thaler@shorty.CS.WISC.EDU (Maurice Thaler) (10/02/89)
I have been using a DELL 310 20Mhz 386 for about 13 months now. When I
first got it, I thought I was getting a real hot machine that would have
few limits. As I have used it for a while and tried to expand it, that
thought has soured. First off, I feel that Dell tech support is
generally "ok" but aimed at less experienced users and tries to act as a
filter when more exprerienced users try to get "higher level" info on
their systems.
My case in point is the bus. I do desktop publishing and have needs for
lots of disk space, and as I have discovered good disk i/o. Through the
last 13 months I have experimented with different disk controllers
trying to get a good combination of disk speed and space. The Dell bus
and BIOS has been a real problem however. The PERSTOR 16 bit controller
did not work with it until about last December. I tried the Adaptec RLL
controller and discovered that although friends of mine with 12mhz AT's
were getting 700K/sec (coretest) throughput without a cache, I could not
manage better than about 550K.
Finally I bit the bullet and sprang for a BIG FAST ESDI drive, a
MINISCRIBE 9380E, 307 Meg (aproximately), 17ms ESDI. I also tried one of
the Compuadd Caching controllers. If you don't know about this, it looks
to be a really nifty controller, it will allow you to install 1-4 SIMMS,
either 256K or 1Meg (you can't mix sizes of course. Well the DELL would
not even BOOT with this controller, I could get it to boot if I ran
through the SETUP in ROM and then hit F10 to continue, but the data
transfer was not good, only about 750K with 256K of cache installed. I
saw this same card do double that throughput on a Mylex 20Mhz 386. When
I asked Dell about this and if they would support any other caching ESDI
controller, their only response was "we don't support anything besides
our original equipment. You can hang whatever you want off of it, but we
won't help you". Well, I could run this new ESDI drive I have with a
Western Digital 1007a WAH (standard ESDI) card, but all I was getting,
was 566K/sec (650K with buffers set up higher), and this friend of mine
who builds clones in town was checking out DTC controllers from the QUME
company, so I tried that in my DELL. Well it seemed quite compatable,
great setup in the ROM of the controller, but the DELL just could not
handle 1:1 interleave with this sucker. I got 54K (thats right)/sec. I
set it to 1:2 interleave and got 495K/sec (gosh its fun doing low level
formats of 300+meg hard disks repeatedly :-) ). I took this hard disk
and controller over to my friend's 20Mhz Mylex 386 and did the 1:1
interleave and saw 990K/sec without buffers!! Well I called up Dell
again (they must just love me) and begged the tech support guy to please
explain why my computer performs so much worse that a clone worth less
than 1/2 the price. I asked about the bus speed. Is it really 8mhz like
they said or perhaps maybe it is running a little slow? They just stone
walled me. I got him to find the chip that controls the bus speed, the
82c301, but I seriously doubt that the situation can be improved. So, It
looks like I am going to sell my Dell, and switch to some cheap clone
with twice the performance. Since I am a consultant for desktop
publishing, I will be forced to recommend to my clients that they should
not buy Dell if they are interested it high-performance i/o. In my
opinion, once you get up to about 20mhz with 386's, if you don't have
disk access to match, it is just a waste of juice, unless you are just
doing spreadsheets all day long. If you are doing database or desktop
publishing applications, the i/o is CRITICAL.
Any opinions on this. Am I being unfair to DELL. Is there something
obvious I am not doing? Oh, by the way, when I did my hard disk tests,
I pulled ALL my cards out of my machine except the Paradise VGA-Pro.
Maurice Thaler SYSOP Audio Projects BBS (608) 836-9473
SYSOP Power Board BBS (608) 222-8842 Jim.Birken@bigtime.fidonet.org (Jim Birken) (10/04/89)
I also use a DELL 310. I have liked it a lot, but then I haven't
attempted your i/o struggles. My question to you is two fold. Did
DELL have the disks available that met your requirements, at any cost?
I use mine for spreadsheet work, primarily. Do you have a problem with
one as a file server?
--
Jim Birken
Domain: Jim.Birken@bigtime.fidonet.org
UUCP: ...!{tektronix, hplabs!hp-pcd}!orstcs!bigtime!Jim.Birken
via Big Time Television (bigtime.fidonet.org, 1:152/201)jrv@demon.siemens.com (James R Vallino) (10/05/89)
In article <1090@cbnewsj.ATT.COM> jwi@cbnewsj.ATT.COM (Jim Winer @ AT&T, Middletown, NJ) writes: >DELL builds some of the fastest, most reliable machines on the market. >The key here is *most reliable.* DELL sells to the corporate market >and to the small business market. They usually have the best overall >performance for a *stock* machine. They do not have the best price/ >performance ratio. > >If your business depends on the machine working >absolutely perfectly day-after-day-after-day, by all means get a DELL. There some folks here who would definitely differ with that opinion of Dell machines. We have about 30 Dell machines some dating back to when the company was called PC's Limited. Of these 30 there are 3 or 4 machines which have been continual problems. One 310 had just about everything changed out in the course of its one year warranty. When Dell sent the renewal form for the service contract the user of the machine called Dell instead and complained. He asked for a new machine. Our account representative checked the repair record then called back later that day. She said that the machine sure was a dog and a new one was on the way. Another 310 had about 3 service calls on it in the first few months. In the last month we have had the hard disks on 4 12 MHz 286 boxes give up the ghost. These machines are all about 2 years old so we may be getting to the life expectancy of the disk. It was rather frustrating for them all to go in such a short period of time though. The one thing which Dell is real good about is customer service. Our account representative responds when we have any problems. The service contract people also do a good job. That's good because on about 10 to 15% of the machines we sure needed them. Our policy now is still to buy Dell's but if a machine has more than 2 service calls we will probably ask for a new one to be sent to us. They have also gotten their act together about delivering machines according to the first date specified. We did have an instance when the two machines which would be shipped in 2 weeks finally arrived 4 months later. Jim Vallino Siemens Corporate Research, Princeton, NJ jrv@demon.siemens.com princeton!siemens!demon!jrv (609) 734-3331
thaler@shorty.CS.WISC.EDU (Maurice Thaler) (10/05/89)
In article <1075.2529AB21@bigtime.fidonet.org> Jim.Birken@bigtime.fidonet.org (Jim Birken) writes: >I also use a DELL 310. I have liked it a lot, but then I haven't >attempted your i/o struggles. My question to you is two fold. Did >DELL have the disks available that met your requirements, at any cost? >I use mine for spreadsheet work, primarily. Do you have a problem with >one as a file server? > >-- >Jim Birken >Domain: Jim.Birken@bigtime.fidonet.org >UUCP: ...!{tektronix, hplabs!hp-pcd}!orstcs!bigtime!Jim.Birken >via Big Time Television (bigtime.fidonet.org, 1:152/201) The problem is not the disks per se. The problem is in the bus of the DELL310. The hard disk performance is "throttled" by the bus performance. No matter how fast a hard disk is, you are limited by the speed the bus will accept your data. I have never seen the DELL310 get better than about 675K/sec with coretest. I have tried quite a few different controller/hard disk combinations. I consider this to be a major limitation. If you are doing spreadsheet work, you will never notice the difference. I had problems with the serial I/O port as well. I eventually solved that one by installing a seperate serial card with one of those national semiconductor UARTS (16550an I think) that chuck forsberg recommends in DSZ documentation. I use high speed modem communication at >9600 baud, and had charactors dropping like flies before I did this mod. Once again, I think that generally, the DELL310 is a sound machine, but if you are into high performance (which it was advertised as having when I bought it), you would be well advised to test the DELL310 before plunking down $6K for a "full system". I have a feeling that the same problems might be encountered with the DELL 325 as well, but who knows. Maurice Thaler SYSOP Audio Projects BBS (608) 836-9473 SYSOP Power Board BBS (608) 222-8842
rob@prism.TMC.COM (10/05/89)
Jumping into the fray... A question - have you actually compared the Dell 310 with other machines at doing 'real world' tasks, or just on hardware benchmarks? I have a Dell 325 (a 25Mhz version of the 310, basically), and on disk bound tasks, I have yet to find a machine that can beat it. The disk's low level measurements are good-but-not-great in some areas (transfer rate of 550K/sec) and excellent in others (avg. seek time of 14ms), but the overall performance is outstanding. When a cache is thrown in (which I'd think anyone doing disk intensive work would want), the measurements naturally change (yielding an impressive, though meaningless, transfer rate of 4MB/sec), but it still outperforms other machines by a fair margin. I wouldn't hesitate to recommend it to someone wanting a high-performance machine.
Jim.Birken@bigtime.fidonet.org (Jim Birken) (10/08/89)
The question re: Dell 310 (or 325 for that matter) was not for the
desktop machine of mine that I now use, but a recommendation for a new
file server for the network I am installing in a couple of months. It
will need a fairly big disk drive. I like my Dell now, and would have
recommended it for the file server. You seem to be saying that I can
do much better elsewhere. Where, with what if not the very reliable
Dell?
--
Jim Birken
Domain: Jim.Birken@bigtime.fidonet.org
UUCP: ...!{tektronix, hplabs!hp-pcd}!orstcs!bigtime!Jim.Birken
via Big Time Television (bigtime.fidonet.org, 1:152/201)thaler@shorty.CS.WISC.EDU (Maurice Thaler) (10/10/89)
In article <206900128@prism> rob@prism.TMC.COM writes: > > Jumping into the fray... > > A question - have you actually compared the Dell 310 with other machines >at doing 'real world' tasks, or just on hardware benchmarks? I have a Dell >325 (a 25Mhz version of the 310, basically), and on disk bound tasks, I have >yet to find a machine that can beat it. The disk's low level measurements >are good-but-not-great in some areas (transfer rate of 550K/sec) and >excellent in others (avg. seek time of 14ms), but the overall performance >is outstanding. > > When a cache is thrown in (which I'd think anyone doing disk intensive >work would want), the measurements naturally change (yielding an impressive, >though meaningless, transfer rate of 4MB/sec), but it still outperforms >other machines by a fair margin. I wouldn't hesitate to recommend it to >someone wanting a high-performance machine. I did do some "real world" testing. The bottom line is that the data transfer rate is critical to great performance. I would venture to say that the 20Mhz mylex 386 that I tested on would have a "snappier" feel than your 25mhz because of disk i/o. What first brought this to my attention was when I was visiting a friend's office and I saw him boot up Ventura (a program I use all day long) on his 16Mhz 286 and it came up much faster than on my 20Mhz Dell. The difference between our machines really came down to disk i/o. The DTC controller WITHOUT A CACHE does 990K/sec. which is almost double the rate of the DELL w/ a Western Digital controller. Now if you think about it, a cache will never do you any good untill you have loaded the data once. The bottem line is that you will never do better than the data transfer rate for a first read. Now don't get me wrong, I believe in using a cache, I generally use smartdrv.sys which I have found to be the least prone to conflicts with other software (for example, SUPERPCK is not great with high speed communications), and on my next machine I am going to get an even larger memory configuration so I can use a larger cache, but I do feel that the bus limiting my i/o is extremely frustrating. I am basically saying that the DELL is a fine machine if you expect fast computing, but mediocre i/o. The Dell is generally reliable, uses reasonably generic parts (like the WD controller) and is not a machine to be ashamed of, but they have risen to the higher levels in prestige and gotten a little more like IBM in not supporting equipment they don't supply. If you think the DELL325's disk performance is the best available, it is simply from not having tested the waters lately :-). Maurice Thaler SYSOP Audio Projects BBS (608) 836-9473 SYSOP Power Board BBS (608) 222-8842
rob@prism.TMC.COM (10/10/89)
I suppose the lesson here is that even within a fairly specific area like 'disk performance' there is lots of fuzziness in speed measurements. Take two disks, one with fast access and slow transfer, the other with slow access and fast transfer. Which will be faster? The disk in the Dell 325, which has a very fast (14ms) access time but only a moderately fast transfer rate (550Ksec to 800K/sec, depending on which benchmark you use), does well in database applications, program compiles, and other applications which tend to use disk data repetitively. A disk with a higher transfer rate, even if its access time were slower, would probably do better on loading large programs or data files. As always, the choice comes down to the task in question. On the question of which machines I had compared it to, I've tested it against 25Mhz machines from Compaq, IBM, and ALR. Running database applications, it's faster than any of them by a comfortable margin. Interestingly, the disk in the Compaq had performance characteristics like those mentioned above - slower (19ms) access, but higher transfer (1Mb/sec). Though the Dell won in the benchmarks I was doing, there would undoubtedly be some tasks that would favor the Compaq. Enough of this nonsense. I want a 100Mb RAM-disk. :-)
wek@point.UUCP (Bill Kuykendall) (10/13/89)
>company, so I tried that in my DELL. Well it seemed quite compatable, >great setup in the ROM of the controller, but the DELL just could not >handle 1:1 interleave with this sucker. I got 54K (thats right)/sec. I >set it to 1:2 interleave and got 495K/sec (gosh its fun doing low level >formats of 300+meg hard disks repeatedly :-) ). I took this hard disk >and controller over to my friend's 20Mhz Mylex 386 and did the 1:1 >interleave and saw 990K/sec without buffers!! Well I called up Dell I have 30 or 40 Dell 310s and a *bunch* of older PC's LTD 286 machines. The oldest stuff from PC's LTD was a little flakey, but I have no complaints about their pcs built in the last 2 years or so. They are fast and reliable. The key to the discussion above is probably the Mylex board's bus speed, not the Dell's. Many clone makers run the bus at the clock speed of the CPU. Dell limits the bus speed to 8mz, to ensure compatibility with add-in cards. The chances are pretty good that if you have a high end DTP monitor and video card, and perhaps a LaserMaster card or flatbed scanner attached to your system, that one or all of them will choke on the bus speed of the cheap clone. Or worse still, they might seem to work just fine except on alternate Thursday evenings when you're working late to meet a deadline. It isn't fair to judge a machine on it's performance with one high performance peripheral. Drive controllers are designed to handle faster buses so that they can be used in unix boxes that really *are* io intensive and probably don't have a lot of other exotic peripherals. Before you send the Dell back, I'd suggest plugging your entire configuration into the Mylex clone and see if it all works. If that's OK, then ask yourself if you're ever going to add anything new to the system. If your answer is no, then go for the gusto, and build your own. ---- Bill K. wek@point.UUCP
karl@ddsw1.MCS.COM (Karl Denninger) (10/13/89)
>----- >Response 6 of 6 (5977) by wek at point.UUCP on Thu 12 Oct 89 09:33 >[Bill Kuykendall] >(<Unknown> lines) > >>company, so I tried that in my DELL. Well it seemed quite compatable, >>great setup in the ROM of the controller, but the DELL just could not >>handle 1:1 interleave with this sucker. I got 54K (thats right)/sec. I >>set it to 1:2 interleave and got 495K/sec (gosh its fun doing low level >>formats of 300+meg hard disks repeatedly :-) ). I took this hard disk >>and controller over to my friend's 20Mhz Mylex 386 and did the 1:1 >>interleave and saw 990K/sec without buffers!! Well I called up Dell > >The key to the discussion above is probably the Mylex board's bus speed, not >the Dell's. Many clone makers run the bus at the clock speed of the CPU. >Dell limits the bus speed to 8mz, to ensure compatibility with add-in cards. .... >It isn't fair to judge a machine on it's performance with one high >performance peripheral. Drive controllers are designed to handle faster >buses so that they can be used in unix boxes that really *are* io intensive >and probably don't have a lot of other exotic peripherals. Sure it is! The best systems, from any manufacturer, allow you to >choose< your bus wait states and the like. Our machines have this option in their CMOS setup, and we (and our customers) use it to great advantage. You simply set up for the fastest speed which your cards are rated for, or if you feel lucky crank it wide open and then turn it down as problems appear. What makes this work well is that the bus settings are individual by type -- 8 bit, 16 bit, and 32 bit slots all have their own, independant wait state configuration. This is a wonderful feature, and makes for >excellent< performance. We don't sell systems missing it (at least on the '386 side of things; I haven't seen 80286 machines with this option yet). Most C&T chipset machines can do this, IF you can figure out what to diddle. The AMI BIOS has it built into the "advanced" setup pages. Our boards honor the "0 wait state" line on the 16-bit bus, which tells the system to ignore waits on that transfer. Fast peripherals can yank on that to get a 0 wait state transfer regardless of the CMOS settings. -- Karl Denninger (karl@ddsw1.MCS.COM, <well-connected>!ddsw1!karl) Public Access Data Line: [+1 312 566-8911], Voice: [+1 312 566-8910] Macro Computer Solutions, Inc. "Quality Solutions at a Fair Price"
davidsen@crdos1.crd.ge.COM (Wm E Davidsen Jr) (10/16/89)
In article <[2406.5]comp.ibmpc;1@point.UUCP>, wek@point.UUCP (Bill Kuykendall) writes: | The key to the discussion above is probably the Mylex board's bus speed, not | the Dell's. Many clone makers run the bus at the clock speed of the CPU. | Dell limits the bus speed to 8mz, to ensure compatibility with add-in cards. I agree that the problem being addressed is probably the wrong one. Someone in a non-business environment can look at really high bus speeds, and live with the possibility of failuer of some boards. A machine designed for reliability and targeted at the business market is more likely to have an 8MHz bus. I am not sure what's holding back the transfer rate in the original poster's machine, but let's look at what isn't... namely the bus. An RLL controller has 26 sectors, or 512 bytes, rotating at 60 rps, for a maximum transfer rate (1:1) of 798720 bytes/sec. At 6 cyc/byte that takes a bus speed of 4.8MHz, at 8 cyc/byte 6.4MHz. You say ESDI? Okay, 32 sectors times 512 bytes times 60 rev/sec is 983040 bytes/sec max at 1:1. Assuming 8 cyc/byte that takes a bus speed of 7.9 MHz. But wait! All those assume a 8 bit controller! Cut the required bus speed in half for anything made for the AT bus. Sure doesn't look as if the bus is a problem, unless someone has the system configured for more than 8 wait states! Here's the point: the AT bus is plenty fast enough to keep up with the transfer rates of any hard disk. If there's a problem it's in the controller or the setup. Figuring 8 cyc/transfer, the AT bus will do 2MB/s transfer, and that is more than the sustained one revolution rate of any popular disk type. Systems like Mylex, which allow the controller to sit on a 32 bit bus are a good implementation. While the AT bus will keep up with popular disks, there is nothing wrong with saving cycles, particularly on non-DOS operating systems which don't sit and wiat for the disk i/o to end before running a process. This allows the AT bus to run at a nice reliable rate, adequate for low speed peripherals, such as serial and printers. -- bill davidsen (davidsen@crdos1.crd.GE.COM -or- uunet!crdgw1!crdos1!davidsen) "The world is filled with fools. They blindly follow their so-called 'reason' in the face of the church and common sense. Any fool can see that the world is flat!" - anon
thaler@shorty.CS.WISC.EDU (Maurice Thaler) (10/25/89)
Well, after posting my mild "Dell Flame", I got a message from a Dell
employee who is on the net. He passed my message along to the powers
that be and I was invited to send in my motherboard for an upgrade!
I was told that some PAL chips (I think) were telling the BUS to slow
down so certain old fashioned hard disk controllers would not bog down
on the machine.
Well, after getting the machine back, I inserted the WD 1007 WAH with a
Miniscribe 150Meg ESDI drive (16.5ms) and did a low level format. Ran
coretest and got 675K/sec. Better but not outrageous. Inserted the
DTC (Qume) ESDI controller with the same drive, did a low level format
at 1:1 (just like the WD controller) and formated and ran coretest, this
time I get 54K/sec which is what you can expect from a controller that
is feeding data faster than the bus can handle it. As I understand it,
the it is taking multiple passes to get the data through from one track.
Last time I went through the hassle of reformatting at 1:2 and got about
500K/sec, but what is the point after all, since I know I can get better
from the WD controller which is just plain slower and does not overrun
the bus.
I just bought a 33Mhz 386 w/ Award BIOS, board made by AMC. It has a
switchable BUS speed. I have found that with the DTC controller I get
990K/sec at either 8Mhz or 10Mhz bus speed.
Now I don't know what the story is, but the fact is that my girlfriend
has a 386sx (16Mhz) clone with a DTC RLL card and a Japanese (perhaps
Mitzubishi or something) ST251 look alike (gets 65Meg) that does a
1:1 interleave. She is getting 675/Sec from that setup. Is it
unreasonable of me to expect that with a very expensive ESDI drive on a
faster, more expensive Dell 310 20Mhz machine I should get better
performance?
At any rate, I think that the Dell people were more than fair letting me
swap motherboards to try to improve performance. Too bad the Dell just
can't keep up. Oh well.
This is not a difinitive statement about Dell hardware. The fact is that
I am not savy enough about hardware to know just where the problem is,
but I do know that at this point, it appears that it would be a waste of
time and money to try and get a high ESDI performance via the use of
souped up cacheing controllers. I seriously doubt that the DPT
controller would work in the Dell. I do know that the DTC and Compuadd
controllers don't work. If anybody discovers a work around, let me
know.
Maurice Thaler SYSOP Audio Projects BBS (608) 836-9473
SYSOP Power Board BBS (608) 222-8842