19.09.2007 03:15:34|Einstein@Home|Scheduler RPC succeeded [server version 601]
Can I expect with this server version, that I can use for example 64bit Vista with a 64bit BOINC client for Einstein? I mean, that the server sends the 32bit application to me?
All I know is that the feature is now supported, not sure it is enabled. Unfortunately I lack a 64 bit OS system to test this on. Anybody want to try?
19.09.2007 03:15:34|Einstein@Home|Scheduler RPC succeeded [server version 601]
Can I expect with this server version, that I can use for example 64bit Vista with a 64bit BOINC client for Einstein? I mean, that the server sends the 32bit application to me?
All I know is that the feature is now supported, not sure it is enabled. Unfortunately I lack a 64 bit OS system to test this on. Anybody want to try?
CU
I have had a 64bit OS for a while and been unable to get any work; I just downloaded my first unit.
Sun 23 Sep 2007 12:33:12 PM PDT|Einstein@Home|Message from server: platform 'x86_64-pc-linux-gnu' not found
My system: Intel Q6600. Ubuntu 7.04. I'm using the boinc software from the ubuntu repositories.
Help? I was crunching along fine with an app_info.xml file.
FYI for others. You need the latest client to get the 32 bit app sent to a 64 bit boinc. Ubuntu's client in the repositories is too old for that function. I downloaded 5.10.8 from the boinc site and I'm up and running.
Not necessarily. The FPU is still the same size, and since most projects are FPU intensive, the 64 bit processor will give very little extra. In fact it may hinder at times, because of the extra memory overhead needed.
So, do not assume things will run faster just because you got the new software.
There is plenty of ways to run the software on your 64 bit machines, just read the "Problems and Bug Reports" area and it's explained in there (ad nausea).
Going to 64 bits will actually help, but not 2x as much performance. Because the AMD64 architecture requires SSE2, requires that all FPU work be done in the XMM (a.k.a. SSE) registers instead of the x87 floating point stack, and it doubles the number of XXM registers, floating point will get a performance boost. First, a good compiler will try to vectorize as much work as possible to use SSE2 effectively without having to check to see if SSE2 is present (and therefore generating time and space wasting branch instructions). Second, the old x87 floating point stack architecture is the cause of so much slowdown that x87 (32-bit and 16-bit floating point) is the joke of the floating point world. Banning x87 FPU work in 64-bit mode and moving all FPU work to the cleaner XMM register file provides a noticeable boost in floating point performance, even for code that cannot be vectorized. Doubling the number of XMM registers allows the FPU to hold more numbers internally before having to waste time fetching more data from the cache or memory.
RE: BOINC shows under
)
All I know is that the feature is now supported, not sure it is enabled. Unfortunately I lack a 64 bit OS system to test this on. Anybody want to try?
CU
H-BE
RE: RE: BOINC shows under
)
Tested my self, works fine!
)
Tested my self, works fine! :)
Yeah, now E@H gives work
)
Yeah, now E@H gives work units to my Linux AMD64 box.
It's not working for me.
)
It's not working for me.
Sun 23 Sep 2007 12:33:12 PM PDT|Einstein@Home|Message from server: platform 'x86_64-pc-linux-gnu' not found
My system: Intel Q6600. Ubuntu 7.04. I'm using the boinc software from the ubuntu repositories.
Help? I was crunching along fine with an app_info.xml file.
RE: It's not working for
)
FYI for others. You need the latest client to get the 32 bit app sent to a 64 bit boinc. Ubuntu's client in the repositories is too old for that function. I downloaded 5.10.8 from the boinc site and I'm up and running.
RE: Not necessarily. The
)
Going to 64 bits will actually help, but not 2x as much performance. Because the AMD64 architecture requires SSE2, requires that all FPU work be done in the XMM (a.k.a. SSE) registers instead of the x87 floating point stack, and it doubles the number of XXM registers, floating point will get a performance boost. First, a good compiler will try to vectorize as much work as possible to use SSE2 effectively without having to check to see if SSE2 is present (and therefore generating time and space wasting branch instructions). Second, the old x87 floating point stack architecture is the cause of so much slowdown that x87 (32-bit and 16-bit floating point) is the joke of the floating point world. Banning x87 FPU work in 64-bit mode and moving all FPU work to the cleaner XMM register file provides a noticeable boost in floating point performance, even for code that cannot be vectorized. Doubling the number of XMM registers allows the FPU to hold more numbers internally before having to waste time fetching more data from the cache or memory.
Got my first WU through my 64
)
Got my first WU through my 64 Bit BOINC on Ubuntu 7.04 64 Bit. Works good! :-)
BOINC 5.10.20 (64 bit) works
)
BOINC 5.10.20 (64 bit) works fine on Vista Ultimate 64 bit.
Intel Q9300 Quadcore, 2500 Mhz, 4096 MB RAM, GeForce 9800 GT, Vista Ultimate 64-bit, Ubuntu 10.10 64 bit