CompuServe Thread

#Haynie Conference

18 messages in this thread
#93371From: Richard S. Rose,M.D.Nov 14, 1987 12:52 AM
Don, I agree with you in general. And I'm going to buy a 2000 because it is a the problem. The 1000 became a dead machine as soon as the 2000 was announced despite the fact that the OS is still compatible. Everyone (developers of both hardware and software) will be developing for the 2000. The progression of MS DOS machines (which was _not_ and orderly process) never hurt owners of the earlier machines because of the huge support by developers of the older machines. That support does not exist for the Amiga. And the smaller developer support base will move on to the newer machines for very good reason (I am especially referring to hardware support) For example, I think that something like a card cage for the 1000 will fail miserably. It will be just to costly to produce in the small quantities that will be in demand. For the cost of a card cage and a little memory one can upgrade to a 2000. The same thing may happen with the 2000 vs a 68020 or 68030 native machine. I just wish that I could say that my 1000 (a $2400) investment in Oct, 85) ever was more than just a pretty face sitting on my desk. As far as useful work, an ancient Kaypro and plain vanilla PC have been churning out word processing, DBM and spreadsheet jobs 12 hours a day for years. The 2000 better be able to join the group! But it sure is one sexy piece of equipment. And oh mi gosh what potential! Hmmm Where have I heard that before? Regards, Rich
#93384From: Don Curtis/SYSOPNov 14, 1987 2:17 AM
Rich, I really don't want to put down the A1000—-it doesn't deserve it, it's a fine machine, the first in the line, and fully software compatible with the A500/2000 (or properly, the 500/2000 are fully compatible with the 1000). That being said, I do have to agree, that little, if any new hardware will be developed for it. They can be had for bargain prices now (used machines from the trade in deals) and are well worth every penny for someone who wants an Amiga….in fact, if I were just going to buy now, I'd get a used 1000 rather than a new 500 if I were looking in that price range…it's simply a better machine from a hardware standpoint (power supply, detached keyboard, etc.). Add an Insider and an HD to an A1000, and it'll be a very useful and viable machine for a long time. When the next Amiga comes out…I just hope that it will be slot compatabile with the A2000, if it is, then the A2000 is a great machine today. I believe that the next machines will be slot compatible…and thus bought an A2000. Don
#93488From: Charlie HeathNov 14, 1987 6:18 PM
I don't think everyone will be developing on a 2000 … right now, my B2000 is turned off and I'm working on the A1000 because I like the keyboard better, and it is quieter. Also I don't have "slow-fast" ram in my A1000 so it runs a little faster. Probably once I connect up the DMA hard disk on the B2000 and get a 68020 card, I'll switch to it, but for now I'll happier working on the '1K.
#93523From: Don Curtis/SYSOPNov 14, 1987 10:09 PM
Cheath, Just for reference…the 2090 controller/MiniScribe drive combination has diskperf results of reads in the 50+K area and writes in the 13K area which isn't bad compared to the results I've seen for A1000 HD's. (that's with a 60% full 20Megger). Don
#93552From: Charlie HeathNov 15, 1987 3:40 AM
You know where I can find DiskPerf? I don't see it in the DL's ( checked Dl9, 10, &11). Thanks- Charlie
#93559From: John DraperNov 15, 1987 4:31 AM
Charlie, It's there, but isn't called Diskperf… try a BRO/KEY:perf* or a BRO/KEY:HARD in DL 9. Regards, Larry.
#93561From: Charlie HeathNov 15, 1987 4:45 AM
Ah, thanks, I'll check –
#93580From: Don Curtis/SYSOPNov 15, 1987 12:29 PM
Charlie, Try DSKTST.ARC in DL 9. Don
#93587From: Ben BlishNov 15, 1987 12:52 PM
Hey Don — We got 55k read, 39k writes under diskperf using the C LTD SCSI interface and our driver for it; That's not even DMA… it's programmed I/o. The DOS manages the data so badly I'd call it broken… the 2090 is supposed to be a DMA design, isn't it? You'd think they'd be able to do better than 13k writes. Sheesh! –Ben–
#93599From: Don Curtis/SYSOPNov 15, 1987 2:15 PM
Ben, Yup, the 2090 is a DMA setup…and I presume that the new fastdos (if and when it appears) will take care of the write problem. I presume you bypassed DOS to get the 39K writes because the 15K writes (I've gotten it that high with the 2090 and 0% full drive) seem good compared to other diskperf results I've seen (on the A1000…which shouldn't make any difference). The one interesting thing is that there is little dropoff as the drive fills…the results with a 50-60% full drive aren't much different that the results with a 0% full drive. One other thing…their new hddisk driver is a bit slower in reads than the older one, but faster in writes. Fragmentation does slow it down a bit also…but that's to be expected the way I see it. Don
#93716From: Ben BlishNov 16, 1987 1:54 AM
We did _not_ bypass DOS to get the write speeds up. And frankly, I have _no_ idea why no one elses drivers, including DMA drivers, seem to be able to reach up towards the read speeds. Not that I'm complaining. <grin>. –Ben–
#93686From: Bill AtkinsonNov 15, 1987 10:22 PM
Ben, I get 73K/sec reads and 30K/sec writes with the 2090 and a ST-251 40Meg which is faster, and DMA. I heard that someone is beta testing the new FFS with the same setup as mine and is getting about 300K/sec. I was also told that CBM is working on a much improved hddisk driver. I think this is the first generation driver for the 2090 where the Cltd has gone through 10^N revisions and has reached its limits. Back to your bike again. <grin> Bill.
#93719From: Ben BlishNov 16, 1987 2:03 AM
That's ok — I don't _expect_ to beat DMA speeds… I'm just continually amused that we usually do! I think that we are utilizing the programmed I/o controller pretty well, considering that amigados spends a LOT of effort getting in the way of efficient data processing; And one thing you should know; This is NOT the same driver C LTD has been using all along; it's a brand new design, and this is it's first time in the field. We were not involved with any of their previous efforts, and so bear no responsibility for those generations, or the bugs contained therin. I suspect that your performance under _our_ fast file system, will be substantially better than that achievable with CBMs… we will see. –Ben–
#94004From: Charlie HeathNov 18, 1987 4:42 AM
I've been playing with the new FFS, and though it is great for large block reads a lot of that speed advantage won't be borne out with the current DOS because of the way LoadSeg() works – when doing LoadSeg() and resolving addresses, a lot of small block reads are done rather than loading a large block and passing back over the data after reading. The biggest advantage of the FFS comes when reading large contiguous blocks. Not to say it's a slouch, it will be lots faster than the current filesystem if only because the directory seeks are better organized, and when you do perform large block IO's, the performance is quite nice.
#94179From: Ben BlishNov 19, 1987 8:18 AM
Yeah — IF they ever release it. <grump!> –Ben–
#94209From: Charlie HeathNov 19, 1987 1:38 PM
They will release the FFS eventually, Commodore isn't very fast about getting out software releases but it is imminent. It still has some bugs now, so it's better for it not to get out to the general public yet. But the performance is much better – large block transfers with my 20 meg HD show 218453 bytes read, 154202 bytes written per second. Also it is much more intelligent with directories, so there isn't as much thrashing. Interestingly, this works very well with BlitzDisk so far, because the new FFS prefetches more directory blocks that BlitzDisk can preserve; and the large data block reads & writes get passed through automajikly since BlitzDisk only looks at 512 byte transfers. . BTW, are you in contact with CLTD? I've been waiting for several weeks to get my HD back from service… …cheath
#94325From: Ben BlishNov 20, 1987 12:48 AM
I'm in contact with them, and will see if I can't nudge them 4 u. The performance sounds good… we will see. –Ben–
#94350From: Charlie HeathNov 20, 1987 1:58 AM
Thanks – guess I should give 'em another call too.