mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Hardware > GPU Computing

Reply
 
Thread Tools
Old 2016-06-01, 07:10   #1
UBR47K
 
UBR47K's Avatar
 
Aug 2015

22·17 Posts
Default Nvidia 364.xx drivers returning incorrect results

Has anyone experienced getting wrong residues with Nvidia 364.xx with CUDALucas?
I recently went investigating for the reason why my Titan Z decided suddenly crapping wrong results out and the timeframe coincides with a driver upgrade (I am using Linux here).

Googling around, I found this thread https://www.primegrid.com/forum_thread.php?id=6769 which describes issues not too different from what I am experiencing right now.
https://forums.geforce.com/default/t...71356/#4871356 shows that Nvidia acknowledges some issues with compute with their driver.

I have no way to rollback drivers currently so I can't be 100% sure if my card could have crapped out.

If the driver turns out to be bad, could it mean that the "recent" CUDALucas and mfaktc results be wrong? The driver is already out for a few months now.

EDIT:
One important note: The results I get some of them were correct but quite a sizable portion were mismatchings which turned out to be wrong.

Last fiddled with by UBR47K on 2016-06-01 at 07:12
UBR47K is offline   Reply With Quote
Old 2016-06-11, 16:58   #2
mognuts
 
mognuts's Avatar
 
Sep 2008
Bromley, England

628 Posts
Default

BOINC/PrimeGrid is reporting this:

PrimeGrid: Please upgrade your Nvidia drivers!
Recently, Nvidia released driver 368.22 which fixes the serious error introduced with their 364.xx driver. Please upgrade to 368.22 or later, unless you are running Windows XP or Windows Vista.

If you are running the 364.xx or 365.xx drivers, it is imperative that you upgrade to 368.22 or later. The 364.xx and 365.xx drivers can produce incorrect calculation results.

However, if you are running Windows XP or Windows Vista, these versions of Windows are no longer supported by Nvidia's drivers as of 368.22. If you're running XP or Vista, you must revert to an earlier version of the driver (363.xx or earlier).
mognuts is online now   Reply With Quote
Old 2016-06-11, 22:38   #3
bgbeuning
 
Dec 2014

3·5·17 Posts
Default

When I was running CUDAlucas and got Linux updates to nvidia,
I had to recompile CUDALucas or I would get all 0 residues.
bgbeuning is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Nvidia unlocking Titan X (P) through drivers Karl M Johnson GPU Computing 3 2017-08-11 08:40
NVIDIA Quadro K4000 speed results benchmark sixblueboxes GPU Computing 3 2014-07-17 00:25
Latest Nvidia drivers 331.40 & Titans nucleon GPU Computing 11 2013-10-28 21:28
ecm_factor returning the same number as input Capone GMP-ECM 17 2007-06-17 09:19
PRP v2.3, incorrect results for k > 2^32 Angular Math 6 2002-09-26 02:28

All times are UTC. The time now is 19:24.


Tue Mar 28 19:24:24 UTC 2023 up 222 days, 16:52, 0 users, load averages: 0.54, 0.72, 0.85

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2023, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.

≠ ± ∓ ÷ × · − √ ‰ ⊗ ⊕ ⊖ ⊘ ⊙ ≤ ≥ ≦ ≧ ≨ ≩ ≺ ≻ ≼ ≽ ⊏ ⊐ ⊑ ⊒ ² ³ °
∠ ∟ ° ≅ ~ ‖ ⟂ ⫛
≡ ≜ ≈ ∝ ∞ ≪ ≫ ⌊⌋ ⌈⌉ ∘ ∏ ∐ ∑ ∧ ∨ ∩ ∪ ⨀ ⊕ ⊗ 𝖕 𝖖 𝖗 ⊲ ⊳
∅ ∖ ∁ ↦ ↣ ∩ ∪ ⊆ ⊂ ⊄ ⊊ ⊇ ⊃ ⊅ ⊋ ⊖ ∈ ∉ ∋ ∌ ℕ ℤ ℚ ℝ ℂ ℵ ℶ ℷ ℸ 𝓟
¬ ∨ ∧ ⊕ → ← ⇒ ⇐ ⇔ ∀ ∃ ∄ ∴ ∵ ⊤ ⊥ ⊢ ⊨ ⫤ ⊣ … ⋯ ⋮ ⋰ ⋱
∫ ∬ ∭ ∮ ∯ ∰ ∇ ∆ δ ∂ ℱ ℒ ℓ
𝛢𝛼 𝛣𝛽 𝛤𝛾 𝛥𝛿 𝛦𝜀𝜖 𝛧𝜁 𝛨𝜂 𝛩𝜃𝜗 𝛪𝜄 𝛫𝜅 𝛬𝜆 𝛭𝜇 𝛮𝜈 𝛯𝜉 𝛰𝜊 𝛱𝜋 𝛲𝜌 𝛴𝜎𝜍 𝛵𝜏 𝛶𝜐 𝛷𝜙𝜑 𝛸𝜒 𝛹𝜓 𝛺𝜔