don@chopin.udel.edu (Donald R Lloyd) (03/14/91)
Two questions:
I've run seceral benchmark programs which tell me that the speed of
the CHIP ram in my A3000/25/50 is running at only 85% or so of what it
should be. I popped open the case to take a look at the chips in there,
and found that the ones soldered on were 100ns chips! (The ones I moved
over from the fast RAM section were 80ns, though). Seems to me that
100ns CHIP RAM on a 3000 would cause a bit of a bottleneck... why does my
machine have these slower chips (apparently some 3000's have 80ns CHIP RAM,
and the benchmarks I've run are based on that...)
Also, I've heard claims of measuring 1.2 mbytes/sec with DiskSpeed.
With what settings was this achieved? On my 2M chip 2M fast system I've yet to
get anything that high.
--
Gibberish May the Publications Editor, AmigaNetwork
is spoken fork() be Contact don@brahms.udel.edu for more information.
here. with you. DISCLAIMER: It's all YOUR fault.jap@convex.cl.msu.edu (Joe Porkka) (03/15/91)
The speed rating of the RAM chips has nothing to do with how fast the computer accesses them. The computers design decides how fast it will access memory, so the chips you put in need to be at least that fast. Why is your chip mem slow? You probly have either a deep screen, say 8 or 16 colors, or an overscanned screen. Set you display to be 640x400 4 colors, then run your tests. Why can't you get >1Meg/sec xfer rates? Your hard disk isn't expensive enough :-)
vrr@cbnewsj.att.com (veenu.r.rashid) (03/15/91)
In article <1991Mar15.000705.23258@msuinfo.cl.msu.edu> jap@convex.cl.msu.edu (Joe Porkka) writes: > >The speed rating of the RAM chips has nothing to do with how >fast the computer accesses them. The computers design decides >how fast it will access memory, so the chips you put in need to >be at least that fast. > >Why is your chip mem slow? You probly have either a deep screen, say >8 or 16 colors, or an overscanned screen. Set you display to be >640x400 4 colors, then run your tests. > >Why can't you get >1Meg/sec xfer rates? Your hard disk isn't expensive >enough :-) I have a Quantum LP52S and a Quantum 105S and I still can't get rates that high. I get rates of ~650k/s on the LP52 and about 400k/s on the 105. I have tried everything. The drives are formatted ffs. I set the mask using advanced options in HDToolBox for the 105S. It was originally set to '0x7ffffffe' by the auto read parameters when I formatted it. I changed it to '0x7ffffffc' to be the same as the parameters on the 52S which came with my drive. This made access a bit faster, but not much. What could the problem be? Is it the format or the filesystem or the mask or something that I still haven't considered? Please email any suggestion, tips, etc. to vrr@cbnewsj.att.com. For those of you with the original LP52S in the A3000, what are your transfer rates (use DPerf2 or Diskspeed) Thanks very much, Veenu vrr@cbnewsj.att.com
mwm@pa.dec.com (Mike (My Watch Has Windows) Meyer) (03/16/91)
FIn article <1991Mar15.174212.22988@javelin.es.com> blgardne@javelin.es.com (Blaine Gardner) writes:
Fujitsu 1.2 GB drive = 1.9 MB/sec
Hewlitt-Packard 1 GB drive = 1.5 MB/sec.
With the stock Quantum 105 I was getting a peak of 800-900 KB/sec, and
with the Quantum 170 I replaced the 105 with I get 1 MB/sec, and the CDC
(aka Imprimis, now aka Seagate) Wren III gets 975 KB/sec.
Just out of curiosity, how big are those partition, and where they
newly formatted? Going from a small, empty partition to a large, full
one I see a 30% drop in speed.
<mike
--
Must have walked those streets for hours, Mike Meyer
In the dark and in the cold, mwm@pa.dec.com
Before I really could accept, decwrl!mwm
There's no place called hope road.blgardne@javelin.es.com (Blaine Gardner) (03/16/91)
don@chopin.udel.edu (Donald R Lloyd) writes: > Also, I've heard claims of measuring 1.2 mbytes/sec with DiskSpeed. >With what settings was this achieved? On my 2M chip 2M fast system I've yet to >get anything that high. I've measured 1.5 megabytes/sec and 1.9 megabytes/sec, both with DiskSpeed, and the Intensity and Stress options hardly slowed it down at all. The settings used were: A3000/25 2M Chip, 4M Fast. Oh, and this probably made some difference too: :-) Fujitsu 1.2 GB drive = 1.9 MB/sec Hewlitt-Packard 1 GB drive = 1.5 MB/sec. With the stock Quantum 105 I was getting a peak of 800-900 KB/sec, and with the Quantum 170 I replaced the 105 with I get 1 MB/sec, and the CDC (aka Imprimis, now aka Seagate) Wren III gets 975 KB/sec. What I've seen from testing a dozen different drives on the A3000, the limiting factor is always the hard drive. The A3000's SCSI controller is far faster than any hard drive you can buy, but it won't work miracles on a slow drive either. (I tried a Teac 40 meg, designed for the laptop market, that peaked at 250KB/sec on the A3000). -- Blaine Gardner @ Evans & Sutherland 580 Arapeen Drive, SLC, Utah 84108 blgardne@javelin.sim.es.com or ...dsd.es.com!javelin!blgardne DoD #0046 My other motorcycle is a Quadracer. BIX: blaine_g Anticipation, anticipation, is making me late, is keeping me waiting.
sschaem@starnet.uucp (Stephan Schaem) (03/17/91)
The slow done is not cause by how full is it or not, but Head move. The head move is cause because the free block are scatered. Nothing more nothing less, you can make test look really bad if anyone one wanted to put drive down that way!
blgardne@javelin.es.com (Blaine Gardner) (03/17/91)
mwm@pa.dec.com (Mike (My Watch Has Windows) Meyer) writes: >FIn article <1991Mar15.174212.22988@javelin.es.com> blgardne@javelin.es.com (Blaine Gardner) writes: > Fujitsu 1.2 GB drive = 1.9 MB/sec > Hewlitt-Packard 1 GB drive = 1.5 MB/sec. > With the stock Quantum 105 I was getting a peak of 800-900 KB/sec, and > with the Quantum 170 I replaced the 105 with I get 1 MB/sec, and the CDC > (aka Imprimis, now aka Seagate) Wren III gets 975 KB/sec. >Just out of curiosity, how big are those partition, and where they >newly formatted? Going from a small, empty partition to a large, full >one I see a 30% drop in speed. On the Fujitsu and H-P they were freshly formatted 512 meg partitions (the largest possible under 2.02 due to a bug that has since been fixed). The Quantum and Wren were on freshly formatted partitions of between 70 and 150 meg. I haven't done any empty vs. full comparisons on the A3000. But with an A2000/A2620/Hardframe/Wren III combination I measured about a 50% decrease in speed between a freshly formatted partition, and one about 60-70% full and moderatly fragmented. -- Blaine Gardner @ Evans & Sutherland 580 Arapeen Drive, SLC, Utah 84108 blgardne@javelin.sim.es.com or ...dsd.es.com!javelin!blgardne DoD #0046 My other motorcycle is a Quadracer. BIX: blaine_g Anticipation, anticipation, is making me late, is keeping me waiting.
daveh@cbmvax.commodore.com (Dave Haynie) (03/20/91)
In article <16647@chopin.udel.edu> don@chopin.udel.edu (Donald R Lloyd) writes: > I've run seceral benchmark programs which tell me that the speed of >the CHIP ram in my A3000/25/50 is running at only 85% or so of what it >should be. How do you know what it should be? >I popped open the case to take a look at the chips in there, and found that >the ones soldered on were 100ns chips! (The ones I moved over from the fast >RAM section were 80ns, though). Seems to me that 100ns CHIP RAM on a 3000 >would cause a bit of a bottleneck... I guess you don't understand how computer memory works. The memory rating of any chip is a _potential_. It has nothing to do with how fast a memory system work, only whether or not it will work. The memory controller, in this case the Agnus chip, is responsible for driving the DRAM chip at an appropriate speed. An analogy would be found in your stereo system. You have speakers rated for a certain wattage. This is your DRAM. And you have an amplifier. This is your DRAM controller. If you crank more watts out of the amp than the speakers are rated for, they may blow. If your DRAM controller tries to run DRAM faster than its rated for, you may get memory failures. On the other hand, you may have a 50W speaker and only a 40W amp -- that works fine. Similarly, you may have a 100ns DRAM in a system that only need 120ns parts; that generally works fine too, and so would an 80ns part, but the memory doesn't go any faster than a 120ns part in that system. In the case of your Amiga, all Agnus chips run their memories at the same speed, which is a 280ns cycle time. Typically, a DRAM's minimum cycle time is a little less than twice its access time (the number stamped on it). So an Agnus should work fine with a 100ns part, which can cycle in 200ns or less, though it should also work OK with a 120ns part, which cycles in 240ns or less. That's a guideline, not a hard rule -- other features of a memory system may cause it to require faster parts than you'd expect. But you can rarely succeed using slower parts. You would know if your chip memory wasn't fast enough, since the computer would not work in such a case. Microprocessors may tolerate wait states, Agnus does not. -- Dave Haynie Commodore-Amiga (Amiga 3000) "The Crew That Never Rests" {uunet|pyramid|rutgers}!cbmvax!daveh PLINK: hazy BIX: hazy "What works for me might work for you" -Jimmy Buffett