home *** CD-ROM | disk | FTP | other *** search
- Xref: sparky comp.unix.bsd:10758 comp.os.linux:22137 comp.unix.pc-clone.32bit:926
- Newsgroups: comp.unix.bsd,comp.os.linux,comp.unix.pc-clone.32bit
- Path: sparky!uunet!math.fu-berlin.de!informatik.tu-muenchen.de!roell
- From: roell@informatik.tu-muenchen.de (Thomas Roell)
- Subject: Re: ET4000/W32 and VESA VL-Bus
- In-Reply-To: hasty@netcom.com's message of 29 Dec 92 22:42:41 GMT
- References: <1992Dec28.190316.22747@Informatik.TU-Muenchen.DE>
- <1992Dec28.223727.3569@netcom.com>
- <1992Dec29.191833.1401@Informatik.TU-Muenchen.DE>
- <1992Dec29.224241.21987@netcom.com>
- Sender: news@Informatik.TU-Muenchen.DE (USENET Newssystem)
- Organization: Inst. fuer Informatik, Technische Univ. Muenchen, Germany
- Date: Wed, 30 Dec 1992 09:56:45 GMT
- Message-ID: <1992Dec30.095645.29100@Informatik.TU-Muenchen.DE>
- Lines: 72
-
- >They do a good DRAM design according to S3 they simulate a VRAM in the
- >8c801/8c805 design. I don't know the specifics of their design and I doubt
- >that they are going to make it public.
-
- I assume that they simply read the video data for one complete scanline
- (1024 or 2048 bytes) within one page cycle (tRAS = 90ns, tCAS= 30ns, but
- just guesses, please could someone look the exact thing up in the databook),
- which means with a 1024 bytes buffer they would have around 77.6 MB/sec
- available (with tRAS = 120ns & tCAS = 40ns, this would be 44.3 MB/sec) at
- 70Hz refresh rate and 1024x768. This might be actually over the limit the
- graphics engine could make use of. The big BUT is still, how are things
- looking at 1280x1024, and how are they compared to the 928 which has fully
- 100MB/sec available ?
-
- >>Will be intresting. Have another benchmark number for people who like
- >>really hot boards ... This is the Nth Engine/250, a VRAM based board
- >>with a 82C481 on it... Maybe the 928 will be faster ;-)
- >
- >My expectation is that is going be faster than 120k xstones :-)
-
- Hmmm .... I make no guesses here, but going that high, you'll hit some
- other limits, which are generally called io-bus limits. On the
- Nth Engine/250 I computed a xstones rating based upon earlier tests I ran
- on other 481 boards and the measured rect500 / scroll500 throughput.
- I expected 128k. The reason why this was not reached was that in some
- operations, the host to 481 communitcation was simply the limit. Since
- most commands go exclusively over io-ports (which are slower than memory
- accesses, and awfully slow on the ISA bus), you can do only around
- 1 million read/writes per second. For some other operations clipping again
- got a cpu performance issue. The local bus seems to have solved some
- things regarding this io-problem, but some other problems will still
- prevent you from getting the top results by just simply linearly
- interpolate the xbench results based upon the maximum graphics chip
- throuput.
-
- >Who is the chip manufacturer for the 82c481 and do you have the phone number
- >for the board manufacturer of the "Nth Engine/250"?
-
- The 82C481 is from Chips & Technologies (I thought Amanicio would be the
- ultimate expert on these things ;-)). The Nth Engine/250 is of course from
- Nth. Don't have their number here. This board is way out a reach for most
- people since it is a 2MB VRAM solution with a Bt485 RAMDAC on it. I think
- it sells for over $1500 ... Anyway I see the prices dropping if 928 boards
- with 2MB VRAM will actually sell (my personal favourite 928 board is still
- the ELSA WINNER 1000, which does 1280x1024 non-interlaced for DM 950 in
- Germany which is about $600).
-
- >and at 1024x768. I can't because I am stuck with a low resolution monitor.
-
- You can run these higher refresh rates as well. Nobody said that you have
- to watch xbench running. Just run XS3 and xbench, but without monitor. The
- results will be correct.
-
- >All I want to say is that at $199 the Actix GraphicEngine32 DRAM card is not
- >a bad card at all. Also, as much as I want to get my hands on a 928 it was
- >delayed two months and I still have to get hold of one to see how it works.
- >have the databook and I am all set to got at it.
-
- You are right with the point that a 801/805 is right now generally a better
- buy than a 911/924 buy. And since unluckly the 928 boards seem not to be
- selling within 2 or 3 months from now they still will be for some time.
- But if someone can wait this long, he/she will get a better buy with the
- 928, which also will do 1280x1024 and 1024x768x16bpp and 800x600x32bpp.
- I predict that this will be much more attractive for people that do not
- exchange their cards every other week.
-
- - Thomas
- --
- -------------------------------------------------------------------------------
- Das Reh springt hoch, e-mail: roell@sgcs.com
- das Reh springt weit, #include <sys/pizza.h>
- was soll es tun, es hat ja Zeit ...
-