what does 512 bit really mean?

[borat]

just a genuine question here no, matrox 512 nvid 256 therefor nvidia must die rubbish, what is the point in having a 512bit card over a 256 bit one?it doesnt seem to give any more performance that i can see so why?what does it mean?

[Greebe]

First off I'd like to thank you for that tasty morsel of engrish

(hopefully I can remove my decoder ring now

Parhelia doesn't have any form of Occulsion culling thus requires greater bandwidth to function at max performance.

Now this would have been far more detrimental to it's operation IF the core ran at, lets say 300mHz. Seeing that this speed (for whatever reason) couldn't be produced, so the neccessity of having a mem bus so wide is fairly negated, but to have removed it would have come at great expense and more delays.

BTW if it had clocked to 300mHz it would compete with the best of the best currently out dispite the fact they have to run at substancially higher clocks (~100mHz)

That proves just how powerful Parhelia's design really is

[Dogbert]

Greebe, I think Borat will need a decoder ring for your post as well

[Greebe]

Well he can go buy his own box of CrackerJack cause this one be MINE!

[Kurt]

Originally posted by Greebe
First off I'd like to thank you for that tasty morsel of engrish

(hopefully I can remove my decoder ring now

Parhelia doesn't have any form of Occulsion culling thus requires greater bandwidth to function at max performance.

Now this would have been far more detrimental to it's operation IF the core ran at, lets say 300mHz. Seeing that this speed (for whatever reason) couldn't be produced, so the neccessity of having a mem bus so wide is fairly negated, but to have removed it would have come at great expense and more delays.

BTW if it had clocked to 300mHz it would compete with the best of the best currently out dispite the fact they have to run at substancially higher clocks (~100mHz)

That proves just how powerful Parhelia's design really is

what of latency? bus efficiency? you're talking about theoretical peak bandwith only....

[Greebe]

No I'm talking about real world facts

[OasisPlaces]

512 Dual Bus.. marketing jargon.. it just means 2 256bit
uni directional bus's (1 way).. if it actually connected to the memory with separate read/write lines it would be useful..
you'd have two separate streams.. but in reality it's
still like having 2 lanes turn into a 4 lane highway then
back to 2 lanes again.. for the most part wasted pavement.
But it is really good marketing.. a bigger number.. like
"but.. this amp goes to eleven.."

[borat]

hmm in that case i presume someone has had a parhelia somewhere running at close to those speeds for that comment to be made.so does 512bit mean it can process 512bits per clock or something different?apologies for my grammar was very tired when i wrote that after drinking some local brew 15% proof cider the evening before.wonder what you could get out of p with a pelt, good chip and some more power. if i have an agp pro slot can that deliver more power to the cparhelia than standard agp?or does the card have to be specifically agp pro to take advantage of this?

[WayneHu]

If the Parhelia-512 still uses the Dual bus design inside the core, Matrox would probably call it 1024-bit dual bus according to the design on G200/400/550. The Dual bus width of Matrox's design is always double data rate per cycle of the external memory interface. Ex:

G200: 128bit Dual bus with 64-bit SDR SGRAM.
G400: 256bit Dual bus with 128-bit SDR SGRAM.
G450: 256bit Dual bus with 64-bit DDR SGRAM.
(G550)

The external data bus of Parhelia-512 can transfer 256bit data per half cycle (DDR mode). Thus each queue of the external memory interface's read/write buffer should hold the full 512-bit data to deal with the burst mode of DDRSGRAM in 256-bit data bus width.

The next generation GPU which uses DDR II interface might take the big Latency advantage of Dual bus design. DDR II has the separate read/write data pins. Probably it is able to do the concurrent read/write access in one cycle on page hit if the address bus is also bi-edged stroking.

[Chrono_Wanderer]

I don't think so... it will be more like Dual 256-bit bus for the Parhelia.

If i remember correctly G200 has 2 64-bit bus, while the G4/5xx has 2 128-bit bus

[OasisPlaces]

Originally posted by borat
hmm in that case i presume someone has had a parhelia somewhere running at close to those speeds for that comment to be made.so does 512bit mean it can process 512bits per clock or something different?apologies for my grammar was very tired when i wrote that after drinking some local brew 15% proof cider the evening before.wonder what you could get out of p with a pelt, good chip and some more power. if i have an agp pro slot can that deliver more power to the cparhelia than standard agp?or does the card have to be specifically agp pro to take advantage of this?

Nah.. it just has to do with how they wire the internal bus
to the memory, and the agp interface, it doesn't make
a difference.. just reads and writes are running on independant
lines inside the chip. outside you're back to sharing.. so
the gain is just a few %points.. probably about the same as
overclocking by a few Mhz

here's an article

http://www.lostcircuits.com/video/mga400/3.shtml

[ElDonAntonio]

From anandtech:

We already explained where the Parhelia part of the name comes from but what about the '512' suffix? Luckily this is a bit easier to explain; as you'll soon see, there are a number of situations where the number 512 appears when looking at the architecture of the Parhelia. Matrox's decision to name the GPU the Parhelia-512 is akin to NVIDIA calling the NV10 the GeForce256.


At the very start of the Parhelia-512's pipeline (after the AGP interface) you have the vertex processors. Matrox has outfitted the Parhelia-512 with four vertex shader units that they refer to as "128-bit Vertex Shader Engines." If you multiply 128 by the 4 units you'll get that magical 512 number. The 128-bits comes from the fact that each one of these units can work on four 32-bit floating point numbers at the same time provided that they are packaged as a 4-operand vector. Each one of these vertex shader engines is comparable to each one of the two vertex units engines in the GeForce4 or four of the vertex processors in the 3DLabs P10. This means that right off the bat, the Parhelia-512 has twice the vertex throughput of the GeForce4 at equivalent clock speeds.