P8X core speed

[Kurt]

Since this is the Crystall Ball...

[dbdg]

Said 250 cause 225 wasn't an option.

[GNEP]

3000000+ cos this is gonna whip everything else out there

[Nappe1]

depends on many things.
I'd say between 250 and 300 or between 300 and 3000000 depending if it has been respinned to 0.13µm...

[Kurt]

That's the question, innit? with all the time they had and a part that could do 290 on the old process, it might well go beyond 350 on a now-well-tuned 0.13 process.

But a die-shrink costs money...

[magog]

Originally posted by Kurt
That's the question, innit? with all the time they had and a part that could do 290 on the old process, it might well go beyond 350 on a now-well-tuned 0.13 process.

But a die-shrink costs money...

If they were really smart, they also used the optimised core, that Cadence routing program has delivered in Nov/Dez 2003 (+20% speed).
[Edit]
Removed link to Haig's comment, because it wasn't linking to Cadence routing program...so time will tell.

I also think with the restructured core there should be 300(+) MHz possible for the chip clock. Well we will see when the first buyer has this 8x baby.


For those who don't know....

Source: http://www.cadence.com/company/succe...?xml=matrox_ss
============================
NanoRoute Ultra Routing Boosts Chip Performance Significantly
"We obtained a NanoRoute Ultra license on a Thursday, immediately installed it into our flow, and scheduled Cadence to come in the following week to help us get going with it," reported David Chiappini, ASIC Director. "However, on the following day, Friday, one of our engineers reviewed the users' manual and decided to give it a try by himself on one of our designs that we had already released as a chip. Over the weekend, with no assistance at all, he wrote a script that succeeded in rerouting the chip using NanoRoute Ultra. As if that weren't impressive enough, the new design ran 20% faster than the old one. We'd made some changes to the design so I can't say for sure the entire performance increase was due to NanoRoute Ultra, but clearly it was a major contributor."

The design was very substantial and quite complicated according to Chiappini. It had approximately 80 million transistors, seven million nets, and 20 clock domains.
============================

[Nicram]

Ahhhh... this chip was made the same time that agpx4 was made so... 250

[Kurt]

@magog: a lot of wishful thinking. they might not be working on the card of your dreams. who knows if it's not an obscure project for a no less obscure market as seems to be where they're heading now...

[mabraham]

When we are speculating and dreaming I might as well throw another thing into the fire (although this is probably TOTALLY unrelated to Matrox it is always fun to speculate).

The article " Game Developer's Conference 2004: Carmack Addresses the Masses" at AnandTech ends with the following: "The second generation of DX9 hardware, coming down the pipe Real Soon Now^{<font size="-1">TM</font>}, is also bringing some much needed speed to the table". Although Parhelia hardly qualified as a first generation DX9 chip the "Real Soon Now^{<font size="-1">TM</font>}" could be interpreted as a Matrox hint (with all the Soon^{<font size="-1">TM</font>} stuff we have been throwing around).

As I stated above this is speculation and probably without any relation to Matrox at all...

[GNEP]

I doubt it but RSN is also something else Matrox-esqe

[borat]

well from haigs comments over at matrox forums we will be getting a DX9 chip capable of >30FPS in all situations with all eye candy from this "summer" project. Sounds pretty exiting considering all i am anticipating is nothing.

[Kruzin]

For the 10000000th time.
We WILL NOT see a new chip this summer.
There will be a new card. It will be slow, expensive, non-dx9, non-gaming.

[mdhome]

Originally posted by Kruzin
For the 10000000th time.
We WILL NOT see a new chip this summer.
There will be a new card. It will be slow, expensive, non-dx9, non-gaming.

Apart from a possible PCI-Express card what will be the difference between this and the current cards that Matrox produce?

Regards MD

[rylan]

It comes on a funky tiedye colored pcb!

[Kurt]

Originally posted by borat
well from haigs comments over at matrox forums we will be getting a DX9 chip capable of >30FPS in all situations with all eye candy from this "summer" project. Sounds pretty exiting considering all i am anticipating is nothing.

Matrox "Summer" Project: this summer, go to Québec, take a bus to the wilderness, find a nice place with a lake or a big pond and start smashing the maringouins, sumulis and other bugs with defective P cores.