RTX3090 Benchmarks and Temps

Updated: Feb 11

So last week was very exciting with the launch of the 3080 cards and besides the stock availability it looks like a great buy. Well, today we will check out the Titan of the 3000 series.

Here we have the RTX 3090 Gaming Pro OC from Palit, it is the top of the line card for the very few that can afford it. In this article we will check out where this card is doing well and where it is doing exceptionally well. We have already published an unboxing video and post, if you’re interested, go check them out.

Let’s start with the specs. This card has 24GB of the new GDDR6X VRAM and 10,496 Cuda cores. It is featuring the new 3rd gen Tensor cores and 2nd gen RT cores. This Palit card uses NVIDIA Reference PCB which is surprisingly narrow while still being 2.7 slots high. On the side we find two 8 pin power connections which are upside down. That’s something new but it actually makes the cable management in the case look neater. Palit has also included an anti-sag bracket in the box, it is just a clear plexiglass sheet, however together with the rest of the lights it looks really nice.

Talking about lights – for all the LED lovers, there is ARGB lighting which can be controlled using ThunderMaster utility together with fan speeds and overclocking. I personally like the armored look with the lights, it’s a shame that you need a vertical bracket to properly enjoy it though.

Right, before covering the results let’s talk about our test bench – we are using our 10700k bench. This is because at higher clock speeds we are getting less CPU bottlenecks as compared to AMD. By the way a full review of Intel vs AMD on NVIDIA 3000 series is coming soon, don’t miss it.

GP 1 – Palit RTX 3090 GamingPro OC GPU 2 – ASUS RTX 3080 TUF Gaming OC GPU 3 – ASUS ROG-STRIX-RTX2080Super GAMING Intel i7-10700k ASUS Z490 Maximus XII Extreme motherboard KLEVV CRAS X RGB (8GB x2 KLEVV CRAS C700 RGB NVME SSD Fractal SFX 650w power supply Fractal Design Celsius+ S24 Dynamic Cooler Streacom BC1 Open Benchtable

With that out of the way – we can now check out the benchmarks. We will be looking at the performance difference between RTX 3090 and 3080 specifically at 1440p and 4K resolutions as 1080p is very much bottlenecked by the CPU and provides us a lot of false results.

Starting with the good old Shadow of the Tomb Raider and we see about 9% improvement on the average scores on both 1440p and 4k. Furthermore if we look at 1 percentiles we see 1440p is 5% lower and 4k is 6% higher. This is where that aforementioned CPU bottleneck is showing.

Moving on to Total War Three Kingdoms which has always been a very GPU taxing game – here we see 12% improvement on 1440p at average scores and 10% improvement on the 1 percentiles. When it comes to 4k, the average score is 11% higher but one percentiles are up almost 40%. When investigating this further we found the CPU utilisation on RTX3080 was only 18% while on RTX3090 it averaged over 25%, indicating that the GPU was able to push further.

Next up we have an even tougher game – Horizon Zero Dawn where RTX3080 was only able to get into 1440p gaming. Here we see 9% improvement on the 1440p averages and 10% on the 1 percentiles. When we look at the 4K the difference is a whopping 33% on average scores and incredible 65% on the 1 percentiles. This result was very surprising so we re-ran the tests to double-check and we have confirmed this as correct. When investigating this further we found the VRAM usage on the RTX3080 was at 9.7GB and on RTX3090 it was 11.6GB. This indicates that the game was probably not just bottlenecked by the computational power but also lacking VRAM. Either way with RTX3090 it is possible to play this game at 4k resolution with 70+ FPS which is pretty awesome.

Now let’s jump into Metro Exodus and here we yet again see similar improvements. At 1440p we get a 12% bump on the average FPS and 7% on 1 percentiles. When looking at 4k it is 12% on the average FPS and 10% on the 1 percentiles.

When enabling ray tracing, we see 13 to 14% improvement on the averages and 14 to 15% improvement on the one percentiles, but unfortunately these are still unplayable frame rates.

Turning on DLSS provides a 13% improvement on 1440p average FPS and 6% improvement on 1 percentiles. While in 4k we see 13% improvement on average FPS and 12% on 1 percentiles. On this game we get a more solid 1440p experience with ray tracing while DLSS is enabled and if we had another 15% performance improvement we would be in the 4k territory.

To quickly sum up gaming – we have about 9 to 14% improvement in most scenarios and then a few very large improvements in selected titles which is pretty good and in most cases this card solidifies 4K gameplay. We did stumble onto some problems with Ghost Recon Breakpoint where this card was performing worse than RTX 3080, in fact at 4k it was almost performing at 2080 Super level. This is most likely due to driver issues so results have been classed as invalid, we will do a follow up on that in the future – just thought it’s best to share it with you as there may be other titles that have similar issues upon launch while NVIDIA optimises things with the developers.

When having a closer look at temperatures, we found this card is a little hotter than 3080 but it is rather expected as it has more horsepower. Here are the averages from our benchmarking: In Shadow of the Tomb Raider it was 67C In Total War Three Kingdoms - 68C In Horizon Zero Dawn - 64C In Metro Exodus - 65C

I don’t think the fans ever went up past 50% anyway.

Talking about fans – we have carried out tests to see what the relative loudness is in comparison to the other two cards. This test is done in an open bench so depending on your case the final results will be different for everybody, but this should provide you with a good oversight on the amount of noise generated. Just a note – our room noise floor was 36.3 decibels.

In testing while at 60% fan speed RTX3090 hits 49 decibels, which is actually the loudest of the three here. When we pushed it to 80% it hit 55.6 decibels and at full blast it was 63.6. So as you can see the card is not exactly quiet when you ramp it up, however during the whole time of reviewing and benching it, there were only a handful of times the card really had to ramp up to an audible level so I wouldn’t be too worried about it.

Now heading over to productivity benchmarks and here we start with Blender BMW Render where we see 18 percent improvement over RTX 3080.

In the longer classroom test the difference in render time is 12 percent from RTX3080 which is still a nice little bump.

Next we ran Davinci Resolve benchmark by Puget Systems to see how additional computational power and VRAM helps out, and it is about 11% improvement on the overall score with the highest improvements being the temporal noise reduction at 18%. The thing to consider though – VRAM capacity opens up a whole different workflow and many creative programs will take as much as you can give it to speed things up, so 24GB is going to work like magic here.

With all the benchmarks out of the way, we were planning to make a price-to-performance section but as we have seen with the 3080 launch, there may be availability issues and prices will shift back and forth for the next few months, plus this card is not exactly aimed for the budget buyers but rather for the few who can afford it. We would simply recommend checking out the performance numbers we have covered today and also compare those against other reviewers before you make the final decision.

To conclude – what do we really think about this card? Well, it is a titan all right! It does after all have 20% more Cuda, Tensor and RT cores over RTX3080 and in some cases it even has the performance to match. While it is not linear scaling, it is still about 10% faster than RTX3080 in most applications and the added 24GB of VRAM is amazing. Now we need game developers to catch up and really optimise for it so the VRAM is not just sitting idle. If you are the kind of person who can afford it, I have a feeling you are already buying it, maybe even two. To everyone else I would say – consider if you actually need this performance or if RTX 3080 is good enough.

Affiliate disclosure: as an Amazon Associate, we may earn commissions from qualifying purchases from Amazon.

#palit3090 #rtx3090benchmarks #palitgeforcertx3090gamingprooc #rtx3090 #rtx3090vs3080 #nvidiartx3090