Topic: 7970 hot hot hot! (Read 6329 times)

Cool.... I don't think those all in 1 Eco cooler things would be all that great for cooling a mining card...

I have to give all credit to this man:

http://www.overclock.net/t/1237219/tript-cc-620-920-h50-h70-gpu-brackets-fan-grills-custom-case-badges

Otherwise I'd be using zip ties and probably making a mess of things.   

The VRMs are still fairly warm.  The Twin Frozr III's have a metal plate that covers most of them with a thermal pad underneath.  They weren't cool to begin with, usually 80°C or more at times.  But right now they're reporting 51°C/77°C with a 120mm blowing at the side.  My reference 7970 has cooler VRMs -- 51°C/68°C.

The biggest pain is the copper shim you have to put between the block and GPU.  I went cheap and got a 5-pack of 20x20x0.5mm shims for $2 on eBay.  FrozenCPU sells something similar for $3.99+shipping.

That's great, 1) I didn't know you could put a CPU cooler on a GPU! 2) How are the VRM temps on that?

A few of my MSI 7950 Twin Frozr III's hit 82°C on hot days (+90°F).  I recently added a Thermaltake Water 2.0 Performer to one of them and it's now at 57°C.  Was hoping for a better result after hooking up MSI GTX 670 up with an Antec KUHLER H2O 620 (max temp is 46°C) but then I realized that the TDP on the 7950/7970 is a bit higher than Kepler.  I do have a Water 2.0 Performer Pro which is similar to an Antec KUHLER H2O 920 which I'm going to try on the other 7950.  Hopefully it will drop the temps a little more.

Over the years, I've always criticized my uncle anytime he ran his computer with the side panel off, but I've noticed significantly lower temps on my home computer when mining by taking it off. This is with a Gigabyte 7970.

you running yours in a case? you probably need more cool air coming in if your seeing that high of temps. Also even a couple more mm between the two cards can make a huge difference if you can separate them. I run caseless though

575 is about right for 975 core...

Find somewhere else for your rigs..... My rigs are loud..... Multiple scyte high RPM fans... Lucky me it's not located at my house

That was it! I was reading the wrong side of the hash rate info. I changed the Intensity to 11 and restarted the card. I'm back up to 700 Mh/s. THANKS!

For 7970 and linux...

Use I of 11

I still keep mine at 70 target temp and just run them slower during the summer.

Lastly - 7970 mem can only run at core less 150 in linux.  Windows plays some games to run lower but not such luck with cgminer and linux.

I have been running 2 - 4x7970 rigs for 7 months now.  They are great cards!

Sometimes driver crashes with the 7970's cause ram speed, or both gpu and ram to not change when cgminer tries to set them. More than likely you just need to restart. Use GPU-z to make sure settings are being applied correctly. Heat should not cause a hashrate decrease, unless your cards are overheating (at least 80 degrees probably more like 90 though) you will see the hash rate start and stop when that happens as the card throttles itself

one more thought if you have the crossfire cable connected, its not needed unless you want to game too. I have mine connected and sometimes if I don't start mining within a minute or so of bootup the non display card will go in low power mode and not have overclocking ability at all

I run my 7970 all the way up to about 77 degrees without any stability issues

Yeah - It's a hash rate issue because card0 - the card I did nothing to, dropped 50Mh/s. This happened when I set the fan on Card 1 to 85% and the temps on both cards dropped. I'm running with an -I 9 setting. I guess I'll have to chalk it up to the big unknown...

yeah those are the default settings

I use 9 on dedicated mining cards and 6 on a desktop card.

So this is a hashrate issue right?
Are you using Dynamic Intensity? I reccomend that you Do Not Do So, because when running under Dynamic instead of a set intensity, Will cause your card to only use One mining 'thread' instead of the normal amount (usually Two threads) Resulting in a drop of mh/s

Now, Even if Both the cards are on dynamic, Just try taking them off dynamic, And setting them to Intensity 7 or 9, If you wanna use the desktop, use 6.
During dynamic, CGminer automatically adjusts the instensity (but still with one thread) to find the "best hashrate", Now thats great and all, But it's a little errorprone...
Sometimes it'll try an intensity>10  wich will cause Massive cpu usage, When this massive usage occurs, You may lose hashrate, So CGminer will go back to <10 But then again goes "i got higher mh when i was at intensity >10", Then the CPU loads up again, Hashrate drops, Repeat.

Try setting both your cards to intensity 7 or 9, 6 if you want desktop, 5 if you wanna game (IMO for higher fps)

You can set the threshold in cgminer.conf (config file)

I tried to deal with the big 120mm Scythe 3200rpm fan on the thing but it annoyed me.... ended up fitting a 5770 fan on to a 5850.... It's been working fine since....

i don't know about 7970's specifically, but ppl tend to give figures way too low for "safe" temps  (esp on overclocking sites)

if the 7970 is anything like a 5830, 5850, 5870, 80o (or probably even 85o) or lower for the GPU is fine.. as long as all GPU sensors are below 85-90o and all VRMs are below 100 (well, these I think are fine up to 110-120, but...)

until about two months ago, i kept all my cards below 70o.  but not really possible now in this heat.  i'm in texas and the comps are in a sealed off room, it's well-ventilated but temps still get to be about 5-10o above outside temp.  so if it's 100o outside, it's 110o in there.   it's cooled down a bit lately, but back when the outside temps were 105o, i'd have a few cards that would get close to 80oC during the day, lately they've been topping out around 75o...

a lot of times if you have old thermal paste on there or someone gunked it on there, what you see as your main GPU temp could be rather misleading.  there have been several cards I've gotten off of eBay that'll read 70oC but have gpu 3 sensor at 90 or 95oC.    that's not good, it usually means 'take this pos apart and re-apply some thermal paste', on rare occasions it means 'take off this pos fan and shroud and ziptie a super-loud badass 120mm fan on this mofo'.  i have one of the latter.  i put it at the bottom of my stack since it takes up 3 slots w/ this massive 120mm fan, but it keeps the middle card a lot cooler too.  

you could shave 2-5°C off by dropping the memory clock (it might hurt your hashrate depending on your kernal, On others you might get a boost)

PowerTune doesn't change the clock, it's much finer grained than that.  Think of it as adding little pauses (choppiness).

PT monitors the different parts of the GPU and *anticipates* when something might be about to overheat (exceed the allowed TTD).

PT responds proactively, throttling the GPU just enough to keep it within the allowed thermal envelope before exceeding it.

PT 0 is the default, PT 20 raises the max allowed temp by 20%.  You can use negative values too.

If you're running too hot, leave it at 0 or 5.