Support NeoGAF

[pawel86ck]

According to techpowerup benchmark summary 3090 is 56% faster on average compared to 2080ti and that's a lot if we talking about 4K gaming.

49fps RTX2080ti, yet 78-80fps, that's huge difference.

In "control" 3090 with "RTX on" is as fast as 2080ti without RTX, yet people here still say 3090 is not big upgrade over 2080ti. I'm not going to buy 3090 for myself (I'm more interested in 3070 16GB for my needs), but I can admit, 3090 eats 2080ti for breakfast at 4K. I bet many rich people will buy it, especially in richer countries where average people earn 2000-3000$. My friend engineer can buy 4x 3090 with his single salary , whats funny he can spend 1000$ just on good bottle of wine not to mention his hobby (some people just cant understand 1500$ can be expensive for some, but cheap for others).

 

[evanft]

Before the last week, nVidia could make the claim of having the seven fastest GPUs in the world, with eight in the top ten. Today they have the top nine spots.

 

[Sophist]

3090 is more interesting for ray tracing (Blender, Daz3D, ...). Power consumption is getting crazy.

NVIDIA GeForce RTX 3090 Performance In Blender, Octane, V-Ray, & More

NVIDIA launched its first Ampere-powered GPU just over one week ago, and on schedule, the second has arrived: GeForce RTX 3090. This GPU targets users who demand uncompromising performance and capabilities out of their graphics hardware, and with 24GB of GDDR6X, this GPU has plenty of breathing...

techgage.com

 

[ZywyPL]

Call me crazy but I'd buy a 12GB 3090 instead of the upcoming 20GB 3080 is there was such an option, those extra FPS are tempting as hell, while double the RAM will most likely never be used, bare in mind it's the consoles that are catching up with 4K and ultra textures, they're not moving any baseline forward in that regards, so the PCs will just continue to do whatever they have been doing in the past years. I think the "insufficient" RAM issue is blown up of proportion and whoever got the 8-10GB cards will be way more than happy wit hthe performance they provide.

36 Tera-fuckin-flops . Don’t even think PS6 will have that much power. You PC gamers are really way ahead.

Only about 65% of them are FP32 TFlops tho, so around 23 "real" TF. Same applies to other Ampere cards as well obviously. So in reality, there's a maximum of 15% difference between 3080 vs 3090 in on-paper performance, and also while actually running the games.

 

[Sentenza]

Based on the leaked AIB shot, and just scaling from the 3070, the 3060 is looking to hit 2080 levels, which would put it in line with the XSX GPU or a bit slower.

Yeah, I'm sure the low budget, console-friendly solution from AMD will match a 3070 toe-to-toe and their PC GPUs will run circles around the 3080 and 3090 at half the price.
Following their long history of results matching and exceeding their original claims, they will once again wipe the floor with Nvidia.

 

[Zathalus]

Yeah, I'm sure the low budget, console-friendly solution from AMD will match a 3070 toe-to-toe and their PC GPUs will run circles around the 3080 and 3090 at half the price.
Following their long history of results matching and exceeding their original claims, they will once again wipe the floor with Nvidia.

Do you have a reading comprehension problem? I said 3060, not 3070. We already known that the XSX can match the 2080 in a unoptimized title with a weaker CPU (Gears 5), putting it firmly at around 2080 Super levels of power. Now the 2080ti (and hence the 3070) is only about 20% faster then the 2080 Super. The 3080 on the other hand is around 55% faster then the 2080 Super. Lastly the beastly 3090 is around 70% faster then the 2080 Super.

Thus the 3060 would be as powerful as the XSX GPU, the 3070 will be around 20% faster, the 3080 will be around 55% faster, and the 3090 will be around 70% faster.

To put that into perspective, if a game runs at a locked 30FPS on the XSX then a 3070 will run it at around 36 FPS, a 3080 will run it at 46 FPS, and a 3090 will run it at 51 FPS.

We don't need to believe anybody's claims about anything, we know how powerful the XSX GPU is, and we know how the 3000 series cards compare to its PC equivalent.

 

[Ascend]

Even if it does ( Big if ), there's a near zero percent chance it will have raytracing performance anywhere close and no response to DLSS 2.0 in sight.

We know nothing about the ray tracing performance. Your near zero percent chance metric is just something you're spouting due to your own bias. Nobody knows how the ray tracing performance of AMD's cards are. The only reference we have is Minecraft running on the Xbox Series X, but that did not give enough data to know how it actually runs.
As for DLSS 2.0, as long as only a handful of games use it, I would not base my choice on a graphics card on just that. If you really love that feature, go for it. I wouldn't be surprised that AMD will ultimately have its own solution that's not a proprietary thing, but will use DirectML instead. Sort of like Hairworks vs TressFX.

There's a big question on the table of how well more CUs will scale. Double the CUs of the 5700XT is not likely to equal double the performance.

If the IPC and clocks were equal, you would be right. But the expected IPC increase is around 15%, and, the clocks are expected to be around 30% higher than the 5700XT.

Are you expecting a 3080 competitor from AMD at $499? Nvidia will just adjust their prices accordingly anyway.

nVidia rarely (if ever) drop prices after cards have launched. They are quicker to release another card instead with the new price. If nVidia does react on pricing, it is a testament to AMD's advances.
Besides, who said it will cost $499?

I would be very very happy to see AMD compete on the high end again, but I have my doubts.

Your post history indicates otherwise. Every chance you get to trash on AMD, you do.

They represent real world performance just as much as they always have.

Yeah... Just wait till the 20TF+ Big Navi is equal to or faster than the 30TF 3080.

 

[wordslaughter]

So you're optimistic about AMD GPUs and I'm pessimistic. Feels like I have every reason to be. They have been consistently disappointing for many years now. Usually a new AMD GPU launch represents AMD catching up to where Nvidia was 18 months to 2 years prior.

Yeah, I don't think the raytracing will be very good in big Navi. It will be AMD's first try at hardware accelerated raytracing and we know next to nothing about it. If it were something special the consoles would be talking more about it. "Class leading ray tracing" or some other buzz words but instead the next gen consoles simple HAVE "hardware accelerated raytracing" - I'm not expecting miracles here.

You like to talk a lot about how Ampere's TFLOPs are fake or misleading without really understanding what they are. Only to try to point out that Ampere "isn't as powerful as it seems" but in reality with DLSS 2.0 it's actually more powerful than it seems. It will literally perform at 4k where a similarly powerful GPU (but lacking DLSS 2.0) would perform at 1440p.

I don't trash talk AMD, only Radeon. AMD makes awesome CPUs now - I personally use a 3900X.

Big Navi is not going to be faster than the 3080 and the consoles don't have GPUs on the same level as a 3070.

If Big Navi is better than what Nvidia is offering then I will buy Big Navi. It's really as simple as that ...but I would not expect it.

 

[Revas]

Even if it does ( Big if ), there's a near zero percent chance it will have raytracing performance anywhere close and no response to DLSS 2.0 in sight.

There's a big question on the table of how well more CUs will scale. Double the CUs of the 5700XT is not likely to equal double the performance.

Are you expecting a 3080 competitor from AMD at $499? Nvidia will just adjust their prices accordingly anyway.

I would be very very happy to see AMD compete on the high end again, but I have my doubts.

I doubt it'll match Ampere ray tracing performance, I've seen leaks that RDNA2 is closer to Turing. The price will have to reflect it's level of ray tracing ability - my gut says $599 for NAVI 21 but I wouldn't be surprised if it were a little lower. I think many players would be interested in a card offering top tier rasterized performance at a mid range price point even if it's weak in terms of RT.

 

[llien]

If Big Navi is better than what Nvidia is offering then I will buy Big Navi.

No, you won't.

You'll find lame "RT" excuse (something that even 2080Ti users turn off in the handful of games that supports it and there has been no major improvement on this front).
If that doesn't work, you'll find lame upscaling excuse.
If even that doesn't work, there is always that lovely old "stability" and "drivers" song.

On the other hand, you don't have to!
Please, buy only green stuff.

So you're optimistic about AMD GPUs and I'm pessimistic.

There are reasons to be optimistic (major Vega to Navi jump after Radeon group got modest R&D budget), while pessimism is alike "when I was standing on that street last week, it was rainy" kind of "logic". Even RDNA1 is enough to roll out competitive cards with performance way beyond 3070.

 

[wordslaughter]

No, you won't.

You'll find lame "RT" excuse (something that even 2080Ti users turn off in the handful of games that supports it and there has been no major improvement on this front).
If that doesn't work, you'll find lame upscaling excuse.
If even that doesn't work, there is always that lovely old "stability" and "drivers" song.

On the other hand, you don't have to!
Please, buy only green stuff.



There are reasons to be optimistic (major Vega to Navi jump after Radeon group got modest R&D budget), while pessimism is alike "when I was standing on that street last week, it was rainy" kind of "logic". Even RDNA1 is enough to roll out competitive cards with performance way beyond 3070.

Your response makes no sense.

I said I'd buy Big Navi if it was better, you then proceed to imagine what I'd say if Big Navi is worse in both ray tracing and had no response to DLSS 2.0.

Hate to break it to you, but if that's true, that Big Navi has both worse raytracing AND no answer to DLSS 2.0 (2.1 really) ... then Big Navi is worse, so why would I bother buying it? Why would anyone?

I've owned AMD GPU's in the past and would do so again if AMD can beat (or even match) Nvidia. I loved my 7970 Ghz Edition and bought a 280X for my wife a couple years later. Remember, back when AMD was still competing on the high end? I do, but that was a LONG time ago.

You say even "RDNA1 is enough to roll out competitive cards with performance way beyond 3070." ... BUT DID THEY?! No, they didn't.

If AMD "could have" released an RDNA1 GPU that could beat a 2080ti then why the fuck didn't they? Maybe you know more about RDNA 1 then AMD does?

The best RDNA1 card was the 5700XT (July 2019), which wasn't even as good as a 1080ti (March 2017) - This is EXACTLY my problem with Radeon on the high end. A 1080ti is my current GPU. It took AMD 28 months to release a GPU that, even then, couldn't match what Nvidia released more than 2 years earlier.

So don't be surprised when people who shop for high end GPUs are less than optimistic when it comes to Radeon. Nvidia has literally been our only option for many many years. If AMD can change all that, great. But I'm not hyping up ANY of their GPUs. Been disappointed too many times.

 

[Ascend]

Lessons from trading/investing have just popped up in my head... When everyone is the most fearful is the time to buy, when everyone is the most greedy, is the time to sell.
Right now, everyone is extremely greedy for the 3000 series, and extremely fearful of Big Navi. I know where I'm placing my bets. Just need a confirmation candle to buy

 

[wordslaughter]

"extremely fearful of Big Navi" LOL

 

[AquaticSquirrel]

"extremely fearful of Big Navi" LOL

I think by fearful he means uncertain of their chances so fearful of customers to buy one rather than fearful they will be a runaway success.

The comparison was between buying stocks on the market.

 

[Ritsumei2020]

Can it play Crysis?

 

[Kenpachii]

Why are people suprised by this? its the same gpu chip which means its a binned 3080 chip with 24gb of v-ram and probably higher clocks. that's about it.

10% better... until games need more than 10GB of VRAM.

Exactly.

 

[Ascend]

"extremely fearful of Big Navi" LOL

Do not misinterpret it. I do not mean that people are afraid of Big Navi in the sense that they are expecting it to be competitive and afraid hey are going to beat nVidia or whatever. I mean people have so little faith in AMD's graphics division at this point in time, that the sentiment pretty much mirrors when there's capitulation in the stock market. nVidia's 3000 series sentiment pretty much mirrors a bubble at this point. And bubbles inevitably pop.

 

[llien]

I said

then Big Navi is worse

Why did you even bother writing that up, FFS?
RT is a gimmick kept alive by hype.
DLSS 2.0 is a little more than TAA, available in a handful of peculiar games, hyped by the types known for shilling.
It is guaranteed AMD won't roll out either.

So why don't you stop finding lame excuses and simply STFU about unrealistic scenarios in which your majesty would bother buying something from another manufacturer? It's a rhetorical question. It means I'm not really asking you anything.

If AMD "could have" released an RDNA1 GPU that could beat a 2080ti then why the fuck didn't they?

Because financial and engineering resources are limited. It made no sense to roll out 7nm oversized DUV chip, when 7nm EUV is around the corner. It is also the case of "two years older" argument not being braindead, the market for chips of that size/price was already saturated a bigger leap is needed to sell there.

 

[Mod of War]

Dial it back on the hostility and antagonistic posts.

Thank you.

 

[Allnamestakenlol]

In "control" 3090 with "RTX on" is as fast as 2080ti without RTX, yet people here still say 3090 is not big upgrade over 2080ti.

Are people saying that, tho? I thought all the talk was comparing it to the 3080.

 

[danielJackson]

People whom buy 1500€ GPUs wont care if it is "only 10%" faster.

Price is crazy even if it would be 50% or 100% faster. Aka it cant be explained by logic, some people just want the best no matter the price.

IMO if anything, people should talk about the prices of GPUs today overal, they are all too damn high

 

[Siri]

Nvidia marketed it as a gaming card too.
Watch gamers nexus review.

Are you referring to that Tech Jesus review everyone’s on their knees about?

It’s actually a little insulting to say that Nvidia shouldn’t have marketed the 3090 as a ‘gaming card’. It‘s their job to sell cards by creating hype. This is what companies selling product do,

As consumers it’s our job not to fall for hype and to do a modicum of research and not rush out and spend an insane amount of money on an un-reviewed card. If you purchased a 3090 without reading any reviews, and you were expecting a superior card, then you have no one to blame but yourself.

But honestly, I’d be surprised if there’s even one person right now who‘s sitting there crying over his 3090, repeatedly saying, “But... Nvidia said this was a gaming card! It’s not a gaming card! They... they lied to me!!”

 

[wordslaughter]

Why did you even bother writing that up, FFS?
RT is a gimmick kept alive by hype.
DLSS 2.0 is a little more than TAA, available in a handful of peculiar games, hyped by the types known for shilling.
It is guaranteed AMD won't roll out either.

So why don't you stop finding lame excuses and simply STFU about unrealistic scenarios in which your majesty would bother buying something from another manufacturer? It's a rhetorical question. It means I'm not really asking you anything.


Because financial and engineering resources are limited. It made no sense to roll out 7nm oversized DUV chip, when 7nm EUV is around the corner. It is also the case of "two years older" argument not being braindead, the market for chips of that size/price was already saturated a bigger leap is needed to sell there.

You are seriously tieing yourself in knots to make excuses for AMDs GPU failings. Why?

 

[DonMigs85]

all that shading power but the memory bandwidth, ROPs and TMUs didn't scale along with it.