RTX 3090 Hash Rates & Settings

[booze]

I just picked up a RTX 3090 and interested in other people's hash rates they can achieve with a 3090.

I've read people achieving over 120mhs but I've had no such luck!

100% power
Core Clock -100
Mem Clock +1500
Fan 70%
Temp 70 cels
Power 340W

Hash rate fluctuates between 105 and 108mhs which is quite underwhelming TBH.

Is this a silicone lottery kind of thing or am I doing something wrong?

Cheers

[1713929600]

1713929600

[1713929600]

1713929600

[mak013]

First of all - try to touch backplate. 3090 may throttle due to high memory temperature. If its ok - try to add core clock several times per 100MH one time.

[th00ber]

Add FAN to the backplate. It's mandatory because 3090 are really hot in the back while mining and will thermal throttle even if the GPU is @60°c and FAN @100%

My personnal settings are
CORE 1050 MHz@0.725V
MEM 10000 MHz
112 MH/s @ 290W

[booze]

First of all - try to touch backplate. 3090 may throttle due to high memory temperature. If its ok - try to add core clock several times per 100MH one time.

I have checked HWInfo and no thermal throttling, only power throttle!

In MSI Afterburner its not even displaying the mV, why would that be do you think?

Also the fan for the backplate that comes with the unit doesnt fit into my case because of another cable

Thanks for replies

[Wotan Wipeout]

Memclock + 1500 is much too high, try +1000 or 1100.
The RAM Temp is not shown and this is the problem.

[STSMiner]

Here's some info for you guys that have that shiny new RTX 30x0 card .... this is mostly for the RTX 3080 / 90 cards.

The test below was not carried out by me, but I have managed to confirm the findings with my Zotac RTX 3090 Trinity I have.

The card used in this test -- MSI RTX 3090 Gaming (you can find the video source of this test at the end of this post).

Picture 1  --   Two heat sensors are placed on the memory chips that are on the back of the card.

Picture 2  --   This shows the card running in idle ..... nothing running, notice the temperature   33C

Picture 3  --   This shows the card mining Dagger / Ethash, the monitor on the left shows the card mining, the temperature on the memory chips climb to over 76C, this is what causes the cards to throttle down, the 3080 cards also suffer with this issue with the memory chips on the other side of the card (heat sink / fan side).
In order to get the most out of these cards, the memory chips and the VRM's need to be cooled down as much as possible.
Make sure you have plenty of air flow around the card(s) and give them space to breath.




This is my RTX3090 ..... with extra cooling (fan blowing cold air on the back plate).


Below is a link to an RTX 3080 video showing where the card gets hot, same applies for the RTX 3090.
https://www.youtube.com/watch?v=dAsQ0o6BsEk

Source for the pictures used above from the video below.
https://www.youtube.com/watch?v=82sf5Lk2lqA

[mak013]

First of all - try to touch backplate. 3090 may throttle due to high memory temperature. If its ok - try to add core clock several times per 100MH one time.

I have checked HWInfo and no thermal throttling, only power throttle!

In MSI Afterburner its not even displaying the mV, why would that be do you think?

Also the fan for the backplate that comes with the unit doesnt fit into my case because of another cable

Thanks for replies

You will not see memory throttle, that`s why try to touch. I can`t say anything about MSI Afterburner - i just don`t remember it. Try HiveOS or RaveOS. It`s simple and helps to configure your GPUs easily.

[falsealarm_bf1]

I am going crazy with my two 3090s. Both run at about 105Mh/s. The MSI Triple is the primary and runs without issues, but the Gigabyte secondary either wants to spin the fans up to 100%, or throttle GPU down to 800Mhz, and crank out only 80MH/s. I have tried power limiting, GPU downclocking, and memory overclocking, however could not resolve the issue. I have a healthy Corsair 1200W power supply which previously ran 4x1070s and a single 1080 with no issues. Power draw shows 350W on the MSI, and 250W on the Gigabyte (on a riser). Both have their PCIE power inputs fully populated with clean power. Any ideas?

[Hackintoshihope]

I am going crazy with my two 3090s. Both run at about 105Mh/s. The MSI Triple is the primary and runs without issues, but the Gigabyte secondary either wants to spin the fans up to 100%, or throttle GPU down to 800Mhz, and crank out only 80MH/s. I have tried power limiting, GPU downclocking, and memory overclocking, however could not resolve the issue. I have a healthy Corsair 1200W power supply which previously ran 4x1070s and a single 1080 with no issues. Power draw shows 350W on the MSI, and 250W on the Gigabyte (on a riser). Both have their PCIE power inputs fully populated with clean power. Any ideas?

You have to have a huge memory downclock with a fan blasting air on the backplate... Also you must make sure that whatever 3090 you are using has memory cooling pads that touch the backplate. The issue is the GDDR6X gets extremely hot and overwhelmed with an overclock passed 500MHz and then the core is downclocked to try and cool it.

[th00ber]

My 3090 settings :
Core 1050 Mhz@0.725v
Mem 10100 Mhz
Fan 80%

I'm getting ~113-115MH/s @ 295-300W

/!\ THE 3090 NEED FANS ON THE BACKPLATE /!\ because the back of the card is the weak point and even with a GPU reported @60°c the components on the back will throttle the graphic card because they are too hot.
Look --> https://imgur.com/a/2ONFgGw (That's my PC case mod )

Without fans, the card will mine for some seconds or minutes then decrease hashrate to around 90MH/s in my desktop

[falsealarm_bf1]

What did you use to undervolt? If Afterburner, can you share an image of your curve?

[Sledge0001]

I've been running the PNY XLR8 3090's stable at 120Mh/s using Phoenix Miner 5.5c

Power Limit 88 to 91 which ends up pulling 308W to 318W It really depends on each card.
Memory Clock is +1020  to +1022
Custom fan curve.

They have supplemental cooling from front to back but they do run hot regardless.

[FloppyPurpleGherkin]

Yeah these 3090 with their 24gb (lol) vram are a complete nightmare to cool, the 3080's as well to some extent but not as bad as the 3090.

My way of thinking for my 4x 3090 is fully water-cooled or don't bother.. 3080's are a bit better on air but not massively.

[Hackintoshihope]

Yeah these 3090 with their 24gb (lol) vram are a complete nightmare to cool, the 3080's as well to some extent but not as bad as the 3090.

My way of thinking for my 4x 3090 is fully water-cooled or don't bother.. 3080's are a bit better on air but not massively.

All you have to do is replace thermal pads with gelid extreme... fixed my memory heat problems on 3090.

[archerdotpm]

check the latest HWInfo application, it shows mem temperatures now for the 3090s..

[falsealarm_bf1]

I see GPU memory junction temperatures of 102C on one of the cards, and 108 on the other. Are these OK for continuous operating temperatures? Oddly, I have triple fan cooler less than an inch away from the back of the one card that shows 108 for memory junction temperatures.

[reb0rn21]

its not ok try to have it at 90C or so, I think maybe best thermal pad you can find might help.....
but let see some ppl test them! as you need to remove the cooler to change them

[booze]

Any links to a tear down on how to replace/add thermal pads? I have never gone further than sticking a GPU in the PCIE slot haha

[falsealarm_bf1]

Yeah these 3090 with their 24gb (lol) vram are a complete nightmare to cool, the 3080's as well to some extent but not as bad as the 3090.

My way of thinking for my 4x 3090 is fully water-cooled or don't bother.. 3080's are a bit better on air but not massively.

All you have to do is replace thermal pads with gelid extreme... fixed my memory heat problems on 3090.

How do you replace the thermal pads with Gelid? There is quite the gap back there to fill with paste. I think it's 1mm, or 2mm depending on where the pad is. Also, I am not sure if the problem VRAM are the ones on the back under the plate.

[falsealarm_bf1]

Yeah these 3090 with their 24gb (lol) vram are a complete nightmare to cool, the 3080's as well to some extent but not as bad as the 3090.

My way of thinking for my 4x 3090 is fully water-cooled or don't bother.. 3080's are a bit better on air but not massively.

All you have to do is replace thermal pads with gelid extreme... fixed my memory heat problems on 3090.

How do you replace the thermal pads with Gelid? There is quite the gap back there to fill with paste. I think it's 1mm, or 2mm depending on where the pad is. Also, I am not sure if the problem VRAM are the ones on the back under the plate.

I just realized you were referring to the Gelid GC pads, and not the GP paste. How many did you need in total, and what thickness? I am sure it won't be the same as my Gigabyte Gaming OC 3090 but it would at least give me an idea.