CompuServe Thread

#4000 Deinterlacing

17 messages in this thread
#73163From: Paul GerhardtOct 11, 1992 10:26 AM
On the subject of deinterlacing, Ben (Black Belt) posted that there was significant performance costs to deinterlacing in software. I put the C= display inhancer from my 2500 in the 4000 and did some tests: 1. Yes it does work! In fact better than the software method. On my 4000 the software deinterlacing produces a very slight but visible vertical pattern of light & dark regions about 8 pixels wide. There is no such pattern using the hardware deinterlacer. It does not like the higher color resolution modes. That is not surprising seeing that the board is not long enough to fill the new longer video slot. I use the second monitor that I had for viewing DCTV, connected to the regular RGB port to view native graphics & DCTV now. 2. There DOES seem to be a penalty for using software deinterlacing. My informal test was to simply run a hirez 256 WB with two 256 color pics as workbench and window backdrops. With software deinterlacing there was a significant slowdown. You could easily see widows being rendered from the bottom up. With software deinterlacing off performance improved to the point where window rendering look "normal". 3. By using hardware deinterlacing life is much simpler. There is none of the "multisyc jump" caused by switching from interlace to non-interlace (even with mode promotion on, some programs still force interlace screens) since my 1950 always sees deinterlaced screens. And my DCTV is happy since it now always sees interlaced (DCTV does NOT like software deinterlaced screens) screens. For now my display enhancer stays in the 4000! BTW I also installed a 2091 (my dealer will let me trade it in as soon a the true SCSI II cards come out), and I have my Syquest and 80 meg Quantum back. I am a happy camper now! But there goes two slots! …AutoPilot Engaged… -Paul-
#73173From: Black Belt SystemsOct 11, 1992 2:16 PM
Paul… How would you describe the performance of the 2091 as compared to the internal IDE drive? –Ben …via AutoPilot
#73188From: Paul GerhardtOct 11, 1992 5:47 PM
Ben, I just ran a few quick tests. With an 80K picture file: Loading and viewing with superview: IDE 2sec 2091 4sec Copying from ram to: IDE 8sec 2091 5sec Now these are just "one one thousand, two one thousand, etc" seconds and not real accurate. But it appears that the IDE is faster when reading data, and the 2091 wins on saving data. I am going to use the IDE drive just for storing programs and my SCSI Quantum & syquest for data. Once I get a SCSI II card I should be all set. BTW While I have your ear. I was playing with AIBB and this puppy is AMAZINGLY fast with math code that written for the 040's internal math processor. Will there be a version of IM that is optomized for the 040? …AutoPilot Engaged… -Paul-
#73199From: Black Belt SystemsOct 11, 1992 8:34 PM
Paul, We do plan to add an '040 version of Imagemaster to the release disks, however, it does depend on the SAS C compiler, which we just received a week ago. We've got to get familiar with the new compile options and then run a LOT of tests so we can be sure that the extra cost of including an 040 version provides some results. Compiling for the 030 is close to compiling for the 040 with the exception of floating point operations, which actually very few operations in Imagemaster require – Morph, and a few others. Thanks for the insight on the speed; I kind of had a feeling that the writes were way off, for some reason. Hey – what's "AIBB"? –Ben …via AutoPilot
#73205From: Steve SetzerOct 11, 1992 9:35 PM
Ben, AIBB is a very good benchmarking program, I believe it is on the Fish disks, if not I can up a copy. While I am no expert, the test on AIBB seem to be very well thought out and useful. I use everytime I add hardware or am just checking to see if my system is running ok, it saves the results of each test. I know some of the earlier version come with source, so YOU could look and see what it is doing, (it's greek to me). …Steve On AutoPilot! (at last)
#73226From: Black Belt SystemsOct 12, 1992 1:26 AM
Aha. Thanks for telling me. –Ben …via AutoPilot
#73241From: Joanne DowOct 12, 1992 2:24 AM
Are these both in the A4000? And from those speeds I would suggest you run dspeed4 that Mike Sinz wrote. The figures you get MAY be pretty bad. {O.O}
#73281From: Paul GerhardtOct 12, 1992 10:41 AM
Yes I mounted an 80 meg Quatum on the 2091, so both are "in" the 4000. OK I'll run some more formal tests and post the results, but basically they ARE slow. Can't wait for real SCSI II :). …AutoPilot Engaged… -Paul-
#73427From: Denny Atkin [COMPUTE/AmiOct 14, 1992 12:15 AM
Note that it's been discovered that many A4000s seem to have been misconfigured at the factory with (ack!) old filesystem partitions! Do an INFO on all your drives and if any of them show 488 bytes per block, you should back them up and reformat them with FFS. Should increase speed by a factor of 10 overall…
#73456From: Forrest McKinneyOct 14, 1992 10:20 AM
Well it is actually formated with a Custom file system… I dont understand the reasing behind it myself… but I went in and reformated it with the FFS with Directory Cache, but there is no speed increase in the drive… Forrest
#73486From: Denny Atkin [COMPUTE/AmiOct 14, 1992 6:22 PM
Directory Cache will slow writes down something horrible on hard drives. Only really recommended for really pokey HDs, such as old A1000 parallel port drives, and for floppies.
#73554From: Forrest McKinneyOct 15, 1992 9:41 AM
When I reformated with FFS and Directory cache, there was no speed change at all.. but it does cache the Directories.. Forrest
#73582From: Dean BrownOct 15, 1992 8:16 PM
Forrest, Do not use directory caching on HD's. It's generally a loss. The DCFileSystem is really intended for floppies where the caching can overcome the extreme slowness of the drive. Dean DKB Software
#73483From: Paul GerhardtOct 14, 1992 5:16 PM
Yep, 488 bytes per block! 🙁 …AutoPilot Engaged… -Paul-
#73240From: Joanne DowOct 12, 1992 2:23 AM
Varied. Some people seem to be getting drives that are HIDEOUSLY slow. Others report speeds in the range that you'd expect. C= folks are now aware of this and are trying to nail down why. Tomas Rockiki (Radical Eye/TeX) has such a slow machine and is helping nail down the cause. A working IDE drive will be slower than an equivalent SCSI. But that is with drives in the 2.2meg/sec class in SCSI which on the A4000 come out at speeds on the order of 1meg/sec. {@_@} Um, the 2091 is by reports also hideously slow but not as slow as some of the IDE drives in a few machines that "escaped" or something.
#73310From: Greg MalcangiOct 12, 1992 5:19 PM
If anymore info is useful. I`ve just got an A4000 in the UK and it`s giving me around 1,057,000/sec, which is noticeably slower than my A2000 GVP HD, which is more like 1,675,000/sec. It does seem strange to have such a powerful tool in many areas and for it to load and save considerably slower than it`s little brother. Greg :^)
#73312From: Black Belt SystemsOct 12, 1992 5:40 PM
JoAnne, Has anyone (besides people AT Commodore) reported a "fast" drive? –Ben …via AutoPilot