Monday, April 10th 2023

Official NVIDIA RTX 4070 Performance Claims Leak Online

Just a few days before the official launch scheduled for April 13th, the first official performance figures for the RTX 4070 have found their way online. As expected, NVIDIA is showing the performance with DLSS 3 and Frame Generation, comparing it to three graphics cards in the RTX 30 series, the RTX 3080, the RTX 3070 Ti, and the RTX 3070.

According to NVIDIA, the GeForce RTX 4070 is targeting 1440p performance at above 100 FPS, with ray tracing and DLSS 3 enabled, of course. NVIDIA officially did not reveal any performance numbers without DLSS 3, but earlier rumors puts the RTX 4070 at around the same performance as the RTX 3080. With DLSS 3 and according to NVIDIA, the RTX 4070 is 1.4x, 1.7x, and 1.8x times faster, compared to the aforementioned RTX 30 series graphics cards.
NVIDIA has a pretty decent list of games, with and without Frame Generation, showing the RTX 4070 pulling way ahead of the RTX 3070 Ti and RTX 2070 Super, as expected. NVIDIA was keen to note that the RTX 3080 launched at $699, and the RTX 3070 Ti launched at $599, which should be the MSRP of the RTX 4070.

As detailed earlier, the GeForce RTX 4070 should be around 15 percent slower than the RTX 4070 Ti in gaming, with the gap closing on higher resolutions. Of course, these are all numbers with DLSS 2/DLSS 3, so you should wait for reviews to show up to get a better idea on the actual performance, performance per Watt, and performance per dollar. The first slide also confirms that the GeForce RTX 4070 will be available on April 13th at $599.
Source: Videocardz
Add your own comment

113 Comments on Official NVIDIA RTX 4070 Performance Claims Leak Online

#1
Frick
Fishfaced Nincompoop
The 4080 is about $1500 here, so I assume this will be $1000 minimum.
Posted on Reply
#2
P4-630
Nice, I do 1440 gaming, but I'm not in a hurry, I'll wait for Blackwell 2x~2.6x the performance..

Posted on Reply
#3
john_
So, Nvidia is marketing Frame Generation as a performance improvement.
Posted on Reply
#4
Calmmo
P4-630Nice, I do 1440 gaming, but I'm not in a hurry, I'll wait for Blackwell 2x~2.6x the performance..

Press x for doubt.

AD brought a noticeable improvement over ampere because ampere was very limited by Samsung's node.
Posted on Reply
#5
hsew
P4-630Nice, I do 1440 gaming, but I'm not in a hurry, I'll wait for Blackwell 2x~2.6x the performance..

For 2x the price, probably!
Posted on Reply
#6
P4-630
CalmmoPress x for doubt.
Sure, still rumors.
Posted on Reply
#7
napata
CalmmoPress x for doubt.

AD brought a noticeable improvement over ampere because ampere was very limited by Samsung's node.
It's highly unlikely but AD is also basically copy paste Ampere with almost no changes. AD also scales terribly at the top end so something like AD102 heavily underperforms based on specs alone. With a big change in architectures that scales better you could see big gains.
Posted on Reply
#9
dirtyferret
Just a few days before the official launch scheduled for April 13th, the first official performance figures for the RTX 4070 have found their way online.

Just a few days before the official launch scheduled for April 13th, Nvidia sent official performance figures to their typical "leakers" in order to build up buzz for the RTX 4070 launch like they always do.

There i fixed it for you
Posted on Reply
#10
BSim500
So the RTX 4070 is around 15-20% faster than the RTX 3070Ti without FFMB (Fake Frame Marketing Bullsh*t)? Still, at least that means they'll be affordable, right?... ;)
Posted on Reply
#11
N/A
BSim500So the RTX 4070 is around 15-20% faster than the RTX 3070Ti without FFMB (Fake Frame Marketing Bullsh*t)? Still, at least that means they'll be affordable, right?... ;)
Not if you run the Last of us on 1440p max settings where 3070 is slower than a 3060. Or 30 vs 60 fps truly a 2x.

Affordability is debatable. compared to a 4080 that is 50% faster for 70% overcharge. This is 20% better for equal perf$. I can't wait for the real press release on 12 th.
kondaminwill it be a 16GB card?
You wish, sadly Nvidia decided to push the real 4060 and 4060 Ti as 70 class. What a joke.
Posted on Reply
#12
Pumper
Again comparing to older GPUs with frame generation ON for 4000, and that price comparison using 2 year old MSRPs for 3000 cards is just a fuck you to every customer.

It's obvious that the frame generation was locked out on 2000 and 3000 not because of some hardware limitation, but just as a PR stunt for 4000 series.
Posted on Reply
#13
oxrufiioxo
P4-630Nice, I do 1440 gaming, but I'm not in a hurry, I'll wait for Blackwell 2x~2.6x the performance..

While I don't doubt that a 5090 could be 2x the 4090 I still think anything 5070 and below will be trash especially considering 3nm will be another massive increase in wafer cost.
napataIt's highly unlikely but AD is also basically copy paste Ampere with almost no changes. AD also scales terribly at the top end so something like AD102 heavily underperforms based on specs alone. With a big change in architectures that scales better you could see big gains.
You could also say the cut down versions AD103 and AD104 perform better than expected. Regardless they are priced almost exactly how they perform in comparison of each other. I expected the 4080 to be a lot worse given it's specs but it's still 30-40% less performance depending on game at 4k for 25% less money FE vs FE at least here in the states I know that varies by region.
N/AYou wish, sadly Nvidia decided to push the real 4060 and 4060 Ti as 70 class. What a joke.
At least this has 12GB the 4060/4060ti look like they are going to have 8GB of Vram making them DOA for anything but 1080p and even then likely trash. I personally don't think the 4070 is going to be very impressive but if you just account for inflation this is like a 25 usd increase over the 3070 another card I really didn't like due to the 8GB of vram. Even today it's hard to get a 3070 for less than 530 and at a min this will be a much better buy than it.
PumperAgain comparing to older GPUs with frame generation ON for 4000, and that price comparison using 2 year old MSRPs for 3000 cards is just a fuck you to every customer.

It's obvious that the frame generation was locked out on 2000 and 3000 not because of some hardware limitation, but just as a PR stunt for 4000 series.
I really like Frame generation but I agree Nvidia needs to stop using it in bar graphs unless they are comparing 40 series cards only. As to whether older generations could support it idk even if my 3080ti supported it I still would have grabbed a 4090 partly because I look at it just as a bonus to have not something that impacts my buying decision It's also only useful in games like CP2077 and Witcher 3 next gen making it very limited to how I would use it at least at a min it's definitely useless for MP gaming.
Posted on Reply
#14
R0H1T
CalmmoPress x for doubt.

AD brought a noticeable improvement over ampere because ampere was very limited by Samsung's node.
Press F for F U to you know who :nutkick:
oxrufiioxoI personally don't think the 4070 is going to be very impressive
Without frame gen this thing's a daylight robbery! On par with vanilla(?) 3080 & likely way more expensive? The 3080 12 GB (on deal) went for ~600 USD or lower IIRC.
Posted on Reply
#15
oxrufiioxo
R0H1TPress F for F U to you know who!

Without frame gen this thing's a daylight robbery! On par with vanilla(?) 3080 & likely way more expensive? The 3080 12 GB (on deal) went for ~600 USD or lower IIRC.
At the same time anyone who want's a new gpu right now has the terrible 3070/3070ti for almost the same price as an alternative. People have short memories but both these cards were over 1000 usd most their usable life to begin with. 3080 10G was what 3x-4x the cost..... But people rather just moan and complain. Do I like the card no but comparing this to end of like ampere cards that consume twice the power isn't really useful to me.

The 3080 12G spent less than half it's life at a decent price and even it I wouldn't buy over this that card started at like 1200 usd to begin with.

In two years the same people will be crying about Blackwell and comparing it to end of life ada pricing and a 1000 usd ish 4090 lmao.
Posted on Reply
#16
dirtyferret
oxrufiioxoAt least this has 12GB the 4060/4060ti look like they are going to have 8GB of Vram making them DOA for anything but 1080p and even then likely trash.
The thing is Nvidia is hell bent on protecting their high price positioning so in order to keep the 4070 "justified" they are gimping the 4060/ti so it won't make sense as a "good enough" card to fall back to

Part of the problem (and part of the reason for Nvidia's high prices) is actually AMD. Unlike with their CPUs, AMD couldn't go for the jugular with their GPUs if you literally put their mouth to Nvidia's neck. They act as if 20% market share is good enough for them and no need to rock the boat. As consumers, we desperately need Intel to be competitive in the GPU industry.
Posted on Reply
#17
R0H1T
Nothing to do with short memories, Ampere debuted at great price in fact it was probably the best priced Nvidia gen in the last decade or so. Then yeah Covid happened & prices went awry for everyone ~ but still doesn't take away from the fact that it had great initial pricing. ADA is like Turing 2.0 all over again!
oxrufiioxoIn two years the same people will be crying about Blackwell and comparing it to end of life ampere pricing and a 1000 usd ish 4090 lmao.
Maybe people with memory span of a goldfish.
dirtyferretPart of the problem (and part of the reason for Nvidia's high prices) is actually AMD.
A bigger part of the problem is that a lot of users continue to support them at these prices, just like Apple! AMD's a distant second in that list.
Posted on Reply
#18
oxrufiioxo
R0H1TNothing to do with short memories, Ampere debuted at great price in fact it was probably the best priced Nvidia gen in the last decade or so. Then yeah Covid happened & prices went awry for everyone ~ but still doesn't take away from the fact that it had great initial pricing. ADA is like Turing 2.0 all over again!


Maybe people with memory span of a goldfish.


A bigger part of the problem is that a lot of users continue to support them at these prices, just like Apple! AMD's a distant second in that list.
Other than the 3080 10G and only at launch was ampere pricing decent every launch that followed that MSRP were just made up. To this day a lot of ampere cards never even hit MSRP. As for this being a Turing like generation while I do see similarities especially when it comes to price the only card in the Turing lineup that offered anything useful at launch was the 2080ti for like a 70% price increase over the 1080ti but even then it was only like a 30% performance increase as bad as the 4080 is it isn't that bad offering like 46% more performance for the same price increase over it's predecessor
also the 4090 can double the performance of the 3090 in RT heavy games for the massive price increase of 100 usd.

Even the 4070 in this article is meant to replace 1070/2070s for them in the current market this card probably isn't too bad. It will likely be a 2x-2.5x performance increase for them.

As far as goldfish memory you see this every generation even Pascal people complained about and that ended up being one of Nvidia's best in hindsight.



AMD is pretty much irrelevant at this point mostly due to them thinking only offering slightly better value cards is going to work for them I know plenty of people who spent more money on 3070/3070ti vs the likely safer options in the 6700XT/6800 due to vram to know the mInd share Nvidia has means they can charge whatever they want.
Posted on Reply
#19
evernessince
oxrufiioxoWhile I don't doubt that a 5090 could be 2x the 4090 I still think anything 5070 and below will be trash especially considering 3nm will be another massive increase in wafer cost.
?? 3nm is is a mere 17.64% more expensive than 5nm. 17K for 5nm and 20K for 3nm. That's half the typical price increase.

The 5090 could be 2x the price but it will be due to greed, not wafer cost.

It's a flawed comparison to imply that wafer costs dictates product pricing to begin with, most of Nvidia's expenses are in it's software and R&D. As Nvidia itself points out, it's a software company.
Posted on Reply
#20
Why_Me
R0H1TPress F for F U to you know who :nutkick:

Without frame gen this thing's a daylight robbery! On par with vanilla(?) 3080 & likely way more expensive? The 3080 12 GB (on deal) went for ~600 USD or lower IIRC.
The 3080 10GB has a MSRP of $700. The 4070 12GB has a MSRP of $600.

$700 - $600 = $100

Posted on Reply
#21
oxrufiioxo
evernessince?? 3nm is is a mere 17.64% more expensive than 5nm. 17K for 5nm and 20K for 3nm. That's half the typical price increase.

The 5090 could be 2x the price but it will be due to greed, not wafer cost.

It's a flawed comparison to imply that wafer costs dictates product pricing to begin with, most of Nvidia's expenses are in it's software and R&D. As Nvidia itself points out, it's a software company.
I mean if the wafers cost more and it uses an 800mm² die what would likely need to happen for a 2x performance increases do you think Nvidia is going to eat that cost lol

Same with anything and even more so the further down the stack.

Even if it's just a 15-20% difference 100% of that is going into the increased cost. The perfomance increases they are likely to get going from 4n to whatever they call the 3nm nvidia customized version of it is going to be much smaller than the terrible samsung 10nm vs the current node meaning unless they work some magic the die sizes are going to go up for any decent performance increases over ada.

Who knows maybe this will only be another 30% at the top generation like Turing was and die sizes will be smaller I guess in two years we will know.
Posted on Reply
#22
evernessince
R0H1TNothing to do with short memories, Ampere debuted at great price in fact it was probably the best priced Nvidia gen in the last decade or so. Then yeah Covid happened & prices went awry for everyone ~ but still doesn't take away from the fact that it had great initial pricing. ADA is like Turing 2.0 all over again!
There was what, a whole 5 second window when the cards launched where you were able to buy at MSRP. I'm not really sure how relevant Nvidia's MSRP was when that's the only time you were able to purchase at your supposed "great price". 99.9% of Ampere cards were sold far above MSRP. It also ignores the fact that cards like the 3080 ti and 3090 Ti had pretty ridiculous MSRP to boot.

The only thing "good" about the pricing of the 3000 series is the way it looked on paper, in effect it was one of the worst GPU generations price wise. On top of that the entire generation was memory crippled (aside for the 3090 and 3060) and power hungry. Transients on higher end cards were also an issue. Imagine paying $1,200 during the pandemic for a 3070 to already have to lower graphics options and be forced to turn off RT due to the tiny VRAM allotment.
Posted on Reply
#23
oxrufiioxo
evernessinceThere was what, a whole 5 second window when the cards launched where you were able to buy at MSRP. I'm not really sure how relevant Nvidia's MSRP was when that's the only time you were able to purchase at your supposed "great price". 99.9% of Ampere cards were sold far above MSRP. It also ignores the fact that cards like the 3080 ti and 3090 Ti had pretty ridiculous MSRP to boot.

The only thing "good" about the pricing of the 3000 series is the way it looked on paper, in effect it was one of the worst GPU generations price wise. On top of that the entire generation was memory crippled (aside for the 3090 and 3060) and power hungry. Transients on higher end cards were also an issue. Imagine paying $1,200 during the pandemic for a 3070 to already have to lower graphics options and be forced to turn off RT due to the tiny VRAM allotment.
I'd be willing to bet even for a lot of 3070 owners this card doesn't look too bad at least they will be able to turn up settings on it in the latest games lol, The downside for them is they likely paid 300-500usd more vs this cards MSRP that it will likely actually hit.

Don't get me wrong I don't like this card the 4070ti or even the 4080 but that has nothing to do with Ampere of the 3 this one has the least crap pricing.
Posted on Reply
#24
rv8000
oxrufiioxoOther than the 3080 10G and only at launch was ampere pricing decent every launch that followed that MSRP were just made up. To this day a lot of ampere cards never even hit MSRP. As for this being a Turing like generation while I do see similarities especially when it comes to price the only card in the Turing lineup that offered anything useful at launch was the 2080ti for like a 70% price increase over the 1080ti but even then it was only like a 30% performance increase as bad as the 4080 is it isn't that bad offering like 46% more performance for the same price increase over it's predecessor
also the 4090 can double the performance of the 3090 in RT heavy games for the massive price increase of 100 usd.

Even the 4070 in this article is meant to replace 1070/2070s for them in the current market this card probably isn't too bad. It will likely be a 2x-2.5x performance increase for them.

As far as goldfish memory you see this every generation even Pascal people complained about and that ended up being one of Nvidia's best in hindsight.



AMD is pretty much irrelevant at this point mostly due to them thinking only offering slightly better value cards is going to work for them I know plenty of people who spent more money on 3070/3070ti vs the likely safer options in the 6700XT/6800 due to vram to know the mInd share Nvidia has means they can charge whatever they want.
The only 4000 series card that has any “value”, and it’s difficult to say when then card is very expensive, is the 4090.

The 4080 downward offer nothing that wasn’t or is already available price to performance wise. I mean unless you want deep learning picture degradation super sampling 3, DLPDSS 3 for short.
Posted on Reply
#25
Why_Me
rv8000The only 4000 series card that has any “value”, and it’s difficult to say when then card is very expensive, is the 4090.

The 4080 downward offer nothing that wasn’t or is already available price to performance wise. I mean unless you want deep learning picture degradation super sampling 3, DLPDSS 3 for short.
RTX 3080 10GB MSRP: $700
RTX 3080 Ti 12GB MSRP: $1200
RTX 4070 Ti 12GB MSRP: $800
RTX 4070 12GB MSRP: $600



Posted on Reply
Add your own comment
Dec 18th, 2024 18:18 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts