7990s at 100C - am I pushing them too hard?

[CJxD]

Before you instantly say 'yes', hear me out.

Not gonna lie, I'm mining scrypt-based coins. I started with 1 MSI 7990 GPU. This was reaching 90C and dropping back to 85 when the mining program throttled it. The hash rate was fine though.

Overclocking and underclocking didn't do anything to help. I'm using MSI Afterburner, but it seems it doesn't make any difference to the hash rates and GPU temperatures if I play about with the core and memory clocks.
With my last 7970, this made all the difference.

(Oddly, with my first GPU, I have to change the memory clock to something other than 1500 to get core 0 to do its work properly, core 1 is fine. My second card doesn't do this. Silicon lottery maybe?)

Now I have 2 cards in. And their thermal properties mess with the mining program. Card 1 Core 2 reaches 100 degrees nearly straight away. When it gets throttled by the program, it stays near 100 degrees.
As for the rest, the mining intensity is lowered from 13 to 12. This makes a massive difference to the hash rates, but not a lot of difference to the temperature.

So I decided to scrap throttling entirely as I realised 100C is the hard-set maximum.

Now I'm just letting this maximum take care of things. I'm seeing acceptable hash rates for both cards. The top card is constantly 100C, and the bottom card is between 95C and 100C.

It's been running for 10 minutes now - I don't seem to have any other issues so far.

So is this safe?

 

[benjii]

It will reduce the life of your cards. How quickly they deteriorate is anyones guess. The question I suggest you ask yourself is, can you afford to lose them if they burn out within a year or two?

I'm not saying they will die in that timeframe, it's more of a worse case scenario.

 

[Silent_Scone]

I'm not sure on the threshold on those cards, I'm sure someone will be along shortly to give you a clearer answer. Either way at those temps you'll reduce the life span by a fair wedge.

However, I'd imagine if you're at 100c whilst mining as a constant, then eventually something is going to give. The VRM won't be thanking you for it at those temps. Just keep an eye on it, if it goes any higher than 110C I'd call it a night.

 

[CJxD]

I've set a software cutout at 105, just in case.

I only intend to keep these cards for 2 months before the mining difficulty goes above what I can afford.

If only I could watercool these things cheaply...

I don't suppose there's a way of setting a lower hard limit? Maybe 90C?

 

[LtMatt]

7990 has superb Volterra vrms that can operate up to 120c. The core however should be kept at or below 90c ideally. 95c should be ok. Would keep below that if possible.

 

[Diagro]

I would Stop.. And Watercool them.

I would Not run anything over 90c imo

 

[Nelly]

Guess they not 'under volted'?

 

[Uncle Petey]

I had a single 7990 mining yesterday set to 1.1v @1000/1200 and it only reached 56/65C on the cores and 65C on the VRMs.

Two 7990's at the same clocks/volts did reach about 97C even with a side fan pumping loads of heat out before I shut it down.

I'd suggest undervolting, as at stock clocks/1.2V, the single card would push into the 80's or more.

Anyway, I think I've managed to secure the last two 7990 blocks in the land unless they get back to me with some stock error crap.

 

[soundb0y]

I had a single 7990 mining yesterday set to 1.1v @1000/1200 and it only reached 56/65C on the cores and 65C on the VRMs.

Two 7990's at the same clocks/volts did reach about 97C even with a side fan pumping loads of heat out before I shut it down.

I'd suggest undervolting, as at stock clocks/1.2V, the single card would push into the 80's or more.

Anyway, I think I've managed to secure the last two 7990 blocks in the land unless they get back to me with some stock error crap.

What hashrates were you getting?

I also have the problem of core 0 only hashing at about 60k. What memory clock did you find worked best?

 

[Uncle Petey]

Around 500k per core IIRC. The settings weren't tweaked properly as we're just having a little go here and there while waiting for WC bits n bobs to arrive.

Haven't experimented much with memory clocks either tbh.

 

[Jono8]

I wouldn't want to run them at 100 degrees on a regular basis.

 

[Paradiso]

I run my 7990's at 1.050V 1000/1500 for mining at the temp barely getting into the 80's on vrms' and core on stock cooler. I did stick 2 in together at one point but that's when scary temps started to occur haha. So I had to water cool them both, now the computer is barely audible whilst core is 50-60C and vrms' are maybe 60-70C.

In short no I would not run them over 90C despite if its safe or not, wouldn't want to chance it, especially 2 in the same machine. Maybe try undervolting again and buy a cheap mobo/cpu/4gb ram/psu to run another card?