Silicon Graphics w/ 12 x i860xp
Goto page Previous  1, 2

Post new topic   Reply to topic    CPU-World.com forums Forum Index -> Modern Chips (Collectible Chips only)
View previous topic :: View next topic  
Author Message
ItsMeOnly



Joined: 06 Jun 2006
Posts: 173
Location: Warszawa, Poland

PostPosted: Mon Jun 11, 2007 4:16 pm    Post subject: Reply with quote

swaaye wrote:
Just think....that whole board is now days compressed into just a portion of a graphics chip. Smile

Well, there's a little more to it- the RealityEngine 2 beats the crap of even contemporary graphics by one factor: reliability. AFAIR in 1995, the internal data transfer was about 1 Gbit(!) per second, something today's consumer grade memory achieved just recently. Besides, RE can render broadcast-quality scenes in real time with guaranteed frame rate (so no slowdowns when more details appear in foreground or there's just some environmental mapping), you have 8 graphic pipelines- something that consumer cards started to have a year or so ago.

And besides- the today's cores run at 450 MHz, 860's here run at 50 MHz each...
Back to top
View user's profile Send private message   Visit poster's website Yahoo Messenger MSN Messenger
swaaye



Joined: 02 Sep 2005
Posts: 47
Location: WI, USA

PostPosted: Wed Jun 13, 2007 3:46 pm    Post subject: Reply with quote

I wonder how one of G80's or R600's stream processors, at 1.35 GHz and 740 Mhz respectively, compares to a 50 MHz i860? Smile And they have over a hundred of each in those GPUs.

1 Gigabit per second is only about 125 MB/s bandwidth.. That's nothing. Even if it was 1 gigabyte per second, that was exceeded by Voodoo3/TNT back in the '90s with external commodity SDRAM chips.

8 pipelines is great for the time, but I'm sure they are very basic relative to what was in the first 8 pipeline fully-integrated GPUs like ATI R300. Playstation 2 technically has a 16-pipeline GPU, by the way, but you wouldn't know it from the graphics because the GPU is very simplistic.

I also would imagine that considering both NVIDIA and ATI make professional-grade products out of these processors (Quadro and FireGL) that they are fully capable of delivering reliably smooth rendering.

Ex-SGI engineers are the basis for a lot of the engineering talent at NV and ATI.
Back to top
View user's profile Send private message   AIM Address
Windmiller



Joined: 24 Jun 2005
Posts: 1716
Location: Chapel Hill, NC

PostPosted: Wed Jun 13, 2007 3:56 pm    Post subject: Reply with quote

swaaye wrote:


Ex-SGI engineers are the basis for a lot of the engineering talent at NV and ATI.


I think you're right. Are you familar with the Pixel-Plane project from UNC? Many of those that were involved with it went to either SGI, Nvidia and places like Nintendo, etc.
Back to top
View user's profile Send private message [ Hidden ]
ItsMeOnly



Joined: 06 Jun 2006
Posts: 173
Location: Warszawa, Poland

PostPosted: Wed Jun 13, 2007 4:07 pm    Post subject: Reply with quote

swaaye wrote:
I wonder how one of G80's or R600's stream processors, at 1.35 GHz and 740 Mhz respectively, compares to a 50 MHz i860? Smile And they have over a hundred of each in those GPUs.

In that it's superscalar, superpipelined processor with really good FPU performance? And you have 12 of'em?
Quote:
1 Gigabit per second is only about 125 MB/s bandwidth.. That's nothing. Even if it was 1 gigabyte per second, that was exceeded by Voodoo3/TNT back in the '90s with external commodity SDRAM chips.

Actually I was wrong: the referenced Reality Engine memory bandwidth was 8,3 GB/s.
Quote:

8 pipelines is great for the time, but I'm sure they are very basic relative to what was in the first 8 pipeline fully-integrated GPUs like ATI R300.

Using 8 monitors simultaneously? 30fps HDTV? on one graphic pipe, for that matter
Quote:

that they are fully capable of delivering reliably smooth rendering.

Well, again, we're talking about 10 years of difference! And still, my point is valid- the reliability stays, there are still fully blown RE's working in graphic studios, how long does your graphics last?

Btw, some benchmarks:
Code:
RealityEngine2:

  2M t-mesh triangles/second
  930K textured t-mesh triangles/second
  80/160/320M textured, anti-aliased pixels/second
  Hardware texture mapping
  Real-time anti-aliasing
  VGA up to 1600x1200 and HDTV display
  Advanced stereo modes
  Hardware image processing acceleration
  48-bit RGBA color, quad-buffered (192 bits total)
  256 to 1024 bits per pixel
  40 to 160MB frame buffer
  NTSC/PAL/S-Video output

Back to top
View user's profile Send private message   Visit poster's website Yahoo Messenger MSN Messenger
swaaye



Joined: 02 Sep 2005
Posts: 47
Location: WI, USA

PostPosted: Thu Jun 14, 2007 4:06 pm    Post subject: Reply with quote

Those are, of course, embarrassingly slow results. It's ancient technology. It was neat at the time, but it's a ridiculous waste of electricity today and isn't going to provide anything remotely like reasonable performance compared to something recent. The only reason to have one around is to keep reaching for some more return on investment on that beast.

As for this nebulous "reliability" claim, what makes you think a vastly smaller and more powerful and less proprietary machine from the past 5 years wouldn't be more reliable?

The slowest current Quadro seems to offer multiples more performance than an entire Reality Engine 2. While using 21 watts. Smile
http://www.nvidia.com/object/IO_11761.html
Back to top
View user's profile Send private message   AIM Address
Display posts from previous:   
Post new topic   Reply to topic    CPU-World.com forums Forum Index -> Modern Chips (Collectible Chips only) All times are GMT - 5 Hours
Goto page Previous  1, 2
Page 2 of 2
Jump to:  
You can post new topics in this forum
You can reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You cannot attach files in this forum
You cannot download files in this forum

Powered by phpBB © 2001 phpBB Group