We are taking a look at not one but two GeForce rtx 3090 graphics cards. Because why not? I guess I'm just ripping the Band-Aid off with this one and condensing old up. What's the point of this review? We cannot buy them anyway comments into a single article. So on hand for testing, we have the absurdly massive gigabytes or rtx 3090 extreme. This is basically the same design as the 3080 model well looked at recently, then put that down. We have the gigabytes rtx 3090 gaming OC it again. This is basically the same design as the 3080 gaming OC that I looked at not that long ago.
Let's quickly take a look at each model before jumping into the results. I'd normally call the rtx 3090 card as the gaming OC a very large high-end 3090 ti card because by normal high-end graphics because standards well it very much is. However, sitting next to aorus Xtreme certainly doesn't look impressive that high-end, not very massive, especially when you compare them side-on.
A Closely Look Rtx 3090 Card
In terms of dimensions, they both measure 320 millimeters long, but where's the gaming OC stands 129 millimeters tall, the extreme is 140 millimeters tall. So that's almost 10 percent taller. The real difference, though, as I noted a moment ago, is the width. The gaming OC is a fairly typical 55-millimeter wide triple slot design. The extreme, though, is an insane four-scored card measuring 70 millimeters wide.
So almost 30 percent wider, making it the beefiest graphics card I have ever seen. All in all that means that the volume of the rtx 3090 Xtreme 38 percent larger, and therefore probably wasn't surprise you to learn that it weighs exactly 38 percent more at 1894 grams versus 1370 grams.
Now, normally I'd spend the next 15 minutes or so going over each rtx 3090 cards to design the cooling and PCB, for example, but since we've kind of done that already with both models using the rtx 3080 versions but skip that and just get on with the testing. The coolers, in particular, are exactly the same, with only minor changes being made to the PCB. For example, there are a few extra power stages to help deal with the added power draw, and of course, there are many more gddr6 memory chips to achieve that 24-gigabyte capacity.
But beyond that, if you want to learn more about the cooling and design of these cards, please refer to the rtx 3090 reviews. In terms of clock specifications, rtx 3090 lists a core clock frequency of 1755 megahertz for the gaming OC and 1860 megahertz for the or aurus nvidia 3090 extremes. Out of the box, I'm expecting the full slot monster to clock just 6 percent higher as for the gddr6 memory. Both run at the default 19.5 gigabits per second specs by nvidia 3090. This is very typical as AIB is based on an allowed to play around with the memory spec.
Gigabyte Nvidia 3090 Gaming OC
Kicking the testing off around the gaming OC in the shadow of the tomb raider for 30 minutes with the graphics card installed inside the corsair obsidian 500d, which has been fully populated with case fans this three 120 millimeter models in the front then two 140 millimeter exhaust fans at the top as well as a real 120-millimeter exhaust fan. Also, plays are all testing has been conducted with a closely monitored room temperature of 21 degrees under these conditions, the gaming OC pick just 66 degrees which are remarkable really especially given the 1800 RPM fan speed, which saw the cot operate at just 41 decibels.
The saw the nvidia 3090 cards maintain a core clock frequency of 1875 megahertz, so high than that of the large foundation model, yeah for the massive or nvidia 3090 Xtreme. This model fatherless impressive, at least in terms of operating temperature and volume. The bigger four-slot nvidia 3090 cards actually ran a 4 degrees hot up peaking at 70 degrees, and for me, this was made all the more disappointing given how much louder its walls operating 48 decibels with a fan spate of just over 2400 RPM.
The cores did manage to clock higher 1890 megahertz. However, that is less than a 1 percent increase, so I find it unlikely that many of you will find the increased fan speed and operating temperatures justifiable. So having been disappointed with the stock OC bias, I switched over to the silent bias, and he the card ran even hotter, hitting 74 degrees, so that was to be expected and is, of course, due to the low fan speed, which was dropped down to 2000 RPM.
Here the car generated 41 decibels of noise and was noticeably quiet up that said the core clock frequency was also reduced to 1755 megahertz on average, a 7 percent drop in frequency, so again quite a disappointing result of role. Now when manually overclocking, I was able to push the gaming OC cores up to 1830 megahertz, which admittedly is a very poor overclock and just a 4 percent boost over the factory oc.
In-game, this resulted in an average score clock frequency of 1965 megahertz. That's a 5 percent increase from the same previously. Certainly not an amazing nvidia 3090 overclock but also not really that far off the very best results we've seen. As you might have expected to see, the nvidia 3090 Xtreme, which uses been silicon, was capable of a higher clock heating in 1995 megahertz, which translated to an average in-game GPU boost frequency of 2055 megahertz. So a further 5 percent over what the gaming OC achieved again not amazing, but that's about all you're able to extract from these new AMP Nvidia 3090 GPU.
Let's move into the benchmark graphs; as usual, be testing with our nvidia 3090. GPU test rig with 32 gigabytes of DDR4 3200 CL 14 memory they please not for this testing there is no difference in the performance between 3950x or 5950 X. as we are entirely GPU limited.
Now, as usual with these custom AIB graphics card reviews, look at loads of gaming benchmarks in-fact of a shot of the tomb rider will do if you want loads of benchmarks and please do watch on day one review or any of the head to head bench my reviews we've done since then.
The focus here is on thermals power and overclocking starting with a 1440p data in a show of the tomb rider we say that out of the box, the or nvidia rtx 3090 extreme was just 1 percent faster than gaming OC, which is roughly in line with the observed frequency difference between these two models.
So I'd say in terms of stock performance, the nvidia rtx 3090 extreme is very disappointing given that it runs not just harder but also allowed a then the gaming OC. Now when it came time to overclock these 3090 ti cards manually, we did see the nvidia rtx 3090 clocking 5 percent higher, and surprise that resulted in a 5 percent performance advantage again not exactly a shattering difference.
The 4K data is very similar to yours. Xtreme is 2 percent faster out of the box and 5 percent faster once both models I manually overclocked to the Max, so in terms of FPS performance, both models are very similar. However, I do suspect I've received a particularly bad gaming OC sample in terms of OC headroom. So this is likely going to be a best-case scenario for the aurus nvidia rtx 3090.
So now for a quick look at power consumption, I feel this is the main issue with the aorus nvidia rtx 3090 extreme stock out of the box the aorus Xtreme seems to be using 17 percent more power when compared to the gaming OC. At the same time, that might not sound like a lot.
A 63-watt increase is substantial. I'll talk more about that towards another article with both overclocked to aorus nvidia rtx 3090 Xtreme still consume 15 percent more power. That's a big increase when you consider just how extreme the power requirements up free stock nvidia rtx 3090.
It's also well worth noting that when now dumping what is essentially 455 watts of heat into the computer case from just the 3090 ti card as we move on to look at the thermal results again, keep in mind that we are testing inside an nvidia rtx 3090 case and not an open test bench.
So I stock nvidia rtx 3090 Xtreme saw a peak GPU temperature 70 degrees which is quite good but remember the fan speed he was also very high the gaming OC on the other hand, so it's fan spending 25 percent slower and despite that the car around 4 degrees cooler peaking at just 66 degrees. The real issue here can be seen when noise normalizing these 3090 ti cards. We see no change in temperature for the gaming OC as it was already operating 41 decibels, so the slight drop in fan speed to achieve 40 decibels didn't increase the GPU temperature.
The nvidia rtx 3090 Xtreme is so a massive 13-degree increase in operating temperature. This is much worse than what we saw with the silent bias, and the reason for this being that we're not lowering the power target. The cores are still operating at 1890 megahertz. So when you consider the aorus extreme is really faster than the Asus Strix gaming MSI gaming x trio, and those models run even cool in the gaming OC, it's hard not to be disappointed with the aorus extreme.
Moving on to look at the VRM temperatures, which I measured using A series of k-type thermo couples, we can find that out of the box, the gaming OC and aorus Xtreme are very similar in fact to the gaming OC VRM, around 3 degrees hotter, but again, keep in mind that there is a big difference in terms of operating volume.
So with the 3090 ti card noise normalized, we see that the gaming OC VRM around 8 degrees cooler, which isn't a particularly great result relative to the other models, but it's certainly better than the more premium nvidia rtx 3090 aorus extreme this is perhaps the most concerning data yet they are out of the box with its higher fan speed the nvidia rtx 3090 aorus Xtreme did work just fine.
It was much louder than the other models tested to stop the memory still ran 3 degrees cooler on the gaming OC. Still, it's the noise normalized results that are most concerning he the memory peaked at 97 degrees in the nvidia rtx 3090 aorus Xtreme, about 17 degrees hotter than what we saw with nvidia's founder's edition model at 97 degrees the internal temperature of the gddr6 chips has to be near critical levels which are probably why the fans spin so fast by default.
Yeah, you might be wondering why is the memory on the aorus extreme so hot relative to the other models especially given that they all use the exact same gddr6X memory operating at the same frequency. The problem here is again the increased power usage of the GPU. Much of that heat is dumped directly into the PCB, where it's very difficult to extract the spikes of the surrounding memory chips, especially those on the rear side of the PCB where much of the heat ends up. The only real county here is to rapidly increase the fan spade pushing more and over the base of the heat sink and the PCB itself.
So, in summary, out of the box, the aorus Xtreme is just one to 2 percent faster than the gaming OC, but it also makes a lot more noise due to a 25 percent increase in fan speed to deal with the 15 to 20 percent increase in power usage. I was still compared to the more premium models from the likes of an Asus and MSI, the aorus nvidia GeForce rtx 3090 extreme is no faster, but again it also makes more noise. It consumes a good bit more power.
Unfortunately, even overclocked the nvidia GeForce rtx 3090 aorus Xtreme doesn't exactly end up in a field of its own. Here was just 5 percent faster than the gaming OC. I think it consumed much more power resulting in high operating temperatures and greater noise output.
The problem with nvidia GeForce rtx 3090 aorus Xtreme was really highlighted with the noise normalized testing, which doesn't neutralize the difference in power usage as we can't really do that without taking clock speeds. So you might be wondering why does pushing power consumption up by further 15 to 20 percent results in much high operating temperatures when noise normalized shouldn't this 38 percent bigger and heavier heatsink offset the increased power usage.
You did think so, but there is a thermal bottleneck here. The surface area of the GPU doesn't increase with the larger cooler. So we are dealing with the exact same size thermal hotspot, and there is a point where you simply cannot extract the heat fast enough, at least using conventional materials. Both models make contact with the silicon using a copper base and, of course, a thermal compound to take up any air gaps.
Thermal dead's arms, let's say the aorus extreme does include a vapour chamber which will help spread the thermal output across a large section of the base improving efficiency, but that will only get you so far. The interface material used to suck the heat away from the die remains the same. It's, of course, still copper.
So with GPU like the nvidia geforce, rtx 3090 by default already pushed well beyond her efficiency window, increasing the power usage by for 15 to 20 percent makes it extremely difficult to extract that additional heat output from the relatively small 628-millimetre square die.
Again what we're looking at he is a thermal bottleneck, and no matter how large you make the heatsink itself, the contact point with the die is going to be the primary bottleneck. The next step here be use a water block. While you are still using copper to extract the heat, it's much easier to directly cool the cooper base with water as you no longer relying on a series of the heat pipe to transfer heat away from the base and move it through a large array of fins.
In short, while the. Or so extreme heat sink is 38 percent larger it's not 38 percent better as it still has to suck heat away from the same size surface area I without significantly improving thermal extraction you know going to see significantly better thermal results this then makes the aorus nvidia geforce rtx 3090 Xtreme a bit of a point was prompt in my opinion yes it is designed for the overclocker. Yes, it does have a clocking better than the gaming OC, but for most of you, the trade-offs simply would be worth a 5 percent increase in performance.
Therefore I'd much prefer the nvidia geforce rtx 3090 aorus extreme as it used to be gaming OC bias and have the same power limits. It would still be 38 percent better, but there would certainly be some thermal and noise improvements with the larger cooler if you really want an extreme nvidia geforce rtx 3090 on a 3090 ti card from a clocking I'd be looking to force extreme waffles it uses a 240 millimetre AIO. I suspect it'll make far more sense otherwise.
You just far better off buying something than the gaming OC, in my opinion, and I've got to say this much more modest graphics caught impressed me. In terms of quality, it's every bit as good as MSI's gaming extract of the full I wouldn't hesitate to recommend to the gigabytes nvidia geforce rtx 3090 gaming OC, assuming you can find one. You want to dump just an absolutely absurd amount of money on a 3090 nvidia card. Actually, on that note, for games, I strongly suggest you hold out and just buy the nvidia geforce rtx 3090 gaming OC. It is an excellent 3090 ti card.