But I do have the same promblem on a newly buit [HOST#12619532]

Moreover I have MD5 problem with lots of other files, like FGRP OpenCL NVidia-1k.exe.

Will try to change BOINC version. Let's see what happens.

Unlikely I think.  There are no other reports I'm aware of.  I would expect to see other reports if it was happening to all new installs.  Last time, the problem was not caused by bad data or tasks - just a bad stored value for the checksum.

I'm also a bit surprised that with a GPU plus 32 CPUs, there is a daily limit of only 40.

Have you compared the file on the new host to the same file on one of your other hosts?  Have you checked the MD5sum stored in client_state.xml on each?  The previous occurrence was a bad MD5sum stored in the state file.  Do you have a bad MD5sum on the new host?  It should be the same on all hosts.  If you have, can you try editing the state file to change the bad value into the correct value?  That was the workaround used last time.

If all of those things turn out negative, I have seen examples of checksum failures being caused by bad RAM.  Since it's a new host, are you confident the RAM is good?

I have also been getting download errors on all my WU's here the past couple days. :( 

Task 789416778

<core_client_version>7.12.1</core_client_version>
<![CDATA[
<message>
WU download error: couldn't get input files:
<file_xfer_error>
  <file_name>JPLEPH.405</file_name>
  <error_code>-119 (md5 checksum failed for file)</error_code>
  <error_message>MD5 check failed</error_message>
</file_xfer_error>
</message>
]]>

I dug through some of last year's posts and seem to have it working. Here's what I did:

1. Exit BOINC manager, making sure all work has stopped.

2. Open your BOINC program files and look for "client_state.xml" (Mine was in C > Program Data > BOINC)

3. Open the file "client_state.xml" with a simple text editor like Notepad.

4. Search for the line:

<file>
    <name>hsgamma_FGRP5_XXX.exe</name>

The "XXX" part doesn't matter, it's just version number info. I did a Ctrl+F search for occurrences of "hsgamma" and found the right line almost immediately.

5. JUST BEFORE that ^ line you just found, paste in all this:

<file>
    <name>JPLEPH.405</name>
    <nbytes>9319680.000000</nbytes>
    <max_nbytes>0.000000</max_nbytes>
    <md5_cksum>d6ce12bacd2a81a56423f5f238ba84eb</md5_cksum>
    <status>1</status>
    <sticky/>
    <download_url>http://einstein8.aei.uni-hannover.de/EinsteinAtHome/download/29/JPLEPH.405</download_url>
</file>

 6. Save your newly edited "client_state.xml" file, overwriting the old one. Make sure it does NOT save as a Text Document (*.txt),  use the All Files (*.*) dropdown option. Also make sure it DOES save with ANSI coding (which seems pretty standard).

7. Restart BOINC

Thanks to everyone who had to work this out for themselves last time. Hope it helps somebody else!

Sorry for the trouble, and thanks for the notification.

This is a result of a race condition between the two FGRP workunit generators. This should be fixed now (i.e. not happen again). Also too the broken md5sums were fixed in the database, i.e. new tasks should be sent out with the correct checksum even if they are instantiated from old workunits.