Core Throttling on 7970, VRM 1 temps a possible culprit? Causes unplayable FPS dips.

Toggle sidebar Toggle sidebar
F

Fanjita

Honorable
Jan 17, 2013
15
0
10,510
Hi all,

So I was merrily playing BF4 yesterday (and have fine for many, many months) when i witnessed a drop in FPS for a short while with no explainable trigger. The FPS climbed to a more manageable condition and I carried on. It then started a cycle, every 30 or so seconds my FPS would drop quite significantly.

I'd be happily playing at 70-90 fps then suddenly it drops to 30-40 for a few seconds then back again. Tried a few other games and the drop is now present in all of them too.

Since this problem started I've tried a few things;
Opened PC and reseated CPU with new thermal compound, cleaned out all dust too.
Complete reinstall of graphics drivers
Reset BIOS and reinstated overclock settings
Checked CPU throttling and park settings

This is all to no avail, the FPS drops still occur. More interestingly, I've noticed it doesn't immediately do it, I have to be playing for a few minutes before it begins. I'm becoming more concerned that due to the sudden appearance of this problem and the fact the game has to be running for a few minutes that it may be a hardware problem.

My system temps all seem acceptable, hottest my graphics gets is 87 degrees centigrade whilst the hottest CPU core is 67 degrees.

PC:
CPU: I5 2500K (Scan bundle - overclocked by their engineers to 4.5 gHz with memory and mobo)
RAM: Corsair DDR3 1600Mhz
GPU: XFX 7970 Black Edition
Motherboard: ASUS P8Z68-V LE
CASE: Coolermaster Elite 330U

I have owned the overclocked bundle for around 3 years now, and the GPU for less than a year but bought second hand.
 
Sort by date Sort by votes
laviniuc

laviniuc

Judicious
Nov 29, 2013
5,505
18
32,415
open up gpuz for example and check out the gpu frequencies while the fps drop. if there's a direct link you have your answer...

if this is the case you can most likely downclock the gpu about 50mhz without noticing a lot of difference and it would help decrease temps a few degrees at least. i do say it may be the gpu but those temperatures are normal. the danger zone starts at 90-95.
 
Upvote 0 Downvote
F

Fanjita

Honorable
Jan 17, 2013
15
0
10,510


Sorry I'm getting my head around this forum, i should've replied rather than answered i think....

Check the link below

https://www.dropbox.com/s/e9ewl67i76nve1r/2.gif

Well, looks like you guys are on to something, check out the core clock, it's throttling alright.

Also, the VRM Tempurature 1, it seems to hit 114 degrees, is this a possible cause?

Might have to look at cleaning the dust out as maybe it's blocking one of the memory cooling blocks?
 
Upvote 0 Downvote
F

Fanjita

Honorable
Jan 17, 2013
15
0
10,510


Ok to test I decided to manually increase the fan rpms, as the standard control seems to do a terrible job of keeping the temps down (it hardly goes over 50%rpm)

The spikes have stopped and the VRM temp doesnt exceed 100 degrees. Looks like i just need to clean the heatsink channels and this should be the end of it.

I presume taking it apart to reapply thermal compound invalidates my lifetime warranty?
 
Upvote 0 Downvote
F

Fanjita

Honorable
Jan 17, 2013
15
0
10,510
Sorry to resurrect this...

Been having some issues. This problem was getting more and more frequent until I contacted XFX who refused an RMA and basically I don't have a leg to stand on with them.

So I took the heatsink off and was met with a terrible excuse for thermal compound on the core. Replaced this and reseated all the thermal pads (which look fine), temperature problems still there. After countless strips of the unit and adjusting the pads, the problem still occurs. On a standard clock setting, the card will increase to 115 degrees VRM 1 and just sit there, spiking as the core throttling occurs and then climbing back when it attempts to recover.

The GPU core is very good now, and VRM 2 is fine also. I cannot see any damage on the single VRM 1 ( sat towards the DVI port end of the card, right?). I managed to salvage some thermal pads from other areas of the card where there was a clear excess, and seated this on the VRM 1 module.

The temperature still climbs to critical and just sits there....

Can anyone offer any suggestions? Anything I havent considered? Do damaged VRMs have heat problems?
 
Upvote 0 Downvote
F

Fanjita

Honorable
Jan 17, 2013
15
0
10,510


They are saying I'm out of the warranty period, I don't have a lifetime warranty according to them.

It's an 800W Thermaltake PSU
 
Upvote 0 Downvote
F

Fanjita

Honorable
Jan 17, 2013
15
0
10,510


I am actually the second owner, the first owner didn't register for the extended warranty annoyingly. it appears to be more than 2 years old (just) now too, so the normal warranty is out the door.

My only chance is through the person I bought it from, but although cooperative, there's very little this person can do aside from a refund. I can't rely on that happening either.

I just feel it should be 'fixable' as the card boots up fine and only starts to get too hot in games. I have tried everything bar replacing the thermal pads entirely, I have however reseated them all. Am I right in saying VRM 1 is the bank of VRMs to the right of the core?

The thermal pads don't look especially tired, and I flipped the strip over on the VRM 1 section, made very little difference.

Can a faulty VRM be absolutely fine until the card starts to take on more load? Then have heat problems?

 
Upvote 0 Downvote
S

swifty_morgan

Illustrious
Feb 10, 2006
13,566
1
42,960
run an app that will let you under clock the cards processor and the cards ram.

as a matter of fact there's that option in AMD's CCC under over drive............

and while the thought comes to mind............ I believe some of those cards had some kind of BIOS switch>>>>????? not sure.......... if you flip the switch to one side or the other or maybe it's "stuck" half way between?
 
Upvote 0 Downvote
F

Fanjita

Honorable
Jan 17, 2013
15
0
10,510


I have underclocked and undervolted it, and the card responds well to it. What seems to happen now though is the card just takes longer to get to critical temp, but still eventually gets there. The frequency of the throttling is less too as a result.

I've tried both positions on the BIOS, and also tried flashing new bios. The bios I've tried aren't 100% compatible though, and I lose my second monitor as a side effect. It's running the switch position for the read only bios.

While I'm thinking about it, there are quite a few threads of people hitting 122 degrees centigrade or higher without problem. Am I singling the VRM temp out wrongly? Is this just a red herring and the real problem lies elsewhere?

Why else would it throttle?

Is my screenshot above from GPU-Z showing anything else amiss? (voltages, current draw, etc)

I'm considering looking at 3rd party heatsink setups as mine is the reference card. There is no point however if this actually isn't the culprit. Since doing the thermal compound on the GPU it's temps have plummeted.
 
Upvote 0 Downvote
S

swifty_morgan

Illustrious
Feb 10, 2006
13,566
1
42,960
thought about this last night. the ram on the graphics card can be another problem. have you had the heat sink off to reapply new thermal paste and if you did what do they have to cool the ram chips?

if 1 of those chips is bad the whole card is finished. is the card under some kind of warranty?
 
Upvote 0 Downvote
F

Fanjita

Honorable
Jan 17, 2013
15
0
10,510


Thanks for your commitment to my conundrum swifty, unfortunately this is now a closed case. My warranty (was mentioned in a previous comment) is non existent. It's outside the 2 years standard, and the previous owner (who assured me it's covered by a lifetime one) didn't register in the first 30 days. I'm not holding out much hope the seller will assist me however, even though it has only been around 3 months of ownership...

Unfortunately, on attempting to show the symptoms to a knowledgable family member, the system booted up and I ramped the fan speeds up in preparation for the 3d load test. Minutes later the system turned off suddenly (before the game had even loaded) and the displays would not come back online. The smell then confirmed the card has somewhere had a component failure. One dead 7970. Tested with an ancient gtx 260 confirms my system is still healthy and boots up fine, but the 7970 is now dead.

The smell seems to originate towards the VRM1 area but could be any number of components. Looks like no end of moving and reseating cooling parts would have prevented this. The memory was covered with thermal pads, as are the VRMs. No thermal pads had yet been replaced however, and all of them looked undamaged and healthy when reseated.

It might be worth adding no changes to the clock/core/voltage were in place at the time the card went, nor have I or the previous owner (or so he says) ever clocked it more than factory.
 
Upvote 0 Downvote
Z

Zvonimir Kokorovic

Reputable
Sep 3, 2014
1
0
4,510
This is seriously sad that no one in here has told this person anything useful and is suggessting he replace a card that will work just fine if it has NEW thermal pads put on the VRM 1 strip which is the LONG strip between the GPU and the PEG power connectors.


http://www.overclock.net/t/1455768/290x-vrm1-high-temps-ek-waterblock

this issue has not just been isolated to the 290xs, as my 7970s also had VRM1 heat issues.

Switching thermal pads to FUJIPOLY EXTREME dropped my VRM1 temps over 40 degrees.

I was astonished, still am, and hope its not too late for the OP and he still has the card to perform this relatively cheap and simple ($20) fix, with enough TIM left over to do a bunch of other cards too should he have to.

Good luck
 
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom