Alpha 1000 4/233 Performance ?

Alpha 1000 4/233 Performance ?

Post by SteFet » Tue, 08 Feb 2005 23:27:53


Hi

Opinion needed.

How fast this server run if I compare with computers of today ?

For example a Penthium 3GHZ 32bits computer ???

Thanks
 
 
 

Alpha 1000 4/233 Performance ?

Post by Sylvain Ro » Wed, 09 Feb 2005 03:50:56


It runs at 233MHz, regardless of what you compare it to.


The Pentium operates nearly 13 times as fast.

I'm not being sarcastic, what you're asking for isn't a fair comparison.

If you want to compare an AS-1000 4/233 to anything, compare it to
either a computer that was current at around the same time, or at least
to one with similar processor characteristics (either a different Alpha,
or a Pentium operating at 233 MHz), then establish what you would like to
compare the two systems doing. There are things that Alphas do "faster"
or "better" than Pentiums, and there are things that Pentiums are better
at. Likewise, there are some circumstances where you're better off with
the 64 bit chip, and others where you're better off with the 32 bit chip.

If you're looking at benchmarks, are the benchmarks compiled with the
same compiler on the same OS in both cases? Is the compiler optimzed
more for one of the processors than for the other? Is the OS optimized
more for one than for the other?

I hope I've helped ...

--
----------------------------------------------------------------------
Sylvain Robitaille XXXX@XXXXX.COM

Systems analyst Concordia University
Instructional & Information Technology Montreal, Quebec, Canada
----------------------------------------------------------------------

 
 
 

Alpha 1000 4/233 Performance ?

Post by SteFet » Wed, 09 Feb 2005 05:01:26

Sylvain Robitaille < XXXX@XXXXX.COM > wrote in



Oui c'est vrai c pas fort comme comparaison.

J'aimerais savoir si un serveur Web mont?en Apache / Linux / mySql et je
va exuter mon application plus rapidement sur un AS-1000 que sur un PC
standard.

J'ai da ce serveur rouler 50 programmes avec 30 usagers pendant une
semaine dans l'usine ou je travaille.

J'imagine qu'il peut peut-etre ger 100 usagers dans une application Web
 
 
 

Alpha 1000 4/233 Performance ?

Post by Sylvain Ro » Wed, 09 Feb 2005 07:56:26

(I'm going to stick to replying in English; my written French is not
something I willingly subject others to ...)




The Alpha will likely be "fine" as a web server, but if you have a 3GHz
Pentium as your "other" choice, it will very likely give you better
performance. Consider that Apache, Linux, and MySQL are generally
developped on 32-bit Intel systems, and the compiler you would use to
install your software is more advanced on 32-bit Intel chips than on the
Alpha. In my own experience, these packages run better on 32-bit Intel
than on Alpha, even with DEC/OSF and DEC's own compiler.


I'm sure it's "fine" that way, but if you put up the Pentium along with
it and asked users to report back on relative performance, I'm quite
certain that you'll find the users feel the Pentium responds faster.
(assuming both systems have reasonably fast disks, equivalent memory,
similar network connection, etc...) The performance will of course also
be very dependant on the application itself.


Both systems should be acceptable for that. If you have the Alpha already
running, and it isn't causing you any performance problems, you probably
can hold off spending money for a new system. If, on the other hand,
you're wondering whether a new Pentium-based system will improve the
performance of your application compared to the Alpha, the answer is
"probably, yes" (given the software that you've indicated you're using;
only experimentation will provide a definite answer for you, but I would
say that chances are very good you'll see a remarkable improvement).

If you can pick up a system to use for proof-of-concept, it would be
worth doing so. At the very worst, that system would become available
for redeployment in another application if you don't end up using it here.

The Alpha, of course would still be usable for less performance-critical
applications, or for something that's more optimized for a 64-bit RISC
processor, depending on what your organization's needs are.

--
----------------------------------------------------------------------
Sylvain Robitaille XXXX@XXXXX.COM

Systems analyst Concordia University
Instructional & Information Technology Montreal, Quebec, Canada
----------------------------------------------------------------------
 
 
 

Alpha 1000 4/233 Performance ?

Post by Ms Rullg » Wed, 09 Feb 2005 08:05:29

SteFetS < XXXX@XXXXX.COM > writes:



Switching to French isn't likely to give you any better answers.
However, what you describe sounds like it will depend at least as much
on the amount of RAM in the machine as on the CPU speed.

--
Ms Rullgd
XXXX@XXXXX.COM
 
 
 

Alpha 1000 4/233 Performance ?

Post by Christophe » Wed, 09 Feb 2005 09:30:52


It's wimpy by now.

You'll find that it's easy to get Intel/AMD hardware that _way_
outperforms any Alpha hardware you can get economically.

Unfortunately, Digital essentially gave up on the line 3 years ago, so
that there's not newer, greater hardware, at least not at reasonable
prices.

If you're doing a lot of floating point work, it'll look a bit better
than if you're just doing I/O or integer stuff, but it won't be real
fast in any case.
--
select 'cbbrowne' || '@' || 'ntlug.org';
http://www.yqcomputer.com/
Rules of the Evil Overlord #118. "If I have equipment which performs
an important function, it will not be activated by a lever that
someone could trigger by accidentally falling on when fatally
wounded." < http://www.yqcomputer.com/ ;
 
 
 

Alpha 1000 4/233 Performance ?

Post by David J Da » Wed, 09 Feb 2005 10:53:38


I'd watch for ES40 Alphas, even ES45s coming out of service, etc. They
are cosndierably faster than 233MHz and can even support Fibre Channel
HBAs.

--
David J Dachtera
dba DJE Systems
http://www.yqcomputer.com/

Unofficial OpenVMS Hobbyist Support Page:
http://www.yqcomputer.com/

Unofficial Affordable OpenVMS Home Page:
http://www.yqcomputer.com/

Unofficial OpenVMS-IA32 Home Page:
http://www.yqcomputer.com/

Coming soon:
Unofficial OpenVMS Marketing Home Page
 
 
 

Alpha 1000 4/233 Performance ?

Post by DiskMa » Thu, 10 Feb 2005 01:48:38

Its not just the mhz that matter but also, something I didn't see mentioned
here, the chip version.
The Alpha Server 4/233 is a EV4, which translates in to SLOW...Speedwise,
the Alpha EV4 233mhz cpu is about the equivelent to an Intel Pentium 200mhz.
A Pentium 200mmx who's motherboard will accept all the latest hardware [ ***
the SRM console] and will run Linux, Solaris, Windows, or NT4 [wouldn't dare
try WinXP, it would run in slow motion]

Here are some comparison tests of the AlphaServer versus a Pentium 90mhz,
its nothing to be get e *** d over, trust me on this:
http://www.yqcomputer.com/

The Alpha EV56's aren't much better. I went with a AlphaPC 164SX several
years ago and when it comes to speed, it barely bucks a Pentium/AMD 600mhz,
and thats with tweaking to the high hell! About the only thing it places
good [64bit vs 32bit and mhz to mhz] is in the area of floating point. You
really don't see that 2x boost [64bit+mhz = 2x32bit+mhz ratio until you hit
the EV6's.]

Personally Alpha's are way over priced. I wouldn't pay more than $200 for
ANY alpha box out there, thats the way I am. For $200, you can get a AMD
Dual Motherboard + cpu's. What would that get you:

Asus A7M266-D motherboard $100
Two Athlon 1700+ $10 ea.
Memory extra

For well under $200, you could have a rocking DUAL AMD 1700+ for next to
nothing! If you toss in another $20, you could change that to DUAL 2000+mhz
cpu's...get my drift? Alpha's have pretty much lost the 'speed to cost'
ratio long ago :-(

Tell you something funny, I had a snapperhead try to sell me a UP1500
w/800mhz EV67 for $1400 [not incl. $200 shipping]. I about fell out of my
chair laughing! I guess he was thinking it was something like a 'Mac
Classic' or something. He kept trying to equate the price to the WAY newer
EV7 1200mhz selling for 10grand!!! I had to point out, that they only sold
about 5 of them WORLDWIDE since its intro a year or so ago! Will L G
 
 
 

Alpha 1000 4/233 Performance ?

Post by John Beard » Thu, 10 Feb 2005 18:10:11

In message <B%5Od.4711$ XXXX@XXXXX.COM >, DiskMan
< XXXX@XXXXX.COM > writes


Broadly I agree, though one benchmark we did a few years ago with a C++
simulation running on linux console and w2k command prompt made the
190MHz EV4 about 1.6 times faster than a PII/366.

We've just brought the Alpha box back into service as an office server.

When I started the process all I knew was that it needed a new power
supply and that a couple of its 2 gig disks were ill.

If I'd also know that I'd have to spend ages cleaning the memory and CPU
cards to make it reliable, some time sorting out a mysterious SCSI
error, and still wouldn't have found time to have built a kernel, I'd
probably have used an old PII box instead ! None the less I quite like
the Alpha, though it's a bit on the noisy side to have in an office
where people need to talk. Keeps it warm though !


Cheers, J/.
--
John Beardmore
 
 
 

Alpha 1000 4/233 Performance ?

Post by Atro Tossa » Thu, 10 Feb 2005 21:08:22

ohn Beardmore < XXXX@XXXXX.COM > writes:


It's all in what you need to do. Take OpenSSL "speed", for example.

500 MHz 21264 in a six year old XP1000 running Tru64 vs. 3 GHz P4 in
a current PC running Linux. SHA1, MD5 and RMD160 are done in hand-
optimised assembly on x86. There are no assy optimisations on the Alpha.

The long DSA and RSA results are noteworthy. The rest is not so glorious...

Alpha:
sign verify sign/s verify/s
dsa 512 bits 0.0008s 0.0009s 1316.8 1118.1 87%/90% of P4
dsa 1024 bits 0.0019s 0.0023s 539.6 434.3 104%/103%
dsa 2048 bits 0.0055s 0.0068s 183.1 148.0 118%/115%

P4:
sign verify sign/s verify/s
dsa 512 bits 0.0007s 0.0008s 1499.2 1241.8
dsa 1024 bits 0.0019s 0.0024s 515.9 419.0
dsa 2048 bits 0.0065s 0.0078s 153.8 128.3

Alpha:
sign verify sign/s verify/s
rsa 512 bits 0.0012s 0.0001s 869.0 9330.3 63%/67% of P4
rsa 1024 bits 0.0038s 0.0002s 261.8 4273.2 107%/87%
rsa 2048 bits 0.0202s 0.0006s 49.5 1585.5 117%/102%
rsa 4096 bits 0.1327s 0.0020s 7.5 497.8 119%/119%

P4:
sign verify sign/s verify/s
rsa 512 bits 0.0007s 0.0001s 1379.4 13756.7
rsa 1024 bits 0.0041s 0.0002s 242.8 4905.2
rsa 2048 bits 0.0238s 0.0006s 42.0 1553.9
rsa 4096 bits 0.1575s 0.0024s 6.3 415.7

type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes

md2 (Alpha) 326.70k 692.50k 963.65k 1071.25k 1097.09k
md2 (P4) 2967.98k 6404.57k 9074.13k 10151.59k 10452.99k

mdc2 (Alpha) 1372.68k 1646.66k 1759.23k 1801.32k 1802.24k
mdc2 (P4) 7285.78k 8423.53k 8900.95k 8989.01k 9087.66k

md4 (Alpha) 3238.27k 10730.84k 28462.42k 48420.86k 64001.37k
md4 (P4) 22229.20k 74494.44k 202068.65k 355311.96k 455953.07k

md5 (Alpha) 2669.76k 8428.05k 21160.70k 34425.51k 44440.00k
md5 (P4) 19019.83k 63750.59k 168593.41k 286971.90k 362463.23k

hmac(md5) (Alpha)4099.89k 12058.43k 26732.49k 38147.07k 44998.66k
hmac(md5) (P4) 19598.27k 67915.05k 176754.26k 294326.27k 366439.94k

sha1 (Alpha) 2814.07k 8122.05k 19526.40k 30393.34k 38073.69k
sha1 (P4) 19533.10k 59483.86k 146098.26k 231772.70k 279972.52k

rmd160 (Alpha) 1959.74k 5431.36k 11434.84k 16448.17k 19243.42k
rmd160 (P4) 13415.96k 35985.32k 73150.55k 98276.01k 108672.34k

rc4 (Alpha) 3608.24k 56849.34k 57121.88k 57754.62k 58273.95k
rc4 (P4) 98102.41k 108134.23k 111079.25k 112207.19k 112679.73k

des cbc (Alpha) 8160.36k 8551.53k 8761.81k 8878.42k 8921.09k
des cbc (P4) 50602.54k 50963.26k 50911.49k 51214.38k 51190.41k

des ede3 (Alpha) 3350.09k 3439.59k 3479.72k 3512.66k 3522.56k
des ede3 (P4) 19841.59k 20019.91k 19719.25k 19894.61k 19917.48k

idea cbc (Alpha) 7704.99k 8156.05k 8229.38k 8349.70k 8396.80k
idea cbc (P4) 28745.28k 29795.22k 30034.52k 30206.97k 30127.45k

rc2 cbc (Alpha) 6152.74k 6450.65k 6571.76k 6638.25k 6671.02k
rc2 cbc (P4) 27028.58k 28179.86k 28519.68k 27838.46k 28355.24k

rc5-32/12 cbc
 
 
 

Alpha 1000 4/233 Performance ?

Post by DiskMa » Fri, 11 Feb 2005 02:30:50

ere you go, I have the latest [or close to it] on the Alpha and a 'semi'
working [the best of 3 different distros] for the AMD 1.4+WinXP optimized
for the i486!

AlphaPC 164SX 533mhz [21164 pca56]
256mb ram
Kernel-2.6.10
compiler:
gcc -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -DOPENSSL_NO_K
RB5 -O3 -mtune=ev56 -mcpu=ev56 -mieee -mmax -ffast-math -funroll-loops -fstr
ict-aliasing -finline -fomit-frame-pointers -O3 -DL_ENDIAN -DTERMIO
available timing options: TIMES TIMEB HZ=1024 [sysconf value]
timing function used: times
sign verify sign/s verify/s
rsa 512 bits 0.0042s 0.0004s 240.1 2442.9
rsa 1024 bits 0.0138s 0.0008s 72.3 1261.8
rsa 2048 bits 0.0754s 0.0022s 13.3 456.9
rsa 4096 bits 0.4627s 0.0071s 2.2 140.2

timing function used: times
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192
bytes
md5 1114.64k 4140.59k 12908.90k 26378.88k
38916.18k

The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192
bytes
aes-128 cbc 7418.33k 7769.91k 8064.17k 8156.39k
8150.18k
aes-192 cbc 6347.12k 6641.00k 6785.82k 6830.34k
6820.13k
aes-256 cbc 6439.38k 6265.19k 6378.92k 6425.50k
6411.87k

AMD 1700 [1.4ghz]
512mb PC2100ram
WinXP SP1
compiler:
gcc -DL_ENDIAN -DDSO_WIN32 -fomit-frame-pointer -O3 -mcpu=i486 -Wall -
DBN_ASM -DMD5_ASM -DSHA1_ASM -DOPENSSL_NO_IDEA -DOPENSSL_NO_KRB5
available timing options: TIMEB HZ=1000
timing function used: ftime
sign verify sign/s verify/s
rsa 512 bits 0.0009s 0.0001s 1167.1 13313.4
rsa 1024 bits 0.0040s 0.0002s 251.8 4766.2
rsa 2048 bits 0.0229s 0.0007s 43.7 1456.0
rsa 4096 bits 0.1539s 0.0023s 6.5 426.1

The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192
bytes
md4 11484.16k 40064.99k 114676.80k 212638.99k
285387.47k

The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192
bytes
aes-128 cbc 0.00 0.00 0.00 0.00
0.00
aes-192 cbc 0.00 0.00 0.00 0.00
0.00
aes-256 cbc 0.00 0.00 0.00 0.00
0.00


"Atro Tossavainen" <Atro.Tossavainen+ XXXX@XXXXX.COM > wrote in
message news: XXXX@XXXXX.COM ...
glorious...
bytes
1097.09k
10452.99k
1802.24k
9087.66k
64001.37k
455953.07k
44440.00k
362463.23k
44998.66k
366439.94k
38073.69k
279972.52k
19243.42k
108672.34k
58273.95k
112679.73k
8921.09k
51190.41k
3522.56k
19917.48k
8396.80k
30127.45k
6671.02k
28355.24k
17197.71k
105580.91k
15179.78k
89986.39k
11886.59k
74757.46k
18496.53k
75669.50k
16459.06k
67518.46k
14999.55k
60896.60k
own.
ATTACHMENTS


 
 
 

Alpha 1000 4/233 Performance ?

Post by jalyv » Tue, 15 Feb 2005 12:19:18


Is this right? 60 megabytes/sec on a 3Gig PC?

I tried some simple Itanium2 (1.5GHz) software (SSH) and got only 1/4 of that.

What is your secret sauce?

Yes this may be ot.


--J
 
 
 

Alpha 1000 4/233 Performance ?

Post by Atro Tossa » Tue, 15 Feb 2005 17:10:31


XXXX@XXXXX.COM (John) writes:


I'll repeat the test for you.

OpenSSL> speed aes-256-cbc
Doing aes-256 cbc for 3s on 16 size blocks: 11989878 aes-256 cbc's in 3.00s
Doing aes-256 cbc for 3s on 64 size blocks: 2783465 aes-256 cbc's in 3.00s
Doing aes-256 cbc for 3s on 256 size blocks: 689818 aes-256 cbc's in 3.00s
Doing aes-256 cbc for 3s on 1024 size blocks: 178203 aes-256 cbc's in 3.00s
Doing aes-256 cbc for 3s on 8192 size blocks: 21937 aes-256 cbc's in 3.00s
OpenSSL 0.9.7e 25 Oct 2004
built on: Mon Jan 3 15:51:10 EET 2005
options:bn(32,32) md2(int) rc4(ptr,int) des(ptr,risc1,16,long) aes(partial) idea(int) blowfish(idx)
compiler: gcc -DOPENSSL_THREADS -D_REENTRANT -DDSO_DLFCN -DHAVE_DLFCN_H -DOPENSSL_NO_KRB5 -DL_ENDIAN -DTERMIO -O3 -fomit-frame-pointer -m486 -Wall -DSHA1_ASM -DMD5_ASM -DRMD160_ASM
available timing options: TIMES TIMEB HZ=100 [sysconf value]
timing function used: times
The 'numbers' are in 1000s of bytes per second processed.
type 16 bytes 64 bytes 256 bytes 1024 bytes 8192 bytes
aes-256 cbc 63946.02k 59380.59k 58864.47k 60826.62k 59902.63k


SSH does a lot more than just encrypt traffic with AES (or any given
cipher, for that matter), which means you're comparing apples with
oranges. Run the corresponding test to see corresponding values.

Also, there are likely no assy optimisations at all for the Itanic in
either OpenSSL or any brand of SSH. Also, your software may have been
compiled with gcc, whose Itanic performance is unknown to me, but at
any rate, you probably have access to better compilers, so you should
try recompiling speed-hungry apps with them.

--
Atro Tossavainen (Mr.) / The Institute of Biotechnology at
Systems Analyst, Techno-Amish & / the University of Helsinki, Finland,
+358-9-19158939 UNIX Dinosaur / employs me, but my opinions are my own.
< URL : http : / / www . helsinki . fi / %7E atossava / > NO FILE ATTACHMENTS