x86-64 or dual MP

Alexander Dalloz ad+lists at uni-x.org
Fri Apr 1 16:22:12 UTC 2005 

Am Fr, den 01.04.2005 schrieb Jon Shorie um 16:21:

> We are upgrading one of our servers this year that is used to run perl apps in 
> apache and mysql databases.  I am looking at two different options that all 
> cost about the same money.  They all have 1GB ECC Ram and dual 80 GB hard 
> drives.
> 
> The only real differences are the cpus.
> 
> One has Dual Athlon MP 2200's

The AMD rating? Then it runs with real 1800 MHz.

> The other has a single Socket 754 Athlon 64 3700+

Has real 2400 MHz, but is the old, dying  socket 754. Socket 939 should
be preferred IMHO.

> Which should give me better performance if I am running Fedora Core 3 on the 
> box?
> 
> Jon Shorie

Not that easy to give a good answer to this question. A multi CPU system
can have advantages against a single CPU system. On a server that can
count.

To both system architectures: the Athlon MP is already some years old.
It is a dying platform. AMD does not let follow any processor type of
that build. Maximum CPU available is an AMD Athlon MP 2800+, 2133MHz,
133MHz FSB, 512kB Cache. (I even think it should be not really more
expensive than the MP 2200.) It is way faster because of doubled 2nd
level cache. The frontside bus speed of the MP is limited to 133/266 MHz
and memory speed to max. 2100 DDR SDRAM. For a dual CPU system meanwhile
a limiting factor. (Though i.e. my board and the still available Tyan
boards have 64bit PCI slots.)
Don't take a board with MP chipset (predecessor), only the MPX one.

The AMD64 is the current platform by AMD with an FSB of 200/400 supporting newer
and faster DDR SDRAM types. Current boards support newer interface types like SATA (II),
PCI-X/PCIe or USB2. You have more choices regarding motherboards. Actually the 64bits do
not count that much, it is the design of the CPU at all which makes it impressing fast. A nice
feature too is the Cool 'n Quite function of the AMD64, throttling it down to 1000 MHz if load is
low with 2 steps between up to maximum speed of 2200 MHz. It keeps the CPU nicely cool at
< 30 °C. That can be a factor if using a 1U 19" case - big advantage over the Athlon MP which
you can't cool down that good, especially not in 1U.

Here at home I am running a dual Athlon MP system (ASUS A7M266-D with 2 Athlon XP-M
- yes they are mobiles - 2500+ at ~ 2150 MHz). Info from /proc/cpuinfo:
model name      : AMD Athlon(TM) MP 2800+
cpu MHz         : 2151.017
The system runs rock solid 24/7, CPU temperature between 40 and 50°C, depending on load.
Running multiple applications in parallel works smoothly.
On the other hand I have a new AMD64 system which I set up to run as a server for mail and
web hosting. It is a socket 939 AMD64 3500+ (Winchester core) on an ASUS A8N-SLI Deluxe.
I have it now for 3 days doing installation and some testing and behaves very well (compared to
the nightmares I had with the Intel SE7221BK entry server board with an Intel P IV 630 it is fun).

Now a small comparison between both systems running an OpenSSL test. It has to do with real
life as it shows the speed the CPUs can handle for instance HTTPS sessions on a webserver.

a) My dual Athlon MP ~3000 on Fedora Core 3 (32bit)
- single
$ openssl speed rsa1024
Doing 1024 bit private rsa's for 10s: 3685 1024 bit private RSA's in 9.97s
Doing 1024 bit public rsa's for 10s: 70175 1024 bit public RSA's in 9.99s
OpenSSL 0.9.7a Feb 19 2003
built on: Thu Mar 25 10:34:04 EST 2004
options:bn(64,32) md2(int) rc4(idx,int) des(ptr,risc1,16,long) aes(partial) blowfish(idx)
compiler: gcc -fPIC -DZLIB -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -DKRB5_MIT -DOPENSSL_NO_IDEA -DOPENSSL_NO_MDC2 -DOPENSSL_NO_RC5 -DOPENSSL_NO_EC -I/usr/kerberos/include -DL_ENDIAN -DTERMIO -Wall -O2 -g -pipe -march=i686 -Wa,--noexecstack -DSHA1_ASM -DMD5_ASM -DRMD160_ASM
available timing options: TIMES TIMEB HZ=100 [sysconf value]
timing function used: times
                  sign    verify    sign/s verify/s
rsa 1024 bits   0.0027s   0.0001s    369.6   7024.5
- dual
$ openssl speed rsa1024 -multi 2
Forked child 0
+DTP:1024:private:rsa:10
+DTP:1024:private:rsa:10
Forked child 1
+R1:3693:1024:10.00
+DTP:1024:public:rsa:10
+R1:2792:1024:10.00
+DTP:1024:public:rsa:10
+R2:70524:1024:10.00
Got: +F2:1:1024:0.002708:0.000142 from 0
+R2:51510:1024:10.00
Got: +F2:1:1024:0.003582:0.000194 from 1
OpenSSL 0.9.7a Feb 19 2003
built on: Thu Mar 25 10:34:04 EST 2004
options:bn(64,32) md2(int) rc4(idx,int) des(ptr,risc1,16,long) aes(partial) blowfish(idx)
compiler: gcc -fPIC -DZLIB -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -DKRB5_MIT -DOPENSSL_NO_IDEA -DOPENSSL_NO_MDC2 -DOPENSSL_NO_RC5 -DOPENSSL_NO_EC -I/usr/kerberos/include -DL_ENDIAN -DTERMIO -Wall -O2 -g -pipe -march=i686 -Wa,--noexecstack -DSHA1_ASM -DMD5_ASM -DRMD160_ASM
available timing options: TIMES TIMEB HZ=100 [sysconf value]
timing function used:
                  sign    verify    sign/s verify/s
rsa 1024 bits   0.0015s   0.0001s    648.4  12196.9

b) AMD64 3500+ on CentOS4 x86_64
$ openssl speed rsa1024
Doing 1024 bit private rsa's for 10s: 8958 1024 bit private RSA's in 9.99s
Doing 1024 bit public rsa's for 10s: 164103 1024 bit public RSA's in 10.00s
OpenSSL 0.9.7a Feb 19 2003
built on: Thu Feb 17 16:14:37 CST 2005
options:bn(64,64) md2(int) rc4(ptr,char) des(idx,cisc,16,int) aes(partial) blowfish(ptr2)
compiler: gcc -fPIC -DZLIB -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -DKRB5_MIT -DOPENSSL_NO_IDEA -DOPENSSL_NO_MDC2 -DOPENSSL_NO_RC5 -DOPENSSL_NO_EC -I/usr/kerberos/include -DL_ENDIAN -DTERMIO -Wall -DMD32_REG_T=int -O2 -g -pipe -m64 -Wa,--noexecstack
available timing options: TIMES TIMEB HZ=100 [sysconf value]
timing function used: times
                  sign    verify    sign/s verify/s
rsa 1024 bits   0.0011s   0.0001s    896.7  16410.3

Compare the values at end:
Athlon MP single: 7.024,5 
Athlon MP dual: 12.196,9
AMD64 (single): 16.410,3   <-- just with the single CPU!

Alexander


-- 
Alexander Dalloz | Enger, Germany | GPG http://pgp.mit.edu 0xB366A773
legal statement: http://www.uni-x.org/legal.html
Fedora Core 2 GNU/Linux on Athlon with kernel 2.6.10-1.771_FC2smp 
Serendipity 18:22:05 up 2 days, 15:48, load average: 0.30, 0.44, 0.53 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: Dies ist ein digital signierter Nachrichtenteil
URL: <http://listman.redhat.com/archives/fedora-list/attachments/20050401/4ebfdd8e/attachment-0001.sig>

More information about the fedora-list mailing list