3090FE, 17xxMHz average in Valley? (Monitoring screenshot included)

[RipGroove]

Does this look right?

No overclock
Valley Preset = Ultra
Res = 3440x1440

GPU was 74-77°C after about 20 mins because it's hot today (about 26°C outside) and I have a custom curve to keep it quiet. With the fan speed set to 100% the temp went down to 65°C but the core clock didn't increase so I assume at 74-77°C it wasn't thermal throttling.

FPS 160-260
GPU Clock 1575-1890MHz (averaged about 1700MHz)
CPU Clock - Constant 4805MHz

Giant image so you can see all the numbers, click to enlarge.

I was going to ditch my fan curve and try a little OC just for fun but Valley seems to make my core clock just fluctuate wildly so I can't really see what max MHz it's actually running at. Is there a better stress test to run that will just pin the core clock to it's max constantly?

Seen a few YT videos of people running Valley with OC's and their GPU's sit at a constant clock of 2xxx, am I doing something wrong? I tried upping the power and temp sliders as usual and increasing the core clock but it had no effect on the reported core clock (yes I did press apply) as it just fluctuates a lot.

I would post my 3dMark (demo) score but their site doesn't seem to be working correctly right now.

 

[Maldoror]

In that screenshot you have Temp limit flagged, meaning it's pulling back your clocks whether you are on the 'limit' itself or not.

Could you take a screenshot of it running with the 'sensor' tab open on GPU-Z?

 

[RipGroove]

In that screenshot you have Temp limit flagged, meaning it's pulling back your clocks whether you are on the 'limit' itself or not.

Could you take a screenshot of it running with the 'sensor' tab open on GPU-Z?

Ooh, well spotted, that's obviously a mistake on my part then. I'll redo the same test with the temp limit off and the sensor tab open on GPU-Z, I may not have time to do that until tomorrow though so stick around

 

[Maldoror]

Ooh, well spotted, that's obviously a mistake on my part then. I'll redo the same test with the temp limit off and the sensor tab open on GPU-Z, I may not have time to do that until tomorrow though so stick around

The core there is running with a voltage of 0.850v, which is very low, i.e. pulled back due to the throttling. When you see pictures online, most people will be sharing situations where they have worked on conditions such that the core is being fed 1.093v or around that. 0.850v is deep into throttle territory. I note that your memory temperature there is very high (108c) which can also contribute to throttling.

Valley is not a particularly power hungry benchmark although I do notice that your Power flag has been triggered at some point (Max value:1). Still, in the immediate sense in the moment of the screenshot, I would say that you are thermal throttling.

Let's have a look at GPU-Z next time and we can check the temps carefully and the throttling flag reason.

I'm guessing you might need a little more airflow around the back of the card (rear memory modules) to get the performance up. It is a hot day though.

 

[RipGroove]

The core there is running with a voltage of 0.850v, which is very low, i.e. pulled back due to the throttling. When you see pictures online, most people will be sharing situations where they have worked on conditions such that the core is being fed 1.093v or around that. 0.850v is deep into throttle territory. I note that your memory temperature there is very high (108c) which can also contribute to throttling.

Valley is not a particularly power hungry benchmark although I do notice that your Power flag has been triggered at some point (Max value:1). Still, in the immediate sense in the moment of the screenshot, I would say that you are thermal throttling.

Let's have a look at GPU-Z next time and we can check the temps carefully and the throttling flag reason.

I'm guessing you might need a little more airflow around the back of the card (rear memory modules) to get the performance up. It is a hot day though.

Awesome thanks, I'll get the required info as soon as I get a min.

 

[RipGroove]

I'll try and do it early morning when it's cooler.

 

[Grim5]

Clock are ok if that's a entry level "cheap" 3090

id be more concerned with your vram sitting at 112c and throttling performance

 

[RipGroove]

Clock are ok if that's a entry level "cheap" 3090

id be more concerned with your vram sitting at 112c and throttling performance

It's an FE card, yes I did see those temps, it was very hot in the room tbf, I checked those temps out when I first got the card and IIRC it was 90 something degrees on a normal temperature type day.

 

[Grim5]

It's an FE card, yes I did see those temps, it was very hot in the room tbf, I checked those temps out when I first got the card and IIRC it was 90 something degrees on a normal temperature type day.

Component temperature can only move with ambient temperature assuming everything else is saturated. Meaning even if your card was 92c before - your room would have to be 20c hotter today to get 112c. It's unlikely that your room had such a large temperature swing though so probably your 90 odd reading was when the gpu temperature was not fully saturated so it could have still gone higher with enough time like a long gaming session - as long as it's under 105c it's ok otherwise the card will throttle and you will lose performance.

What case are you using and how many fans?

 

[RipGroove]

Component temperature can only move with ambient temperature assuming everything else is saturated. Meaning even if your card was 92c before - your room would have to be 20c hotter today to get 112c. It's unlikely that your room had such a large temperature swing though so probably your 90 odd reading was when the gpu temperature was not fully saturated so it could have still gone higher with enough time like a long gaming session - as long as it's under 105c it's ok otherwise the card will throttle and you will lose performance.

What case are you using and how many fans?

It was high 90's last time and it definitely was a lot cooler then so I can definitely see a 10-15°C swing in the last week regarding ambient temps, today's test was done later afternoon when it was super hot and IIRC the previous test was done early morning.

So it was definitely throttling earlier today then. I've had the same case for over 4 years, Phanteks Enthoo Pro, with 2x 140mm exhaust fans in the top and a 140mm exhaust fan at the rear. Now for the scary bit, I have a 240mm AIO rad up front for my CPU acting as the intake. I know this isn't ideal but when the system was originally designed and built I had an AIO cooler on the GPU as well so all was well, happy cool low noise gaming for 4 years up until recently when I upgraded the CPU and GPU and of course with things the way they are I needed up with an air cooled GPU so it's something I know I need to sort out, cool air coming in front and maybe even fans at the bottom blowing cool air right up into the rad, like a Lian Li 011 air case or whatever it's called.

I don't believe I can fit my 240mm AIO rad up top without modding the case so I either have to do that or get a new case. I've had the 3090 for a few weeks now and it's actually all been fine up until today where it's been unusually hot so it's exacerbated the issue. Ideally I want a case where I can fit 3x 140mm fans up front (I much prefer the low noise that 140mm's make over 120mm's) so I've been doing a lot of research to see which ones I like the look of and all that so it will be happening soon.

 

[RipGroove]

This is how it was before the recent CPU/GPU upgrade but the case and fans are still the same now, except there is a 140mm exhaust fan where the GPU rad is in that photo.

 

[Maldoror]

There are a lot of ways to throw money at the problem of cooling the back of the card. But not everything need be expensive. Simple options would be:

1) re-pad the back to something like Gelid Extreme. No need to go crazy with Fujipoly and the associated expense. The Gelid will still knock a few degrees off.
2) Alternatively (or in addition), there are a lot of ghetto solutions to cooling the backplate. Most of these take the form of putting an old-school CPU cooler on the back of it (with some thermal paste between). A fan pointing at the back and blowing air across the back of the card can also make a huge difference.

The FE is a great card. It uses the same MP2888B controller as the Strix and the Kingpin, has a premium PCB and there seem to be a disproportionately high number of unicorn FE cores around. If you can get the back of the card cooled a bit I've no doubt you'll get some nice clocks out of it (and plenty of people on here will help you do that).

 

[RipGroove]

There are a lot of ways to throw money at the problem of cooling the back of the card. But not everything need be expensive. Simple options would be:

1) re-pad the back to something like Gelid Extreme. No need to go crazy with Fujipoly and the associated expense. The Gelid will still knock a few degrees off.
2) Alternatively (or in addition), there are a lot of ghetto solutions to cooling the backplate. Most of these take the form of putting an old-school CPU cooler on the back of it (with some thermal paste between). A fan pointing at the back and blowing air across the back of the card can also make a huge difference.

The FE is a great card. It uses the same MP2888B controller as the Strix and the Kingpin, has a premium PCB and there seem to be a disproportionately high number of unicorn FE cores around. If you can get the back of the card cooled a bit I've no doubt you'll get some nice clocks out of it (and plenty of people on here will help you do that).

Yes definitely one way or the other I'll get it cooled down a bit, I just like a silent system so it'll take a bit of doing primarily with air cooling. I'll do another Valley run in the morning when it's cooler with GPU-Z sensors displayed and also uncheck that temp button in Afterburner just so we know where we're at.

I know heat likes to rise but maybe I could reverse my fans temporarily so the top is the intake and the front is the exhaust, might help for now as a temporary measure.

 

[Maldoror]

Yes definitely one way or the other I'll get it cooled down a bit, I just like a silent system so it'll take a bit of doing primarily with air cooling. I'll do another Valley run in the morning when it's cooler with GPU-Z sensors displayed and also uncheck that temp button in Afterburner just so we know where we're at.

I know heat likes to rise but maybe I could reverse my fans temporarily so the top is the intake and the front is the exhaust, might help for now as a temporary measure.

Reversing the airflow is definitely worth a go. I wouldn't put too much emphasis on the hot air rising dynamic when it comes to active airflow like this - do what makes sense for your case. And it's easy enough to switch back if the temps don't improve.

You also don't necessarily need a very high rpm fan pointed at the back of the card - just something to keep the air moving a bit. In the long run that may even improve acoustics by having less of the 'revving' sound as the 3090 cooler attempts to ramp up repeatedly to cool the ram via the minimal airflow the rear fan provides for the surface.

 

[RipGroove]

Yes I'll give it a go. Just looking at your comment about the temp limit being flagged, I thought the way I ahd it set was for prioritzing power as I can see the lightening bolt logo, temp priority would be like this below surely:

 

[RipGroove]

FYI I'm just doing that Valley run again but with the temp logo displayed in Afterburner as per the photo above.

 

[RipGroove]

I'll run it for another 10mins and stick the screenshot up but so far the memory is about 100-103°C, the core is about 65°C and I'm still only seeing 17xxMHz average core clock and that's with the stock GPU fan curve so it's pretty loud as the GPU fans have settled at about 77% speed. Valley is reporting my GPU clock as being pinned at 2100MHz but I assume that for some reason it is incorrect and that Afterburner is the more accurate reading?

 

[RipGroove]

Here we go, set to prioritize temp I beleive (as the temp logo is shown), notice how the core clock pretty much stayed at 17xxMHz the whole time:

 

[Sarge2255]

Clock are ok if that's a entry level "cheap" 3090

id be more concerned with your vram sitting at 112c and throttling performance

I wouldn't call that card cheap lol

 

[RipGroove]

Just doing some gaming and the core clock is spending a lot of time at 1500MHz at 69°C ish average core and 100°C ish average memory, soz forgot to open GPU-Z for this one but you can see GPU-Z for Valley above: