4890 running hot

[medmn]

I have been running two HIS 4890 with EK full cover blocks for about a year. Recently had an overheat due to an air bubble apparently. Computer shut itself down. Checked it out found card really hot. Bled out the air, topped off coolant and restarted. Since then the top card - the one that overheated runs about 17 to 18 degrees C hotter than the lower card - lower card is where the monitor is connected. I have removed the block, disassembled, cleaned and inspected - no apparent problems found. Installed new thermal pads on the regs etc., and new thermal paste on the GPU. Card ran cooler for about a week and now its running hot again. Presently lower card at 38c upper at 57.5c. Am I missing something?

Thanks for any help offered.

 

[rubix_1011]

Are you running them in serial or parallel?

If serial, it doesn't seem you would have blockage or the 2nd card would also suffer in temps. If parallel...could be blockage in your first block, but if you disassembled and cleaned, likely not.

Sounds more like either a warped block or a bad seat on the hot card. A razor blade diagonally across the GPU contact should tell if it is warped; you shouldn't see any space between the blade and seat.

Confirmed your flow is OK? Was the air bubble in a rad or how did you discover it? You usually don't see air stop a loop like this after being running for that long, unless you had a component leak coolant out, and air in. I've been running at least 2.5 years on the same water...haven't had to add any yet.

Strange that it runs cool for a week and then gets hot again...any other info that might help?

 

[medmn]

running parallel with EK bridge. I'll try your suggestion with razor blade. Interestingly enough tonite it got over 68c and then started cooling down - currently 38c on the bottom and 49c on the top??? makin me nuts.

 

[rubix_1011]

Parallel...is there something wrong with the bridge? The fact that you are running parallel helps isolate the issue to the single card either with the block, the card or the bridge. A good test would be to swap cards (keeping same blocks) to see if position on the bridge is the issue. If the 2nd card remains cool in the 1st slot, its the 2nd block/card. If the 2nd card gets hot in the 1st slot- its something with your bridge.

 

[nforce4max]

I have noticed that some cards tend to warp over time and some cards from that era tend to fail that way with unexplained rises in temp even though everything else is normal especially when everything else is kept up in great shape. The second card is ok but if you are will to try everything you can take apart the block to inspect the seals that nothing is dry rotted or the block has been corroded. I have seen some blocks that looked more like Swiss cheese than metal after a few years of use.

 

[rubix_1011]

If he had a bad seal, its likely he had a leak...which could potentially explain the air bubble, but he should still have flow keeping the card cool. If it was that big of a temp difference due to a bad seal and bad flow, you should be able to see visible signs of a leak from that block.

 

[boiler1990]

A good test would be to swap cards (keeping same blocks) to see if position on the bridge is the issue. If the 2nd card remains cool in the 1st slot, its the 2nd block/card. If the 2nd card gets hot in the 1st slot- its something with your bridge.

Other way around, right?

 

[rubix_1011]

He was saying the top card gets hot, bottom stays cool...if you flip them and get the same result, I'd say it has to do with the bridge. If the cards stay cool and hot like before...its the card or block causing the issue.

 

[boiler1990]

I understand what you're saying, I just think you meant it the other way around (not saying you're wrong - just don't want it to be misinterpreted).

If he swaps the cards and the 2nd card is still hot it's the block.

If he swaps the cards and the 1st card is now hot, then it's the bridge.

 

[rubix_1011]

Yeah, that's what I meant.

I haven't had enough coffee today.

 

[medmn]

Hey guys talked with an engineer friend who suggested it may be a simple matter of balanced flow. Meaning I need to put returns at both ends of the bridge on the outlet side - he thinks the outflow of top card isn't merging with the outflow on the lower one. Also said I could have a pressure differential due to a slight restriction in the bottom card circuit - which would increase the outlet pressure - making it impossible for the top card outflow to overcome the higher pressure - thereby stopping the flow. Sheesh!! I'll keep you posted as I work through this. 1st thing is to disassemble bottom block to eliminate anything obvious re an obstruction. 2nd thing will be to plumb outlets from top and bottom to a Y fitting downstream. Thanks for your help and suggestions so far.

Regards

 

[rubix_1011]

Is it not possible to run them serial instead of parallel or is the bridge block only capable of parallel?

 

[boiler1990]

Is it not possible to run them serial instead of parallel or is the bridge block only capable of parallel?

The EK bridges are either serial or parallel. The design is internal and bulit-in, so it can't be changed.

 

[nforce4max]

I prefer serial anyway but that has its own problems and risks.

 

[boiler1990]

Yeah, my 6950 2GB temps are pretty good in serial...40/45 max for GPUs 1 and 2.

 

[medmn]

System now running normal!! Did nothing except add a little fluid to replace some I lost when trying to bleed the bridge!! I'll repost if I find or learn anything else. Thanks to all

 

[boiler1990]

Any idea what it was?

 

[medmn]

2 possibilities - one system was overfull on coolant-no place for the air to go.
second - just a freak air bubble that got hung up in the top card. As I said in post above didn't really do anything except optimize the coolant level. I had been trying to bleed the bridge on the top of the outlet side, so had to keep adding coolant. Overfilling may have caused my problem - honestly don't know - I'll repost if I figure it out definitively. Thanks to all for the suggestions and responses.

 

[rubix_1011]

That's really odd...sounds like air somewhere in the top card and it wasn't creating enough vacuum to pull the water through the parallel block somehow?

Not sure on this one...

 

[nforce4max]

Take the system and set it on it's side then move it around till you see any air bubble up in one of the lines.

 

[rubix_1011]

He got it fixed...however, wasn't sure what the actual problem was...topping off water was the only apparent solution, but wasn't running 'low' (according to OP) but who knows...still sounds like air causing an issue with the vacuum on the water flow through the parallel block.

 

[medmn]

Best answer selected by medmn.

 

[medmn]

Thanks for all the advice folks. Incidentally, I run the pump for many hours using a separate power supply, and very carefully bleed the system before I power it up. I'm using a Thermaltake Sword case, a little too heavy to be rasslin' around regarding turning on its side etal. Thanks again to all.