Titan's Best Choice - Page 2

Madpoo · 2016-06-02, 14:21

Quote:

Originally Posted by chalsall

Or, more specifically, the dew point.

Here in Bimshire our rum and ice and coconut water drinks sweat water as condensation.

So do our servers, if not properly managed....

I experience this sensation from time to time when I'm in a server room working on something and standing right over/under the incoming cool air vent. It gets chilly.

Then when I leave the room and enter the real world, almost instantly (if it's humid enough), I'm covered in a sheen of condensation. Fortunately I'm mostly water proof and I haven't short circuited.

ATH · 2016-06-02, 20:27

You also have to figure out at what temperature the GPU is most efficient. Either in terms of total GFLOPs or in GFLOPs per Watt.

I would think this might be higher than room temperature?

VBCurtis · 2016-06-03, 02:17

Lower temperature -> less vibration -> less noise -> less voltage needed for signal to overcome noise -> less power needed for a given speed.

Mild overclocking (or stock speeds) with good cooling can manage substantial undervolting of CPUs, I assume GPUs are the same. It's a nice loop- lower temp requires lower voltage for signals produces less heat.

Anyway, I can't think of a reason that a higher temp would be more stable or more efficient, among temp choices above 10C. Super-cold settings could possibly inflict instability when only some parts of a card are cold while others are background temp.

thyw · 2016-06-03, 09:58

Cannot speak from experience, but there was something about room temp and pc temp: the bigger the difference, the faster the cooling is(the temps are going to a "thermal balance"...). So the cooling "slows" down, if it's close enough to room temp.
So maybe geting too close to room temp isn't efficent, since it'll need more cooling (room temp air) to get lower.
Warning: this comment ~~probably~~ may be total bs nonsense.

TObject · 2016-06-06, 18:31

My Titan, my beautiful Titan has burned. There is a burn mark on the back of the board opposite the fan, and it is shorting out power.

Has anyone had a Titan replaced on warrantee recently? What are they replacing it with, another Titan or something not CudaLucas worthy?

I am debating whether to take it apart and see if I can fix it myself; maybe it is just a blown cap.

Edit: never mind about my warrantee question; turns out I am behind the curve on EVGA policies; they do not cover anything off eBay anymore. Now, where is my Torx set...

TObject · 2016-06-06, 19:51

Well, it looks like where there are two R33 inductors next to each other, the whole area around looks nasty not sure what exactly got damaged.

A PCB picture I found on the Internet shows R47 inductors, where where there are R33 are on mine. Here is a link to the internet pictue (not mine):

http://hardzone.es/app/uploads/2013/...-gk110-PCB.jpg

TObject · 2016-06-06, 20:45

It appears R33 inductors blowing up is a common problem on these cards. Some folks on the Internet recommend putting thermal pads on the inductors (as is they are sitting in a cutout and must heat up quite a bit). So when you have your Titan apart consider doing that.

LaurV · 2016-06-07, 04:18

I have already two Titans with this problem, and someone else here (was it airsquirrel?) had one. I repaired one of the two (see my posts in the hardware thread) by changing the coils (also "33" on mine all 6 cards) and the mosfet which was driving it (that is why you have the short). This is quite a common design problem with Titans, especially air cooled, where the heatsink has a hole in that place, covering those mosfets only partially. The second card was water cooled, I didn't repair it yet. You will need a hot-tin-bay or hot air blower to take out that mosfet, it has a very strange soldering pattern under it, with quite sensible tracks...

TObject · 2016-06-08, 21:10

LaurV,

Thank you for sharing the wisdom. I think this is the post you are talking about:
http://mersenneforum.org/showthread....542&post426542

What is the datasheet you are referring to in the post?

Also, have you happened to have the MOSFETs part numbers; the markings are completely burned off of mines... Is 4901NF from airsquirrels' post it?

LaurV · 2016-06-09, 06:05

Quote:

Originally Posted by TObject

What is the datasheet you are referring to in the post?
<snip>
4901NF ?

Yes, I was referring to this datasheet, publicly available from the vendor. The copper pattern under that mosfet is easy to destroy if you pull it hard without melting the tin properly.

TObject · 2016-06-09, 17:07

Gotcha, thanks.

So, the shorted power line is the indication of blown MOSFET or possibly traces under it get cross-connected?

I am wondering if it is worth it trying to replace just the coils, and see what happens. As I will need to purchase a soldering heat gun to deal with the MOSFETs.

Just a blown coil, with everything else intact, would not short out the power line, or would it?

2016-06-02, 20:27	#13
ATH Einyen Dec 2003 Denmark 2²×863 Posts	You also have to figure out at what temperature the GPU is most efficient. Either in terms of total GFLOPs or in GFLOPs per Watt. I would think this might be higher than room temperature? Last fiddled with by ATH on 2016-06-02 at 20:29

2016-06-03, 09:58	#15
thyw Feb 2016 ! North_America 2³×11 Posts	Cannot speak from experience, but there was something about room temp and pc temp: the bigger the difference, the faster the cooling is(the temps are going to a "thermal balance"...). So the cooling "slows" down, if it's close enough to room temp. So maybe geting too close to room temp isn't efficent, since it'll need more cooling (room temp air) to get lower. Warning: this comment ~~probably~~ may be total bs nonsense. Last fiddled with by thyw on 2016-06-03 at 09:59

2016-06-06, 18:31	#16
TObject Feb 2012 3⁴·5 Posts	My Titan, my beautiful Titan has burned. There is a burn mark on the back of the board opposite the fan, and it is shorting out power. Has anyone had a Titan replaced on warrantee recently? What are they replacing it with, another Titan or something not CudaLucas worthy? I am debating whether to take it apart and see if I can fix it myself; maybe it is just a blown cap. Edit: never mind about my warrantee question; turns out I am behind the curve on EVGA policies; they do not cover anything off eBay anymore. Now, where is my Torx set... Last fiddled with by TObject on 2016-06-06 at 18:46

2016-06-07, 04:18	#19
LaurV Romulan Interpreter "name field" Jun 2011 Thailand 2833₁₆ Posts	I have already two Titans with this problem, and someone else here (was it airsquirrel?) had one. I repaired one of the two (see my posts in the hardware thread) by changing the coils (also "33" on mine all 6 cards) and the mosfet which was driving it (that is why you have the short). This is quite a common design problem with Titans, especially air cooled, where the heatsink has a hole in that place, covering those mosfets only partially. The second card was water cooled, I didn't repair it yet. You will need a hot-tin-bay or hot air blower to take out that mosfet, it has a very strange soldering pattern under it, with quite sensible tracks... Last fiddled with by LaurV on 2016-06-07 at 04:20 Reason: s/is/it/

2016-06-08, 21:10	#20
TObject Feb 2012 3⁴·5 Posts	LaurV, Thank you for sharing the wisdom. I think this is the post you are talking about: http://mersenneforum.org/showthread....542&post426542 What is the datasheet you are referring to in the post? Also, have you happened to have the MOSFETs part numbers; the markings are completely burned off of mines... Is 4901NF from airsquirrels' post it? Last fiddled with by TObject on 2016-06-08 at 21:12

2016-06-03, 02:17	#14
VBCurtis "Curtis" Feb 2005 Riverside, CA 1011011011110₂ Posts	Lower temperature -> less vibration -> less noise -> less voltage needed for signal to overcome noise -> less power needed for a given speed. Mild overclocking (or stock speeds) with good cooling can manage substantial undervolting of CPUs, I assume GPUs are the same. It's a nice loop- lower temp requires lower voltage for signals produces less heat. Anyway, I can't think of a reason that a higher temp would be more stable or more efficient, among temp choices above 10C. Super-cold settings could possibly inflict instability when only some parts of a card are cold while others are background temp.

2016-06-06, 19:51	#17
TObject Feb 2012 3⁴×5 Posts	Well, it looks like where there are two R33 inductors next to each other, the whole area around looks nasty not sure what exactly got damaged. A PCB picture I found on the Internet shows R47 inductors, where where there are R33 are on mine. Here is a link to the internet pictue (not mine): http://hardzone.es/app/uploads/2013/...-gk110-PCB.jpg

2016-06-06, 20:45	#18
TObject Feb 2012 195₁₆ Posts	It appears R33 inductors blowing up is a common problem on these cards. Some folks on the Internet recommend putting thermal pads on the inductors (as is they are sitting in a cutout and must heat up quite a bit). So when you have your Titan apart consider doing that.

2016-06-09, 17:07	#22
TObject Feb 2012 3⁴×5 Posts	Gotcha, thanks. So, the shorted power line is the indication of blown MOSFET or possibly traces under it get cross-connected? I am wondering if it is worth it trying to replace just the coils, and see what happens. As I will need to purchase a soldering heat gun to deal with the MOSFETs. Just a blown coil, with everything else intact, would not short out the power line, or would it?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Titan Black	ATH	Hardware	15	2017-05-27 22:38
Nvidia announces Titan Xp card	ixfd64	GPU Computing	10	2017-05-17 15:19
Nvidia announces Titan X	ixfd64	GPU Computing	20	2015-04-28 00:27
Geforce GTX Titan 6GB	ATH	GPU Computing	295	2013-05-12 21:35
2x AMD 7990 or 2x Nvidia Titan ??	Manpowre	GPU Computing	27	2013-05-12 10:00