Vista VS Win7; A Couple Benches

[johnz]

I've held the position that Vista and Win7 were just about equal in performance. I was challenged on this, so now I have a couple of benchmarks. I used Everest, and Geekbench. PCmark05 and CrystalMark wouldn't run. wPrime_155 was taking too long so I bailed on that. If any of you really want to see it, I guess I can run that, but it may take a day or so.

Both O/Ss were in VMs using MS VirtualPC07. They were installed using default settings, but not activated, and virtual machine additions were installed. Vista32 Business was updated to SP1, and all essential updates were installed. Win7-32 Professional had all essential updates installed. Both VMs were defragged after update using the built in defragger. I gave them both 1gb of ram for the VMs. I thought it would be interesting seeing the results on a low spec machine. I disabled the sidebar in Vista, but otherwise it was an "as installed" system.

I haven't compared the benches myself yet, so as soon as I get them here, it'll be news to me also :^D

So here we go...

Windows 7

Everest(I trimmed the comparison machines)

--------[ EVEREST Ultimate Edition ]------------------------------------------------------------------------------------

Version EVEREST v4.60.1500
Benchmark Module 2.3.237.0
Homepage http://www.lavalys.com/
Report Type Report Wizard [ TRIAL VERSION ]
Computer WIN7VM-PC
Generator win7vm
Operating System Windows 7 Professional Media Center Edition 6.1.7600
Date 2009-10-28
Time 19:25


--------[ Summary ]-----------------------------------------------------------------------------------------------------

Computer:
Computer Type Advanced Configuration and Power Interface (ACPI) PC
Operating System Windows 7 Professional Media Center Edition
OS Service Pack [ TRIAL VERSION ]
Internet Explorer 8.0.7600.16385
DirectX DirectX 10.0
Computer Name WIN7VM-PC
User Name win7vm
Logon Domain [ TRIAL VERSION ]
Date / Time 2009-10-28 / 19:25

Motherboard:
CPU Type DualCore Intel Core 2 Duo, 3100 MHz
Motherboard Name Microsoft Virtual Machine
Motherboard Chipset Intel 82440BX/ZX
System Memory [ TRIAL VERSION ]
BIOS Type AMI (02/22/06)
Communication Port Communications Port (COM1)
Communication Port Communications Port (COM2)
Communication Port Printer Port (LPT1)

Display:
Video Adapter VM Additions S3 Trio32/64 (8 MB)

Multimedia:
Audio Adapter Multimedia Audio Controller [1414-0007] [NoDB]

Storage:
IDE Controller Intel(R) 82371AB/EB PCI Bus Master IDE Controller
Floppy Drive Floppy disk drive
Disk Drive Virtual HD ATA Device (15 GB, IDE)
Optical Drive MS C/DVD-ROM ATA Device (Virtual DVD-ROM)
SMART Hard Disks Status FAIL

Partitions:
C: (NTFS) [ TRIAL VERSION ]
Total Size [ TRIAL VERSION ]

Input:
Keyboard Standard PS/2 Keyboard
Mouse Microsoft PS/2 Mouse

Network:
Primary IP Address [ TRIAL VERSION ]
Primary MAC Address 00-03-FF-05-40-23
Network Adapter Intel 21140-Based PCI Fast Ethernet Adapter (Emulated) (192. [ TRIAL VERSION ])

Peripherals:
Printer Fax
Printer Microsoft XPS Document Writer

DMI:
DMI BIOS Vendor American Megatrends Inc.
DMI BIOS Version 080002
DMI System Manufacturer Microsoft Corporation
DMI System Product Virtual Machine
DMI System Version VS2005R2
DMI System Serial Number [ TRIAL VERSION ]
DMI System UUID [ TRIAL VERSION ]
DMI Motherboard Manufacturer Microsoft Corporation
DMI Motherboard Product Virtual Machine
DMI Motherboard Version 5.0
DMI Motherboard Serial Number [ TRIAL VERSION ]
DMI Chassis Manufacturer Microsoft Corporation
DMI Chassis Version 5.0
DMI Chassis Serial Number [ TRIAL VERSION ]
DMI Chassis Asset Tag [ TRIAL VERSION ]
DMI Chassis Type Desktop Case
DMI Total / Free Memory Sockets 4 / 0


--------[ Memory Read ]-------------------------------------------------------------------------------------------------


Core 2 Duo 3133 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 7376 MB/s



--------[ Memory Write ]------------------------------------------------------------------------------------------------


Core 2 Duo 3100 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 5980 MB/s



--------[ Memory Copy ]-------------------------------------------------------------------------------------------------


Core 2 Duo 3118 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 5745 MB/s



--------[ Memory Latency ]----------------------------------------------------------------------------------------------


Core 2 Duo 3133 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 61.7 ns



--------[ CPU Queen ]---------------------------------------------------------------------------------------------------


Core 2 Duo 3133 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 0


--------[ CPU PhotoWorxx ]----------------------------------------------------------------------------------------------


Core 2 Duo 3133 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 4796



--------[ CPU ZLib ]----------------------------------------------------------------------------------------------------


Core 2 Duo 3100 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 17542 KB/s



--------[ CPU AES ]-----------------------------------------------------------------------------------------------------

8x
Core 2 Duo 3100 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 4858



--------[ FPU Julia ]---------------------------------------------------------------------------------------------------


Core 2 Duo 3133 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 2338



--------[ FPU Mandel ]--------------------------------------------------------------------------------------------------



Core 2 Duo 3100 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 1188



--------[ FPU SinJulia ]------------------------------------------------------------------------------------------------


Core 2 Duo 2966 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 604


GeekBench

Summary
Section Description Score Geekbench Score
Geekbench 2.1.4 for Windows x86 (32-bit)
Integer Processor integer performance 2687 2131
Floating Point Processor floating point performance 2389
Memory Memory performance 1128
Stream Memory bandwidth performance 1289

System Information
Operating System Microsoft Windows 7 Professional
Model Microsoft Corporation Virtual Machine Motherboard Microsoft Corporation Virtual Machine
Processor Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz
Processor ID GenuineIntel Family 6 Model 15 Stepping 6
Processors 1 Threads 1
Cores 1 Memory 1024 MB SDRAM 0 MHz
Processor Frequency 5.00 MHz Bus Frequency 100.0 MHz
L1 Instruction Cache 0.00 B L1 Data Cache 0.00 B
L2 Cache 4.00 MB L3 Cache 0.00 B
BIOS American Megatrends Inc. 080002

Integer Section
Section Score 2687
Blowfish
single-threaded scalar 1776
78.1 MB/sec
Blowfish
multi-threaded scalar 1885
77.3 MB/sec
Text Compress
single-threaded scalar 2473
7.91 MB/sec
Text Compress
multi-threaded scalar 2258
7.41 MB/sec
Text Decompress
single-threaded scalar 2904
11.9 MB/sec
Text Decompress
multi-threaded scalar 2778
11.1 MB/sec
Image Compress
single-threaded scalar 2566
21.2 Mpixels/sec
Image Compress
multi-threaded scalar 2335
19.6 Mpixels/sec
Image Decompress
single-threaded scalar 2240
37.6 Mpixels/sec
Image Decompress
multi-threaded scalar 2169
35.4 Mpixels/sec
Lua
single-threaded scalar 4465
1.72 Mnodes/sec
Lua
multi-threaded scalar 4406
1.69 Mnodes/sec

Floating Point Section
Section Score 2389
Mandelbrot
single-threaded scalar 1983
1.32 Gflops
Mandelbrot
multi-threaded scalar 2177
1.42 Gflops
Dot Product
single-threaded scalar 1345
650.3 Mflops
Dot Product
multi-threaded scalar 1391
634.0 Mflops
Dot Product
single-threaded vector 3370
4.04 Gflops
Dot Product
multi-threaded vector 4405
4.58 Gflops
LU Decomposition
single-threaded scalar 2565
2.28 Gflops
LU Decomposition
multi-threaded scalar 2640
2.32 Gflops
Primality Test
single-threaded scalar 4019
600.2 Mflops
Primality Test
multi-threaded scalar 3080
571.6 Mflops
Sharpen Image
single-threaded scalar 651
1.52 Mpixels/sec
Sharpen Image
multi-threaded scalar 636
1.47 Mpixels/sec
Blur Image
single-threaded scalar 2568
2.03 Mpixels/sec
Blur Image
multi-threaded scalar 2617
2.06 Mpixels/sec

Memory Section
Section Score 1128
Read Sequential
single-threaded scalar 1377
1.69 GB/sec
Write Sequential
single-threaded scalar 1764
1.21 GB/sec
Stdlib Allocate
single-threaded scalar 882
3.29 Mallocs/sec
Stdlib Write
single-threaded scalar 534
1.11 GB/sec
Stdlib Copy
single-threaded scalar 1086
1.12 GB/sec

Stream Section
Section Score 1289
Stream Copy
single-threaded scalar 1328
1.82 GB/sec
Stream Copy
single-threaded vector 1620
2.10 GB/sec
Stream Scale
single-threaded scalar 995
1.29 GB/sec
Stream Scale
single-threaded vector 979
1.32 GB/sec
Stream Add
single-threaded scalar 1115
1.68 GB/sec
Stream Add
single-threaded vector 1180
1.64 GB/sec
Stream Triad
single-threaded scalar 1251
1.73 GB/sec
Stream Triad
single-threaded vector 1849
3.46 GB/sec

http://browse.geekbench.ca/geekbench2/view/179280


Vista

Everest(trimmed comparisons again)

--------[ EVEREST Ultimate Edition ]------------------------------------------------------------------------------------

Version EVEREST v4.60.1500
Benchmark Module 2.3.237.0
Homepage http://www.lavalys.com/
Report Type Report Wizard [ TRIAL VERSION ]
Computer VISTAVM-PC
Generator vistavm
Operating System Microsoft Windows Vista Business 6.0.6000 (Vista Retail)
Date 2009-10-28
Time 20:53


--------[ Summary ]-----------------------------------------------------------------------------------------------------

Computer:
Computer Type Advanced Configuration and Power Interface (ACPI) PC
Operating System Microsoft Windows Vista Business
OS Service Pack [ TRIAL VERSION ]
Internet Explorer 7.0.6000.16916
DirectX DirectX 10.0
Computer Name VISTAVM-PC
User Name vistavm
Logon Domain [ TRIAL VERSION ]
Date / Time 2009-10-28 / 20:53

Motherboard:
CPU Type DualCore Intel Core 2 Duo, 3166 MHz
Motherboard Name Microsoft Virtual Machine
Motherboard Chipset Intel 82440BX/ZX
System Memory [ TRIAL VERSION ]
BIOS Type AMI (02/22/06)
Communication Port Communications Port (COM1)
Communication Port Communications Port (COM2)
Communication Port Printer Port (LPT1)

Display:
Video Adapter VM Additions S3 Trio32/64 (8 MB)

Multimedia:
Audio Adapter Microsoft Virtual Machine Audio Device Driver [1414-0007] [NoDB]

Storage:
IDE Controller Intel(R) 82371AB/EB PCI Bus Master IDE Controller
Storage Controller Microsoft iSCSI Initiator
Floppy Drive Floppy disk drive
Disk Drive Virtual HD ATA Device (63 GB, IDE)
Optical Drive MS C/DVD-ROM ATA Device (Virtual DVD-ROM)
SMART Hard Disks Status FAIL

Partitions:
C: (NTFS) [ TRIAL VERSION ]
Total Size [ TRIAL VERSION ]

Input:
Keyboard Standard PS/2 Keyboard
Mouse Microsoft PS/2 Mouse

Network:
Primary IP Address [ TRIAL VERSION ]
Primary MAC Address 00-03-FF-04-40-23
Network Adapter Intel 21140-Based PCI Fast Ethernet Adapter (Emulated) (192. [ TRIAL VERSION ])

Peripherals:
Printer Fax
Printer Microsoft XPS Document Writer

DMI:
DMI BIOS Vendor American Megatrends Inc.
DMI BIOS Version 080002
DMI System Manufacturer Microsoft Corporation
DMI System Product Virtual Machine
DMI System Version VS2005R2
DMI System Serial Number [ TRIAL VERSION ]
DMI System UUID [ TRIAL VERSION ]
DMI Motherboard Manufacturer Microsoft Corporation
DMI Motherboard Product Virtual Machine
DMI Motherboard Version 5.0
DMI Motherboard Serial Number [ TRIAL VERSION ]
DMI Chassis Manufacturer Microsoft Corporation
DMI Chassis Version 5.0
DMI Chassis Serial Number [ TRIAL VERSION ]
DMI Chassis Asset Tag [ TRIAL VERSION ]
DMI Chassis Type Desktop Case
DMI Total / Free Memory Sockets 4 / 0


--------[ Memory Read ]-------------------------------------------------------------------------------------------------


Core 2 Duo 3183 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 7876 MB/s


--------[ Memory Write ]------------------------------------------------------------------------------------------------


Core 2 Duo 3166 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 6237 MB/s



--------[ Memory Copy ]-------------------------------------------------------------------------------------------------


Core 2 Duo 3200 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 6188 MB/s



--------[ Memory Latency ]----------------------------------------------------------------------------------------------


Core 2 Duo 3183 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 58.9 ns



--------[ CPU Queen ]---------------------------------------------------------------------------------------------------


Core 2 Duo 3184 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 0


--------[ CPU PhotoWorxx ]----------------------------------------------------------------------------------------------


Core 2 Duo 3166 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 6926



--------[ CPU ZLib ]----------------------------------------------------------------------------------------------------


Core 2 Duo 3200 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 19238 KB/s



--------[ CPU AES ]-----------------------------------------------------------------------------------------------------


Core 2 Duo 3166 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 5451



--------[ FPU Julia ]---------------------------------------------------------------------------------------------------


Core 2 Duo 3052 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 2423


--------[ FPU Mandel ]--------------------------------------------------------------------------------------------------


Core 2 Duo 3181 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 1312



--------[ FPU SinJulia ]------------------------------------------------------------------------------------------------


Core 2 Duo 2300 MHz [ TRIAL VERSION ] i440BX/ZX Dual EDO 483




GeekBench

Summary
Section Description Score Geekbench Score
Geekbench 2.1.4 for Windows x86 (32-bit)
Integer Processor integer performance 2656 2127
Floating Point Processor floating point performance 2345
Memory Memory performance 1309
Stream Memory bandwidth performance 1151

System Information
Operating System Microsoft Windows Vista Business
Model Microsoft Corporation Virtual Machine Motherboard Microsoft Corporation Virtual Machine
Processor Intel(R) Core(TM)2 CPU 6600 @ 2.40GHz
Processor ID GenuineIntel Family 6 Model 15 Stepping 6
Processors 1 Threads 1
Cores 1 Memory 1023 MB SDRAM 0 MHz
Processor Frequency 3.05 GHz Bus Frequency 100.0 MHz
L1 Instruction Cache 0.00 B L1 Data Cache 0.00 B
L2 Cache 4.00 MB L3 Cache 0.00 B
BIOS American Megatrends Inc. 080002

Integer Section
Section Score 2656
Blowfish
single-threaded scalar 1787
78.5 MB/sec
Blowfish
multi-threaded scalar 1923
78.8 MB/sec
Text Compress
single-threaded scalar 2445
7.82 MB/sec
Text Compress
multi-threaded scalar 2434
7.99 MB/sec
Text Decompress
single-threaded scalar 2691
11.1 MB/sec
Text Decompress
multi-threaded scalar 2791
11.1 MB/sec
Image Compress
single-threaded scalar 2368
19.6 Mpixels/sec
Image Compress
multi-threaded scalar 2416
20.3 Mpixels/sec
Image Decompress
single-threaded scalar 1716
28.8 Mpixels/sec
Image Decompress
multi-threaded scalar 2012
32.8 Mpixels/sec
Lua
single-threaded scalar 4698
1.81 Mnodes/sec
Lua
multi-threaded scalar 4591
1.77 Mnodes/sec

Floating Point Section
Section Score 2345
Mandelbrot
single-threaded scalar 2454
1.63 Gflops
Mandelbrot
multi-threaded scalar 1991
1.30 Gflops
Dot Product
single-threaded scalar 1339
647.0 Mflops
Dot Product
multi-threaded scalar 1391
633.9 Mflops
Dot Product
single-threaded vector 3109
3.73 Gflops
Dot Product
multi-threaded vector 4033
4.19 Gflops
LU Decomposition
single-threaded scalar 2445
2.18 Gflops
LU Decomposition
multi-threaded scalar 2503
2.20 Gflops
Primality Test
single-threaded scalar 3830
572.0 Mflops
Primality Test
multi-threaded scalar 3129
580.8 Mflops
Sharpen Image
single-threaded scalar 633
1.48 Mpixels/sec
Sharpen Image
multi-threaded scalar 641
1.48 Mpixels/sec
Blur Image
single-threaded scalar 2699
2.14 Mpixels/sec
Blur Image
multi-threaded scalar 2639
2.08 Mpixels/sec

Memory Section
Section Score 1309
Read Sequential
single-threaded scalar 1601
1.96 GB/sec
Write Sequential
single-threaded scalar 1707
1.17 GB/sec
Stdlib Allocate
single-threaded scalar 950
3.55 Mallocs/sec
Stdlib Write
single-threaded scalar 952
1.97 GB/sec
Stdlib Copy
single-threaded scalar 1335
1.38 GB/sec

Stream Section
Section Score 1151
Stream Copy
single-threaded scalar 964
1.32 GB/sec
Stream Copy
single-threaded vector 1059
1.37 GB/sec
Stream Scale
single-threaded scalar 1003
1.30 GB/sec
Stream Scale
single-threaded vector 1264
1.71 GB/sec
Stream Add
single-threaded scalar 873
1.32 GB/sec
Stream Add
single-threaded vector 1038
1.44 GB/sec
Stream Triad
single-threaded scalar 2258
3.12 GB/sec
Stream Triad
single-threaded vector 751
1.41 GB/sec

http://browse.geekbench.ca/geekbench2/view/179297

 

[johnz]

Ok, there's a discrepancy in clock speeds in Everest. The VMs were setup exactly the same. Do any of you know why that would be?

Edit:
I'm trying to clean this up so they're easily comparable. Here's the GeekBench runs in .png form...



I'll try to do something with Everest. Sorry for the chaos :^D

Edit2:
Here's a PNG of the Everest benches together...

 

[hokiealumnus]

Nice comparison johnz; I like the low-spec angle.

For those interested and that don't want to click on the links, the Geekbench total scores were:

Vista - 2,127 & Win7 - 2,131.

From previous experience running GB, those are well within any margin of error. Were the tests run more than once, they may have switched places.

 

[johnz]

What's up with the Everest clock speeds?

It's hard to compare the numbers because the clock speeds vary. Just guessimating for the speed difference, the numbers are about equal, but the reason you bench is so you don't have to guesstimate :^D

I'll look for another interesting bench. Maybe I can find something that's easier to compare side by side....

Edit:
There's some trade off back and forth in the GB bench, but it looks like Vista tends to be on the losing side more often. Maybe I'll bench them both a couple more times, and see how much variance there is.

 

[johnz]

I got CrystalMark to run on Vista, but not Win7. Here's the benches for Wprime. It seemed to run a good bit faster this time. I'm not sure what was going on before :^/

Windows 7



Vista

 

[hokiealumnus]

Looking very close so far. Will SiSoft's Sandra work with your installs?

 

[johnz]

Looking very close so far. Will SiSoft's Sandra work with your installs?

I'll give it a try later. I'm currently running the Wprime 1024m on Vista. The longer times might be more revealing. Incidentally, I think that's what I started the very first time, and that's why it was taking so long. I haven't really benched my systems in awhile, and I forgot how all this stuff works :^D

The PCmark05 bench seemed to run fine(I /think/ I tried Win7 first), but when it was over it didn't display any results. Neither online or local showed anything other than 0s :^/

 

[Mr.Guvernment]

i think anyone with a modern rig will see almost no performance gain going to windows 7, going to windows 7 would be purely for features.

good testing though.

 

[DaveHCYJ]

I wonder if benching in a VM negates some of the lower level optimizations that were put in Windows 7 since it isn't interfacing with actual hardware anymore?

I agree with your theory that Vista and 7 are roughly equal in performance for benchmark type of things. If you are encoding video on your machine etc you will probably notice no difference between Vista and 7 in the time it takes for the video to encode.

But benchmarks aren't everything, Windows 7 has moved a lot of things around so they happen in different order. The net effect is that it still takes the same total amount of time as vista did, however it is important to note that from the user's perspective Win7 appears more "snappy". One major "speed up" I've noticed is when starting up the computer or resuming from sleep/hibernate I get to the type your password in screen much faster in Win7 and it then continues to load in the background while I type my password in, while with Vista unless I'm ready to pounce on the login screen there is time wasted because I don't see the login screen until windows is basically done loading. So if you are ready to pounce on the Vista login screen loading times might be the same, but otherwise the user interaction unbenchable part of the startup process usually causes Vista to be slower.

So yeah, from the time I press the video encode button until it finishes encoding in Vista and Win7 are probably not much different, where are the "benchmarks" on how much time Win7 saved me not having to wait for the start menu to pop up or how much time did Win7 save me with xxxx user experience?

 

[doublejack]

i think anyone with a modern rig will see almost no performance gain going to windows 7, going to windows 7 would be purely for features.

good testing though.

+1

I've run Vista for a long time now, and have been playing with Win7 for almost a year. They're nearly the same OS. To be sure there are some differences, but not in the area of performance.

I really don't see much reason to upgrade from Vista to 7. That gets into the enigma of why Vista has a horrible reputation while 7 appears to be much more accepted, but I don't want to derail the thread.

Anyway, good dataset John. It really backs up the feel that the two OS's are so similar.

 

[johnz]

I ran a couple more benches. For Sandra, I just ran the performance index test. That gives you 1 number that's easy to compare. The individual benchmarks looked like they duplicated functionality from the other benchmark programs. I also ran Gprime1024m so they'd have a longer test period.

Windows 7




Vista

I wonder if benching in a VM negates some of the lower level optimizations that were put in Windows 7 since it isn't interfacing with actual hardware anymore?

I agree with your theory that Vista and 7 are roughly equal in performance for benchmark type of things. If you are encoding video on your machine etc you will probably notice no difference between Vista and 7 in the time it takes for the video to encode.

But benchmarks aren't everything, Windows 7 has moved a lot of things around so they happen in different order. The net effect is that it still takes the same total amount of time as vista did, however it is important to note that from the user's perspective Win7 appears more "snappy". One major "speed up" I've noticed is when starting up the computer or resuming from sleep/hibernate I get to the type your password in screen much faster in Win7 and it then continues to load in the background while I type my password in, while with Vista unless I'm ready to pounce on the login screen there is time wasted because I don't see the login screen until windows is basically done loading. So if you are ready to pounce on the Vista login screen loading times might be the same, but otherwise the user interaction unbenchable part of the startup process usually causes Vista to be slower.

So yeah, from the time I press the video encode button until it finishes encoding in Vista and Win7 are probably not much different, where are the "benchmarks" on how much time Win7 saved me not having to wait for the start menu to pop up or how much time did Win7 save me with xxxx user experience?

I don't know what affect it would have running in a VM as opposed to real hardware. It could very well make a difference though.

I agree regarding the startup and shutdown. It's hard to test Windows in this regard, as it seems variable from 1 time to the next. It'll be pokey one time, then fast the next. Also, I don't agree with their builtin boot timer, at least in Vista. They stopped the clock before I considered the system usable. I stop watched Vista before, and it was significantly longer to a usable system than Vista had stated.

After several reboots doing this stuff, I think Win7 is ACTUALLY faster. It's purely subjective, but the total time to a usable desktop feels quicker than Vista. I don't think Win7 loads as much into memory at startup, and I'm guessing that's where a lot of the savings comes from. MS also probably tweaked some settings to get it going a bit quicker.

Edit:
Disregard the blue bars in the Sandra benches. Those are for comparative systems, and aren't related to what I tested.

 

[EarthDog]

Very thorough, thank you!

But my question is also regarding testing one actual OS vs a VM OS. Not sure this is valid really now . Try doing a dual boot instead of a VM and see if the results change to insure the validity of your already established results.

 

[johnz]

Very thorough, thank you!

But my question is also regarding testing one actual OS vs a VM OS. Not sure this is valid really now . Try doing a dual boot instead of a VM and see if the results change to insure the validity of your already established results.

I don't have the space to load a real dual boot. I had to delete a bunch of files(mostly Linux .isos) to get room for 2 VMs.

I /think/ the results are valid here. The "hardware" should appear identical to both O/Ss, but it's hard to tell without knowing exactly what MS did under the hood.

What it looks like to me is Win7 IS a bit faster. There was some trade off back and forth, but on a couple of tests Win7 was significantly ahead(> than the margin of error). Is that extra speed really noticeable on the desktop? That I don't know. My stance the whole time has been that Win7 only APPEARS faster, but in reality it's mostly the same as Vista. Other than startup and shutdown, which do appear truly faster to me, I think I'll have to stick with my original contention; that Win7 and Vista are for all intents and purposes equal.

I'll run 1 more set of benches just for kicks. I'll stopwatch time the startup and shutdown of each 3 times in a row, and take the average. This will be somewhat rough as I have to type in a password for each, but that shouldn't add more than a couple seconds of potential error, and we're looking for more than a couple seconds here.

 

[EarthDog]

http://en.wikipedia.org/wiki/Comparison_of_platform_virtual_machines

Check out the "More Details" table, the "Guest OS speed relative to host OS" column for w/e VM app you used. Not many are "native" which is the same speed as host OS.

Virtual PC 2007 = Near native with Virtual Machine additions

That tells me you should see a bit better results across the board as its not native.

 

[johnz]

http://en.wikipedia.org/wiki/Comparison_of_platform_virtual_machines

Check out the "More Details" table, the "Guest OS speed relative to host OS" column for w/e VM app you used. Not many are "native" which is the same speed as host OS.

Virtual PC 2007 = Near native with Virtual Machine additions

That tells me you should see a bit better results across the board as its not native.

If I'm following you right... Both O/Ss I tested were virtual machines. The host was Vista64 Business. Other than potential optimizations for running in a VM, both tested O/Ss should have been on equal footing.

Here's my stopwatch timed startup and shutdown times. Win7 was significantly faster on startup, but shutdowns were about equal. I timed startup from button press to internet connection at desktop. I timed shutdown from button push to the close of the VM window. I had my finger on the first key for my simple 7 character password. I didn't have any major fumbles, so that shouldn't have affected time much in either direction.

Windows 7

Startup
43.90
39.11
38.95
Avg- 40.65

Shutdown
12.10
10.52
9.11
Avg- 10.58

Vista

Startup
53.92
54.40
53.13
Avg- 53.82

Shutdown
11.24
10.03
10.16
Avg- 10.48

 

[EarthDog]

Oh crap. I thought you have one of the OS's as 'native' and the other OS as virutal. If you had both on virtual, then the testing is valid. My apologies.

Darn, its in the first post, I missed it. My bad.

 

[DaveHCYJ]

Tomshardware did a bunch of benchmarks a while back and as I recall their conclusion was that Win7 was negligibly faster but also consumed more power.

I /think/ the results are valid here. The "hardware" should appear identical to both O/Ss, but it's hard to tell without knowing exactly what MS did under the hood.

Well the VM's are just presenting virtual hardware to the OS's. So if Win7 added some trickery with optimized hardware interfaces you won't actually see that, because most virtual hardware is generic. For example, I thought Win7 was supposed to make better use of SSD's, but in a VM Win7 doesn't see it as an Intel Model abc SSD, Win7 just sees it as an unidentified generic hard drive.

Something else to consider is that you're virtual hard drives are just files stored on your real hard drive, and depending on where each is stored one of them is going to have an inherent advantage, which depending on the placement could be sizable, but most likely isn't that big.

Your benchmarks are certainly still meaningful, I was just sort of getting a little more philosophical and wondering how much of an effect doing them in a VM has. In the end for me if there is less than 5% difference it is too close to call and I don't really care. If one OS encodes a video and 30 minutes and the other does it in 31 minutes, it doesn't matter, because I'm in the other room watching TV and I'm not going to come check to see if its done for an hour anyway...