Website update, new client, Hi Louie

[tqft]

Website update, new client, Hi Louie
" SPEEDUP: NEW SB CLIENT v27.7 AVAILABLE
(posted by Louie Helm)
Wednesday, 16 May 2012

UPDATED Prime95 client available for SB

There's a new version of SB available thanks to even more great work by George Woltman.


Major Enhancements

25% speed increase on Intel i5/i7 processors
Better multi-threaded performance
Faster FFT for Core 2 w/ 1MB L2 cache
Mac OS X client now GUI instead of CLI
"

If I remember correctly, if I have an existing setup, open the .gz, extract mprime to working directory and fire off as usual

[Warped]

Thanks for the heads-up.
I rarely look at the SoB home page.

[engracio]

Anybody else having possible hardware but it was reproducible error message on 27.7 build 2 on AMD cpus???? No error message on core 2 cpus.

[prime95]

Anybody else having possible hardware but it was reproducible error message on 27.7 build 2 on AMD cpus???? No error message on core 2 cpus.

If you are getting small reproducible roundoff error values like 0.40625 and 0.4375 then it is likely due to testing a value near the limits of what that FFT size can test. In other words, nothing to worry about.

If you are getting roundoff errors like 0.49999 then it is a hardware issue.

P.S. The FFT limits changed ever so slightly between 26.6 and 27.7. So it is entirely possible that the same number would use a different FFT size in the two versions.

[engracio]

If you are getting small reproducible roundoff error values like 0.40625 and 0.4375 then it is likely due to testing a value near the limits of what that FFT size can test. In other words, nothing to worry about.

If you are getting roundoff errors like 0.49999 then it is a hardware issue.

P.S. The FFT limits changed ever so slightly between 26.6 and 27.7. So it is entirely possible that the same number would use a different FFT size in the two versions.

THanks George you are correct, it was small reproducible roundoff error values like 0.40625 and 0.4375. It also states to disregard the error. Your expanation seems to match the problem which was nothing compare to the speedup added on the new build. Thank you again.

[enderak]

Yep, I noticed this on mine too shortly after upgrading (Intel i7 here, not AMD)

One of my tests seems to keep getting errors. It says:

18 ROUNDOFF > 0.4 of which 8 were repeatable (not hardware errors).
Confidence in final result is very poor.

Does this mean 10 of the 18 errors it thinks are hardware errors? Yesterday it said confidence was "fair", then "poor", now "very poor" so it's getting worse. I am thinking maybe I just need to abort this test and start a new one. (This test was started in v. 26.6) I'm not overclocked, I'm not overheating, and I've never had issues with this computer before, so I am hesitant to call it a hardware issue. I am running two workers (each with 2 threads) and the 2nd worker does not have any problems. The 2nd worker was very near the end of the test when I upgraded, so it completed the test that was started in 26.6 and started a new test in 27.7 within an hour or two of upgrading, with no errors.