Windows SSE4-Core 2 32-bit V5.15, R-1.41

[Dirk Sadowski]

Hello Simon!
Hello optimizers!

I read that there are sometimes probs with the:
Windows SSE4-Core 2 32-bit V5.15 'Chicken Good!' (R-1.41|+more_vec|xT)

Have a look here:
http://setiathome.berkeley.edu/forum_thread.php?id=35728&nowrap=true#473429


This happen only when you overclock?
Or it's a little bug?


Greetings!

[Dirk Sadowski]

If I have the QX6700...

If maybe every 10th WU is invalid...

...then maybe it's better to use the
SSE3-Intel P4/PD optimized Windows S@H app (Rev.2.0)
and not the
Intel Core 2 SSSE3 optimized 5.15 app (Rev.1.41)?

How much less performance have this SSE3 to the SSSE3?


Greetings!

[Simon]

My Woodcrest system doesn't seem to have this problem, really.

Dirk, do whatever you feel comfortable with - in my case, I'm comfortable with leaving it running on the Woodcrest until something quicker comes along.

Regards,
Simon.

[Dirk Sadowski]

Hello Simon!

I looked at some Core 2 Duo, Core 2 Quad and Xeon too, and sometimes they have invalid results...
They crunch "normal" but didn't got Credits...

I found three invalid WUs from your C2Dx Xeon 51xx (Woodcrest) too :

http://setiathome.berkeley.edu/result.php?resultid=427847127
http://setiathome.berkeley.edu/result.php?resultid=427082185
http://setiathome.berkeley.edu/result.php?resultid=426441006


You can imagine what could be wrong?


Greetings!

[Simon]

Again:
No!

Also:
I'm not especially anxious about 3 invalid results - currently, that host has hundreds online, so it's less than one percent.

As is the case with any optimized app - YMMV (your mileage may vary, oder auf deutsch: Benutzung auf eigene Gefahr). Do what you feel comfortable with, please don't expect me to tell you what that is

Regards,
Simon.

[Fivestar Crashtest]

My Core 2 Duo went from 1700 RAC to flirting with 1900 using the SSE4 optimized app.  I am very happy with it. 

One thing keeps me away from the idea of building a quad core- the price and what it might do to my electric bill.  TWO things keep me from building a quad core, the price, what it might do to my electric bill and I have no more room left for computers.  THREE things ...

[chapellane69]

This host of mine has returned two invalid results, here and here.  It has run the optimised application for all but a very few of its returned results, but as it's new it hasn't returned that many.  I think it's probably failed to validate these 2 out of about 60 results.

I'll keep an eye on the failure rate.  If it stays realitvely high, I'll switch applications.  It's a shame because this one is so fast.

[I used the relevant Chicken application for my AthlonXP which was also great, and never failed to validate once that I can remember].

Neil

[Josef W. Segur]

This host of mine has returned two invalid results, here and here.  It has run the optimised application for all but a very few of its returned results, but as it's new it hasn't returned that many.  I think it's probably failed to validate these 2 out of about 60 results.

I'll keep an eye on the failure rate.  If it stays realitvely high, I'll switch applications.  It's a shame because this one is so fast.

[I used the relevant Chicken application for my AthlonXP which was also great, and never failed to validate once that I can remember].

Neil

You may want to look at my post to the S@H NC forum for more detail. The next versions will have improved accuracy and better speed than 1.41, meanwhile the SSE2 versions of 2.0 come close to 1.41 in speed on Core 2 systems and have not exhibited any validation problems.

Many users with Core 2 systems have concluded that since 1.41 gives them higher average daily credits than 2.0, they're doing more useful work for the project by using 1.41 even if there's an occasional invalid result.
                                                                            Joe