In the last two days, I upgraded two single core AMD64 processors to dual core. In the first case, the drive had failed and I did a new Suse 10.2 install, a Linux 5.8.16 BOINC download, and the dual core was instantly recognized as: 2 A64 cores [Family 15 Model 35 Stepping 2]. The first system DL two sets of files and runs both.

The second system did not have a failed drive. I installed a SCSI drive and installed a new copy of Suse 10.2. I copied the current BOINC subdirectory over to the new drive and started it. It only recognized one cpu. I checked, found an old (5.4.8) copy of BOINC. I DL the 5.8.16, installed and restarted BOINC, yet it still only sees *one* cpu. It recognizes as: 1 AMD Hammer Family processor [Family 15 Model 35 Stepping 2].

It sees the right ID, but doesn\'t recognize it. Looks like something else is still left over from the earlier version of BOINC.

System differences:
First: MSI MB NEO4-F, PCIe video. Old CPU A64 3800 (512K cache, 2.4 GHz) single core. Suse 10.0 original OS, BOINC 5.4.8.
Second: Ecs PCI Extreme MB, AGP 8X graphics. Old CPU Opteron 148 (1MB cache, 2.2 GHz), BOINC 5.8.16.

How do I fix the second system to run both cores?

Thanks.

---

Hi Bill

What a lot of computers and cpdn models!

When you update your version of boinc, you just install the new version over the old one. But if you go back to an earlier version, you have to uninstall the newer version first. I\'d guess that just deleting it isn\'t clean enough.

If you have difficulty uninstalling the remnants of the newer version, AFAIK the advice is to reinstall the newer version to make it complete again, then uninstall it. Then download the older version.

If you have difficulty with this, post back because I think I could probably find the page on the boinc website about extreme or troublesome boinc uninstalls.

I only use Windows, but I\'m pretty sure that what I\'ve said is also the case with Linux.

---

Cpdn news

I\'m not positive, but I THINK that 5.8.16 has the local configuration over-ride file, whatever it\'s called. If so, then it may have picked up the \"On multiprocessors, use at most\" setting of 1 from the old BOINC folder.

If the file exists, you can either edit it, or use \"Clear\" in the Manager (somewhere) to remove the lot.
Then BOINC should follow what you have set in preferences on the server.

But it could be something else. BOINC is getting far too complex for it\'s own good.

---

Backups: Here

Bill:

AMD did not make a dual-core version of the Opteron 148. BOINC detected it correctly.

http://www.amd.com/us-en/Processors/ProductInformation/0,,30_118_8796_9240,00.html

I\'m not positive, but I THINK that 5.8.16 has the local configuration over-ride file, whatever it\'s called. If so, then it may have picked up the \"On multiprocessors, use at most\" setting of 1 from the old BOINC folder.

If the file exists, you can either edit it, or use \"Clear\" in the Manager (somewhere) to remove the lot.
Then BOINC should follow what you have set in preferences on the server.

But it could be something else. BOINC is getting far too complex for it\'s own good.

I agree that BOINC has gone over the top in complexity. The graphics could easily be a separate package, maybe stats as well. This would be an improvement plus making changes easier. And the CPU recognition strings ought to be refreshed from the CPDN site when BOINC is installed.

As to the local file, auth_prefs.xml, it has 2 cpus as the max. Should work.
I actually think the problem is in recognizing the processor. It is not a \'Hammer\' processor (the original A64), but a late model dual core 4200, code name \'Toledo\'. It did get the proper string, but did not identify correctly. Weird.

PS. Funny how few people read my post carefully. :-)

---

Graphics is going to be a separate package in version BOINC 6.*.
And it\'s NOT going to be backwards compatible. Those who upgrade to 6.* in the middle of a model will lose all the model graphics.

---

Backups: Here

I agree that BOINC has gone over the top in complexity. The graphics could easily be a separate package, maybe stats as well. This would be an improvement plus making changes easier. And the CPU recognition strings ought to be refreshed from the CPDN site when BOINC is installed.

As to the local file, auth_prefs.xml, it has 2 cpus as the max. Should work.
I actually think the problem is in recognizing the processor. It is not a \'Hammer\' processor (the original A64), but a late model dual core 4200, code name \'Toledo\'. It did get the proper string, but did not identify correctly. Weird.

PS. Funny how few people read my post carefully. :-)

Well, OK so you have an Athlon X2 Toledo. I assume you\'ve already checked task manager in windows and you see two processors in the graph? If so, then you can force BOINC to use both CPUs. In the BOINC folder, create or append to a file called cc_config.xml as below. Note the ncpus tag in the XML; that is the only part you need to add if this file already exists. Of course, this will not make it go above your max CPUs in general preferences, so make sure that\'s at least 2.

<?xml version=\"1.0\" encoding=\"UTF-8\" ?>
<cc_config>
<log_flags>
<task>1</task>
<file_xfer>1</file_xfer>
<sched_ops>1</sched_ops>
</log_flags>
<options>
<save_stats_days>30</save_stats_days>
<dont_check_file_sizes>0</dont_check_file_sizes>
<http_1_0>0</http_1_0>
<ncpus>2</ncpus>
<max_file_xfers>1</max_file_xfers>
<max_file_xfers_per_project>1</max_file_xfers_per_project>
<work_request_factor>1</work_request_factor>
</options>
</cc_config>


For details, see:
http://boincfaq.mundayweb.com/index.php?language=1&view=91

Does the operating system itself recognise and use the second core? If not, Boinc won\'t be able to do any better, regardless of what the various configuration files say.

Try updating the Bios if this is the case.

PS. Funny how few people read my post carefully. :-)

I think the reason people weren\'t sure about the processors is that you didn\'t actually say what the new processor was in your original post :-)

---

I'm a volunteer and my views are my own.
News and Announcements and FAQ

Does the operating system itself recognise and use the second core? If not, Boinc won\'t be able to do any better, regardless of what the various configuration files say.

Try updating the Bios if this is the case.

PS. Funny how few people read my post carefully. :-)

I think the reason people weren\'t sure about the processors is that you didn\'t actually say what the new processor was in your original post :-)

Well, this one\'s for the books. I did chase down the number of CPUs detected by the OS, Suse 10.2, which defaults to SMP. It only detected one, and checking the startup log against an identical X2 4200 in another system, running dual cores, it was clear that the two systems were seeing different cpus, and #2 saw only one core.

It just happened that I had another X2 4200 for another project, which I swapped into system #2. Surprise! Exactly the same readings, exactly the same result - one cpu.

Now how likely is it that I got *two* bad cpus in two separate orders?
Well, how close to zero can you count? So the real cause is now reduced to the old MB I was using in system 2. It is an old ECS system with AGP, and was probably designed before dual cores were shipping. Apparently, the MB doesn\'t have all of the circuits needed to register dual cores. <sigh>

I will shortly be swapping in a new ASUS MB, which should solve the weirdness I\'ve run into. But wotta PITA!

Thanks for all of the help and suggestions - they are much appreciated.

BillN

---

All socket 939 motherboards *in theory* should be able to handle dual cores given a Bios update (even if designed prior to the introduction of dual cores), but only if the motherboard manufacturer followed AMD\'s specification to the letter. Some didn\'t...

---

I'm a volunteer and my views are my own.
News and Announcements and FAQ

All socket 939 motherboards *in theory* should be able to handle dual cores given a Bios update (even if designed prior to the introduction of dual cores), but only if the motherboard manufacturer followed AMD\'s specification to the letter. Some didn\'t...

In theory, yes. However, this is the second ECS board I have discarded. The first ECS board failed in less than two years, and this one needed the replacement BIOS after the first few months. Maybe it\'s just a fluke, but I\'ve used my last ECS board. I\'ll let you know what happens with the new ASUS board.

BillN

---

All socket 939 motherboards *in theory* should be able to handle dual cores given a Bios update (even if designed prior to the introduction of dual cores), but only if the motherboard manufacturer followed AMD\'s specification to the letter. Some didn\'t...

In theory, yes. However, this is the second ECS board I have discarded. The first ECS board failed in less than two years, and this one needed the replacement BIOS after the first few months. Maybe it\'s just a fluke, but I\'ve used my last ECS board. I\'ll let you know what happens with the new ASUS board.

BillN

Okay, system is back up with a new MB and new install. Dual core recognized by OS, BOINC was downloading the second simulation by the time I got the screen up. It *was* the ECS MB not handling the dual core.

BillN

---

Nice to hear that you solved it, even if it did cost a bit more.