Bitcoin Forum
May 28, 2024, 02:57:46 AM *
News: Latest Bitcoin Core release: 27.0 [Torrent]
 
   Home   Help Search Login Register More  
Pages: [1]
  Print  
Author Topic: What variables at play to allow some titans to undervolt far more than other?  (Read 361 times)
jacobmayes94 (OP)
Sr. Member
****
Offline Offline

Activity: 364
Merit: 250



View Profile
December 21, 2016, 06:55:16 PM
Last edit: December 21, 2016, 07:06:07 PM by jacobmayes94
 #1

I got a titan today, and compared to the other one, this one has been able to undervolt to an INSANE degree as compared to the other. While the other did not allow most dies to go below -439v etc, this has allowed most dies to go into the -8XX / -9XX range and below at 300MHz. I have not yet pushed them to 325 MHz as I am currently happy with the insanely low power usage at 300. One cube refuses to go below -9XX range on ANY die without HW errors getting stupid, but that's still good.


I understand component quality, silicon quality / individual chip and its tolerances are at play and even how good the power supply input you are feeding in, ripple among other things. Both titans were batch 1 ( think)

This puppy is able to go much lower than the other, big variance. Power supplies are simple corsair HX platinum.

I consider myself reasonably versed in electronics, the variance is big enough for me to mention it.

Anyone have any ideas to chime in? HW errors are still below 1.2%.

Have I just simply got a 'lucky' titan?

Worth the £800 I paid for this titan, at any rate.

lightfoot
Legendary
*
Offline Offline

Activity: 3108
Merit: 2240


I fix broken miners. And make holes in teeth :-)


View Profile
January 06, 2017, 12:36:08 AM
 #2

Well, having taken a number of titan dies apart I can say there were at least two die releases. It's possible the later release handled under-clocking a lot better than the older release did, but for practical purposes they always seemed similar to me.

Some hints on keeping it running:

Do not run all the dies at full speed.

Watch the temps at the PCIe power plug. If that plug gets warm it will eventually develop resistance on the grounds. Which will heat it up, melt it, then cause a ground fault that will ground the cube through the die and the controller. If you are lucky you will blow up the cube. If not you will blow up all cubes and the controller.

Always plug the ribbon cable into the cube first THEN the controller. If you plug into the controller then plug into the cube so you are on only half the pins you will blow up the controller and the cube. This is fixable but costs money to fix.

Run it at a reasonable speed, around 60-70mh tops, and mine like hell.

C
dltrading
Newbie
*
Offline Offline

Activity: 5
Merit: 0


View Profile
January 07, 2017, 05:38:18 PM
 #3

All my 3 controllers works fine. I put 6 cubes per controller now I have 2 KNC TITAN controller for sale.

Thank you again
lightfoot
Legendary
*
Offline Offline

Activity: 3108
Merit: 2240


I fix broken miners. And make holes in teeth :-)


View Profile
January 07, 2017, 05:40:13 PM
 #4

All my 3 controllers works fine. I put 6 cubes per controller now I have 2 KNC TITAN controller for sale.

Thank you again
Indeed. Glad to hear it.

C
Pages: [1]
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2006-2009, Simple Machines Valid XHTML 1.0! Valid CSS!