My last post on the shoot out between Windows 3.1 apps running under SoftWindows 3 on a Mac and the same Windows 3.1 apps running on native hardware seemed to warrant a postscript I thought. This post is that postscript.
Last time, we looked at Windows 3.1 apps running under SoftWindows 3.0 on a 200 MHz Power Macintosh 7300/200 vs. the same apps running natively on a 100 MHz 486DX4100. My naïve expectation at the time I set this testing up was that the SoftWindows emulation would exact about a 2 to 1 performance hit, and so it should be a fair contest. As it turned out, this could not have been farther from the truth.
SoftWindows 3.0 emulates an Intel 80486, and the native Windows machine used in the testing was also an 80486, and so that much at least was fair, but the performance impact of the emulation turned out to be significantly higher than 2X… it was more like 6X! So, comparing machines of dissimilar clock rates was not such a good idea after all. What I needed to do was compare machines with the SAME clock rate, and observe those results. That would be an Apples to apples (pun intended) comparison and would directly measure the performance impact of PC emulation on a Mac… now that would be interesting.
To accomplish this, I pitted two 200 MHz machines against each other: the above mentioned 200 MHz Power Macintosh 7300/200 and a 200 MHz Pentium Pro PC. To keep it as fair as possible, I upgraded the Mac to Virtual PC 3.0, which emulates an Intel Pentium. Was the Pentium vs. Pentium Pro difference a significant one for performance? Would the Pentium Pro have an unfair advantage?
Well, given the mixture of 16 bit and 32 bit software in Windows 95, the Pentium was actually supposed to be the superior chip, clock for clock, vs. the Pentium Pro, which was reported to be optimized for pure 32 bit processing. Hence, if anything, I was giving a bit of a handicap to the Power Macintosh. Fair enough.
As you may have guessed from the above, Windows 95 was the test vehicle this time around, vs. the Windows 3.1 of the last post.
So, we had Windows 95 running on a 200 MHz Power Macintosh under Virtual PC 3.0 vs. Windows 95 running natively on a 200 MHz Pentium Pro. No matter what apps I ran, the native hardware would clearly outperform the emulated hardware, but by how much? What is the real world performance penalty for PC emulation on a Mac OS machine?
Like last time, I chose the CPU intensive task of image decoding as the test vehicle of choice, and like last time, I stayed with the same four images: abyss.jpg, arnold.jpg, beer.jpg and shark.jpg. I also used three very familiar image viewers (from previous posts) on both machines, but in each case the 32 bit Windows 95 versions this time: IrfanView 4.23, ACDSee 2.4 and finally, LViewPro 1.D2.
OK, enough setup! Lets let’s the results speak for themselves. Here is how the two setups fared in timed tests of the same images, viewed with the same image viewers, under the same operating system (Windows 95). All times are in seconds:
What can we tell from the above?
To start with, we can see from these results that on average, the native Pentium hardware turned in a performance of roughly 5X that of the emulated Pentium hardware. This is largely in keeping with the results derived in our last post, where the native hardware was approximately 6X faster.
Now 5X is better than 6X, and thus, we can also see from these results that VPC 3.0 has more efficient emulation than SoftWindows 3: the difference between SoftWindows 3 and VPC 3.0 is an improvement from a 6X native hardware performance advantage to “just” a 5X native hardware performance advantage… not a lot, but every little bit helps!
So, Apples to apples, or any other way you can imagine, I think that we can conclude from these results that if you want to run an emulated PC on your Macintosh, the faster your Macintosh, the better! 🙂