RTX 5070 Founders Edition Review & Benchmarks

Better review in text: https://www.techpowerup.com/review/nvidia-geforce-rtx-5070-founders-edition/39.html

Native vs x4 framegen in Cyberpunk only costs 7ms latency, from 47 to 54ms, but 4x the framerate. Other games are shown as well combined with DLSS upscaling. Amd is going to have a hard time if they don't have a match for transformer dlss and multiframegen.

Even just dlss upscaling vs the
7900XT fsr, it trounces it on quality vs quality mode. The 5070 looks much better too, thanks to the Transformer model. That's if you want to ignore frame gen which they comment works extremely well. Regardless, this is an impressive amount of power for $550!
 
GoldenTiger said:
What do you mean?
Click to expand...
Its a 4070 Super with a theoretical $50 price cut.

Other features Nvidia offers on the software side aside, which are indeed rather impressive and pulls even further ahead of AMD's current software offerings, the hardware of the 5070 just sucks. The 12gb sucks. The gen over gen uplift sucks.

Literally one of the absolute worst generations Nvidia has ever released on a gen over gen uplift standpoint, but here some Nvidia marketing to just pump multi frame gen to make up for it.

Meanwhile: missing ROPs, no supply, fake prices, black screens driver issues. Literally all the crap AMD is always accused of having problems with.

This launch is terrible and I hope people don't reward Nvidia for it.
 
Sir Beregond said:
Its a 4070 Super with a theoretical $50 price cut.
Click to expand...
Not really. It has 16 more ROPs which add quite a bit of perf, plus multiframegen which adds a little latency in exchange for a huge amount of motion clarity and smoothness. Did you read the article? It's comparable to a 4070ti, not super.

Sir Beregond said:
missing ROPs, no supply, fake prices, black screens driver issues. Literally all the crap AMD is always accused of having problems with.
Click to expand...
Missing ROPs is a very rare defect and can be handled with a manufacturer rma. No supply? That's true if every gpu launch the past few generations. Fake prices? That's caused by vendors hiking it despite there being sold at msrp previously at launch.

Driver issues are rare on Nvidia compared to amd on the whole, and you can always just go back a version while Nvidia fixes it up if your rig doesn't like them, but they fixed it within a week or two already for most people. Amd on the other hand has issues constantly. Not even comparable.
 

Attachments

  • 1741101781319.png
    1741101781319.png
    353.5 KB · Views: 0
GoldenTiger said:
Not really. It has 16 more ROPs which add quite a bit of perf, plus multiframegen which adds a little latency in exchange for a huge amount of motion clarity and smoothness. Did you read the article? It's comparable to a 4070ti, not super.


Missing ROPs is a very rare defect and can be handled with a manufacturer rma. No supply? That's true if every gpu launch the past few generations. Fake prices? That's caused by vendors hiking it despite there being sold at msrp previously at launch.

Driver issues are rare on Nvidia compared to amd on the whole, and you can always just go back a version while Nvidia fixes it up if your rig doesn't like them, but they fixed it within a week or two already for most people. Amd on the other hand has issues constantly. Not even comparable.
Click to expand...
If you think the supply problems and fake prices are the exact same as previous launches, I dont think you've been paying attention. Literal definition of a paper launch so far with far lower launch supply and restocks vs previous generations. So no, it's not the "exact same". This is also after Jensen claimed to be pumping out Blackwell massively at his CES keynote. Really? Where are they then?

I did say the software stack is impressive so please re-read what I said. I said the hardware sucks. By previous gens standards this thing is a 5060 at best. The 4070 at least matched a 3080. The 3070 at least actually matched a 2080 Ti. No not with the caveat of "multi frame gen". Actual.

Where's that 5070 = 4090 at?

Anyway my point with all the launch issues is everyone shits on AMD for theirs but here comes the Nvidia fans to make excuses. How do you even let cards with missing ROPs go out.
 
Last edited:
GoldenTiger said:
Not really.
Click to expand...
pretty close, +~7.5% at 4k or so, using 10% more power to do it, it is a 4070 super refresh with a price cut. They often match.

Will have to see one day when HUB will compare dlss transformer performance among a suite of games to see if it is a little bit more than that, but that pretty much it, the upcoming full die version 5070 super that around the 7900xt will be what this should have been at a minimum, it is a small 263mm cutting it down for the 5070 is getting quite greedy.
 
Last edited:
Absalom said:
GN seems to disagree.
Click to expand...
That's because they're starting with too low a framerate... It's a scenario no one would try, starting at 25-30fps, in actual usage. EDIT: Is that at 1440p? Looks like 4k which this card isn't meant for.
 
Last edited:
GoldenTiger said:
That's because they're starting with too low a framerate... It's a scenario no one would try, starting at 25-30fps, in actual usage.
Click to expand...
Do you think the average gamer that buys into this product will know what their rendered fps actually is? You're speaking for yourself, not the masses. The reality is the majority are going to enable this and be oblivious to what's going on. Unless someone educates them.

And then it's just a matter of time when FG becomes defaulted on for every game, thus exacerbating things.

Yes, this type of investigation matters.
 
Absalom said:
Do you think the average gamer that buys into this product will know what their rendered fps actually is? You're speaking for yourself, not the masses. The reality is the majority are going to enable this and be oblivious to what's going on. Unless someone educates them.

And then it's just a matter of time when FG becomes defaulted on for every game, thus exacerbating things.

Yes, this type of investigation matters.
Click to expand...
If the average gamer wouldn't notice it at all then it works well. Which it does, here's some actual data from a much better source rather than a sensational youtuber who can't even be bothered to list settings on a chart: https://www.techpowerup.com/review/nvidia-geforce-rtx-5070-founders-edition/39.html
 
GoldenTiger said:
Better review in text: https://www.techpowerup.com/review/nvidia-geforce-rtx-5070-founders-edition/39.html

Native vs x4 framegen in Cyberpunk only costs 7ms latency, from 47 to 54ms, but 4x the framerate. Other games are shown as well combined with DLSS upscaling. Amd is going to have a hard time if they don't have a match for transformer dlss and multiframegen.

Even just dlss upscaling vs the
7900XT fsr, it trounces it on quality vs quality mode. The 5070 looks much better too, thanks to the Transformer model. That's if you want to ignore frame gen which they comment works extremely well. Regardless, this is an impressive amount of power for $550!
Click to expand...
No on is clamoring for 4 generated frames of oddness. Feel free to buy em up though. The more you buy the more you will save.
Also lets stop playing their game and calling it frame generation. Its frame smoothing.
 
ChadD said:
No on is clamoring for 4 generated frames of oddness. Feel free to buy em up though. The more you buy the more you will save.
Also lets stop playing their game and calling it frame generation. Its frame smoothing.
Click to expand...
Gamers Nexus did a review on DLSS4 and FG and its most honest review of the tech so far. Despite what nVidia's marketing would have you believe DLSS4 is NOT always better than 3 and FG is well FG.

View: https://www.youtube.com/watch?v=Nh1FHR9fkJk
 

View: https://www.youtube.com/watch?v=qPGDVh_cQb0

Some interesting takes with HUB. Looking at the FE, and Gigabyte, and Asus models. The FE seems to report incorrect fan speeds? The Gigabye OC model seems to pull 20 watts LESS and is under clocked. Some oddness in this model.

End of the day though 5070 is a 4070 super - $50 MSRP. If MSRP means anything at all. You could have bought this card for probably less 2 years ago. No doubt you could have bought it for less 6 months ago.
 
ChadD said:
End of the day though 5070 is a 4070 super - $50 MSRP. If MSRP means anything at all. You could have bought this card for probably less 2 years ago.
Click to expand...
The generational gains being nonexistent means that a 5070 being a 4070 Super at -50 MSRP is still pretty much garbage. Because most of the time the new gen X070 should beat the old gen X080 or even X090, especially in the Ti case. Which neither of them do lol. You're essentially giving Nvidia most of the money for the exact same thing. Heck, in the 5070 Ti's case, you're giving them almost all of the price of a 4080 Super at street prices.... for a 4080 Super.

Grats you can buy our gen old tech, with PhysX taken out entirely, in exchange for our AI stuff... out of which only 4x Framegen is new.
 
ChadD said:

View: https://www.youtube.com/watch?v=qPGDVh_cQb0

Some interesting takes with HUB. Looking at the FE, and Gigabyte, and Asus models. The FE seems to report incorrect fan speeds? The Gigabye OC model seems to pull 20 watts LESS and is under clocked. Some oddness in this model.

End of the day though 5070 is a 4070 super - $50 MSRP. If MSRP means anything at all. You could have bought this card for probably less 2 years ago. No doubt you could have bought it for less 6 months ago.
Click to expand...

Yeah I know we have to use msrp as a baseline for comparison but it almost is seeming useless in lots of cases because getting one for msrp might be ridiculously difficult.
 
StoleMyOwnCar said:
The generational gains being nonexistent means that a 5070 being a 4070 Super at -50 MSRP is still pretty much garbage. Because most of the time the new gen X070 should beat the old gen X080 or even X090, especially in the Ti case. Which neither of them do lol. You're essentially giving Nvidia most of the money for the exact same thing. Heck, in the 5070 Ti's case, you're giving them almost all of the price of a 4080 Super at street prices.... for a 4080 Super.

Grats you can buy our gen old tech, with PhysX taken out entirely, in exchange for our AI stuff... out of which only 4x Framegen is new.
Click to expand...
I would argue its worse as this gen sucks more power. They regressed with blackwell and brought it to par by feeding it more juice. On the upside for the 5070 it won't melt cables.
 

View: https://youtu.be/UxvmfEZkvc8?si=52bfdyrx2TpiTwlv&t=1382

Liked by time to his I guess funny conclusion. Whatever you think of Jay, and I think he has a pretty weak testing regiment in the past. He can be funny at times. Tomorrow should be fun. Very disruptive launch tomorrow apparently... according to Jay the 5070 is Trent captain of the polo team. Apparently tomorrow the new kid comes to school and he doesn't give a shit about authority.

Jensen and Lisa in an 80s John Hughes movie. I need AI to make me that now. lol
 
ChadD said:
I would argue its worse as this gen sucks more power. They regressed with blackwell and brought it to par by feeding it more juice. On the upside for the 5070 it won't melt cables.
Click to expand...

This is just what lack of competition does. AMD isn't providing them enough competition in the consumer GPU space, which leaves Nvidia focusing almost entirely on their datacenter side, and all that their new GPUs are getting is trickle down from that. They don't give a shit because people gobble it up anyway.

polonyc2 said:
seems like every 50 series card (outside of the 5090) should have been one tier lower...the 5080 should have been the 5070 and the 5070 should have been the 5060
Click to expand...
Honestly, no, even the 5090 could have been the 5080. I think historically usually the X080 stomps on the previous gen's strongest card. So seeing X090 go to (X+1)080 with a performance uplift isn't really unheard of.
1741107946171.png


3090 to 4080 Super was about a 28% uplift, which is still kind of in the ballpark of this gen's 4090 to 5090. Likewise, there was no 2090, but 2080 Ti to 3080 was an uplift of 31%. Let's not sugercoat just how bad this gen is.

Hell even worse because the 5090's uplift is almost entirely due to just having more chip, fueled by more power. There are no gains.
 
In other news 9070xt’s will be absolutely sold out now.

I’d like to say this is another fumble by NV, but actually it was expected. And it will STILL be sold out. People talk crap all day right now about them, yet they have 90% of the market. Pretty ironic.
 
GoldenTiger said:
If the average gamer wouldn't notice it at all then it works well. Which it does, here's some actual data from a much better source rather than a sensational youtuber who can't even be bothered to list settings on a chart: https://www.techpowerup.com/review/nvidia-geforce-rtx-5070-founders-edition/39.html
Click to expand...
Let's take a look at the different games at 1440p. I picked some at random just to see what we were looking at. Some the 5070 results are around 4070 Super levels. Others its more like 4070 Ti levels. Cyberpunk and Hogwarts it seems to do especially well.

We're not exactly talking a ground shattering average FPS spread here between any of these cards. So all Nvidia has effectively done here is re-release the same performance you could already buy in previous gen, slap 3-4X frame generation on as an exclusive 50-series feature to make up the actually piss poor generational uplift. As for the rest of the software stack, it has no specific bearing on the 5070 as a product since non-50-series cards can also make use of new DLSS4 and transformer model upscaling. What this amounts to is the single worst generation I have ever seen Nvidia release in say the last 10-15 years. I don't know how anyone can look at the 50-series stack below the 5090 and not say its entirely phoned in. A 5080 should have been spec'd to have 12-15k shaders and 24GB. Both the 5070 Ti and 5070 should be 16GB cards with the 5070 Ti spec'd closer to 4080 shaders and 5070 spec'd closer to 4070 Ti specs on shaders and also 16GB.

This is honestly one of the most pathetic showings on a hardware level Nvidia has done in a long time. Even back in Kepler where they dropped the GTX 680 to mid-range silicon, it still beat the GTX 580 by 25-30%. What the hell is this crap?

Alan Wake II:
1741107265496.png


Black Myth: Wukong:

1741107201049.png


Cyberpunk 2077:
1741107320981.png


Hogwarts Legacy:
1741107359440.png


Horizon Forbidden West:
1741107403688.png


Ratchet and Clank:
1741107488256.png


STALKER 2:
1741107460483.png


The Last of Us Pt 1:
1741107592803.png
 

Attachments

  • 1741107148780.png
    1741107148780.png
    41.7 KB · Views: 0
Brackle said:
Oh yeah easily. Don't forget they tried to sell a 4080 12gb model.....Which turned out to be the 4070 on release date lol.

Good ole Nvidia
Click to expand...
That was our fault. Rather, when it was announced there were going to be two 4080 models, presumably differing only on VRAM capacity, we put on our OUTRAGED hats and made drama out of it. So, Nvidia pulled every 12GB 4080 from the warehouse prior to the launch, waited a bit, and relaunched it as the 4070 Ti. The rest is history.

In a way, we fooled ourselves on that one.
 
GoldenTiger said:
It's called marketing ;). But one would have a strong argument when using framegen on both cards, yes. Most new demanding titles have dlss and frame gen, so...
Click to expand...
You can only smear so much lipstick on a pig...

And Nvidia tried to cover the whole pig with lipstick marketing and pretend it has the performance chops of a 4090.

At best, it is just ever so slightly better than a 4070 super, but just under that of a 4070 Ti.

In other words, the 5070 pretty much sucks compared to its last gen counterpart.
 
You must log in or register to reply here.
Back
Top