Benchmarks done by Tai Morris

During a recent discussion in the USENET newsgroup comp.sys.mac.hardware.storage Markus Trenkle (wrongly) suggested that the Quadra 900's external bus is slower that the internal bus, to which reader Tai Morris posted the following, which includes benchmarks he performed.
The answer as posted by Markus Trenkle is here. 


From: tai@physics.utexas.edu (Tai Morris)
Newsgroups: comp.sys.mac.hardware.storage
Subject: Re: FWB HDT driver better than Silverlining and PowerTools?
Date: Sun, 12 Jan 1997 17:51:32 -0600
Organization: The University of Texas at Austin, Austin, Texas
Mime-Version: 1.0
Content-Type: text/plain
Content-Transfer-Encoding: 8bit
Message-ID: <tai-ya02408000R1201971751320001@newshost.cc.utexas.edu>
References: <tai-ya02408000R0901971555470001@newshost.cc.utexas.edu> 
            <5b4tkr$rd9@news.euro.net> 
            <jones.948-ya02408000R1001971405280001@news-stand.acs.ohio-state.edu> 
            <5b694r$gfg@news.euro.net> <32D935C3.1610@krist.uni-erlangen.de>

<You made one error in your tests: You used the external bus of the
<Quadra 900, which is a plain SCSI-2 bus and has a max. speed of 5MB/s
<(I found that Apple normally uses SCSI interfaces that max out at about
<3MB/s, but I don't know about the Quadra 900). The drive you tested
<should have speeds of about 7MB/s. So for me it's pretty clear that 
<the limiting factor in your test is the SCSI bus, not the drive or
<driver.
<
<I did nearly the same test yesterday on the internal Fast-SCSI of a 
<7600 with a similar drive (IBM DFHS1080, same speed, lower capacity),
<and I found that the speed of the drive doubled(!) from 3.5MB/s to
<7MB/s when I changed the driver from Apple to FWB. I have not tried 
<other possibilities, such as Silverlining etc. and I will not do so,
<because the performance is very close to what the drive is capable of.
<
<I don't understand why Apple is not able to write a piece of software
<that makes use of the hardware they ship (they had a lot of time,
<the introduction of the 7/8/9500 with Fast SCSI was more than two years 
<ago)

Which benchmarkers did you use for your measurements? I also did some tests
with the 1.2G internal drive (not sure if it was IBM or Quantum) from
a 7600/120 and my results are fairly similar to that Ernst J. Oud saw.
That is, the Apple HD SC Setup 7.3.5 driver is pretty good. I reinitialized
the drive using Apple's HD SC Setup 7.3.5, APS Powertools 4.0.7, FWB HDT
2.0.5, and Silverlining 5.7. The benchmarks I used were FWB's HDT 2.0.5
benchmarker and Macbench 4.0 disk suite. On the 7600 above, HDT and Apple's
driver were the best, roughly equal to each other. APS was about 10% slower
in most of the tests, and Silverlining 5.7 was next to last, but all were
pretty similar.

I then went home and reinitialized a Quantum Fireball 1.06G I bought from
Lacie. It was on the external scsi bus, attached to a 6100/80. Using the 
above drivers, as well as Silverlining 5.6.4, HDT and Apple's 7.3.5 driver 
(patched to work with a 3rd party drive) also came out tops, though, again, 
not that far from the others. In fact, I would say it was a 3-way tie, with 
a few percent, between APS 4.0.7, HDT 2.0.5, and HD SC 7.3.5. Silverlining 
5.6.4 and 5.7 were very close to each other, and very close to the rest, 
EXCEPT for the sequential reads and writes test where the top 3 were as 
much as 50% faster. I only used HDT's benchmarker, not Macbench 4.0, for 
these results.

From the testing on the internal 1.2G drive on the 7600 and the external 
1.06G drive on the 6100, it seems Apple's driver and FWB are consistently 
the best. A little surprising!

FWIW, here are the numbers from the HDT benchtests:

Volume         Sus Read (KB/sec)       Sus Write (KB/sec)         Average    Average  Reads Writes
Tested         Low   Peak  Avg.        Low   Peak  Avg.           Access(ms) Seek(ms) (sec) (sec)

7600/120 internal 1.2G----------------------------------------------------------------------------

APS 4.0.7      3416  4593  4121        3574  4967  4530           17.5       9.4      787   609
FWB HDT 2.0.5  3043  5189  4500        3882  6261  5549           17.2       9.3      899   700
FWB HDT 2.0.5  3042  5189  4490        3882  6267  5547           17.0       9.4      873   697
Apple HD SC
 setup 7.3.5   3239  5221  4451        3577  6253  4992           17.1       9.4      877   694
Silverlining
   5.7         3075  5297  4468        3516  6241  5057           17.5       9.4      791   619


6100/80 external 1.06G Fireball-------------------------------------------------------------------

APS 4.0.7      3100  4268  3964        3309  4421  4173           17.3       9.9      749   553
HD SC Setup                                                                  
  7.3.5        3120  4311  3933        3537  4435  4237           17.2       9.9      776   563
hdt 2.0.5      2804  4294  3909        3482  4415  4112           17.7       9.8      791   570
hdt 2.0.5*     2805  4287  3919        3322  4437  4120           17.7       9.9      791   572
Silverlining
     5.6.4     3004  4249  3842        3499  4377  4125           17.8       9.9      528   455
Silverlining                                                                 
     5.7       3089  4286  3815        3216  4407  4142           17.4       9.9      508   434


*second run of HDT 2.0.5 to check consistency.

BTW, does anyone know if the 6100's internal bus allows a greater
throughput than the external bus?

JTM

Back to previous page