How to reduce VRM temps on 290X with HG10 & H90?

[4K_shmoorK]

Got a question for you guys.

Recently picked up a reference 290X with a Corsair HG10 and H90 AIO cooler.

I'm able to maintain 1150/1375 @ 62C load on the core with an additional 38 mV. However before I overclock any further, I'd like to get the VRM 1 temps down. Right now with the fan on the card @ 35% (about as loud as I can tolerate), they sit at 93C full load. That seems a bit hot to me.

I searched around and found that I should be shooting for the lows 80s on VRM 1.

I was thinking about picking up an EK backplate to maybe passively bring VRM temps down. However, I haven't been able to find (from google) whether or not I would be able to mount a backplate without the EK waterblock.

So does anyone have any advice on how to get these VRM temps down a bit?

Similar setup:

 

[Face2Face]

Most Mosfets are good for 120c+
As long as you're not running into instability problems, I wound't worry to much.

VRM #1 is harder to cool. You could probably grab an Antec spot fan or zip tie a fan to the bracket some how? Not sure what else you can do?

 

[4K_shmoorK]

Most Mosfets are good for 120c+
As long as you're not running into instability problems, I wound't worry to much.

VRM #1 is harder to cool. You could probably grab an Antec spot fan or zip tie a fan to the bracket some how? Not sure what else you can do?

Yeah good till 120C, just felt like they were a good 10-20C hotter than what I've seen in reviews and forums. Maybe I'll just leave it alone then.

Thanks for the reply. I don't know how anyone could deal with the noise the reference fan puts out. Anything over 40% speed is mind-blowingly loud.

 

[thesmokingman]

Yeah good till 120C, just felt like they were a good 10-20C hotter than what I've seen in reviews and forums. Maybe I'll just leave it alone then.

Thanks for the reply. I don't know how anyone could deal with the noise the reference fan puts out. Anything over 40% speed is mind-blowingly loud.

That doesn't mean you should let it run to 120c. You'll find longevity of any card being run near max for a long period of time to not last nearly as long as a card kept cooler.

To get to 80c on vrm full load, you need more heatsink mass to draw the heat out. Toss out the stock fan too, it's stupid to use a squirrel fan that is blowing at nothing which is what that Corsair dumb design is. The Corsair plate is good, but its vrm design is a lame duck. After you remove the fan, thermal glue some heatsinks to the top over the vrm 1 strip and juryrig a proper fan on top.

http://hexus.net/tech/reviews/cooling/77237-corsair-hydro-series-hg10/?page=4


Better vrm cooling:

 

[4K_shmoorK]

That doesn't mean you should let it run to 120c. You'll find longevity of any card being run near max for a long period of time to not last nearly as long as a card kept cooler.

To get to 80c on vrm full load, you need more heatsink mass to draw the heat out. Toss out the stock fan too, it's stupid to use a squirrel fan that is blowing at nothing which is what that Corsair dumb design is. The Corsair plate is good, but its vrm design is a lame duck. After you remove the fan, thermal glue some heatsinks to the top over the vrm 1 strip and juryrig a proper fan on top.

http://hexus.net/tech/reviews/cooling/77237-corsair-hydro-series-hg10/?page=4


Better vrm cooling:

Thanks for the reply, I'll check out adding some additional heatsinks.

If you noticed in my post, I said they max at 93C not 120C. Raising the fan speed on the card will reduce VRM 1 temps to ~84C @ load, but is too loud. Doesn't exactly do 'nothing'. I think I would have to throw out the entire Corsair setup to apply the stick on heatsinks though.

The review you linked even explains this:

The reason why VRM #1 gets so hot is because the reused stock fan receives its RPM signals from the GPU core temperature. That GPU core temperature is kept very low from the watercooling meaning the stock fan maintains a low RPM even when the VRM is rising in temperature. The result is that the VRM gets more use because thermal throttling is eliminated resulting in more power draw, but the same VRM simultaneously receives less cooling because of the RPM signals being sent to the stock fan.

I got the card and cooler setup in a partial trade for my 980 Ti. I'd like to avoid spending much more to get the VRM temps down a bit.

 

[Face2Face]

Yeah good till 120C, just felt like they were a good 10-20C hotter than what I've seen in reviews and forums. Maybe I'll just leave it alone then.

Thanks for the reply. I don't know how anyone could deal with the noise the reference fan puts out. Anything over 40% speed is mind-blowingly loud.

I haven't owned a reference 290/X, but right now I have a reference 7970 GE, and man, it's really loud when cranked to 60% or higher. My reference cooler on my old GTX 780 was much quieter in comparison.

Back to your card. I would stick with your OC for the time being or bump it down A little if you want lower VRM temps. The difference between 100MHz core clock isn't large enough to justify the extra cost.

 

[4K_shmoorK]

I haven't owned a reference 290/X, but right now I have a reference 7970 GE, and man, it's really loud when cranked to 60% or higher. My reference cooler on my old GTX 780 was much quieter in comparison.

Back to your card. I would stick with your OC for the time being or bump it down A little if you want lower VRM temps. The difference between 100MHz core clock isn't large enough to justify the extra cost.

Right, just wanted to see if there was anything I could do that I wasn't thinking of. Trying to find the right balance between performance, noise, and heat.

I wanted to go up to 1175 or 1200 on the core because I think the card has it, but definitely don't want to go sustained 100C on VRM 1.

Thanks for your input.

 

[thesmokingman]

Thanks for the reply, I'll check out adding some additional heatsinks.

If you noticed in my post, I said they max at 93C not 120C. Raising the fan speed on the card will reduce VRM 1 temps to ~84C @ load, but is too loud. Doesn't exactly do 'nothing'. I think I would have to throw out the entire Corsair setup to apply the stick on heatsinks though.

The review you linked even explains this:


I got the card and cooler setup in a partial trade for my 980 Ti. I'd like to avoid spending much more to get the VRM temps down a bit.

If you're going to run the fan off the pwm lead of the gpu, then you'd best configure the duty cycles with that in mind. That would seem obvious wouldn't it? The problem though with the built in pwm controller of a gpu is that you basically have very little control. With a cpu pwm controller liek your mb's you can choose the curve and DC points to hit. That way you can tailer the cpu's curve to account for the vrm or whatever other part.

That said, I don't get what your point is. You couldn't even run the fan higher if you wanted unless as you said you wanted to make your ears bleed so obviously that is not working. The reference cooler actually does very well on vrm temp but it's so loud that again your ears bleed. Regardless of their disclaimer, neither works.

Running the stock fan at levels that don't bother your ears equates to too high of a temp. I explained how to cool it a lot better without assaulting your ears... the rest is upto you.

 

[Deders]

I managed to knock 20c off my VRM temps just by adding a side fan. Not sure if it will be as effective if the path between the sidefan and the VRM's are blocked like in the pic you posted, but maybe even having cool air for the graphics fan to work with.

I also found that removing the unused drive bays so that the front fans had nothing blocking them greatly improved all my temps.

 

[Silverforce11]

VRMs in the 90C is very common. You don't need to worry at all.

 

[thesmokingman]

VRMs in the 90C is very common. You don't need to worry at all.

It's common however lower temps will slow degradation.

 

[Despoiler]

The VRMs that AMD uses are good to 150C. You are chasing a rainbow on this one.

 

[MrTeal]

Looking at that setup, it seems VRM cooling is pretty suspect. Depending on your clearances, one extremely simple thing you could do is mount some press-on heatsinks on the backside of the PCB underneath the inductors. If you have reasonable case airflow that will probably drop VRM temps several degrees, without having to pull out the hacksaw.

 

[MrTeal]

The VRMs that AMD uses are good to 150C. You are chasing a rainbow on this one.

The mosfets are good to 150C, but many of the other components are likely not. For example, the polymer can caps at the output of the VRM are rated for 2000 hours at 105C. Generally, you can expect a lifetime with a polymer cap 10x longer if you stay 20C under the max rating. The inductors also lose a lot of performance at elevated temperatures, causing more ripple with further heats them and the output capacitors. How hot other components on the board will get can depend on a lot of things, but it's not uncommon to see board temps close to the VRM temps if cooling or airflow isn't good.

90C likely isn't going to be an issue, though it will definitely impact efficiency. Going too much above that is a bad idea long term, especially if you use it a lot (like for 24/7 mining)

 

[4K_shmoorK]

If you're going to run the fan off the pwm lead of the gpu, then you'd best configure the duty cycles with that in mind. That would seem obvious wouldn't it? The problem though with the built in pwm controller of a gpu is that you basically have very little control. With a cpu pwm controller liek your mb's you can choose the curve and DC points to hit. That way you can tailer the cpu's curve to account for the vrm or whatever other part.

That said, I don't get what your point is. You couldn't even run the fan higher if you wanted unless as you said you wanted to make your ears bleed so obviously that is not working. The reference cooler actually does very well on vrm temp but it's so loud that again your ears bleed. Regardless of their disclaimer, neither works.

Running the stock fan at levels that don't bother your ears equates to too high of a temp. I explained how to cool it a lot better without assaulting your ears... the rest is upto you.

Just manually controlling fan speeds right now. I know how fan profiles work...

I'm saying the fan setup DOES cool off the VRMs, albeit, not effectively/efficiently. Again, I acknowledged the fact that I'm unwilling to sustain these fan speeds because of the noise.

I would go with a setup you suggested but as I mentioned earlier, a stick on VRM heatsink would not fit with the HG10 shroud equipped. I also would not like to spend more than ~$30 on additional cooling for this card. Thanks for your input, though.:thumbsup:

Thanks for the additional replies. I most likely won't be keeping the card for more than a year, as long as the card doesn't conk out on me before then... I don't use the card for mining, in fact, I made the trade because of the looming generational graphics price drops and AMDs strong showing in DX12 optimization. Probably play between 15-25 hours a week on PC, if that.

Sounds like I should be okay as long as I can keep temps around or below 90C.

I managed to knock 20c off my VRM temps just by adding a side fan. Not sure if it will be as effective if the path between the sidefan and the VRM's are blocked like in the pic you posted, but maybe even having cool air for the graphics fan to work with.

I also found that removing the unused drive bays so that the front fans had nothing blocking them greatly improved all my temps.

May add an additional ventilation fan the bottom of my case for additional airflow (750D).

Thanks guys.

 

[KaRLiToS]

Custom watercool

/thread

 

[thesmokingman]

Custom watercool

/thread

Aye. 4 and 5 year old 7970s still overclocking to 1350mhz. Have that one that still does 1400mhz on water. The one that I put the stock cooler back on and gave to my nephew to use for two years aircooled? It degraded. Won't clock past 1275mhz anymore even under water. I emded up taking it back, felt sorry for it and swapped a cheaper 7950 gigafail in.

 

[KaRLiToS]

Aye. 4 and 5 year old 7970s still overclocking to 1350mhz. Have that one that still does 1400mhz on water. The one that I put the stock cooler back on and gave to my nephew to use for two years aircooled? It degraded. Won't clock past 1275mhz anymore even under water. I emded up taking it back, felt sorry for it and swapped a cheaper 7950 gigafail in.

I remember those Quad HD 7970s score on 3dmark among the numerous Quad Titans. You had golden card for sure.

But you are right. My Quad R9 290x could all do 1300mhz too, even 2.5 years later even after pumping 1.56v in them with pt1 bios. Those cards never ran on air until I sold them. Never saw temp over 55'C on the highest temperature I could record. Watercool for life.

 

[guskline]

KaRLiToS, did you use EK waterblocks?

 

[KaRLiToS]

KaRLiToS, did you use EK waterblocks?

Yes

 

[thesmokingman]

Yes

Btw, grats on your sale too.