ASUS GeForce RTX 3090 STRIX OC

[EarthDog]

That's a common misconception. GPU load or utilization is only a certain kind of metric over a sample period of time. Specifically:

utilization.gpu

Percent of time over the past sample period during which one or more kernels was executing on the GPU. The sample period may be between 1 second and 1/6 second depending on the product.

It does not indicate exactly the "capacity" used. There are many other parts of the entire graphics system/pipeline that can be a bottleneck yet result in <99% "utilization".

I get it.

GPU use is typically maxed when there aren't other external bottlenecks.

 

[Tisumi]

Lucky bastard. I'm assuming you're saying you were able to order the Asus 3090 Strix O24G.

Where did you find it?

I've only seen it listed as unavailable on BestBuy and Newegg.

Managed to get one on an official ASUS retail store in Italy, Drako.

 

[birdie]

Literally did not mention a thing about AMD....the fact that you felt the need to bring that company up says a lot about your mindset in this little discussion I'm afraid.

Lets also not forget its the Nvidia users who did not jump from the 10 series to the 20 series and that the presentation was very much about convincing them "thats its time to upgrade"....

And are you really saying the stack of games TPU uses for their reviews are bad choices? that they somehow not represent reality out there? like I honestly im kinda taken back but your viewpoint here.

And intelligence and logic, already said man, 50% more performacne for 50% more power consumption is not some impressive leap in tech, 50% performance while using no more power would be, 50% more performance for 20% less power...now that would be something to cheer for.

I've showed you three games which almost do not scale not matter how much GPU horsepower you throw at them. Is there anything else you'd like to say against NVIDIA and how "poorly" the RTX 3090 performs? In games which are properly coded, we see gains which are in line with the increase in CUDA cores/power consumption. Actually, the RTX 3090 performs better in terms of performance/power than the RTX 3080 which could indicate that your criticism is misplaced.

Sorry, for mentioning AMD fans and their way of thinking. AMD has become a sort of cult in the tech world and I've gotten used to seeing the harshest and in most cases either unwarranted or outright false NVIDIA criticism from the cult followers.

The video in which Jensen Huang announced the RTX 3090 he directly mentioned that the card was to become a Titan of this generation, yet in this thread people keep on being mad at NVIDIA that the company is advertising as the card as a gaming product as if NVIDIA is standing behind their backs with a gun pointing at their heads and forcing them to buy it. Are they?

Do not buy it. OK? Vote with your wallet! Let NVIDIA know how insane they are with their pricing strategy. It's expensive as <no swearing here>. Its power consumption is insane. We get it. Now let's discuss something relevant. For instance the fact that it contains 24GB of VRAM and it's great for content creators while being 40% cheaper than previous generation Titans.

On the other hand, why do luxury/sports cars exist? Who on Earth buys this overpriced junk? They cost up to 200 times more than average cars yet being just 2-3 times faster. Pure insanity!

 

[MikeSnow]

This doesn't seem correct:

Setting the limit to 480W should give you better performance, not worse.

Also, in the power consumption analysis, were the results obtained with the power limit increased to 480W, or the default? Why not include both settings in the power analysis as well? If the analysis was done at the default setting, it would be very interesting to know how much power it can draw in Furmark, with the limit increased to 480W, if it can go over 449W or not.

 

[howiec]

I get it.

GPU use is typically maxed when there aren't other external bottlenecks.

Easier if you read this rather than me explaining:
https://developer.nvidia.com/blog/t...lysis-method-for-optimizing-any-gpu-workload/

Also, you completely ignored the fact that I provided real-world numbers on increased GPU load with lower FPS and decreased GPU load with higher FPS (which makes sense) while CPU load stayed constantly low throughout.

Managed to get one on an official ASUS retail store in Italy, Drako.

Nice!

 

[B-Real]

This doesn't seem correct:

View attachment 169696

Setting the limit to 480W should give you better performance, not worse.

Also, in the power consumption analysis, were the results obtained with the power limit increased to 480W, or the default? Why not include both settings in the power analysis as well? If the analysis was done at the default setting, it would be very interesting to know how much power it can draw in Furmark, with the limit increased to 480W, if it can go over 449W or not.

I also don't get the anomalies regarding the efficiency charts in the 4 tested AIB models. ASUS consumes 20W less in gaming than the Zotac while performing 9% better? These are not different GPUs but the same 3090. Then at Furmark it consumes 50W more. There has to be some inaccuracies.

 

[EarthDog]

Easier if you read this rather than me explaining:
https://developer.nvidia.com/blog/t...lysis-method-for-optimizing-any-gpu-workload/

Also, you completely ignored the fact that I provided real-world numbers on increased GPU load with lower FPS and decreased GPU load with higher FPS (which makes sense) while CPU load stayed constantly low throughout.

thanks again. Always exceptions.

 

[Vayra86]

This card (just like I predicted) is a lot more power efficient that the RTX 3080:

So, your criticism is not totally sincere as the RTX 3080 must be an even worse card in your opinion. Also, NVIDIA has a higher performance per watt ratio for this gen vs the previous gen cards, so your comparison to Intel isn't valid.

As for me personally I never buy GPUs with a TDP above 150W (and CPUs above 95W) because I simply don't want that much heat under my desk, so I guess I will skip the entire GeForce 30 series because RTX 3060 has been rumored to have a TDP around 180-200W.

"Good lord" is misplaced. Again, this is a Titan class card. Either use it appropriately or forget about it - it does not justify its cost purely from a gaming perspective.

The cost simply isn't justifiable. If you buy this.... you buy this. Don't attribute too much sensibility to it.

Pure insanity!

Nail > Head

 

[AddSub]

So, grab these at $150 a pop in 2025 off of ebay, when RTX 5090 64GB GDDR7 model rolls out???

...
..
.

 

[Imouto]

How is this a Titan class GPU when it doesn't even have workstation certified drivers? There are plenty of professional software that runs like a dog on these just because of that.

 

[EatingDirt]

Let me disagree with you here, @W1zzard

This is a Titan class card which used to cost over $2500. Now you're getting a massively faster GPU and even more VRAM at a much lower price.

This is not a gaming card unless you have more money than common sense. It's a productivity card which can be used for gaming.

Titan class cards before the Titan RTX & Titan V used to cost $1,200 MSRP. Nvidia doubled the price with the RTX(& V), because they doubled the price of their flagship gaming line.

There's only 1 reason to buy an RTX 3090, and that's productivity work that needs more than 10GB, and you also need the PC to play games. However, rumors of a 3080 with 20GB are true, then this card is only useful in the immediate short term.

This is just a card for rich people, basically.

 

[SystemMechanic]

@W1zzard Can you show us the actual BIOS Powerlimit in from GPU Z please.

This is NOT a Titan Card, Its just a Ti in desguise with more VRAM, if anything, its close to the Dual GPU's like the GTX 590 and 690 Nvidia used to produced. Linus confirmed that many Titan Features are disabled on this card and probably wont be unlocked..so yeah Stop listening to Jensen's Words.

 

[GhostRyder]

Ill be honest, the review almost turned me towards the 3080 but I am still firm on getting a 3090.

The only thing I am royally mad about is trying to buy one of these at 8:00AM this morning. Nvidia's site refused to load at 8:00 and at 8:02 it was sold out... Newegg loaded but every time I checked out it would say its now sold out right at the end, Best Buy I checked too late, and Micro Center had people around the block waiting. I cannot believe how many people must have used bots to buy this again even though they upped their "Protection". There is no way people checked out that fast without a bots help (I cant even get the pages to load that fast on 1gb internet).

 

[swirl09]

Managed to get one on an official ASUS retail store in Italy, Drako.

Oh nice. I ordered it from a few places, who knows how long it will take tho.

Seeing the noise here though, its not what I hoped. Im surprised that the FE might actually have been just fine really. Would like to see it here for comparison.

 

[pandemonium]

Any notes on the memory used in any of the games?

Given the modest +18% performance over the 3080 at twice the cost, that memory pool is going to be the only saving grace for this beast.

 

[swirl09]

Theres been a bunch of gaming performance analysis that show you shouldnt be seeing a VRAM problem at 4K with 10GBs in the vast majority of games. So for 199 games out of 200, you will be fine. Its only if you plan to go beyond 4K, or pump SS in VR that you will see a benefit from more VRAM (which I do.)

 

[turbogear]

I wonder why this gets Highly Recommended award.

It is power hungry card that delivers only 20% more performance for double the price of 3080.

I will rather buy two 3080 for the price of this monster; one for my pc and one for my son but I will wait until end of the year also to see what RDNA2 will offer.

 

[AvrageGamr]

So no one will be buying two for SLI then??

I read somewhere that Nvidia wont be making game drivers for sli after Jan 2021.

 

[Vayra86]

I wonder why this gets Highly Recommended award.

It is power hungry card that delivers only 20% more performance for double the price of 3080.

I will rather buy two 3080 for the price of this monster; one for my pc and one for my son but I will wait until end of the year also to see what RDNA2 will offer.

Those stickers just go out randomly. Don't worry too much about them. Stickers look good. The reviews here are about the contents

It used to be ridiculous numbers. Let it be. I bet companies love them.

 

[Bill'O]

ok I understand we're crossing borders much like with smartphones market prices...
That said as my work tool, I still consider buying a 3090 but wonder if my corsair RM750i will hold up though ?..

 

[ratirt]

Dude... this is AT LEAST the third time I've said this to YOU.............. and in this forum more times than that. How this is still unclear to you bud is beyond me.

1. In the video, Jensen talks Titan, then in the same breath whips out the RTX 3090 from his oven.
2. In the same video, slides state the 3080 is the 'flagship'.

Oh so if NV CEO says it is Titan then it must be it? Well I disagree.
NV's men said a lot of things and it was never true. Don't fall for the marketing dude. Think a bit.

 

[Bill'O]

Oh so if NV CEO says it is Titan than it must be it? Well I disagree.
NV's men said a lot of things and it was never true. Don't fall for the marketing dude. Think a bit.

I actually never understood why everybody was bashing 2080ti owners after seeing 2 powerpoints at nvidia... It's like guys those gpus are more than 2y old and a new gen comes up so what do you expect ? Well it's performing better ofc but wait for real benchmarks with current drivers. And guess now...

 

[ratirt]

I actually never understood why everybody was bashing 2080ti owners after seeing 2 powerpoints at nvidia... It's like guys those gpus are more than 2y old and a new gen comes up so what do you expect ? Well it's performing better ofc but wait for real benchmarks with current drivers. And guess now...

Well there are people who take things as they are presented to them or stop and think and have different conclusions.

 

[W1zzard]

no holds barred

Fixed

This doesn't seem correct:

View attachment 169696

Setting the limit to 480W should give you better performance, not worse.

Fixed

Also, in the power consumption analysis, were the results obtained with the power limit increased to 480W, or the default? Why not include both settings in the power analysis as well? If the analysis was done at the default setting, it would be very interesting to know how much power it can draw in Furmark, with the limit increased to 480W, if it can go over 449W or not.

It was done at default. I added all the 480 W testing only yday with super limited time. Now moving to a new house, so not sure when I can update

I also don't get the anomalies regarding the efficiency charts in the 4 tested AIB models. ASUS consumes 20W less in gaming than the Zotac while performing 9% better? These are not different GPUs but the same 3090. Then at Furmark it consumes 50W more. There has to be some inaccuracies.

It seems the variance in leakage between Ampere GPUs is quite big. The power limit settings play a big role here, too. Also silicon operating temperature in the case of ASUS

@W1zzard Can you show us the actual BIOS Powerlimit in from GPU Z please.

BIOSes have been posted to our database now

 

[EarthDog]

Oh so if NV CEO says it is Titan then it must be it? Well I disagree.
NV's men said a lot of things and it was never true. Don't fall for the marketing dude. Think a bit.

Nice to see you reply and acknowledge it.

I'm just sharing what we know which is better than plating ignorant when you were told the same thing. Think, dude.