---

>I have looked at some of the other links that are telling me to run all these
>memory checks and other stuff. But if Seti works just fine on my AMD machine >than I don't think that is the problem.

Seti is in no way like CPDN. CPDN is MUCH more memory intensive, and since each timestep depends on data from the previous one, and a work unit is on the order of weeks (approximately 800000 timesteps), any small hardware glitch will show up and eventually amplify. Run the checks at:

http://www.climateprediction.net/board/viewtopic.php?t=2126

While it's no guarantee that running the checks error free, for MANY hours, will result in a crashless CPDN, it can eliminate problems. The person who wrote that thread has (until now) only AMD PCs, many overclocked, and has at least a 10 success to 1 error rate.

>It sound like there is a minor programing error somewhere!?

If so, why do so many AMD users run model after model without them erroring out. Looking through the results, only 1 in 7 models that are started complete, yet some users (with AMD or Intel) have completed many models without an errored one. It may have to do with cooling, power supply, memory quality, or just usage of the system.

It is a worthwhile and interesting project. Hope you can solve whatever it is that is causing your PC's problems with CPDN.

I run cpn on around 10 different computers, mostly intel cpus, some G4's and a athlon xp-m. The athlon seem to be not that stable for cpn, i have had model almost 80% compleate but then after a system crash, the model errored out and Boinc downloaded a new one. since then, the athlon has errored out on three more models. I dont think its a problem with my system as i have turn the memory right down to slow settings and put the cpu well below what its capable off and had no problem with stabilty on any other program. There dose seem to be some sort of problem with AMD CPU's.

I will give my athlon another go at these very slow settings just to rule out it being a overclocking issue but if another fails, i will use my athlon for other projects that are better suited to the athlon.

---

As Geophi has already pointed out, any problem with Athlons isn't a general one, though. Plenty of people do run CPDN successfully with them, and although I don't do so at the moment I had no trouble either under classic or BOINC with different Athlon CPUs. Their main drawback is speed, since the program is optimised for Intels.

Doesn't help you, of course.

---

> As Geophi has already pointed out, any problem with Athlons isn't a general
> one, though. Plenty of people do run CPDN successfully with them, and although
> I don't do so at the moment I had no trouble either under classic or BOINC
> with different Athlon CPUs. Their main drawback is speed, since the program is
> optimised for Intels.
>
> Doesn't help you, of course.
>
I had asked this question a while back. But I was told the model was designed to work with both processor types, and that the problem was with my memory, mainboard, or something else.

Now you are saying that the model was designed for Intel CPU's.??

Their main drawback is speed? My XP 2500 processor (actual clock speed is 1.8g)is considerably faster than my Intel 2.0g processor.

I will continue to run my AMD processor in hopes that it will complete a work unit. So far it is working.

---

> Now you are saying that the model was designed for Intel CPU's.??
>

The program was originally developed for Cray supercomputers, so many of us think that it pretty amazing that we can run it on our desks or even laps :)

When the present version was compiled for BOINC, the software compiler was an Intel version and so optimised for their processors. It would be nice to have an AMD version as well, but it was decided to be too expensive.

---

I updated my Bonic program to 4.25.

My AMD processor finally completed a work unit. Just thought I would share that with anybody that was interested.

---