solaris x86 10 與 Linux x86_64 的效能對比(轉)

solaris x86 10 與 Linux x86_64 的效能對比(轉)[@more@]« MacBook Pro Core 2 Duo Performance | Home
Sun Ultra 20 M2 Linux Performance

On Monday I posted Geekbench results for my Sun Ultra 20 M2 running Solaris and Windows. Afterwards, I received a number of requests asking how Linux performed on the same hardware.

Now that I’ve finally managed to download Fedora Core 6 (arguably the “official” Linux of Geek Patrol), here are the Geekbench results for Fedora Core 6 (and Solaris, as a comparison) on a Sun Ultra 20 M2.
Setup

Here’s the configuration of my Sun Ultra 20 M2:

* AMD Dual-Core Opteron 1210
* 512 MB DDR2-667 RAM (1 DIMM)
* Solaris 10 (6/06)
* Fedora Core 6 (x86_64)
* Geekbench 2006 (Build 230)

I’m reporting the baseline score, rather than the raw score, for each benchmark (where a score of 100 is the score a Power Mac G5 1.6GHz would receive). Higher is better.
Results
Overall Score
Sun Ultra 20 M2 (Solaris)
172.8
Sun Ultra 20 M2 (Linux)

160.4
Integer Performance

Emulate 6502 (single-threaded scalar)

Sun Ultra 20 M2 (Solaris)
87.3
Sun Ultra 20 M2 (Linux)
65.3
Emulate 6502 (multi-threaded scalar)

Sun Ultra 20 M2 (Solaris)
174.7
Sun Ultra 20 M2 (Linux)
121.9
Blowfish (single-threaded scalar)

Sun Ultra 20 M2 (Solaris)
154.3
Sun Ultra 20 M2 (Linux)
169.3
Blowfish (multi-threaded scalar)

Sun Ultra 20 M2 (Solaris)
308.7
Sun Ultra 20 M2 (Linux)
334.2
bzip2 Compress (single-threaded scalar)

Sun Ultra 20 M2 (Solaris)
208.2
Sun Ultra 20 M2 (Linux)
159.7
bzip2 Compress (multi-threaded scalar)

Sun Ultra 20 M2 (Solaris)
418.1
Sun Ultra 20 M2 (Linux)
267.6
bzip2 Decompress (single-threaded scalar)

Sun Ultra 20 M2 (Solaris)
144.4
Sun Ultra 20 M2 (Linux)
212.3
bzip2 Decompress (multi-threaded scalar)

Sun Ultra 20 M2 (Solaris)
297.7
Sun Ultra 20 M2 (Linux)
290.0
Floating Point Performance

Mandelbrot (single-threaded scalar)

Sun Ultra 20 M2 (Solaris)
147.5
Sun Ultra 20 M2 (Linux)
119.8
Mandelbrot (multi-threaded scalar)

Sun Ultra 20 M2 (Solaris)
295.0
Sun Ultra 20 M2 (Linux)
237.2
Dot Product (single-threaded scalar)

Sun Ultra 20 M2 (Solaris)
168.2
Sun Ultra 20 M2 (Linux)
171.0
Dot Product (multi-threaded scalar)

Sun Ultra 20 M2 (Solaris)
329.9
Sun Ultra 20 M2 (Linux)
170.3
JPEG Compress (single-threaded scalar)

Sun Ultra 20 M2 (Solaris)
103.3
Sun Ultra 20 M2 (Linux)
90.5
JPEG Compress (multi-threaded scalar)

Sun Ultra 20 M2 (Solaris)
205.8
Sun Ultra 20 M2 (Linux)
165.6
JPEG Decompress (single-threaded scalar)

Sun Ultra 20 M2 (Solaris)
100.7
Sun Ultra 20 M2 (Linux)
84.3
JPEG Decompress (multi-threaded scalar)

Sun Ultra 20 M2 (Solaris)
199.5
Sun Ultra 20 M2 (Linux)
161.8
Memory Performance

Read Sequential (single-threaded scalar)

Sun Ultra 20 M2 (Solaris)
158.9
Sun Ultra 20 M2 (Linux)
160.5
Read Sequential (multi-threaded scalar)

Sun Ultra 20 M2 (Solaris)
170.4
Sun Ultra 20 M2 (Linux)
163.8
Write Sequential (single-threaded scalar)

Sun Ultra 20 M2 (Solaris)
177.2
Sun Ultra 20 M2 (Linux)
155.0
Write Sequential (multi-threaded scalar)

Sun Ultra 20 M2 (Solaris)
240.4
Sun Ultra 20 M2 (Linux)
196.4
Stdlib Allocate (single-threaded scalar)

Sun Ultra 20 M2 (Solaris)
102.1
Sun Ultra 20 M2 (Linux)
211.9
Stdlib Allocate (multi-threaded scalar)

Sun Ultra 20 M2 (Solaris)
43.4
Sun Ultra 20 M2 (Linux)
329.7
Stdlib Write (single-threaded scalar)

Sun Ultra 20 M2 (Solaris)
150.7
Sun Ultra 20 M2 (Linux)
58.7
Stdlib Write (multi-threaded scalar)

Sun Ultra 20 M2 (Solaris)
173.1
Sun Ultra 20 M2 (Linux)
91.7
Stdlib Copy (single-threaded scalar)

Sun Ultra 20 M2 (Solaris)
161.0
Sun Ultra 20 M2 (Linux)
77.1
Stdlib Copy (multi-threaded scalar)

Sun Ultra 20 M2 (Solaris)
167.9
Sun Ultra 20 M2 (Linux)
130.7
Stream Performance

Stream Copy (single-threaded scalar)

Sun Ultra 20 M2 (Solaris)
115.5
Sun Ultra 20 M2 (Linux)
133.4
Stream Copy (multi-threaded scalar)

Sun Ultra 20 M2 (Solaris)
150.1
Sun Ultra 20 M2 (Linux)
171.1
Stream Scale (single-threaded scalar)

Sun Ultra 20 M2 (Solaris)
115.5
Sun Ultra 20 M2 (Linux)
136.9
Stream Scale (multi-threaded scalar)

Sun Ultra 20 M2 (Solaris)
149.3
Sun Ultra 20 M2 (Linux)
178.3
Stream Add (single-threaded scalar)

Sun Ultra 20 M2 (Solaris)
98.8
Sun Ultra 20 M2 (Linux)
148.3
Stream Add (multi-threaded scalar)

Sun Ultra 20 M2 (Solaris)
124.9
Sun Ultra 20 M2 (Linux)
160.6
Stream Triad (single-threaded scalar)

Sun Ultra 20 M2 (Solaris)
101.8
Sun Ultra 20 M2 (Linux)
144.5
Stream Triad (multi-threaded scalar)

Sun Ultra 20 M2 (Solaris)
132.1
Sun Ultra 20 M2 (Linux)
177.9
Conclusion

Solaris, once again, is faster overall (although only 7.5% faster than Linux, as opposed to 15% faster than Windows); Solaris (and the Sun Studio compiler) is faster in the floating point category while Linux (and GCC) is faster in the stream category.

If you’re using Linux for processor-intensive tasks, it might be worth checking out Solaris (and the Sun Studio compiler) to see if your tasks run faster under Solaris rather than Linux.
Digg It | Add to del.icio.us
Trackbacks & Pingbacks

1. Geek Patrol | Sun Ultra 20 M2 Performance pingbacked Posted October 27, 2006, 11:59 am

Comments

1.
1 Openwookie says:

Cool! Nicely done.

Now how about trying a BSD or two, (FreeBSD at least) and put everything on the same graph? It would be cool to get a snapshot of where all the OS families currently stand in respect to performance.
↓ Quote | Posted October 27, 2006, 2:20 pm
2.
2 Diego says:

Mmh, this benchmark seems to be designed to compare “systems” (machines with different hardware and same software), more than OSes - most of the benchmarks can be heavily affected by the compiler used.
↓ Quote | Posted October 27, 2006, 3:04 pm
3.
3 David Pacheco says:

It’s interesting that Solaris is particularly faster for multi-threaded cases.
↓ Quote | Posted October 27, 2006, 3:17 pm
4.
4 David Pacheco says:

I forgot to ask: were you using ZFS on the Solaris box? That might impact the performance of the read/write benchmarks a lot.
↓ Quote | Posted October 27, 2006, 3:23 pm
5.
5 Sebastián Benítez says:

Diego said:

Mmh, this benchmark seems to be designed to compare “systems” (machines with different hardware and same software), more than OSes - most of the benchmarks can be heavily affected by the compiler used.

True. It does a direct comparisson of compilers. Unless you expect to use GCC on Solaris and VStudio on Linux (if it were available), it doesn’t make sense as an Operating System benchmark.

It would be interesting to do a system performance benchmark, like accessing files, transfering bytes through the network, measuring a database and web server scalability, testing multiprocessing performance and high availability.
↓ Quote | Posted October 27, 2006, 3:39 pm
6.
6 Passenger says:

There is an increase in the Solaris’ score with respect to the previous benchmark in the same hardware/OS (170.5 vs 172.. I guess this is caused by the system not being completely idle in one or both benchmarks, or different initial conditions. It would be interesting to have something like 171.6+/-1.1 (maybe running the tests several times?)…
↓ Quote | Posted October 27, 2006, 3:59 pm
7.
7 John Poole says:

Passenger said:

There is an increase in the Solaris’ score with respect to the previous benchmark in the same hardware/OS (170.5 vs 172.. I guess this is caused by the system not being completely idle in one or both benchmarks, or different initial conditions.

I flashed the Ultra 20’s BIOS on Tuesday, so I re-ran the Solaris test just in case the new BIOS made a difference (hence the new Solaris numbers).
↓ Quote | Posted October 27, 2006, 4:16 pm
8.
8 Jeff Schroeder says:

Maybe you should try this with SELinux disabled again. Solaris doesn’t have the Trusted Extensions (mandatory access control) enabled by default where fedora has SELinux (mandatory access control) enabled by default.

The massive amounts of added security mandatory access control come at a 5-8% performance cost.
↓ Quote | Posted October 27, 2006, 7:28 pm