NVIDIA GeForce RTX 3080 Founders Edition

[R0H1T]

A lot i.e. more than 5~15% incommensurate with their performance.

 

[mechtech]

It's amazing you've just signed up to spoil the launch with your very valuable opinion. Also, remember, we live in a free society, so please stop it with "overpriced". Did NVIDIA put a gun against your head and ask you to buy any of their GPUs? No? Then how on Earth are they overpriced? Also, I agree, "0GB VRAM is very low... for 4k". Except this card features 10GB and we have close to zero games which actually require more than 8GB of VRAM at 4K. Also read the rest of my comment.

Speaking of "planned obsolescence" due a lack of VRAM:

See how the GTX 1060 3GB still works relatively OK despite not having enough VRAM even 5 years ago. Yes, it's very slow in fact 33% slower, but not 2 or 3 times as slow as its 6GB brother. Also, see how both cards are unusable at this resolution.

• Game developers target the most popular GPUs and most of them contain 8GB of VRAM or even less. Consoles also won't really feature more than 8GB of VRAM because their memory pool is shared between an OS core, game code and VRAM and that all should fit into 16GB or less. And both consoles feature uber fast textures streaming off their storage.
• Also, it's been shown time and again, that NVIDIA has superb drivers and their GPUs performance is not seriously affected even when there's not "enough" VRAM. E.g. Call of Duty: Modern Warfare eats VRAM for breakfast (over 9.5GB of VRAM use at 4K), yet, are NVIDIA cards with 6GB of VRAM affected? Not at all. Besides with RTX IO it all becomes moot.
• Lastly, by the time 10GB of VRAM is not enough, this GPU performance will be too low even if had twice as much VRAM.

Still you can always buy a future-proof GPU from AMD. BTW, do you remember Radeon R9 Fury? Was released with paltry 4GB of VRAM which was strangely enough for AMD fans.

Most people who have a budget in the GTX 1060 range are also people whose monitor budget is in the 1080p range. Also every game is different, look at Witcher 3.

The Witcher 3: Performance Analysis

In this article, we put the GTX Titan X, R9 295X2, GTX 980, R9 290X, GTX 970, R9 290, GTX 960, and R9 285 through the Witcher 3: Wild Hunt. We do so at resolutions of 1600x900, 1920x1080, 2560x1440, and 4K to assess what hardware you need to play this recently released top-title.

www.techpowerup.com

If all I had was a 1080p and a small budget, I wouldn't get more than 4GB, and maybe even a 2GB card since a GPU's life cycle is so short anyway, compared to that of a monitor which is usually 10 years.

 

[nguyen]

Vega was also on an inferior node & admittedly an inferior uarch, that compounded their problem. But what's clear that there's still bias against AMD & towards Nvidia, JHH can launch a proverbial turd (remember Fermi?) & still get accolades while AMD not only has to please the audience but also pay them to do so

RTG just reap what they sown, the market will response when RTG make a good product like Ryzen.
Vega was too inefficient
Navi is too expensive (compare to Polaris), 5500XT and 5600XT are like sacrificial pawn.
Not to mention the state of their drivers and hardware compatibilities, well you can blame AMD for that because the R&D funding was pooled into CPU developments, that was also why Raja left.

 

[moproblems99]

that was also why Raja left.

We'll find out soon if that matters.

 

[Caring1]

There's no way i'm buying ANYTHING from someone who wears a leather jacket in the kitchen.

Would you rather he wore only an apron?

 

[mechtech]

W1zz, what is the BIOS version for that EVGA Z390 motherboard?

 

[R0H1T]

RTG just reap what they sown, the market will response when RTG make a good product like Ryzen.

Oh sure when AMD fucks up it's their fault, when they don't it's still their fault

Let's see how many GPUs or series (uarch) do you think they didn't botch in the last 5 years?

well you can blame AMD for that because the R&D funding was pooled into CPU developments

For good reason, going EPYC not only helped them stay afloat but also now lead the top end CPU market virtually on all platforms.

 

[xorbe]

Looks good, but I am uneasy with that much wattage for a video card. The 3070 is supposed to be 220W? And only 8GB, why would anyone buy 220W of gpu fire power, and only have 8GB.

 

[z1n0x]

Too much hype man, too much hype, nothing can live up to overhype. (ot) CP2077 will have the same faith.

OMG all those CUDA cores, forgeting the need to have balanced architecture that can feed all those executiion units. Forgot Vega64 vs GTX1080, 4k vs 2.5k ALUs

It's good generational jump.

Nvdia shouldnt have pushed it beyond the efficiency curve. Also 10GB for 4k card, yeee no.

 

[W1zzard]

do you think with v.mod like shunt mods i have done on many cards will help with overclocking headroom

most certainly

Is there any plan to add the Call of Duty series back into reviews any time soon?

No plans, bnet banned me for trying to benchmark the game, and their support didn't want to admit it and chose to ignore my ticket for 3 months.
Always online sucks, because they will force perf-changing patches on you whenever it's most inconvenient.

 

[Dyatlov A]

If it has a manual power limit adjustment range of up to 370 W, from a default of 320 W, is it also possible to lower max power consumption, for example to 250W?

 

[Deleted member 185088]

While the performance is impressive finally a true 4k card and now that it's been confirmed that on LG OLEDs it works at 4k120HDR, this combination is just phenomenal.
But the card is way too expensive, the 700$ is just fake (I assume it doesn't include tax) elsewhere it's much more expensive (even in Asia where the card is built it can cost up to 1000$ or more), why not use a similar pricing as consoles (MSRP), the PS5 and Series X cost around the same in Europe, north America and Japan.
At the end of the day, great performance ruined by stupid pricing (I'm talking about nVidia MSRP).

 

[nguyen]

While the performance is impressive finally a true 4k card and now that it's been confirmed that on LG OLEDs it works at 4k120HDR, this combination is just phenomenal.
But the card is way too expensive, the 700$ is just fake (I assume it doesn't include tax) elsewhere it's much more expensive (even in Asia where the card is built it can cost up to 1000$ or more), why not use a similar pricing as consoles (MSRP), the PS5 and Series X cost around the same in Europe, north America and Japan.
At the end of the day, great performance ruined by stupid pricing (I'm talking about nVidia MSRP).

You are mixing up supply and demand there, retailers just like to jack up the price when stuffs are in short supply.
Just wait few months when price settles, PS5/XBX price will be inflated when they first launch too.

 

[wolf]

If it has a manual power limit adjustment range of up to 370 W, from a default of 320 W, is it also possible to lower max power consumption, for example to 250W?

There are a couple of links on page 9, it seems that this is indeed possible and the results are promising. One site went down the slider to 83% max power limit to match the wattage draw of a 2080Ti at 270w, but it only resulted in a ~4% performance loss, not bad at all! the hypothesis is that Nvidia have had to push the GPU beyond the optimal efficiency sweet spot to hit the performance target they were seeking.

 

[laszlo]

if they would produce it at TSMC on 7nm it will have the projected leap and would be cooler also...; price-wise no matter how i look i'll never pay more than 300 for a card so good purchase to those who'll buy it!

 

[Taraquin]

That's because lower resolutions are CPU limited.

Edit: Actually you're making a great point. I'm using the same 303 W typical power consumption value from the power measurements page on all 3 resolutions, which isn't 100% accurate. Because it's some games are CPU limited, then in those games the power consumption is down, too, which I'm not taking into account

Perhaps all GPU-powerconsumption-testing should be done at 4k so we get a true idea of how much power it uses if not CPU-limited?

 

[Deleted member 185088]

You are mixing up supply and demand there, retailers just like to jack up the price when stuffs are in short supply.
Just wait few months when price settles, PS5/XBX price will be inflated when they first launch too.

No I'm not, this is the MSRP of nVidia, for instance they price the 3080 at 699$ in the US but 1000$+ in Japan the prices are from their website.
So why they didn't price the cards at 699$ or thereabouts everywhere?
The consoles are around the same MSRP in most markets.

 

[hkpolice2]

Under the overclocked performance testing, what settings were used in Unigine Heaven?

Thanks

 

[gridracedriver]

It isn't NVIDIA's fault if you can't read a chart, it's 1.9x at the same performance...

of course I had noticed it, on tom's IT I had also pointed it out to everyone, but still they declared 1.5x with the same power draw! see the picture.
Well, where is this 1.5x in performance/watt?
declare false, there is no excuse.

 

[Dux]

13-24% performance increase over 2080Ti. Depending on resolution. With 28% higher TDP. Don't bother selling your 2080Ti.

 

[londiste]

No I'm not, this is the MSRP of nVidia, for instance they price the 3080 at 699$ in the US but 1000$+ in Japan the prices are from their website.
So why they didn't price the cards at 699$ or thereabouts everywhere?
The consoles are around the same MSRP in most markets.

Consoles are not quire around the same MSRP either. For example, the recently announced price of Xbox Series X is $499/499€/£449 and 49980 yen. Compared to US price Japanese price is reduced as it is a difficult market for Xbox and EU/GB prices reflect included taxes.

There are a number of reasons to vary MSRP across different regions or countries - taxes are a big one and marketing position is another.

13-24% performance increase over 2080Ti. Depending on resolution. With 28% higher TDP. Don't bother selling your 2080Ti.

15-31%. RTX3080 is considerably more CPU limited on 1440p.

 

[Dyatlov A]

There are a couple of links on page 9, it seems that this is indeed possible and the results are promising. One site went down the slider to 83% max power limit to match the wattage draw of a 2080Ti at 270w, but it only resulted in a ~4% performance loss, not bad at all! the hypothesis is that Nvidia have had to push the GPU beyond the optimal efficiency sweet spot to hit the performance target they were seeking.

Combining that with undervolting and maybe no performance lost

 

[nguyen]

No I'm not, this is the MSRP of nVidia, for instance they price the 3080 at 699$ in the US but 1000$+ in Japan the prices are from their website.
So why they didn't price the cards at 699$ or thereabouts everywhere?
The consoles are around the same MSRP in most markets.

US and Japan do not include VAT into their listing price, everywhere else does LOL. Right now it tough for you to avoid VAT when buying online in the US so yeah please add 8-12% VAT depending on where you live.
If you can buy online from Nvidia website without VAT, go right ahead, it's a mighty good deal.

 

[Valantar]

Just check the relative performance of 2080 Ti compare to 3080 across 3 resolution and you will see a change
1080p: 87%
1440p: 81%
4K: 75%
Might as well use DSR to do some 8K testing on this bad boy
People with 1080p and 1440p screen might as well use DSR if they have the 3080, otherwise you are just wasting all performance prowess of 3080.

...if those numbers show anything, it's that there are no CPU bottlenecks at 4k, given that it scales beyond the lower resolutions. Which is exactly what I was pointing out, while @birdie was claiming that this GPU is CPU limited even at 4k:

quite a lot of games being reviewed are severely CPU limited even at 4K!

Which I then asked for some data demonstrating, as this review fails to show any such bottleneck. And, as GN pointed out in their excellent video review, not all seeming CPU limitations are actual CPU limitations - some examples of poor GPU scaling are down to the architecture or layout of the GPU causing GPU bottlenecks that don't show up as 100% load.

That's a strange way to think about it, if it is possible to exceed that amount then it inevitably becomes a limitation.

Only if there is a perceptible difference in graphical quality - without that, the difference is entirely theoretical. And that's the entire point: a lot of games have extremely "expensive" Ultra settings tiers with near imperceptible or even entirely imperceptible differences in quality. If your benchmark for not having a VRAM limitation is the ability to enable all of these, then your benchmark is problematic. If the thing you're worried about for this GPU is that you might at some point in the future need to lower settings an imperceptible amount to maintain performance, then ... what are you worried about, exactly? Stop promoting graphical quality nocebo effects, please. Because at that point, all you are arguing for is the value/security of knowing your GPU can handle everything set to Ultra, no matter what this actually means for the quality of the game. Which is just silly. Not only is it an impossible target, but it's a fundamentally irrational one.

 

[FinneousPJ]

I don't know guys. The 2080 Ti is 19% slower on average in these tests. That doesn't seem super impressive to me -- especially compared to the hype.