512 bit memory bus is here :)

If its not fast at gaming then what is it fast at? And why do we care if it sucks at gaming?

Looks interesting though, thanks for letting us know.:)
 
compres said:
If its not fast at gaming then what is it fast at? And why do we care if it sucks at gaming?

Looks interesting though, thanks for letting us know.:)
Click to expand...

CAD, profesionall apps you know.

Why do we care? well I suspect that with proper drivers it could be a gaming monster.
 
3DLabs specialise in workstation graphics - with this particular board you should pay attention to the fact that it has a very beefy dedicated geometry processor that feeds the two rendering chips (each of which have their own vertex shaders disabled).

And we are interested because this particular forum is the 3D Technology and Hardware forum, which (as the title suggests) is interested in the hardware and technology, not just gaming chips.
 
Dave do you know how many VS units/pipes the seperate geometry chip has and how many the VPUs have (when enabled)?
I think I read it somewhere but now I can't find it.

Clockrates would also be very nice (VPU's, memory and geometry chip) :)
 
Ante P said:
Still only SM2.0-class though.
Click to expand...

I'm trying to feel that one out. The hardware isn't capable of VS2.0 because there are no texture lookups (especially in this multi-chip configuration). However, the fragment pipeline is pretty much fully virtualised (including instruction limits, AFAIK), is FP32 and supports full dynamic branching, so technically this should be PS3.0 capable. If there isn't something else in hardware thats blocking it this might just be a driver thing or possibly some bizarre restriction with HLSL and PS/VS support (?).
 
Ante P said:
Dave do you know how many VS units/pipes the seperate geometry chip has and how many the VPUs have (when enabled)?
I think I read it somewhere but now I can't find it.

Clockrates would also be very nice (VPU's, memory and geometry chip) :)
Click to expand...

Tough question to answer, since they do not appear to be looking at things in a similar fashion to ATI/NVIDIA. Their initial tech documentation states "two vertex shaders" for the VSU and one each for the VPU's, however the pdf documentation states "32 programmable vertex shaders" for the VSU and 16 for the single VPU. Its obvious that they are looking at multiple ALU's for a single VS in the initial documentation - I'm trying to figure out if they are vector or scalar ALU's (P10 listed everything as Scalars but I'm hearing condradictory info on P20).

As for clocks, my initial guess on that will be "go fish" - I've still not tied down clocks for thw Wildcat VP series yet, they are very tight with these.

Bjorn said:
Any particular reason for the 36 bit vertex pipeline ?
Click to expand...

Just for even further increased vertex precision (geometry accuracy is the name of the game in this application).

[edit] Links to the single chip board configs:
http://content.3dlabs.com/Datasheets/realizm100.pdf
http://content.3dlabs.com/Datasheets/realizm200.pdf
 
DaveBaumann said:
Ante P said:
Still only SM2.0-class though.
Click to expand...

I'm trying to feel that one out. The hardware isn't capable of VS2.0 because there are no texture lookups (especially in this multi-chip configuration). However, the fragment pipeline is pretty much fully virtualised (including instruction limits, AFAIK), is FP32 and supports full dynamic branching, so technically this should be PS3.0 capable. If there isn't something else in hardware thats blocking it this might just be a driver thing or possibly some bizarre restriction with HLSL and PS/VS support (?).
Click to expand...

sm3.0 is mentioned here http://www.beyond3d.com/forum/viewtopic.php?topic=13278&forum=10 though. Typo?
 
Nice card but I bet it will be extremely expensive. It not aim at games but professional 3d graphics. Even suppose to have to OpenGL 2.0 support. I was surprise it did not have SMS 3.0 support, because I would expect that OpenGL 2.0 can do what SMS 3.0 has. I wonder how long it will take software to support. I have NewTek's Lightwave 8 and I not sure what OpenGL level it supports.

At first thought I though that this board would take 3 slots by the picture. Is this a trend for boards to take more than one slot. It probable not a big deal when you purchase a higher end board, expect to have a large PS and case configuration. I serious doubt anybody will be sticking one of these GPU's in a shuttle.
 
hstewarth said:
Nice card but I bet it will be extremely expensive. It not aim at games but professional 3d graphics. Even suppose to have to OpenGL 2.0 support. I was surprise it did not have SMS 3.0 support, because I would expect that OpenGL 2.0 can do what SMS 3.0 has. I wonder how long it will take software to support. I have NewTek's Lightwave 8 and I not sure what OpenGL level it supports.

At first thought I though that this board would take 3 slots by the picture. Is this a trend for boards to take more than one slot. It probable not a big deal when you purchase a higher end board, expect to have a large PS and case configuration. I serious doubt anybody will be sticking one of these GPU's in a shuttle.
Click to expand...

Actually it's not that expensive, 2000 EURos or so. Not that bad compared to top of the line Quadros and previous Wildcats.
 
I wonder if two of these will work with Alienware's video array. :p

8k x 8k textures, 32 hardware lights, 256.000 pixel shader instructions. ouch.
 
13.gif


So only VS2.0 but PS3.0 is now confirmed from 3DLabs
 
hstewarth said:
. Even suppose to have to OpenGL 2.0 support. I was surprise it did not have SMS 3.0 support, because I would expect that OpenGL 2.0 can do what SMS 3.0 has.
Click to expand...

I commonly held belief but wrong, OpenGL 2.0 allows 0 texture units in the vertex shader, DirectX vertex shader 3.0 requires a minimum of 4 texture units.

However OpenGL 2.0 can scale upto and beyond vertex shader 3.0
 
DaveBaumann said:
Snip Dave's wealth of info
Click to expand...

=D Does all this info you are sharing by any chance coincide with an immanent review/preview of the card? *crosses fingers* Back when the P10 and Parhelia were launched, I was really impressed with the P10 and very excited at the prospects of its successor. I'd love to see B3D coverage of this beast.
 
Ante P said:
32 GB/s with a 512 bit memory bus
that's 500 Mhz memory (ie 250 "real" MHz)

that simply can't be right now can it?
Click to expand...

The specs are for the single chip 100 & 200 boards, not the full tri-chip 800 board.
 
OICAspork said:
=D Does all this info you are sharing by any chance coincide with an immanent review/preview of the card?
Click to expand...

Nothing on the horizon just yet unfortunatly.
 
You must log in or register to reply here.

Similar threads

S
Memory bandwidth vs memory amount *spin off*
2 3
Replies
42
Views
35K
Kb-Smoker
K
M
Will 512-bit bus + GDDR5 memory have a place in 2009 ?
2 3
Replies
42
Views
11K
CarstenS
CarstenS
M
  • Locked
mobility 9800 benchmarks!
Replies
4
Views
3K
Dave Baumann
Dave Baumann
Back
Top