virtualbox.org

My host is Windows 7 64 bit. I was looking to benchmark some read/write on the disk inside the VM and network transfers and I was curious about using an SSD drive. Using nlite and removing unnecessary features, slipstream in AHCI driver both are created on a sata controller but one has SSD ticked off on the storage part. Both install approximately in the same amount of time and both boot at the same times. About six seconds from start to finish. What could I use to test read/write/seq access times? HDtune and CrystalMark is what I use on my host but inside a VM im not sure if those results are exaggerated or correct. If they are not the SSD version shows a higher burst rate by only maybe 30Mb/sec. The CPU usage is 30% lower. People say that IDE and SATA on XP have no differences in speed but in my opinion they do. Sata controllers are slower but testing this out on SSD drives shows a totally different result: faster (duh its SSD ) I know but what advantages/disadvantages does ticking that option really offer?

Snapshots are like turning a light off and on its pretty fast. I'm not real concerned about SSD wear/lifecycle

Benchmarking inside a VM has numerous difficulties. Basic problem is that the CPU is shared, and the timers are less accurate. IMHO benchmarks are overrated anyway. Best results come from understanding why a particular technology must be better, not in having faith that benchmarks were competently done (rare!) on a particular day.

I'll give you a "for example": My company makes industrial flame detectors. One day we got hold of a test report done by this big, important test house. The detection times were all over the shop, but mostly terrible - our clients were hopping mad that they had already installed these awful detectors. Only saving grace was that competitors detectors had done very badly as well. Of course we understood the technology and were in a position to know that we (and our competitors!) should have done a lot better, regardless of what the numbers said. Then we got hold of some videos of the guys doing the tests. It turned out that their test procedure was to start a stopwatch and then yell to a technician to light the fire... but for some fuels the technician had difficulty getting the thing to ignite, and if he didn't do it right then the fuel might smoulder and take a while to reach the threshold size - and all of that fumbling around time was included in the results - as if it was the detectors fault that these guys were idiots.

This is just one example of why I've learned over the years to pay little attention to benchmarks.

True I understand where you are coming from. I ran several tests last night assigning different amounts of ram, procs, video ram. I know more than one proc slows down everything but tried it anyways but as usual slow. I started out originally with 512 megs of ram, one proc 32 megs of vram. Boot time was 7 seconds on Xp. Change that ram to 192 megs that made it 4 seconds. But this is on a brand new install everytime. If I go back and make any changes to these settings after XP is installed the boot time exceeds 12-15 seconds. Even if I go in and disabled something from services.msc same thing the boot time is decreased. Went into nlite and turned off a couple services that I missed like Windows Audio (forgot the other) and visual themes that you would find on advance tab>let windows decide or custom and it disables everything. It was the same as making a change after installing XP it was slower on boot. Doesn't make sense? Either way just wanted to test some things as I been an avid Vmware user for several years.