kingpincooling.com

Go Back   kingpincooling.com > Overclocking & Modding > TiN's Dungeon

TiN's Dungeon The mad engineer's home at KPC.

Reply
 
Thread Tools Search this Thread Display Modes
  #1  
Old 09-01-2014, 08:07 AM
mushy409's Avatar
mushy409 mushy409 is offline
Ambient
 
Join Date: Sep 2014
Location: Sheffield
Posts: 15
Thumbs up VTX3D HD7950 Repair Project

Hey guys here's my first in-depth post about this VTX3D HD7950 card I got my mits on.

First look at the card, found several components on the back side were almost knocked off. I've resoldered all the SMD caps & resistors without issue, now to turn my attention to the blown MOSFET (and also the cracked coil)...




I've already ordered several replacement MOSFETs, part # BSC050NE2LS (Infineon) and plan on swapping the full set on the card.

Also on the backside the MOSFET driver looks toast, can anyone help identify this part?



Markings on the IC:
8510
75S-10
U1136

I'm not even going to think about applying power to this card until the driver has been replaced. Can anyone help? Thanks in advance!

Last edited by mushy409; 09-02-2014 at 11:52 PM.
Reply With Quote
  #2  
Old 09-01-2014, 09:05 AM
mushy409's Avatar
mushy409 mushy409 is offline
Ambient
 
Join Date: Sep 2014
Location: Sheffield
Posts: 15
Default

T'is ok - info was right under my nose... The MOSFET gate driver is an IR3537 (CHL8510)

Just looking to order some now
Reply With Quote
  #3  
Old 09-06-2014, 02:42 PM
mushy409's Avatar
mushy409 mushy409 is offline
Ambient
 
Join Date: Sep 2014
Location: Sheffield
Posts: 15
Default

Just a quick update... Parts now on order! ETA in 2-3 days apparently but we shall see.

Parts and order #'s from Mouser.com if anyone is interested:

Mouser No: 942-CHL8510CRT
Mfr. #: CHL8510CRT
Desc.: Gate Drivers 12V HI EFFICIENCY MOSFET DRVR

Mouser No: 726-BSC050NE2LS
Mfr. #: BSC050NE2LS
Desc.: MOSFET N-Channel 25V MOSFET

Mouser No: 726-BSC010NE2LS
Mfr. #: BSC010NE2LS
Desc.: MOSFET N-Channel 25V MOSFET

Will update again once they arrive and have been installed
Reply With Quote
  #4  
Old 09-09-2014, 02:08 PM
mushy409's Avatar
mushy409 mushy409 is offline
Ambient
 
Join Date: Sep 2014
Location: Sheffield
Posts: 15
Default

I've done a little bit of experimenting while waiting for those parts. I removed the 2 MOSFETs and the blown gate driver and put the card in a test rig...

Can now power up the board, VCORE reads at just under 1v (0.998v) and VMEM reads at 1.61v

There is no display output or POST beeps, motherboard indicates a fault with GFX card (mobo LED is red for PCIe)
If I remove one of the power plugs the PC gives the usual error beeps but still nothing on display.

So, I decided to pull the card again and look for any more missing/damaged components. Found one small component marked 'B28' which I salvaged from off another card, soldered and retested the card but still the same. This component now reads 1.61v either side so my guess is this is part of the VMEM circuit.

ETA for the parts is Thursday, think I might just wait til then
Reply With Quote
  #5  
Old 09-11-2014, 04:17 PM
mushy409's Avatar
mushy409 mushy409 is offline
Ambient
 
Join Date: Sep 2014
Location: Sheffield
Posts: 15
Default

Another update!

Parts arrived this morning via Fedex. I've already installed the MOSFETs and driver.
Seems there may be more damaged components on this card

Plugged into PC and tested each phase output with a pocket oscilloscope...

Top phase (one with parts replaced) now outputs a nice clean voltage (1.05v) and I can see what appears to be the MOSFETs switching on and off (shown by rising and falling square wave)
The next 3 phases down show quite a rough looking 0.988v and there doesn't appear to be any input back to the CHL8510 from the high-side MOSFETs (there should be a similar looking wave but measuring 12v)



Further testing shows the top phase, and bottom 3 phases are working fine, the middle 3 are not.



Can anyone offer any advice on this? From what I can tell the other 3 CHL8510 are not working correctly and are not switching the high/lowside FETs as they should. The datasheet on the CHL8228 indicates that if any faults are found in the phase circuit (OVP, OCP, UVP etc) it simply shuts down the phase.

Any suggestions on my next move?

My thoughts are to replace the suspected faulty CHL8510 drivers. The MOSFETs attached to these drivers appear fine (no shorts between source & gate)
Reply With Quote
  #6  
Old 09-11-2014, 10:18 PM
mushy409's Avatar
mushy409 mushy409 is offline
Ambient
 
Join Date: Sep 2014
Location: Sheffield
Posts: 15
Default

Could really use some advice on this...

More testing, I'm not sure if the gate drivers are faulty or the actual CHL8228 is faulty?

Testing of the 7 phases (6 Vcore 1 VDDR) gave different results.

The top most CHL8510 I replaced is switching the MOSFETs underneath it as it should. The PWM signal supplied from the CHL8228 reads around 0.2v
The PWM signal going to the 3 suspect gate drivers is a solid 3.3v (no output from MOSFETs)

I have a few questions if anyone could answer them?

Would it be possible to replace the CHL8228 controller or would it require programming? (datasheet states they will not turn on without being programmed first)

Secondly, how can I tell if the PWM signal to the gate driver is good? Could someone post a voltage reading from any of the CHL8228 PWM channel outputs? (HD7950/7970)

I'm swaying more towards the CHiL controller - it isn't controlling the 3 gate drivers which is why 3 of the 6 GPU phases aren't working.

I've already spent on this card, I don't want to buy more parts and be in the same situation again so might end up buying something to make a Zombie, or buy the EPower card (if they are still available?)

Any help would be greatly appreciated!
Reply With Quote
  #7  
Old 09-12-2014, 01:56 AM
Rsnubje's Avatar
Rsnubje Rsnubje is offline
Sub-Ambient
 
Join Date: May 2010
Location: Netherlands
Posts: 81
Default

I like this little project, but unfortunatly I don't have the knowledge to help you out.

Did you already replace the cracked coil and mosfet on top of the card? If the other 3 drivers don't read 12v input, maybe the circuit to those drivers is cut or something?

Last edited by Rsnubje; 09-12-2014 at 01:58 AM.
Reply With Quote
  #8  
Old 09-12-2014, 10:19 AM
mushy409's Avatar
mushy409 mushy409 is offline
Ambient
 
Join Date: Sep 2014
Location: Sheffield
Posts: 15
Default

Hey Rsnubje - thanks for your reply.

Yes I've replaced the coil (R47), the top gate driver (CHL8510) and 2 MOSFETs that it controls. This phase seems to be ok, voltage readings from both hi/low side are good.

Spent a few hours this afternoon butchering a PC motherboard to reclaim a PCIe socket so I can test the card outside the PC. The adaptor works great (thanks TiN!)
At least this way I can probe both sides of the board with my scope.

I've been VERY careful with this card - no tracks have been cut/modified and the only soldering I have done was removing those FETs and driver with a hot air wand & preheater.

What I'm stuck on at the moment is whether the gate drivers are toast (CHL8510) or the actual PWM controller as the PWM signal for each phase differs. I traced the PWM signals back to the CHL8228 and tested at the IC - still the same readings as there are no passive components on the PWM line.

One thing I have noticed is that now 4 of the phases don't seem to be working correctly and when I probe the MODE pin on the CHL8510 (pin 8) of these phases I can hear a faint "click" and the PWM kicks into protection mode by switching off ALL of the phases (reading is 0.1v)

I'm 100000% positive nothing was shorted as I'm using a needle point probe with only 1-2mm of tip showing to prevent any shorts, and also a magnifier to see where I'm at.

If I probe pin 8 of the top (far right in the pic) CHL8510 and also the bottom (far left) this pin reads voltage and the PWM continues to run.
If I probe any in between then the PWM shuts down...

Again, faulty PWM controller or faulty CHL8510? I'm not sure...

I don't want to start pulling out parts just yet, neither do I want to rush into buying more components without being sure. Plus the fact what I've spent so far on the card & parts I could have bought an EVGA Epower card and just have done with the borked VRM circuit.

Damn I wish I finished my A-level electronics! Practical knowledge is great, the theory & maths behind it is another!
Reply With Quote
  #9  
Old 09-14-2014, 05:56 AM
mushy409's Avatar
mushy409 mushy409 is offline
Ambient
 
Join Date: Sep 2014
Location: Sheffield
Posts: 15
Default

FINALLY!!! Some good news on this project, but also some bad news...

After researching on various sites I came across the bios mods/cross-flashing which got me thinking...

Neither bios on the card would allow the PC to boot, so I downloaded around 8 different bios ROMs for this model card and flashed them one by one taking care that the supported memory type and device ID were the same (some ROMs support only Samsung/Hynix, some support all 4 types)

After finding the correct bios for the card, voila! My desktop posted but was presented with lots of grey lines down the screen. I let my PC load up but at the point just after the windows splash screen it just hangs at a black screen. Booting into safe mode got me to the desktop and allowed the drivers to be installed but there was still the grey lines so I'm guessing the RAM is damaged by either overvolting or due to the incorrect bios - probably both



I had a quick look at the DS for the Elpida RAM and the VDDC should be 1.6v. Currently with the card kinda working the RAM measures 1.67v. slightly higher than I'm comfortable with but hey it's a start.
I'm pretty sure that before flashing the various ROMs I measured 1.77v at the RAM.

The only problem with eBay is it tends to be full of liars who will chat s**t to sell you something that was broken not by accident or from wear & tear, but from inexperience.

Now what I plan to do now I've made some progress...

1) Reflow the GPU as it may be due to one or more of the memory IO lines not connected properly. The card is slightly warped so I'll need to clamp it down and preheat it in an attempt to straighten the PCB, then reflow.

2) Reflow the RAM chips (this worked for a HD4870 x2 I had with similar problem)

3) If 1 & 2 fail, I'll attempt to source some new VRAM chips and replace them one by one - Yes it's a long process but I'm not giving up just yet.

I'll keep this thread updated once I have some free time to 'dabble' some more. Right now I'm happy that the card is functioning in a way, but annoyed at the guy on eBay who claimed "It was working fine one day, then all of a sudden stopped working..."
If that was the case how do you explain BOTH bios' incorrectly flashed, a blown MOSFET and driver, and all those missing components? Some people eh?

Reply With Quote
  #10  
Old 09-14-2014, 06:52 AM
mushy409's Avatar
mushy409 mushy409 is offline
Ambient
 
Join Date: Sep 2014
Location: Sheffield
Posts: 15
Default

Just a small update... I've managed to source some replacement DDR5 RAM from eBay, but they are going to cost a fair bit

I've requested a quote for 12 of these chips so I can replace the whole lot

http://www.ebay.co.uk/itm/1911891595...%3AMEBIDX%3AIT
Reply With Quote
Reply


Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT -5. The time now is 07:22 AM.

Contact Us - www.kingpincooling.com - Archive - Top
Powered by vBulletin® Version 3.8.3 Copyright ©2000 - 2017, Jelsoft Enterprises Ltd.