The first and second generation Athlons (Classic and Thunderbird) don't know SSE instructions.
I have several Athlon machines that are 2-3 years old ranging from 1.1 to 1.6. Is there somehow a way to determine what core these chips have? With my luck they're probably all Thunderbirds, but I'd still like to find out.
Thanks for the Cpuid.com. Unfortunately other than a couple of 64 XPs I'm setting on a half dozen T-birds. Is there any hope in the future that these processors will ever be able to get around the SSE instructions, and if not, how big of a performance hit can I expect?
Gary
In the beginning the Universe was created. This has made a lot of people very angry and been widely regarded as a bad move.....Douglas Adams
The first and second generation Athlons (Classic and Thunderbird) don't know SSE instructions.
I have several Athlon machines that are 2-3 years old ranging from 1.1 to 1.6. Is there somehow a way to determine what core these chips have? With my luck they're probably all Thunderbirds, but I'd still like to find out.
Thanks for the Cpuid.com. Unfortunately other than a couple of 64 XPs I'm setting on a half dozen T-birds. Is there any hope in the future that these processors will ever be able to get around the SSE instructions, and if not, how big of a performance hit can I expect?
Gary
You won't be able to process instructions that your processor doesn't have support for. However, your FPU will still be used, albeit, more slowly than the vectorized code run with SSE. As I recall, the SRI and SR1 apps issued 3DNow instructions as well at the same time, which was one of the reasons that AMD chips were faster in the previous apps.
As I recall, the SRI and SR1 apps issued 3DNow instructions as well at the same time, which was one of the reasons that AMD chips were faster in the previous apps.
The speed of the 3DNow! is equal to SSE. (The difference is really small.)
S5R1/S5RI code didn't have any 3DNow! instructions.
As I recall, the SRI and SR1 apps issued 3DNow instructions as well at the same time, which was one of the reasons that AMD chips were faster in the previous apps.
The speed of the 3DNow! is equal to SSE. (The difference is really small.)
S5R1/S5RI code didn't have any 3DNow! instructions.
Then you're saying my T-Birds are still going to be in the "ballpark"? I haven't had a chance to compare any times with any of the T-Bird machines because I still had an inventory of R1s that don't expire until early next month. But the 64 XP is without a doubt slower on the order of at the very least 25%. If this is the case, I wish I had known this before downloading all the Rosetta software.:(
After having run E@H and then E@H on BOINC, for over 3 years I'm a little embarassed to ask this question, but since I now have 2 projects either sharing time automatically or if I choose I just suspend one or the other, I can't seem to get the darn thing to download enough work. I've tried connecting every 10 days on one end to .1 days on the other, and it seems not to make a bit of difference in the least. I should say I changed the settings by going into general pref. and changing the it (default) as well as changing all the home, work, school stuff even though the 5 or 6 machines that will be used are all located at home and are not networked. What am I missing with these setting changes? Running only E@H I had everything set at 10 days and it ran flawlessly on all machines. Now I have a mess.
Gary
In the beginning the Universe was created. This has made a lot of people very angry and been widely regarded as a bad move.....Douglas Adams
[After having run E@H and then E@H on BOINC, for over 3 years I'm a little embarassed to ask this question, but since I now have 2 projects either sharing time automatically or if I choose I just suspend one or the other, I can't seem to get the darn thing to download enough work. I've tried connecting every 10 days on one end to .1 days on the other, and it seems not to make a bit of difference in the least. I should say I changed the settings by going into general pref. and changing the it (default) as well as changing all the home, work, school stuff even though the 5 or 6 machines that will be used are all located at home and are not networked. What am I missing with these setting changes? Running only E@H I had everything set at 10 days and it ran flawlessly on all machines. Now I have a mess.
Gary
Are you remebering to hit the update button after you make the changes to the web pages so that BOINC knows you have made a change and then leaving BOINC for at least an hour so that the scheduler will run, take notice of the change, and calculate a new work fetch quota? Dave.
What might also be part of the problem is the fact that WU crunching time estimates for Einstein are REALLY off atm. My notebook first starts with an estimate of more than 70 hours per WU, while the largest ones I get really take little more than 50. You can imagine this makes a huge difference, so it might confuse the scheduler quite a bit...
What might also be part of the problem is the fact that WU crunching time estimates for Einstein are REALLY off atm. My notebook first starts with an estimate of more than 70 hours per WU, while the largest ones I get really take little more than 50. You can imagine this makes a huge difference, so it might confuse the scheduler quite a bit...
I'm done with my first one after 10 days. Whew. Received more credit than I formerly did for CPDN trickles . . .
What might also be part of the problem is the fact that WU crunching time estimates for Einstein are REALLY off atm. My notebook first starts with an estimate of more than 70 hours per WU, while the largest ones I get really take little more than 50. You can imagine this makes a huge difference, so it might confuse the scheduler quite a bit...
DCF is fairly different than in r1. Rather than wait a week for it to stabilize once my 1st WUs were complete enough to correctly figure the new value i stopped boinc and manually editied it.
Come on people, don't bail. These are big work units - that's all. I've got everything from a 450MHz P3 to an Athlon 4200+ dual core running them, and they run fine - they just take longer.
Taking your computers away from the project only slows things down for everyone, especially if you detach while you still have units.
RE: RE: RE: RE: What
)
Thanks for the Cpuid.com. Unfortunately other than a couple of 64 XPs I'm setting on a half dozen T-birds. Is there any hope in the future that these processors will ever be able to get around the SSE instructions, and if not, how big of a performance hit can I expect?
Gary
In the beginning the Universe was created. This has made a lot of people very angry and been widely regarded as a bad move.....Douglas Adams
RE: RE: RE: RE: Quote
)
You won't be able to process instructions that your processor doesn't have support for. However, your FPU will still be used, albeit, more slowly than the vectorized code run with SSE. As I recall, the SRI and SR1 apps issued 3DNow instructions as well at the same time, which was one of the reasons that AMD chips were faster in the previous apps.
RE: As I recall, the SRI
)
The speed of the 3DNow! is equal to SSE. (The difference is really small.)
S5R1/S5RI code didn't have any 3DNow! instructions.
RE: RE: As I recall, the
)
Then you're saying my T-Birds are still going to be in the "ballpark"? I haven't had a chance to compare any times with any of the T-Bird machines because I still had an inventory of R1s that don't expire until early next month. But the 64 XP is without a doubt slower on the order of at the very least 25%. If this is the case, I wish I had known this before downloading all the Rosetta software.:(
After having run E@H and then E@H on BOINC, for over 3 years I'm a little embarassed to ask this question, but since I now have 2 projects either sharing time automatically or if I choose I just suspend one or the other, I can't seem to get the darn thing to download enough work. I've tried connecting every 10 days on one end to .1 days on the other, and it seems not to make a bit of difference in the least. I should say I changed the settings by going into general pref. and changing the it (default) as well as changing all the home, work, school stuff even though the 5 or 6 machines that will be used are all located at home and are not networked. What am I missing with these setting changes? Running only E@H I had everything set at 10 days and it ran flawlessly on all machines. Now I have a mess.
Gary
In the beginning the Universe was created. This has made a lot of people very angry and been widely regarded as a bad move.....Douglas Adams
RE: [After having run E@H
)
Are you remebering to hit the update button after you make the changes to the web pages so that BOINC knows you have made a change and then leaving BOINC for at least an hour so that the scheduler will run, take notice of the change, and calculate a new work fetch quota? Dave.
What might also be part of
)
What might also be part of the problem is the fact that WU crunching time estimates for Einstein are REALLY off atm. My notebook first starts with an estimate of more than 70 hours per WU, while the largest ones I get really take little more than 50. You can imagine this makes a huge difference, so it might confuse the scheduler quite a bit...
RE: What might also be part
)
I'm done with my first one after 10 days. Whew. Received more credit than I formerly did for CPDN trickles . . .
RE: What might also be part
)
DCF is fairly different than in r1. Rather than wait a week for it to stabilize once my 1st WUs were complete enough to correctly figure the new value i stopped boinc and manually editied it.
Ignore the RAC graph - the
)
Ignore the RAC graph - the active hosts graph is far, far more telling.
Personally I have taken my 4 computers off Einstein and onto Rosetta until E@H gets its' ducks in a row.
Come on people, don't bail.
)
Come on people, don't bail. These are big work units - that's all. I've got everything from a 450MHz P3 to an Athlon 4200+ dual core running them, and they run fine - they just take longer.
Taking your computers away from the project only slows things down for everyone, especially if you detach while you still have units.