Trident XP4 Configurations - Killer Price !

[David G.]

XP4 T1: GPU 250 MHz, RAM 250 MHZ ( 500 MHZ DDR), 64-bits 64MB, US$69

XP4 T2: GPU 250 MHz, RAM 250 MHZ ( 500 MHZ DDR), 128-bits 64MB, US$79

XP4 T3: GPU 300 MHz, RAM 350 MHZ ( 700 MHZ DDR), 128-bits 128/256 Mb, US$99


I don't know what to think about the performances of this chip but having a full DX8.1 chip @ 300/700 Mhz costing less than 100$ is something extraordinary .
They also seem very optimistic about the chip's performance as they say that this fastest version will take up on GF4 4400 .

.... now I ask my self if I shouldn't introduce Trident as a new player on the 3D market for the next year ....

 

[Pete]

When is it going to see release, and what's your source? I haven't been keeping up with Trident.

 

[David G.]

Who knows ... maybe it will get on the market until December .... hopefully .

Source ... ? Sorry ... vr-zone .

 

[KnightBreed]

When is it going to see release, and what's your source? I haven't been keeping up with Trident.

See this thread.

 

[Ailuros]

Anand has a preview on it. It's not a Tiler after all. Just an IMR with a software based tiling mechanism.

 

[Gollum]

Its tiling is not software, its in the hardware - but you're right in that it has nothing to do with deferred rendering (what most people mean when they talk about tiling), it apparently stores information in its framebuffer in tiles and some members of the press immediately though that implied a deferred renderer (like EEtimes).

For that price the XP could be a killer product ... consider me a Trident customer if the drivers are up to the challenge!

 

[Dave Glue]

For that price the XP could be a killer product ... consider me a Trident customer if the drivers are up to the challenge!

Big if with respect to drivers, as most of us are aware. Astounding price/performance however if the XP4 can approach the ti4400.

These are the kind of announcements which really perk my interest, no doubt the NV30/R300 are exciting but at their price point they simply have little impact on the market as a whole, the vast majority simply won't shell out ~$350 for a graphics card no matter how impressive the specs. But offering a DX8 card with ti4400 performance for under $100 really has a chance to propell the PC games market forward, I can't wait to see how the drivers perform.

 

[demalion]

If Trident can deliver without any Xabre type rendering efficiencies, I'd say this would be the hardware release of the year. It would be nice if one of the "come backs" delivered without stumbling (I really do hope Matrox is working hard on a higher clocked part).

 

[alexsok]

Yep, that would be great!

demalion - rumors say that Matrox are indeed working on a 0.13 micron Parhelia chip already!

That means higher core & memory clock speeds, which the Parhelia REALLY needs if it wants to compete with R300/NV30.

What I'm more intersted in though are it's 4 pipelines and their perfomance...

It's obvious that only one of them really works (according to the tests). It's possible that lack of the DirectX 9.0, and therefore, the version 2.0 could cause locking of 3-4 pipelines.

It is quiet possible that DX9 application will "unlock" the full potential of Parhelia 512, but it still sounds rather strange to me...
Besides that, we can probably expect a full dx9 part (Vertex/Pixel Shaders 2.0, etc...).

 

[OVERLORD]

More discussion here http://www.pvr-net.com/cgibin/forum/ikonboard.cgi?s=3d4ef5c51a82ffff;act=ST;f=2;t=453 roadmap et al. Anyone translate chinese? http://www.greatcn.com/hardware/vga/trident/xp4/

 

[Gollum]

OT:
Alexsok, can you point me in the direction of any tests that indicate Parhelia is using only 1 of its pipelines? I hear that for the first time. And if that were to be the case, I wonder how it achieves the performance it does, because if Perhelia performs this well with one pipeline, every other 3D card manufacturer must do something very wrong.

Personally, I think Parhelia's performance "problem" has an entirely different reason. Compared to GF4 and R8500 it mainly seems to be clockspeed and somewhat inefficient use of its impressive bandwidth (memory controler design, no occlusion culling?). That and maybe a less effectively tweaked design overall, Matrox does have significantly less manpower and engineering experience with complex chips than ATI and Nvidia after all. I seriously doubt Matrox would have released Parhelia at all if something were so significantly broken...

 

[ben6]

err Parhelia doesn't use only 1 of it's pixel or vertex pipelines. It uses all 4. That's official word from Matrox btw.

 

[David G.]

Acording to this graph :

the development of XP4 started in Q4 2000 and ended in Q1 2002 with mass production to start 2 quarters later in Q3 2002 .

But , most interesting is that they also show that there is a new chip in development wich should start sampling in January 2003 with mass production in late Q2 2003 :

0.13 & 0.10 micron technology ;
350 Mhz CORE speed ;
1000 Mhz DDR-II Memory speed ;
Full DX9.1 capable .


The graph also shows future developments but I think those are to far to begin talking about them .

 

[Ailuros]

Its tiling is not software, its in the hardware - but you're right in that it has nothing to do with deferred rendering (what most people mean when they talk about tiling), it apparently stores information in its framebuffer in tiles and some members of the press immediately though that implied a deferred renderer (like EEtimes).

For that price the XP could be a killer product ... consider me a Trident customer if the drivers are up to the challenge!

http://www.beyond3d.com/forum/viewtopic.php?t=1972

 

[alexsok]

OT:
Alexsok, can you point me in the direction of any tests that indicate Parhelia is using only 1 of its pipelines? I hear that for the first time. And if that were to be the case, I wonder how it achieves the performance it does, because if Perhelia performs this well with one pipeline, every other 3D card manufacturer must do something very wrong.

Personally, I think Parhelia's performance "problem" has an entirely different reason. Compared to GF4 and R8500 it mainly seems to be clockspeed and somewhat inefficient use of its impressive bandwidth (memory controler design, no occlusion culling?). That and maybe a less effectively tweaked design overall, Matrox does have significantly less manpower and engineering experience with complex chips than ATI and Nvidia after all. I seriously doubt Matrox would have released Parhelia at all if something were so significantly broken...

Sure m8, here you go:

http://www.digit-life.com/articles/parhelia/index.html

 

[PC-Engine]

So the Parhelia 512 will be kicking some major booty when all the units are enabled with new drivers? Are we looking at NV30 and R300 territory as a result?

 

[alexsok]

So the Parhelia 512 will be kicking some major booty when all the units are enabled with new drivers? Are we looking at NV30 and R300 territory as a result?

Nothing is confirmed yet!

NO DX9 applications nor DX9 are avaiable yet, so it's too early to judge...

I'm assuming that it may unleash it's full potential in DX9 applications, but remember that Parhelia 512 is NOT a full dx9 solution, it is a half dx9 soluton, something like that.

Besides, the Parhelia 512 doesn't have any special "bandwitch saving" techniques like R300 & NV30 do.

I predict that the next revision of Parhelia 512 (on 0.13) will be a full dx9 part and if the full potential of the previous chip won't be unleashed when dx9 applications start to appear, then the new chip will definetly unleash it!

 

[David G.]

So the Parhelia 512 will be kicking some major booty when all the units are enabled with new drivers? Are we looking at NV30 and R300 territory as a result?

I've already disscused this problem with Matrox guys and they've said that drivers will only give a 10% speed bost when they will be 100% optimised .

The only "huge" boosts will come in games that experience some bugs with the P512 GPU .

 

[OVERLORD]

Acording to this graph :

the development of XP4 started in Q4 2000 and ended in Q1 2002 with mass production to start 2 quarters later in Q3 2002 .

But , most interesting is that they also show that there is a new chip in development wich should start sampling in January 2003 with mass production in late Q2 2003 :

0.13 & 0.10 micron technology ;
350 Mhz CORE speed ;
1000 Mhz DDR-II Memory speed ;
Full DX9.1 capable .


The graph also shows future developments but I think those are to far to begin talking about them .

Why didn't I think of that? Links obviously not good enough eh?

 

[David G.]

Explications weren't enough . Usualy roadmaps show when will a product be availble but this one ..... the single one that made some sense is showing the whole product manufacturing cycle .