome): Welp, the thing is that this new card would probably still be based on those excess Parhelia cores that they supposedly have lying around... The thing why you would have, for example, less memory, no FAA, slower clock speeds, cut-down buses/shaders and/or no triplehead is that it would run cooler and be more affordable without threatening the position of the "flagship" product.

I don't know anything about it of course, but IMHO artificially crippling the Parhelia would be a bad idea. They can cut the memory bus in half if they like, drop the clock speeds and drop the amount of memory to 64MB, but deliberately disabling any "free" goodies like the FAA is just wrong. Then again, if they intend to replace their G550 with the Sundog and if the P128's price is any indication of its production costs, I think axes would have to swing quite heavily to achieve a decent price level.

[borat]

i think the point is any striped down card would be a parhelia core with features still present but hardware disabled, this is so any one wanting the 'high end' features would need to spend $350 on parhelia but those who dont want all the features can get what they need at a knocked dwn price both shifting cores and maintaining profits as those buying the cheaper cards are unlikely ever to purchase an expensive high end card and those enthusiastic enough to buy a P128 would never want a non flagship product.

[J1NG]

I don't think Triple-Head is removed from the up and coming card. The other stuff is very likely, but not Tripple-Head, not from what's been said lately.

Well, that's my opinion anyway.

J1NG

[fberger]

I think the new card needs to address the high end 2D market, such as video editing. I am not talking about MPEG and NLE, just the display function:

- triple head to enable 2 monitors and a TV simultaneously (only Matrox has it)
- high quality and speed on the TV Out (only Matrox does it well)
- 64 Mb max and lower frequency to differentiate from P and justify lower price

Matrox is now known as the niche market graphics card maker, and there's no market available in the low end 3D. So if they release a new card, it will enhance an existing niche market Matrox product (ie G550) to solidify their position.

The price should be slightly higher than current G550 DH, I'd say in the USD 130 range.

Those are just my opinions (and wishes!) though...

Fred

[Chrono_Wanderer]

Personally, I can live with the speed Parhelia is offering right now. IMO they need to update their core... including...

- Fixed FAA
- Fixed Banding
- 64/128 sample Anisotropic Filtering
- DX9 Baby!
- Some sort of bandwidth saving technologies.

IMO the Parhelia project actually helped Matrox to gain some experiences in making new graphics chipsets. Parhelia already has a good approch (i.e. 256bit DDR interface, 512bit Core logic, effective approch to AA, DX 8.5 (hehe). I don't think implanting DX9 will be too hard for them. (Just look at G550. Although I know it's the remaining of G800, they managed to stuff the TnL and "powerful" Vertex Shaders into a G400 based architecture.

However, the Parhelia's 4x4x4 acrhitecture may be too much of an overkill. Maybe 4x2x8 will help them reduce the die size, so that they can clock the chip at better "ease". (since they are lacking the equiptments to clock the chips) Of course, I won't complain about a 8 pixel pipeline architecture

They seriously need to research some sort of bandwidth technologies though... literally EVERY graphics card has some sort of bandwidth saving technologies to boost its 3D performance, at low transistor cost.

Well, if I remember correctly, the GeForce FX 5200 also has around 80 million transistors. Meanwhile, the FX has all the extra stuff like DX9 pipelines... I am pretty sure they can simplify the Parhelia core a bit, to save some transistors, and therefore easier to clock, which means cheaper production too...

Okay, I don't claim to know anything, but at least that's my ideas.