please post to this new thread instead of the old one...
keep in mind that there are still a lot of people using dialup connections.

Thanks.

---

Join BOINC United now!

Hi Crunch3r,
I updated my P4s and P3 with your latest build and they all seem to be working perfectly.

They are running XP Pro and Windows 2003 Server

Thanks for all your efforts Crunch3r.
Bob
____________

words are the symbols of mental experience

Thanks Bob,

Please post in this thread now.

---

Join BOINC United now!

Reminder to all: Make sure you zip up your boinc dir before updating.

I just updated to 5.2.13 (a64) and 22 worksunits disappeared from my seti dir.

---

Theres a Fine Line between Hobby and Mental Illness.

I don't think its a hardware issue either. My machine is a stock Dell

I'm sorry, but that conjunction of words just had me rolling with laughter... If your Dell is more than a couple of years old, you're right, it's a very stable, decent machine. If it's a newer box, well... you may be lucky.

Just a bit off-topic...

Realy? I wasn't aware that there is problem with newer models. But you're right anyway: my computer is nearly 2 years old, and it never gave me any trouble. I was considering replacing with a newer one, but your comments gave me something to think about.

Cheers

---

Has anyone run the reference WU, to see if the amd64s are really faster with the sse3 version, versus the sse2?

Thanks,
Don

---

Has anyone run the reference WU, to see if the amd64s are really faster with the sse3 version, versus the sse2?

Thanks,
Don

I know that several "beta" testers have reported that the SSE3 implementation for A64 is about 3 to 5 minutes faster then the sse2 version.

Using the client on "live" WUs showed the same results.

---

Join BOINC United now!

Has anyone run the reference WU, to see if the amd64s are really faster with the sse3 version, versus the sse2?

Thanks,
Don

I know that several "beta" testers have reported that the SSE3 implementation for A64 is about 3 to 5 minutes faster then the sse2 version.

Using the client on "live" WUs showed the same results.

Hi Crunch3r, and everybody.

Well, mine is cranking them out in 35 minutes now, with the sse3. I didn't want to stop and run the reference WU, if it has already been done.

I notice about 1 WU every 2-3 days, or less, that says client error, when the rest have gotten credit for the WU. But the speed is clearly worth it.

When it fails, it is due to an memory access voilation, (0xc000...005 windows error). Makes me wonder if the cache is sometimes accessed before being filled up. Maybe on a restart or something, the cache is assumed to be in a state that it is not...

Absolutely not a problem, in the grand scheme of things.

Thanks and happy holidays!

Don

---

http://setiathome.berkeley.edu/result.php?resultid=175386855

Sometime today BOINC totally died and I have a few no heartbeat errors.

Typically (out of a gig of ram) available memory is in the five hundreds and system cache is in the four hundreds, PF Usage in the four hundreds. After I noticed there wasn't any boinc processes running and I had the WU errors I looked at my system performancs and it showed available in the six hundreds, system cache in the six hundreds, and PF usage in the two hundreds. Is it possible to have the sum of the available memory and system cache greater than the total physical memory?


I've gone back to the older version dated 12/8.

Crush3r

Don't know if this might make you feel a little better about your work but, over 400 results and only 1 error. Pretty darn good i think. At least as good if not better success rate than the official App!
I'l give the link if its any help, Probly got more than you want anyway.
Maybe this bit of info might help though. Goes along the same idea of memory leak or whatever. I got that bad result last night and thats the first time i've left the system on running seti constant for over 3 days. Not trying for some record just havent been bothered to turn the thing off. Its just i notice others saying things like every 2-3 day get a bad one ect, maybe just turning of the computer now and then is a solution?
Also just added my old machine to boinc again now things have stabalized a little. Running your app and its sent back its first batch all fine.

Thanks for the hard work you've put into that app and to all involved. Sorry those occasional blips in the results are being a pain but remember that the original official app wasn't perfect.

http://setiathome.berkeley.edu/result.php?resultid=180593729

I know that several "beta" testers have reported that the SSE3 implementation for A64 is about 3 to 5 minutes faster then the sse2 version.

Using the client on "live" WUs showed the same results

I noticed a little increase in crunching speed... with SSE2 version i have got 33.5 minutes average... with SSE3 it's like ~31 maybe 31.5 :)

---

I know that several "beta" testers have reported that the SSE3 implementation for A64 is about 3 to 5 minutes faster then the sse2 version.

Using the client on "live" WUs showed the same results

I noticed a little increase in crunching speed... with SSE2 version i have got 33.5 minutes average... with SSE3 it's like ~31 maybe 31.5 :)

And you just needed to point that out :)
Merry Christams fellows..

//Vyper

---

_________________________________________________________________________
Addicted to SETI crunching!
Founder of GPU Users Group

I noticed a little increase in crunching speed... with SSE2 version i have got 33.5 minutes average... with SSE3 it's like ~31 maybe 31.5 :)

The A64 SSE3 version is an older version, too (dated December 8, 2005), where the A64 SSE2 version is newer (dated December 22, 2005). The older A64 SSE2 version was also dated December 8, 2005, so Crunch3r has not yet updated the A64 SSE3 version yet. I have noticed that the newer versions are a few minutes faster than the older versions, so Crunch3r will hopefully update the A64 SSE3 version with whatever changes he made to the A64 SSE2 version. Right now, his _new_ A64 SSE2 version runs a little faster than the older SSE3 version on my A64 X2 Dual Core 4800+. With those newer updates to the SSE3 version, I would expect it to be even faster...

Regards, Daniel.

---

I see nice times with your X2 =) it's OCed or not? looks like 2600-2700 Mhz...

---

I've been running BOINC since early December on four diverse computers. I installed crunch3r's SETI science ap and BOINC client shortly after their availability was posted (a week or so ago). (roughly 2.5x speedup in actual execution times across the fleet)

Yesterday one of the two Win98 machines seemed to stop BOINCing. On Reboot, BOINCmgr comes up and announces itself not in communcation with a BOINC client. Typing in localhost it says it tried and failed and do I want to try again.

Following advice to another user with a slightly different problem, I deleted BOINCmgr from Windows startup and tried again. No joy on manual launch.

So, after another reboot, I used a DOS window to start BOINC.EXE. With it apparently running in the DOS window, I started BOINCmgr, which is now happy.

Any suggestions on how to patch my installation so it will start normally on reboot again?

Any chance this might be related to my use of crunch3r's client?

Thanks for any assistance.

My two XP machines continue to run healthily.

---

Crunch3r, First thank you for all your hard work and, Merry Christmas my German friend.

Your Boinc client and Seti apps are great and have served me well. Last night I was expermenting with different clients and apps on my 3200+ 64. I found that I could get a little faster times with the newer 5.3.3 Boinc, (Not Optimized) and still using your 5.2.13 seti optimized app on windows XP. Only about 5 minutes faster. About 36 minutes. I have not had any errors yet. Now I am trying the same on my 1800+ laptop. It looks to be faster also. Just thought you should know.

Skip

---

Crunch3r, you're the guy that built the optimized client, correct?

(Or, one model/family of them reoptimized via the Intel C++ compiler)

I went with your client builds, because they seem to produce FASTER floating-point processing abilities because I strongly suspect that the nature of the data this project uses, lends itself to and is processed by the floating point code & ALU functions on the CPU for Fpops, more than it is for iops, as far as how fast you can tear thru a unit.

(ON THAT NOTE? See if you can go with John Carmacks version of the FLOAT function if you can: IIRC, it is not as accurate as math.h FLOAT function from what I understand (not out to as many decimal places of precision) but, many orders of magnitude faster)

I've seen the code for it on Slashdot, & although incredibly small (sub 20 line function) it's fast & I saw it when the Quake III sourcecode released...

(Perhaps it is something you can use for making it even FASTER still!)

:)

* Think about it - food for thought on that function of John Carmack's from the Quake III sourcecode, provided it does NOT adversely affect the accuracy of calculation(s) being performed by the SETI 4.18x client & later ones upcoming still...

APK

P.S.=> IMO, and perhaps YOU can answer me some on this? Floating point calculation is more used & crucial/critical to the operation of SETI units being performed faster than Integer ops are, correct??

I ask that here too, where I compared a build of the SETI 4.18x & BOINC.exe client programs vs. those of other code modders here like Trux... both are great, but yield diff. strengths in diff. areas it seems, see here:

http://setiathome.berkeley.edu/forum_thread.php?id=26012

BOTTOM-LINE - Above all???

Thanks for feedback & thanks for making my SETI@Home 2 boinc managed processing times go WAY faster, 2x if not moreso, because of your reoptimized builds of the client programs for this project... apk

---

http://torry.net/authorsmore.php?id=1781

"The object's hull is made of SOLID neutronium: A single StarShip cannot combat it!" quote Mr. Spock, Star Trek original series, episode title: "The Doomsday Machine"

Crunch3r,

No "- exit code -1073741819 (0xc0000005)" over the last 24 hours with your updated SETI client. Excellent work!

---

I see nice times with your X2 =) it's OCed or not? looks like 2600-2700 Mhz...

No, I never OC any of my machines anymore. It was more of a headache for me than what it was worth. Some people have good success with it, but I never had anything but greif. Add to that the problem of OCing voids the CPU warranty, and no thanks. I quit OCing for good once I started paying $600+ per CPU!

Regards, Daniel.

---

compliments for the good work and... Merry Christmas :)

is good idea to lift the benchmarks (MIPS Dhrystone) for the claimeds credit
because if a lot of people use the client, the granted credits go after all to the beach

Edit:
the problems is on the AMDX2 4400+ with claimed credit < 9
on the AMDXP 2600+ and P4 SSE2 it's OK

Thanks and happy holidays!

Gil

---

(ON THAT NOTE? See if you can go with John Carmacks version of the FLOAT function if you can: IIRC, it is not as accurate as math.h FLOAT function from what I understand (not out to as many decimal places of precision) but, many orders of magnitude faster)

Hello,

I tried your suggestion with John Carmacks math.h on of my linux boxes.
The bottom line is that i don't see any improvement in benchmark nor faster wu processing times here.

I guess nothing beats intels math implementation (mathimf.h). It seems as if the intel compiler is the best you can get if your concerned about speed.
(except Pathscales EKOPath compiler for 64 bit linux which is even faster than intels icc at 64 bit code)

---

Join BOINC United now!