• Please review our updated Terms and Rules here

Faster writes than reads

E

evildragon

Veteran Member
Joined
May 29, 2007
Messages
1,646
Location
Tampa Florida
I'm confused, I always thought reading was faster than writing. In the case of my model 25, it's the other way around.

For the external parallel port, I can easily understand, it's running in nibble mode, so "outputting" would be faster, but the internal MFM 20MB drive is not really pushing much better..

I sent a file over the network via FTP, and these were the max results in speed I found.

MFM
Max Write: 56.9KB/S
Max Read: 46.0KB/S

LPT Uni
Max Write: 26.2KB/S
Max Read: 20.4KB/S

Eh, I'll see what happens when my bi-directional parallel port arrives, but I have a feeling it's not going to be making much of a difference..

The parallel port drive is a 3GB IDE HD..
 
You might be running intoa network bottleneck rather than a disk drive bottleneck...that does seem really unusual. Have you tried copying parallel -> internal?

evildragon said:
The parallel port drive is a 3GB IDE HD..
Click to expand...

Slightly OT, but what parallel case are you using? The case I've got won't work on anything less than a 286.
 
Mine requires a 286 too, but the V30 makes it work..

Definitely not a network bottleneck. I have a network that is entirely gigabit, other than this IBM's 10BaseT, but nothing is communicating, and other devices will share at break neck speeds..

I also ran Eurosofts PC-Check XT, and got similar results too..
 
Don't use FTP as your source of data. I know on some configurations that receiving is 25 to 30% faster than sending. Use a dedicated hard drive benchmark program instead.


Mike
 
evildragon said:
I have a network that is entirely gigabit, other than this IBM's 10BaseT, but nothing is communicating, and other devices will share at break neck speeds...
Click to expand...

My network is also gigabit and yet my Windows 98 machine had some very slow speeds doing anything with TCP/IP, it turned out that my switch was having issues with the 10mbit connection. I switched the card out with a 100mbit NIC and it works great now.

So don't rule out a network issue just yet. Some gigabit switches just can't deal wit a 10mbit client properly.
 
Yes, the above two posts mention what I was really referring to...the cost of running a TCP/IP stack on an 8086 (or V30 in your case), or possibly even the packet driver. I suppose the best test would be from a RAM disk on something like a SixPak Premium, that can allocate extra onboard RAM as a RAM disk and leave conventional memory alone.
 
Ok, well I'll need a pointer for a benchmark utility. Problem is, PC Check never finishes on the MFM drive, it stops around when it hits 50KBps with a controller error.. Syschk does the test completely, but it's numbers are ENTIRELY different, even when PC Check succeeds on a test on the parallel port. So I don't know which one to believe then.
 
Ok, I just HAD to take a picture of this to prove it, my read speeds really ARE slower, no matter what drive I use or testing app.

Internal MFM
mfm.jpg


External Parallel Port
parallel.jpg


Perhaps I can just pay someone to make a XT-IDE for me? ;) I surely don't have an EEPROM programmer, or the will to get the parts myself. Cause at this point, that is sounding like the best idea in my case..
 
The read and write speed for the internal MFM drive are close enough where I would call them "the same."

The read and write speed for the parallel port drive make perfect sense to me. Parallel ports running in "standard" mode (not bi-directional, EPP or ECP) are slower to read than they are to write because they can only read 4 or 5 bits at a time using the status lines, instead of the full 8 bits that a bi-directional or other type of port can do. It's just the nature of the beast - when you get a bi-directional port you will see the read speed come up.

The XT-IDE project does not require you to have an EPROM burner. The chips come with a BIOS on them already, and they are flashable right from the host machine. You just need to be able to do some light soldering.
 
Usually if you are having MFM performance issues like that, it's because the low-level format is using the wrong interleave. I'd use spinrite to get the best interleave for it. Interleave usually doesn't effect write speed more than 20%, but the wrong interleave can slow down reads by as much as 500%.
 
Spinrite says i should use interleave 2 and that the current interleave is 2.

I didnt know xtide already came with a flashed bios, and i can definitely solder.
 
evildragon said:
I didnt know xtide already came with a flashed bios, and i can definitely solder.
Click to expand...

You're most likely better off assembling one on your own then -- certainly cheaper, and assembling projects is fun! Also, the XT-IDE can flash the EEPROM itself, so it is its own EEPROM programmer (unless you do ChuckG's modification...I was unable to get either of my modified XT-IDEs to burn EEPROMs after the mod)
 
I agree with that - but the card can be flashed before modding with the _s version of the modified BIOS using Aitot's utility, then take it out and perform the line switching mod. This is definitely worth it, it literally doubles the throughput on an XT.
 
I'm going to actually wait for the revision 2, which seems that it will include the model 25 power hookup.

What I would like though, is a "kit". Where you get the PCB, pre-flashed BIOS, and all the parts, but you do the soldering, that way I don't have to hunt all the parts down.. I work graveyard for IT right now (and will be working at Apple next week, same hours), and really don't have the time to search for the parts.
 
They do, but it's still a hassle, to order from two locations. They should just offer a kit in my opinion. Cause what if a part is sold out? ;)

I mean, I'm not trying to be hard here, but, when I come home at night, and wakeup in the morning, I want a simple order, not a part by part add to shopping cart. Life for some of us just doesn't give us that kind of time.
 
You must log in or register to reply here.
Back
Top