Yes what largethunder said--- where is everyone-on vacation? And where is Hairy4Paws

On his way to a Cool 1/2 Million Points

in just 82 days !!!!

Great going Esteban69

p.s. Looks like we have another fast mover in warken make way he's on a mission

http://gahood.org/es-hof.jpg


http://gahood.org/e69.jpg
Signature banner for you Sir

Here is the code:
{img]gahood.org/e69.jpg[/img]

Enjoy

Some 2 weeks ago I finally received my HP xw6200 workstation (with 2 CPUs
P4-2.8GHz-XEON and 2 GB RAM). These CPUs support Intel's Hyprerthreading
Technology (HTT) which can be enabled/disabled at boot time. I installed
a multi-boot environment that allows me to boot either:

a) the pre-installed Redhat Enterprise Linux (supports both CPUs and HTT)

b) my preferred Linux distribution Debian (supports both CPUs and HTT)

c) WinXP Professional (supports both CPUs and HTT)


native WinXP
============

I ran the WCG client under native WinXP-Pro for a couple of days to see
the output. Overall score was 179, CPU score was 162, memory 300 (maximum),
net 100 (maximum) and disk the usual 196 (maximum).

If I go to "Device Manager" -> "Device Statistics" and for every line divide
"points generated" by "run time" in hours and multiply by 24 I get about
814 points per 24h for each of the reported 9 days of runtime. Of course
neither the CPU score nor the point output changes with HT enabled or
disabled as the client's "work horse" -- the Rosetta task -- is a single
thread and thus (on an otherwise idle system) benefits neither from
enabled HT nor from the second CPU. This would only change if the
limitation to start only one client under Windows would fall.


Virtual W2K
===========

I also installed VMware Workstation 5.0 in my Debian Linux and set up a
single virtual machine (a virtual PC) running Windows 200 Professional with
512 MB RAM. Will increase this to 1152 MB (memory score 300) today to be
able to directly compare with the native XP score. I already see a constant
600-604 points per 24h for the last 3 days for the 512MB RAM -- might
squeeze out a few more with memory score 300. Didn't check the other scores
yet, will report later. The 600 for the virtual machine are a fine 74% of
the native machine which makes me absolutely sure I will see another 600
points if I were to fire up a second virtual machine which will utilize the
second (now pretty idle) CPU.


VMware virtual machines
-----------------------

A virtual machine is an EMULATED PC running in a Linux window (you can switch
to fullscreen, though!) that exactly behaves like a real PC. It has an
(emulated) BIOS which you can enter to do BIOS settings, it can access your
worksation's devices (keyboard, mouse, Floppy, CD, DVD, burner, sound) by
offering an abstraction of the devices (they look like perfect standard
devices to the virtual PC so you never need drivers to rund them -- as soon
as they run under Linux the virtual PC (or PCs, you can run more than one
virtual machine simultaneously) can use the devices. The "harddisk" of each
virtual machine is also virtual -- each HD simply is a file on the Linux
machine that inside Windows looks like your usual C: etc. The emulation is
intelligent, when the machine is "powered down" it can simply be resized,
backed up etc. It also doesn't use up all the space that you defined as its
size. The file grows as needed. After you create a machine (give is a name,
define HD size and assign RAM size) you "boot" the virtual machine which
opens a Windows on your Linux desktop and this shows a booting PC (with BIOS
report of CPU, RAM, etc -- exactly like a normal PC). It tries to boot from
the (still unformatted) virtual HD, Floppy and CD. So you use your REAL
Windows installation CD/Floppies to install a REAL Windows that NEVER EVER
knows that the PC it runs on is only a fake. This means that EVERYTHING runs
on this machines, it's no emulation of Windows (with possible incompatibilities)
bur an emulation of hardware (PC) that runs the real (licensed!) OS and
Software. You can install everything from Windows 98 to XP, the server version
can also run the server versions like W2K Server, W2K Terminal Services, W2003
and so on. As I already said, you can run multiple machines simultaneously.
You can run a W98 machine running a special lab software that doesn't work
with newer Windows versions, can use W2K as your regular system and fire up
XP from time to time to see if you still hate its look&feel. Or run 4 machines
which each has the WGC client installed -- crunching 4 work units at the same
time. Only limitation is physical RAM. This is also handled intelligently but
a W2K machine needs at least 256 MB RAM to run smoothly and Linux also needs
some 200 MB so for running two 256MB machines you need more than 512MB.



Test 1:
-------
First I will alter the current test with the single machine (512 MB RAM) and
configure it to use 1152 MB (maximum memory score 300). That test will run
from this evening until Monday. It will show if I can do better than the
current 74% ("virtual W2K 512MB" versus "native WinXP 2GB").


Test 2:
-------
After I did that I will run two machines with 768 MB each (memory score 200
each) to test the dual CPU output. As the machine is pretty idle apart from
the two machines and as each machine task is single-threaded (can not use
Hyperthreading) I don't expect much difference for HTT enabled vs. disabled
each machine will use one CPU. Will run this for a couple of days.


Test 3:
-------

Now I will lower the two machines' RAM to 384 MB each (memory score 100 each)
to get some numbers for Test 4. Will run this for a couple of days.


Test 4:
-------
Now I will run four machines with 384 MB each (memory score 100 each) with
HTT disabled to test the quality of the multitasking scheduler. I have only
2 non-HTT CPUs now but 4 competing machines -- 4 competing "workhorse" threads.
I expect the about the same output than with only 2 machines as they should
have the same processing power. Possible influence is by a) the overhead for
constant task switches and b) possible score miscalculations from 4 competing
score benchmarks. But these effects will be the same in Test 5.

Test 5:
-------

Will be like Test 4 but now with HTT enabled. I PERSONALLY expect a substantial
gain vs. Test 4 but I'm unsure how large it will be. We'll see, in 2-3 weeks we
will know more.


Best regards, Stefan.

A HUGE CONGRATULATIONS to
Esteban69 on making it to a cool
1/2 MILLION points.
Nice job Esteban on a job well done!