Prime95 version 29.6/29.7/29.8

[kriesel]

Originally Posted by Prime95

I don't have an AVX-512 machine running Windows here. I use linux for my AVX-512 testing.

If you run just one torture thread instead of 16 or 20 does the torture test begin OK? If one works, does two? etc.

Have you checked the BIOS for AVX-512 offset (voltage)? Mystical has reported many early motherboards had this set improperly or not at all.

No dual-boot on your test AVX512 system? Windows VM?

[harlee]

Madpoo,

It appears that JSON isn't displaying PRP-C results the same way as non-JSON - known factors are not being shown. Please see M5342101 and M7039433 and PrimeNet Most Recent Results page (sort by Type). Even the entry on my Account Results Details page is different:

i5-5250U_1600 7039433 PRP-C 2019-03-02 16:03 0.2 M7039433 is not prime. Res64: 0D75599B41FA5DCE 1.5350

[Comm thread Mar 2 11:03] Sending result to server: UID: harlee/i5-5250U_1600, M7039433/10781201374553 is not prime.  Type-5 RES64: 0D75599B41FA5DCE. Wh10: 8EA45244,6058500,00000000, AID: 271744B661831555BCD839BE279B3753

Entry from results.json.txt:

[Sat Mar  2 11:03:34 2019]
{"status":"C", "exponent":7039433, "known-factors":["10781201374553"], "worktype":"PRP-3", "res64":"0D75599B41FA5DCE", "residue-type":5, "res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fft-length":393216, "shift-count":6058500, "error-code":"00000000", "security-code":"8EA45244", "program":{"name":"Prime95", "version":"29.6", "build":3, "port":10}, "timestamp":"2019-03-02 16:03:34", "errors":{"gerbicz":0}, "user":"harlee", "computer":"i5-5250U_1600", "aid":"271744B661831555BCD839BE279B3753"}

Prime95,

Noticed the PRP-C results was only listed in results.json.txt and not in results.txt as in the previous versions. Is this expected? Resutls for P-1 testing are being reported in both files.

[Cruelty]

Originally Posted by Prime95

If you run just one torture thread instead of 16 or 20 does the torture test begin OK? If one works, does two? etc.

Just tried with one thread - the same result.

Originally Posted by Prime95

Have you checked the BIOS for AVX-512 offset (voltage)? Mystical has reported many early motherboards had this set improperly or not at all.

Tried "Auto" / "0" / "-3" / "-5" - no change in behavior. My motherboard is MSI X299 Tomahawk with latest BIOS.
Just out of curiosity:
29.6.b3 - results in "An unsuported operation was attempted."
29.6.b2 - works fine
29.6.b1 - works fine
29.5.b9 - works fine
Something has happened between b2 and b3.

[ATH]

Originally Posted by harlee

Noticed the PRP-C results was only listed in results.json.txt and not in results.txt as in the previous versions. Is this expected? Resutls for P-1 testing are being reported in both files.

See post #26 + #28:
https://mersenneforum.org/showpost.p...4&postcount=26
https://mersenneforum.org/showpost.p...7&postcount=28

[Prime95]

Originally Posted by Cruelty

Just out of curiosity:
29.6.b3 - results in "An unsuported operation was attempted."
29.6.b2 - works fine

Good info. I'll review all changes from b2 to b4.

[Prime95]

Originally Posted by Cruelty

Any ideas what might be wrong?

Yes! Please try build 5.

[Cruelty]

Originally Posted by Prime95

Yes! Please try build 5.

No change from build 4.

[Prime95]

Originally Posted by Cruelty

No change from build 4.

Crap. Mis-built. Build 5 is the same as build 4 except for the build number.

Build 6 now available.

[Cruelty]

Originally Posted by Prime95

Build 6 now available.

It's working now, thanks!

[Cruelty]

Originally Posted by Cruelty

It's working now, thanks!

"Small FFTs" 12h stress test passed - I'd suggest lowering Max FFT size from 586K to maybe 128K, above this value heat generated is lower (highest temperature reported during run):

120k = 90C
128k = 89C
144k = 84C
192k = 74C
200k = 74C
240k = 71C
280k = 69C

BTW: only LinpackXtreme manages to heat up CPU higher (93C) than Prime95 v.29.6.b6 in my case.

[kriesel]

Originally Posted by Prime95

Crap. Mis-built. Build 5 is the same as build 4 except for the build number.

Build 6 now available.

Thanks for the fast fixes. Stuff happens.
Please update https://www.mersenneforum.org/showpo...42&postcount=2 for builds 4-6. Nice idea, that, one place to look, for a summary of issues and fixes vs. build.