> On Sun, 24 Oct 1999, Allison J Parent wrote:
>
> > <Looks like the 6500 was about 13 VAX-11/780 MIPS. That would make it about
> > <2x to 3x slower than a 486DX2/66.
> >
> > Not even close. More like a stack of a dozen or more of the 486s. It
>
> If you look at:
>
> http://www.digital.com/timeline/1990-4.htm
>
> It states: "The VAX 6500 processor delivered approximately 13 times the
> power of a VAX-11/780 system, per processor."
>
> To me, that means "13 MIPS". 13 MIPS is about 2 to almost 3 times slower
> than a 486dx2/66.
>
>
> But let's take a look at some Dhrystone 2.1 numbers I found:
>
> CPU MIPS MIPS
> System OS CPU (MHz) V1.1 V2.1
> ---------------------- ------------ ----------- ----- ------ ------
> VAX 8650 4.3 BSD ----------- 18 6.3 6.2
> cc -non_shared -DUNIX -O5 -ifo
>
> 80486DX2/66 Linux 2.0.0 80486DX2 66 29.3 26.9
> gcc 2.7.2, gcc -DUNIX -O2
>
> 80486DX2/66 NetBSD 1.4.1 80486DX2 66 ---- 34.4
> egcs -DUNIX -O2
>
>
> For the Dhrystone 2.1 case, 26.9 / 6.2 = 4.33 --> dx2/66 is 4 1/3 times as
> fast as the 8650. (Using the slower Linux numbers, too).
>
> As a sanity check, look at http://www.digital.com/timeline/1984-3.htm
>
> For the 8600, it states: "The VAX 8600, shown here during assembly, offered
> up to 4.2 times the performance of the industry standard VAX-11/780... ".
>
Time for a short lecture in hardware architecture (you can tell
where I'm from :-)). Integer performance is a very misleading
measure of performance when you are talking about system
performance. Not even the supercomputing community uses
raw integer or floating point performance as a benchmark,
there are far too many other issues. For example, all except
the most recent PCs, there is only a single bus. This bus
must be used for all memory transfers, graphics, I/O, etc.
On a single user system, this is sometimes okay, but for
multiple users forget it. Most of the VAXes had multiple
busses, and each was dedicated to a particular function.
This meant that the combined throughput was considerably more
than any existing PC bus. One of the main problems with
all of the PC chips is the limited speed of the FSB. Its
no good having high integer performance if you can't get
the data in or out of the CPU.
I do a lot of high end graphics, both on PCs and SGIs.
Individually the components on the PCs are faster
than the corresponding components on the SGI. The old
195Mhz R10000 do match up to PIII processor speeds, and
I can buy PC graphics cards that have more raw graphics
power than an IR pipe. But, the overall system performance
of the SGI is at least 5x better than the best PC
configuration, and it can bury the PC on any app that
moves a lot of data. Large systems were designed and
built for throughput, so its quite possible that one
of these old systems could outrun a modern PC in high
throughput applications.
--
Dr. Mark Green mark_at_cs.ualberta.ca
Professor (780) 492-4584
Director, Research Institute for Multimedia Systems (RIMS)
Department of Computing Science (780) 492-1071 (FAX)
University of Alberta, Edmonton, Alberta, T6G 2H1, Canada
Received on Sun Oct 24 1999 - 15:46:23 BST