Congratulations to Tomwp on reaching 100,000 MOT points!!!

Yea im in 113 now, was 120 last night, only problem now is my school network is down, something about getting a new router. So my points an research will be delayed

But atleast its still running.

.................A BIG ”My Online Team” WELCOME.................
.......................to our newest team member! ......................
.................................. PioPicone ..............................

Although a W2K machine -- with only the WCG client working -- uses only
about 320KB RAM, the score (and thus the point output) is influenced by
(unused!) additional RAM. Dramatically!

BOTTOM LINE: ADDED RAM EASILY INCREASES POINT GAIN

Here my overdue scores for "the workstation":

physical: 2GB RAM, dual Intel P4 XEON 2.8 GHz, all WCG partitions > 11GB

------------------------------------------------------
SETTING 1
------------
native WinXP, 2GB (WCG counts max. 1152):

CPU score: 161
Memory: 300
Storage: 196
Network: 100
--------------------
Total score: 179

Point output per 24h: 814 points
------------------------------------------------------

In virtual machines (VMware) the CPU score goes down to 155 which is
a loss of 6/161 or 3.7% only:

------------------------------------------------------
SETTING 2
------------
1 x virtual W2K, 512MB RAM assigned

CPU score: 155
Memory: 134
Storage: 196
Network: 100
--------------------
Total score: 135

Point output per 24h: 602 points
------------------------------------------------------

CPU score loss from SETTING 1 to SETTING 2: 6/161 or 3.7%
Total score loss from SETTING 1 to SETTING 2: 44/179 or 24.6%
Points per 24h loss from SETTING 1 to SETTING 2: 212/814 or 26.0%
Memory score loss from SETTING 1 to SETTING 2: 166/300 or 55.3%

Now I gave the virtual machine 1152 MB RAM (more does not count)
although it does NEVER use it and it NEVER runs any faster than
before:

------------------------------------------------------
virtual W2K, 1152MB

CPU score: 155
Memory: 300
Storage: 196
Network: 100
--------------------
Total score: 176

Point output per 24h: 800 points
------------------------------------------------------

First I compare SETTING 2 and SETTING 3 and check the gain after
adding more memory (everything else is identical):

CPU score gain from SETTING 2 to SETTING 3: none, scores identical
Total score gain from SETTING 2 to SETTING 3: 41/135 or 30.4%
Points per 24h gain from SETTING 2 to SETTING 3: 198/602 or 32.9%
Memory score gain from SETTING 2 to SETTING 3: 166/300 or 55.3%

As expected the point gain is roughly about the same as the total score
gain. Although the memory is never needed, adding 640MB of RAM
pushed points gain per 24h by about one third!!! This clearly is a flaw
in the points calculation as the efficiency of the machine is not altered
and crunching is as fast as before.

Now I compare the max-RAM (1152 MB) virtual machine with the
native WinXP machine (assuming that the client is as effective under
W2K Professional and XP Professional on an otherwise idle PC).

CPU score loss from SETTING 1 to SETTING 3: 6/161 or 3.7%
Total score loss from SETTING 1 to SETTING 3: 3/179 or 1.7%
Points per 24h loss from SETTING 1 to SETTING 3: 14/814 or 1.7%
Memory score loss from SETTING 1 to SETTING 3: none, scores identical

A "points per 24h" loss of only 1.7% is negligible in my opinion. This
clearly shows that using VMware imposes only a small loss for CPU
intensive tasks.

Currently the dual-machine test is running. As I don't have enough
physical RAM to start both virtual machines with 1152 MB RAM each,
the benchmark flaw identified above will hit me. The point output will
not show the real working power of the dual-cpu setting but will be
adulterted by the memory score loss. But the numbers above give clear
evidence that the "points per 24h" loss is proportional to the total score
loss. When I have the points output of both virtual machines and can
see their total score, I can calculate a good estimate of the possible
points output if I had enogh RAM to set both machines to 1152 MB. As
I set both to 768MB the estimate will be quite accurate. This max-RAM
output is only dependent on the real CPU throughput and thus I can
directly compare it to the max-RAM output of the single machine to
calculate the dual-CPU gain resp. the overhead of dual-CPU compared to
two comparable separate PCs.

I will calculate the HTT effect the same way as I will have to run the
four virtual machines with some 400MB each only. To be sure that my
assumption of proportion is correct, I will do two additional tests with
one single virtual machine and then two virtual machines -- always using
exactly the same RAM setting for each machine that will be used for the
four concurrent machines in the HTT test.


Best, Stefan.

Hi Esteban

I find this very interesting
Just a suggestion, you may want to start a thread called:
"WCG Benchmarking - Memory is Vital to your Score"
You may want to put it in the Member to Member Support Forum
It will gain a wider audience and a lot more feedback on your tests
I think that it is not being seen by some of the PC Gurus on the Forum which is a shame
I am sure that there are a lot of members here that will want to share your observations