reisert@ricks.enet.dec.com (Jim Reisert) (07/01/90)
Hi, I did some more cache measurements. These were all done on a 20 MHz 386 machine with a 42 KB hard disk running MS-DOS 4.01. I used QEMM as my EXPAnded memory manager. All tests were run at the DOS prompt (as opposed to under Windows or DESQview). This note includes all measurements made to date. The following tests were made using the PC-Labs Benchmark Tests, v5.0. The tests are as follows: TEST1 - 512 records of 512 bytes each TEST2 - 64 records of 4096 bytes each TEST3 - 16 records of 16384 bytes each TEST4 - 8 records of 32768 bytes each Each test had 5 stages: 1) file creation, 2) sequential write, 3) sequential read, 4) random write and 5) random read. The results are in seconds. These tests were made using a 384K cache in EXPAnded memory: TEST1 TEST2 TEST3 TEST4 no cache 58.50 8.86 5.44 3.99 secs. SMARTDRV 35.06 7.26 4.00 3.33 " PC-CACHE 5.5 30.29 6.70 3.73 3.32 " PC-CACHE 6.03 13.13 8.53 3.14 2.75 " SuperPCK 3.56 8.52 4.94 1.83 2.45 " SuperPCK 3.19 - not available - These tests were made using a 384K cache in EXTEnded memory: TEST1 TEST2 TEST3 TEST4 no cache 58.73 9.14 5.50 4.18 SMARTDRV 34.85 7.25 4.01 3.37 PC-CACHE 5.5 29.56 6.51 3.61 3.01 PC-CACHE 6.03 12.50 8.64 4.88 3.87 SuperPCK 3.56 11.07 5.21 3.56 2.36 SuperPCK 3.19 6.00 5.79 1.60 2.54 Here are some results from CORETEST v2.8. CORETEST writes 42KB of data and measures the access speed in KB/sec. All caches are 384K. EXTEnded EXPAnded no cache 537 KB/sec 537 KB/sec PC-CACHE 6.03 2153 " 5369 " SuperPCK 3.56 4193 " 3994 " SuperPCK 3.19 4609 " N/A " PC-CACHE 5.5 4816 " 4368 " SMARTDRV 5479 " 3994 " I don't know what to make of these measurements. These say that if you're using EXTEnded memory, SMARTDRV is the way to go, but if you're using EXPAnded, then use PC-CACHE 6.03. The PC-Labs Benchmark Tests clearly show Super PC-Kwik to be superior in all categories. Furthermore, for the PC-Labs tests, the results were similar for both EXPAnded and EXTEnded memory caches. CORETEST showed much more variation depending on the cache and type of memory. I merely offer this up for discussion, since I'm not sure what to make of it all. I'm going to stick with Super PC-Kwik, though. Anyone care to comment on any of this? jim =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= "The opinions expressed here in no way represent the views of Digital Equipment Corporation." James J. Reisert Internet: reisert@ricks.enet.dec.com Digital Equipment Corp. UUCP: ...decwrl!ricks.enet!reisert 77 Reed Road Hudson, MA 01749-2895
silver@xrtll.uucp (Hi Ho Silver) (07/01/90)
In <12988@shlump.nac.dec.com> reisert@ricks.enet.dec.com (Jim Reisert) writes:
[...]
$These tests were made using a 384K cache in EXPAnded memory:
$SMARTDRV 35.06 7.26 4.00 3.33 "
$PC-CACHE 6.03 13.13 8.53 3.14 2.75 "
$SuperPCK 3.56 8.52 4.94 1.83 2.45 "
[...]
$These tests were made using a 384K cache in EXTEnded memory:
$SMARTDRV 34.85 7.25 4.01 3.37
$PC-CACHE 5.5 29.56 6.51 3.61 3.01
$PC-CACHE 6.03 12.50 8.64 4.88 3.87
$SuperPCK 3.56 11.07 5.21 3.56 2.36
$SuperPCK 3.19 6.00 5.79 1.60 2.54
[CORETEST results deleted]
$I don't know what to make of these measurements. These say that if you're
$using EXTEnded memory, SMARTDRV is the way to go, but if you're using
$EXPAnded, then use PC-CACHE 6.03. The PC-Labs Benchmark Tests clearly show
It's apparent that you don't know what to make of these measurements.
In extended memory, SuperPCK (both versions) beat SMARTDRV in all cases,
so why do you say SMARTDRV is better? And in expanded memory, SuperPCK
beat PC-Cache in all cases, so why do you say PC-Cache is better?
$Super PC-Kwik to be superior in all categories. Furthermore, for the
Your tests also show this.
$I merely offer this up for discussion, since I'm not sure what to make of it
$all. I'm going to stick with Super PC-Kwik, though. Anyone care to comment
$on any of this?
Sure. Here are a few more things to consider:
- Using a utility such as CORETEST to measure the performance of a disk
cache is futile, since the activity it performs (repeatedly reading the
same information) is not a typical task in the real world and, furthermore,
is the situation in which disk caches excel the most.
- The relative performance of various caches will depend on your hardware.
For example, of the caches mentioned above, some may be aware when they
are running on a 386 while others may not. If using extended memory,
this will make a huge performance difference since the 386-aware caches
will realize they don't have to reset the processor to return to real
mode like they do on a 286. Other factors which vary the performance
in varying degrees include the speed of your memory and the data
transfer rate and seek time of your hard drive.
- Changing the size of your cache, as well as tweaking some of the
parameters which better caches allow you to do, will alter the
relative performance levels of the caches.
- One advantage which SMARTDRV has, if you're running Windows, is that
it will dynamically alter the amount of memory it's using in response
to Windows' memory requirements. This is about the only reason I can
think of why anyone would want to use SMARTDRV.
- Every user will benefit differently from a cache, since every user
performs a different mix of activity. As you can see from the
tables, some caches perform better on some tests and worse on others.
The best cache for someone who edits, compiles and tests programs may
or may not be the same as the best cache for someone who does heavy
database work, even on the same hardware.
--
/Nikebo \ Nikebo says "Nikebo knows how to post. Just do it."\silver@xrtll/
/---------\_____________________________________________________\----------/
/yunexus!xrtll!silver (L, not 1)\ Hi Ho Silver \ just silver for short /
/Silver: Ever Searching for SNTF \ Life sucks. \ someone buy me a BEER! /reisert@ricks.enet.dec.com (Jim Reisert) (07/01/90)
In article <1990Jul1.003125.18399@xrtll.uucp>, silver@xrtll.uucp (Hi Ho Silver) writes... >In <12988@shlump.nac.dec.com> reisert@ricks.enet.dec.com (Jim Reisert) writes: > >[CORETEST results deleted] > >$I don't know what to make of these measurements. These say that if you're >$using EXTEnded memory, SMARTDRV is the way to go, but if you're using >$EXPAnded, then use PC-CACHE 6.03. The PC-Labs Benchmark Tests clearly show > > It's apparent that you don't know what to make of these measurements. >In extended memory, SuperPCK (both versions) beat SMARTDRV in all cases, >so why do you say SMARTDRV is better? And in expanded memory, SuperPCK >beat PC-Cache in all cases, so why do you say PC-Cache is better? I knew what I wanted to say, I just didn't say it right. I meant these comments to refer to the results of CORETEST only. Of course it's clear that using the PC Labs benchmarks, Super PC-Kwik comes out on top. Sorry for any confusion this might have caused. The comments about CORETEST being a poor test are informative and noted. jim =-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-= "The opinions expressed here in no way represent the views of Digital Equipment Corporation." James J. Reisert Internet: reisert@ricks.enet.dec.com Digital Equipment Corp. UUCP: ...decwrl!ricks.enet!reisert 77 Reed Road Hudson, MA 01749-2895