keithh@atreus.uucp (Keith Hanlan) (11/28/89)
I'm putting together an A2000 machine... Arggh! I'm going to World of Commodore this weekend with the express purpose of getting a controller/memory/harddisk and would appreciate some informed opionions about: o DMA vs non-DMA I'm almost ready to give up. I continue to read conflicting optinions about DMA vs non-DMA hard-disk controllers. My gut opinion is that DMA is a more elegant solution that off-loads the CPU and hence doesn't degrade the machine as much. Am I right? Then why does anybody build non-DMA? Why do they claim such remarkable through-put with some non-DMA drives? And finally, what the heck is the "quasi" DMA and "true" DMA I occassionally hear about. o ST-506 vs SCSI vs ESDI These are just low-level protocols right? My perception is that ST-506 is what you find in PC's, SCSI is more popular with others, and ESDI is for file-servers etc... Does this mean that each is "tuned" to a different speed and throughput requirements? Does that mean I pay more for SCSI and even more for ESDI? Does that mean each will be progressively "faster"? o importance of access speed That's all a drive manufacturer will give you. Near as I can figure it doesn't matter a tinker's cuss what the access speed is within a few dozen milli-seconds: it seems to be the controller that really matters. o ways of measuring through-put I'm putting together a software development environment. I'm interested in a harddisk that will let my compiler really scoot. Are there somethings I should be more concerned with? What things don't matter as much? Are some controllers more appropriate for this kind of work than others? Dave Haynie (whose opinions I respect) explains that the "main trouble I've seen with the magazine reviews of HD controllers is that the reviews don't quite understand or bother to deal with a real, sound, scientifically accurate comparison. For such a test, you've got to benchmark each controler with exactly the same hard disk, ..., under a variety of system load conditions." I'll buy that! As far as my inexperience helps, every review uses a set of metrics more or less completely orthogonal for every other review. Yes, sure they may give some of the same measurements but if the conditions vary, then I don't know how to compensate. I am leaning to GVP for a couple of reasons: o My dealer supports them and speaks very highly of their support *but* I should be able to get support at least as good for CBM stuff I buy. o They have a good upgrade path to an '030/882 and 32bit RAM *but* maybe everybody else '030 boards are compatible with everybody else's drives, controllers, and RAM add-ons. o Their market share seems re-assuring *but* advertising is hardly a fair evaluation. Your opinions and information will be GREATLY appreciated. Thank you, Keith Hanlan Bell-Northern Research utgpu!bnr-vpa!bnr-fos!bmers58!keithh%atreus or keithh@bnr.ca
himacdonald@grand.waterloo.edu (Hamish Macdonald) (11/28/89)
In article <553@bmers58.UUCP> keithh@atreus.UUCP (Keith Hanlan) writes: >I'm putting together an A2000 machine... >... >o DMA vs non-DMA > I'm almost ready to give up. I continue to read conflicting > optinions about DMA vs non-DMA hard-disk controllers. My gut > opinion is that DMA is a more elegant solution that off-loads the > CPU and hence doesn't degrade the machine as much. Am I right? > Then why does anybody build non-DMA? Why do they claim such > remarkable through-put with some non-DMA drives? And finally, > what the heck is the "quasi" DMA and "true" DMA I occassionally > hear about. >... >Thank you, >Keith Hanlan >Bell-Northern Research >utgpu!bnr-vpa!bnr-fos!bmers58!keithh%atreus or keithh@bnr.ca In effect you are correct. They sell non-DMA controllers because people will buy them... and because they are cheaper to build/buy. A lot of the non-DMA controllers probably do get remarkable throughput, but the point is, you can't do anything else with your CPU while you are getting this remarkable throughput. A DMA controller lets your CPU do other stuff while reading the disk. Don't know about 'quasi' DMA. Hamish. ---------------------------------------------------------------- Hamish Macdonald. himacdonald@grand watmath!grand!himacdonald himacdonald@grand.uwaterloo.ca himacdonald@grand.waterloo.edu
esker@abaa.uucp (Lawrence Esker) (12/01/89)
>In article <553@bmers58.UUCP> keithh@atreus.UUCP (Keith Hanlan) writes: >>o DMA vs non-DMA >> I'm almost ready to give up. I continue to read conflicting >> optinions about DMA vs non-DMA hard-disk controllers. My gut >> opinion is that DMA is a more elegant solution that off-loads the >> CPU and hence doesn't degrade the machine as much. Am I right? Yes. >> Then why does anybody build non-DMA? Faster time to market in design. No benifit to the consumer (IMHO). A non-DMA board just looks like a memory board (if buffer ram) or an I/O port (if FIFO based) via the computer bus. >> Why do they claim such remarkable through-put with some non-DMA drives? Because the predominante speed factor with drives is the seek and access time. The difference in buffer passing is a small percentage of this, so the user doesn't usually notice. However, load down the system in a multitasking Amiga, and that small percentage turns into a large one. As I've mentioned before, a non-DMA slows down for ANY other tasks that may be running. A DMA device slows down only if that other tasks is using overscan hi-res graphics. (Although that no excuss for the A2090 problems.) In article <32033@watmath.waterloo.edu> himacdonald@grand.waterloo.edu (Hamish Macdonald) writes: >In effect you are correct. They sell non-DMA controllers because people >will buy them... and because they are cheaper to build/buy. Anytime I have compared prices, the non-DMA have been close enough to the same price as DMA to not make a difference. They may be cheaper to build, but not buy (ie. more profit). From a hardware designers's viewpoint, I don't believe much in the cheaper to build area either. They require so much more buffer ram and glue logic that its price overcomes the savings for a simple DMA chip. >> And finally, what the heck is the "quasi" DMA and "true" DMA I occassionally >> hear about. >>Keith Hanlan >Don't know about 'quasi' DMA. Quasi-DMA is a marketing buzzword I head from some non-DMA people (I think it GVP). The non-DMA boards have circuitry that directly dumps the data into the onboard buffer ram, which is what DMA is by definition. Its just that it is DMA into the buffer ram using simple glue logic instead of into the system memory by a complex DMA controller chip. >Hamish. -- ---------- Lawrence W. Esker ---------- Modern Amish: Thou shalt not need any \ * * * ******* / computer that is not IBM compatible. \ * * * * * / \ * * * * * ***** / Sr. Hardware/ASIC Design Engineer \ * * * * * * / Allen-Bradley Communications Div. \ ******* * * ******* / Work: (313)668-2500 Home: (313)973-8561 ----------------------------- Compuserve: 76337,2524 UseNet Path: __!mailrus!sharkey!itivax!abaa!esker == esker@abaa.UUCP
GORRIEDE%uregina1.bitnet%ee.udel.edu@cunyvm.cuny.edu (12/07/89)
I can understand you predicament well. The advertizing realy does confuse the whole matter of DMA vs Non-DMA. Supposed benchmarks confuse the matter even more. No member of this newsgroup has currently submitted enough information about this matter for anyone to gain a clear understanding of it. There has not been an overwhelming about of factual info about it at any rate. I've posted this twice before, but maybe this time there will be some response. How about a step by step explanation of DMA vs NON-DMA transfer, showing: 1)Under what conditions is there contention for chip RAM, for both DMA and Non-DMA. 2)Why is there contention for chip RAM? Why is access to chip RAM neccessary? 3)What are the solutions used in a209x, hardframe, and cltd controlers to get around the problem of chip ram contention? What are the resulting speeds (using same drive/mountlist)? 4) where are the device registers and buffers from DMA and NON-DMA devices? Where excactly are they located? What restricts them from being accessed on every cycle? What limits do they have in the size and speed of their transfers? +-----------------------------------------------------------------------+ |Dennis Gorrie 'Chain-Saw Tag... | |GORRIEDE AT UREGINA1.BITNET Try It, You'll Like It!'| +-----------------------------------------------------------------------+