Submitted by A benchmarking comparison of Solaris 10 and Windows XP (64bit) running on a Sun Ultra 20 M2 Dual Core Opteron 1210 workstation. “Overall, I’m pleased with the performance of the Sun Ultra 20 M2. While this model (with an Opteron 1210) isn’t quite as fast as a low-end Mac Pro, it’s less than half the price of a low-end Mac Pro. Models equipped with faster Opterons (like the Opteron 1218) will certainly be more competitive when it comes to performance.”
Solaris and Windows Performance Compared
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
Solaris is faster in most cases, but it would be even faster with more ram, as solaris memory handling is “somewhat” better than microsoft =)
This is actually a Visual C++ vs. Sun Studio compiler test.
If Sun’s compilers are so great, maybe they should sell them on Windows…
More like Sun Studio on Solaris vs Visual C++ on Windows.
Don’t know any differences between the source of the benchies between the two platforms, or if there are other things that differ between how the benchies are run, so I’m not gonna comment the results themselves.
The Geekbench benchmark isn’t that great anyway, and is going to depend fairly heavily on certain compiler and library settings if I remember right. Like how large of chunks memory is allocated in by default – there was a big thread about it quite a while ago.
If Sun’s compilers are so great, maybe they should sell them on Windows…
You can download the Sun Studio compiler for free (If you want you can pay for support). Sorry, no Windows version. I am sure that if they did make a Windows version (unlikely), it would also be free.
Solaris vs XP (server-centric OS vs desktop-centric server)?? It’s unfair, isn’t it?
Solaris vs Win 2003 R2 would be more interesting and relevant…
If they wanted to test just operating systems, they could have used gcc as well. Hopefully that would be showing the differences in OS better.
>> If they wanted to test just operating systems, they could have used gcc as well.
Would they? I think they would be benchmarkiong the gcc implementation as much as the underlying OS.
Edited 2006-10-24 09:02
> Would they? I think they would be benchmarkiong the gcc
> implementation as much as the underlying OS.
Given the processor is the same, there’d be no difference whatsoever in the generated code between the two platforms.
Given the processor is the same, there’d be no difference whatsoever in the generated code between the two platforms.
So you could copy a g++ executable from Solaris to Windows and have it work! The API’s for the 2 OS’s are totally different. Simple things like register assignment (which is different) can dramatically change the compiler optimization choices, therefore the code produced is different.
The API has got nothing to do with code generated for algorithms. Of course, say, opening a window on solaris is done differently than opening a window on windows, but the point is exactly to measure the relative performances of doing similar tasks on the two OS’s.
If they wanted to test just operating systems, they could have used gcc as well. Hopefully that would be showing the differences in OS better.
Why, does gcc do something magical because it is GPL? I don’t how gcc will help. It will just compile slower code.
They should use GCC for a true OS speed comparison. The benchmarks they record are clearly screwed up; they make Windows look way faster than it actually is. I can’t imagine that Sun has let Solaris wither away THAT much.
Edited 2006-10-24 12:42
They should use GCC for a true OS speed comparison. The benchmarks they record are clearly screwed up; they make Windows look way faster than it actually is. I can’t imagine that Sun has let Solaris wither away THAT much.
GCC generally on Solaris produces slower code than Sun Studio. How would using gcc help? What feature of gcc would help un-screw the results?
The argument is that using the same compiler on both platforms would eliminate compiler differences.
Forcing the use of a POSIX based compiler may not be fair to Windows, though.
The argument is that using the same compiler on both platforms would eliminate compiler differences.
Forcing the use of a POSIX based compiler may not be fair to Windows, though.
The problem with this is that gcc produces different code on Solaris and Windows. So it is not quite apples and apples. It is better to use the compilers which most people use to build their applications. On Windows it is certainly not gcc.
you do realize that there was a time when solaris was dog slow…………. so i really don’ think it is an issue of withering away as it is an issue of kickin ass and taking names.
Solaris has really pushed it’s way to the top again in the last couple years.
oh, and one more thing……….. windows slow?…. well, as much as it hurts me to say this, (as i am a linux and bsd user) windows is not slow, not when it is healthy (free of malware). Windows tends to run faster then most linux os’s and is right on par with most bsd’s…. that being said….. untill recently , wiindows has run much faster then solaris in many areas.
Solaris vs XP (server-centric OS vs desktop-centric server)?? It’s unfair, isn’t it?
Solaris vs Win 2003 R2 would be more interesting and relevant…
How so? They’re comparing the two operating systems in a workstation setting – I hardly see it as Sun’s fault that Microsoft see’s the need to split their product line in half, and screw the end user over with inferior performance.
Solaris is the same, whether used for server or workstation; I’m sure they would be tested a “Solaris Workstation Edition” but none exist.
> Solaris vs Win 2003 R2 would be more interesting and relevant…
That won’t make any difference. They tested against the x64 version of XP, which is internally pretty much the same thing as 2003.
Indeed, the x64 version of XP is actually derived from Windows 2003 sourcebase as oppose to the 32bit version of XP, basically took out the server related stuff and added in XP themes etc.
basically took out the server related stuff and added in XP themes etc.
Actually, you can enable Windows XP Luna theme in Windows 2003. I did not do it because I don’t like luna theme. But at several places you can find how to do it (for example check http://www.msfn.org/win2k3/ ).
if Solaris had a BlueTooth stack and sata support for common chipsets with some more polishing for the existing drivers would anyone need windows? I tried Nexenta and I am very very pleased. We need BlueTooth and SATA I/II support and actually support for more common HW. It seems SUN supports HW they sell (not a bad decision).
“Overall, I’m pleased with the performance of the Sun Ultra 20 M2. While this model (with an Opteron 1210) isn’t quite as fast as a low-end Mac Pro, it’s less than half the price of a low-end Mac Pro. Models equipped with faster Opterons (like the Opteron 1218) will certainly be more competitive when it comes to performance.”
This is such a misleading comparison.
1) The benchmarks use a Power Mac G5 as the baseline. This is a totally different architecture than the Mac Pro.
2)While the low-end Sun box does cost half as much as a low-end Mac Pro, it has half as much RAM, half as much hard drive space and 1/4 as many CPU cores
3) Yes, the model with the Opteron 1218 will be more competitive in performance compared to a low-end Mac Pro. A Mac Pro with 2GB of RAM and 2.6GHz CPUs costs $2800. This is the same price as the Sun box he’s referring to. The difference is that the Mac has 4 cores and the Sun has 2. Ok, so the Opteron is faster per clock cycle, but can 2 Opteron cores do as much work as 4 Xeon cores? Probably not.
Geekbench is written in “Platform neutral C++”.
This is hardly a test of optimized VC++.
Geekbench is written in “Platform neutral C++”.
This is hardly a test of optimized VC++.
And this is hardly a test of optimized SunStudio either, so what’s your point?
Look’s like a “Platform neutral C++” benchmark run on the same hardware and the overall performance rating shows Solaris 10 is 14+% faster than Windows XP x64.
Thanks for pointing that out!