Benchmarks: Core i7 vs Core 2 Duo vs Virtual Machine
Posted by Keith Elder | Posted in Computer Hardware | Posted on 09-04-2009
Let’s get ready to rumble! After my new machine arrived I set out to answer some of my questions. How much faster is the new machine than the previous laptop I was using to develop on (anyone that has seen me speak in the last two years has seen this laptop). Is the virtual machine I’ve been developing on for work faster on the new machine? If so is it faster than my old machine?
To get started I downloaded PerformanceTest which gave me a free 30 day trial. Plenty of time to test and gather benchmarks. The program is nice because it allows you to easily load up saved tests and do comparisons as you’ll see. Let’s get started!
The Setup
Updated 4/11/2009 – Several people requested I run some additional benchmarks so this now includes three new benchmarks for the virtual machines.
To try to make things as fair as I could I did a clean boot of each machine and then killed any background processes that were running (utilities, snagit, etc). I then ran the benchmark twice and took the faster run. Here’s what each system is running:
- DellD820, Dell Lattitude D820 Notebook, Core 2 Duo 2.33GHz, 4GB, 100GB 7200RPM HDD, Windows Vista x32
- DellVM2Proc, Dev Virtual Machine, VMWare, 6.5, Two Processors, 1.5GB, Windows Server 2008 x32 (running on D820 Notebook, external eSata drive)
- DevVM1Proc4GB, Dev Virtual Machine, VMWare 6.5, Single Processor, 4GB, Windows Server 2008 x32 (running on new PC below)
- DevVM2Proc3GB, Dev Virtual Machine, VMWare 6.5, Two Processors, 3GB, Windows Server 2008 x32 (running on new PC below)
- DevVM2Proc4GB, Dev Virtual Machine, VMWare 6.5, Two Processors, 4GB, Windows Server 2008 x32 (running on new PC below)
- New PC, Core i7 920, 12GB, 300GB 10,000RPM HDD, nVidia 9400GT, Windows Vista x64
CPU
First up is CPU. Obviously I knew going in the Core i7 was going to take the prize. What I didn’t expect to see is this wide of a margin.
The processor in the Dell is a 2.33GHz Core 2 Duo, so not a bad processor. Still it gets crushed by the Core i7. The VMWare virtual machine took last place in this round. That isn’t too surprising since it is virtualized running only one processor.
Graphics
This is the biggest surprise of the day. The little laptop that could came through and crushed the competition.
Even the virtual machine almost had the same score as the new PC. I am completely puzzled over this one. Maybe some true hardware guru will have some insight, or maybe the benchmark is just broke ?
Since the laptop and the VM don’t have 3D support the only machine that had a score is the new PC, thus no point in showing the benchmark.
Memory
Memory performance starts to give a better idea of how the machine line up in terms of speed. In this round the laptop takes last place with the VM running on the new PC taking second.
Disk Mark
The disk mark shows the VM taking the prize. Why? I have no idea. For some reason the VM was able to read twice as fast and won in every other category as well.
The thing to note in the sequential read mark is the new PC is pushing about 102.8 MB/s throughput. That’s more than twice as fast as the laptop.
Summary
So how did the machines stack up in the end? Well, obviously the new PC is going to take the top prize, but the surprising winner when all the numbers were added up is the VMWare virtual machine took second place! Even more surprising, the VM with 3GB outperformed the 4GB! There is definitely a difference enabling two processors in VMWare, about 100 points difference between the two.
What does this mean? It means that the virtual machine I develop on running on the new PC is faster than the entire laptop I previously used. To be honest, developing on the laptop wasn’t bad at all. It performed very well but the new Core i7 system is literally like running two machines in one.
Words of Wisdom
I know a lot of developers are using laptops to develop on today because the performance of them has gotten to be pretty good. The bottom line though is you will never beat a desktop for speed and performance. My advice is for you to seriously consider ditching your laptop for serious development and use a desktop. You just can’t beat the performance.
Your benchmark is broke. The Nvidia Quadro NVS 120m (at best) in a D820 is only a little better than a
9400GT.
http://www.videocardbenchmark.net/gpu.php?gpu=Quadro+NVS+120M&id=1621
http://www.videocardbenchmark.net/gpu.php?gpu=Geforce+9400GT&id=712
Also, unless you were using an SSD or raid system, no HD in 2009 is going to have sequential read speeds of 200mb/sec. Even this 2012 15000 RPM SAS drive could only attain 150MB/s
http://www.storagereview.com/seagate_constellation2_and_constellation_es2_hard_drive_review
I think the problem is, in virtual machines, the cpu timer doesn’t work correctly. The cpu timer is not actually running continuously and be suspended when the VM is not actively executing. Only the system clock is reliable. A good ole stop watch would work too.
Thanks Keith. This will come in handy (my work laptop is a D820 and I run VMs regularly)
about that VM with 3GB outperform better than VM with 4GB i had this experience before ,higher ram is not always better for VM
Hi Keith,
IMO the Graphics benchmark *must* be broken. Check if it is possible to turn on some diagnostic logging or something in the perf tool.
–larsw
Seems everyone wants more benchmarks, I’ll throw some more together and update this post later on tonight or tomorrow.
Another couple of cool benchmarks would be:
a) running the dev vm on the D820
b) running the dev vm with multiple cores (hyper v does this afaik)
It’s a dev rig right? Why not benchmark compiling some stuff?
Nice post. The disk performance benchmarking of virtual machines has to be inaccurate. The benchmark would only report on the virtual adapter, not the true physical read/write. The virtual adapter would report write success long before it is physically written to disk. Also, reads would be faster since virtual disk do not have physical clusters and tracks, just random reads from one big file that is (probably) cached in virtual adapter. It is likely that true disk benchmarking in virtual machine may only occur using raw disks (for those vm’s that support raw disks).
Hey Dennis,
I started to do that but I had already moved the VM over to the new machine and didn’t feel like copying it back over.
I can tell you that it was painfully slow a lot of the time even though I was running it on an external eSata drive. Having only 3.2GB of available memory on the laptop was the problem. I ran the VM with about 1.3 GBs of memory leaving the OS with about 500MB of free memory to keep things going. It just wasn’t enough memory to go around to truly run it. Now the VM gets whatever it wants on the new machine. It truly feels like a seperate computer and I’m able to run dual screens in the VM flawlessly.
Currently I’m running the VM with 3GB but did run it earlier with 4GB, I haven’t noticed a difference.
I think it would have been interesting to see the VM performance on the old machine vs. the new one. Show a bit of what the development experience is on both.