Announcement

**Umfriend** · 1 March 2006, 12:58

RSN finally.

**lowlifecat** · 1 March 2006, 13:50

Originally posted by Mikko

- Better 3D performance than the current Parhelia-512 chip

Something i would never bet any money on.

Matrox is out of 3D with absolutly no plans for coming back.

I was wrong. TH2G rocks!!!

**schmosef** · 1 March 2006, 16:28

Originally posted by lowlifecat

Something i would never bet any money on.

Matrox is out of 3D with absolutly no plans for coming back.

llc, do you work for Matrox or not?

If you do work for Matrox, are you stating facts, or just opinion.

Vista is designed to rely heavily on the GPU.

What hope does Matrox have if they don't intend to support the basic features of Microsoft's next OS?

**Nicram** · 1 March 2006, 17:09

There will be MS Win Vista version that will not need any GPu power to work fine so... Matrox don't see problem here.

**schmosef** · 1 March 2006, 17:14

"work fine" is a relative term.

if it doesn't support all/any of the bells and whistles then why bother upgrading?

**Nicram** · 1 March 2006, 17:50

Today market for Matrox are graphic professionals.
If someone read Matrox Insight (I do) then he know about AKLTERA FPGA chips, that are programmable FPGAs. I think one of the thing Matrox can do is add such chip to the graphic card,
so the customers may customize some of the graphic card functions, so it will be more specialized in army or labs for example.

**schmosef** · 1 March 2006, 17:56

So anything they produce from here on out will not be something that any of us here on Murc will want.

**Gurm** · 1 March 2006, 20:50

Or... that anyone anywhere will want? I dunno... Matrox is painting themselves into a corner that companies with a lot more history failed to live in.

Cornerstone... well, they still have a business but it's mostly monitors at this point...

Xionics... gone. (And I used to work there... woo!)

I can't recall all the other competitors in the "we only make display solutions for document/image processing professionals", but all of them had to diversify or die. Chips get faster.

**Marshmallowman** · 1 March 2006, 21:53

Matrox's core stuff is imaging and video, the multidisplay and video cards are all kind of designed to work with there imaging/video gear, to the point the embed there graphic chips into the imaging and video stuff.

However with matrox latest video stuff actually leveraging other comanies video cards 3d acceleration hardware we may see the scaling down/demise of any kind of consumer grahics card from them.(just about already happened)

Their multidisplay stuff is still second to none and it is big biscuits to matrox, companies/governments that need video walls,etc kind of stuff pay big $$ becasue the matrox gear "just works" and integrates with other matrox stuff so well.

The company I work for is "matrox only" for graphic and video/imaging hardware with no inkling of trying any other brand. While we don't sell volume gear, we work in provide a complete system (integrated workstations and server setups) that must completely stable and reliable (a bunch of 9's)
and there are plenty of other companies in the same situation

**frankymail** · 3 March 2006, 14:04

Actually, FPGA chips are very expensive, and many are required to emulate complex chip designs.... also, they run at very low speeds (usually less than a tenth of a custom chip). Of course, it would be very nice... but not pratically doable... anyone of you guys interested by a graphic card supporting every possible 3d features? Ok now, if that card were to cost $50 000? an if it ran at the speed of a G200? that'S what I though

**Tjalfe** · 3 March 2006, 15:06

Originally posted by frankymail

Actually, FPGA chips are very expensive, and many are required to emulate complex chip designs.... also, they run at very low speeds (usually less than a tenth of a custom chip). Of course, it would be very nice... but not pratically doable... anyone of you guys interested by a graphic card supporting every possible 3d features? Ok now, if that card were to cost $50 000? an if it ran at the speed of a G200? that'S what I though

price on FPGA's vary hugely depending on the amount of logic in them. Starting at a few $ to well, multiple thousands. I use a 400000 gate Xilinx Spartan 3 at work, I think the cost is around $15-$20
even a 400000 gate chip should be able to do a some special effects, but only a few. I doubt you can squeeze much of full GPU in this space

. Here is an open source video card design for those with too much time on their hands

OpenCores

http://www.opencores.org/browse.cgi/filter/category_video

cores, VHDL, Verilog HDL, ASIC, Synthesizable, standard cell, IP, Intellectual Property, 32-bit RISC, UART, PCI, SDRAM, full custom, system on a chip, SOC, reusable, design, development, synthesis, designs, developers, C, Linux, eCos, open, free, open source cores, RTL code, system-on-a-chip, circuits, digital, GNU, GPL, core, controller, processor, system design, chip design, EDA, design methodology, design tools, ASICs, programmable logic, FPGA's, PLDs, CPLDs, verification, Synthesis, HDL, Simulation, IC design software, semiconductor design, integrated circuits, system designs, chip designs, EDAs, design methodologies, design tool, ASIC, programmable logics, FPGA, PLD, CPLD, Synthesis, circuit, Synopsys, system design, chip design, programmable logic, FPGA's, PLDs, CPLDs, verification, Simulation

as for execution speed, true, a custom chip will run circles around it, but also costs a couple of $millions to start up.

**lowlifecat** · 3 March 2006, 15:11

Originally posted by schmosef

llc, are you stating facts, or just opinion.

i could be wrong

**Nicram** · 3 March 2006, 16:37

Well i ment professional use, not home/office so the price can be high.
Secodn things is that the problem is not the programmable GPU but the fact, that any company will not give access to program it's GPU (because they will have to open documention for this or all). But if You will add few FPGA chips, that will not be very fast, but make some effects few times faster than CPU then it may be intresting. & You will may open all programming options for them & hide programming of main GPU for Yourself.
Ofcourse all that is just my imagination, but it may be intresting

**frankymail** · 17 October 2006, 22:05

Originally posted by Tjalfe

price on FPGA's vary hugely depending on the amount of logic in them. Starting at a few $ to well, multiple thousands. I use a 400000 gate Xilinx Spartan 3 at work, I think the cost is around $15-$20
even a 400000 gate chip should be able to do a some special effects, but only a few. I doubt you can squeeze much of full GPU in this space

. Here is an open source video card design for those with too much time on their hands

OpenCores

http://www.opencores.org/browse.cgi/filter/category_video

cores, VHDL, Verilog HDL, ASIC, Synthesizable, standard cell, IP, Intellectual Property, 32-bit RISC, UART, PCI, SDRAM, full custom, system on a chip, SOC, reusable, design, development, synthesis, designs, developers, C, Linux, eCos, open, free, open source cores, RTL code, system-on-a-chip, circuits, digital, GNU, GPL, core, controller, processor, system design, chip design, EDA, design methodology, design tools, ASICs, programmable logic, FPGA's, PLDs, CPLDs, verification, Synthesis, HDL, Simulation, IC design software, semiconductor design, integrated circuits, system designs, chip designs, EDAs, design methodologies, design tool, ASIC, programmable logics, FPGA, PLD, CPLD, Synthesis, circuit, Synopsys, system design, chip design, programmable logic, FPGA's, PLDs, CPLDs, verification, Simulation

as for execution speed, true, a custom chip will run circles around it, but also costs a couple of $millions to start up.

Of course, there are numbers of low-cost FPGA... but a 400k gates one is far from having the ability to emulate a 300-400 millions transistors custom ASIC...

If you want to be able to do any custom effect real-time (or almost)... it would be simpler to have general purpose microprocessors on-board... wait... isn't it what the CPU is for???

But the biggest problem isn't hardware... it's making a -usable- SDK to make use of the processing abilities... easily... and unfortunately, Matrox' software development team doesn't really have a great track record.

Announcement

What would you expect from the next Matrox GPU?

What would you expect from the next Matrox GPU?

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment