home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!sdd.hp.com!cs.utexas.edu!sun-barr!olivea!sgigate!sgiblab!news.kpc.com!kpc!pcarmack
- From: pcarmack@gimp.kpc.com (Phil Carmack)
- Newsgroups: comp.unix.pc-clone.32bit
- Subject: Re: ET4000/W32 and VESA VL-Bus
- Message-ID: <1992Dec30.201935.6026@kpc.com>
- Date: 30 Dec 92 20:19:35 GMT
- References: <BzBEI1.CH@aeon.in-berlin.de> <1992Dec20.153314.24148@Informatik.TU-Muenchen.DE> <BzKqwn.5vA@wimsey.bc.ca>
- Sender: pcarmack@gimp (Phil Carmack)
- Organization: Kubota Pacific Computer
- Lines: 41
-
- In article <BzKqwn.5vA@wimsey.bc.ca>, bhenning@wimsey.bc.ca (Bill Henning) writes:
- |> The problem with DRAM based accelerators, even if they get 160Mb/sec
- |> bandwidth out of the DRAM's is that the bandwidth is not random-access,
- |> but is rather a result of page mode (or static column, or nibble mode)
- |> DRAM cycles, therefore if you are using say 110Mb of bandwidth to
- |> feed a 1280x1024x8 display, the theoretically 50Mb/sec bandwitdh
- |> left over will actually be more like 10-20Mb/sec, as most graphics
- |> operations will not be able to take full advantage of page mode,
- |> even if the bus interface to the local bus supports it.
- |>
- |> Bill
-
- This is not entirely true. It is true that the actual bandwidth of the
- graphics rendering plus the pixel scan out will be less than the total
- max (because of set up and tear down overhead of paged mode accesses),
- but depending on how much buffering there is during scan out, both
- should be accessing the DRAMs in bursts. Also, during vertical (and
- horizontal for that matter) retrace, the renderer has the complete attention
- of the DRAM.
-
- Of course, as always there are many graphics operations that will break
- page mode accesses because they are not sequential (e.g. drawing vertical
- bars on the screen). But copy performance is one of the most important
- features for running windows since moving a window is just a copy of a
- rectangular region.
-
- I should say that I know absolutely nothing about Tsengs chips or how they
- work. I was just speaking theoretically, and this whole bandwidth issue
- depends entirely on how much buffering the chips do. For example, if the
- chips buffered an entire scan line of data during scan out, total bandwidth
- utilized would approach the max because the overhead of setting up for page
- mode accesses would be amortized of a large chunk of data.
-
-
- Cheers,
-
- Phil
-
- --
- Philip Carmack | pcarmack@kpc.com (408)987-3336
- Kubota Pacific Computers, Inc. |
-