RTX 4070 Ti Super Review: Asus & MSI Results

Summary

In this PCWorld review, Adam and Will compare the GeForce RTX 4070 Ti Super against the original 4070 Ti and Radeon cards. They evaluate gaming performance at 1440p and 4K, ray tracing, content creation workloads, power consumption, and thermals. The 4070 Ti Super boasts increased VRAM and memory bandwidth, but the performance uplift over the non-super variant is subtle. AMD's Radeon cards show potential in non-ray tracing games, but Nvidia leads in ray tracing capabilities.

Takeaways

• 📕 The video reviews the GeForce RTX 40 Super Series, focusing on the 4070 Ti Super against the original 4070 Ti and Radeon cards.
• 📌 The benchmarks include gaming performance in 1440p and 4K resolutions, raytracing, content creation workloads, power draw, and thermals.
• 💵 Both the 4070 and 4070 Ti Super target high-end 1440p gaming with some 4K gaming capabilities and share the same 285-watt TGP and starting price of $800.
• 💲 The 4070 Ti Super upgrades from an AD1 to an AD2 die, increases core count from 7680 to 8448, and significantly improves memory from 12GB to 16GB of GDDR6X with a 256-bit memory bus.
• 💳 The video tests various third-party cards like the Gigabyte GeForce RTX 4070 Ti, ASUS TUF Gaming GeForce RTX 4070 Ti Super, and MSI GeForce RTX 4070 Ti Super Ventus.
• 💱 The benchmarks show the 4070 Ti Super performs closely to the non-Super variant, with AMD cards showing value in certain price points.
• 💡 In 4K gaming, the added VRAM and memory bandwidth of the 4070 Ti Super do not show a significant uplift in raster performance as expected.
• 💥 Raytracing performance sees Nvidia maintaining an advantage over AMD, with the 4070 Ti Super showing playable frame rates in games like Cyberpunk 2077.
• 💫 The power draw and thermals vary by card design, with larger cards generally offering better cooling and lower temperatures.
• 💪 The video suggests that for video editing tasks, more VRAM is beneficial as seen in DaVinci Resolve scores, and the 4070 Ti Super shows a performance increase.

Q & A

What is the main focus of the video review?

The main focus of the video review is to compare the GeForce RTX 4070 Ti Super with the original 4070 Ti and other Radeon cards, evaluating their performance in gaming benchmarks at 1440p and 4K resolutions, as well as their ray tracing capabilities, power draw, thermals, and overall impressions.

What are the key differences between the RTX 4070 Ti and the RTX 4070 Ti Super?

The key differences between the RTX 4070 Ti and the RTX 4070 Ti Super include a move from an AD1 to an AD104 die, an increase in core count from 7680 to 8448, and a significant upgrade in memory from 12GB of DDR6X to 16GB of DDR6X with a bus width increase from 192-bit to 256-bit.

How do the prices of the RTX 4070 Ti Super and the original RTX 4070 Ti compare?

Both the RTX 4070 Ti Super and the original RTX 4070 Ti start at $800, with the 4070 Ti Super not being sold at a price increase.

What gaming benchmarks were conducted at 1440p and 4K resolutions?

Gaming benchmarks at 1440p and 4K resolutions included tests with Cyberpunk 2077, Returnal, Shadow of the Tomb Raider, Rainbow Six Siege, F1 2022, Call of Duty, and Watch Dogs Legion.

How did the RTX 4070 Ti Super perform against AMD's Radeon cards in the benchmarks?

The RTX 4070 Ti Super generally performed well against AMD's Radeon cards, falling between the RX 7800 XT and RX 7900 XT in most benchmarks, and sometimes surpassing them, especially in games known to run better on NVIDIA systems.

What was the reviewer's opinion on the importance of VRAM and memory bandwidth for gaming at 4K resolution?

The reviewer noted that while the addition of extra VRAM and a larger bus width in the RTX 4070 Ti Super didn't show a dramatic improvement in 4K gaming benchmarks, it could be beneficial for future-proofing and for ray tracing performance.

What was the reviewer's approach to testing ray tracing performance?

The reviewer tested ray tracing performance by running games like Cyberpunk 2077 and Returnal with ray tracing settings that allowed for playable frame rates, using FSR upscaling set to balanced for an even comparison across different cards.

How did the power draw and thermals of the different cards vary?

The power draw varied among the tested cards with the ASUS TUF Gaming GeForce RTX 4070 Ti Super being close to the rated 285W, while the MSI GeForce RTX 4070 Ti Super Ventus was slightly lower. The Gigabyte GeForce RTX 4070 Ti had a lower power draw, which might have contributed to it being slightly slower in some benchmarks. Thermals were generally better on larger cards with more extensive cooling solutions.

What preliminary findings did the reviewer mention regarding video editing performance?

The reviewer teased preliminary findings that suggested video editing applications like Premiere Pro and DaVinci Resolve benefited from the increased VRAM of the RTX 4070 Ti Super, showing improved performance in those applications compared to cards with less VRAM.

What advice did the reviewer give on choosing a GPU based on the video content?

The reviewer suggested that one should not spend a lot of extra money on highly overclocked cards and instead consider the base price models. They also mentioned that for those with smaller cases, the size of the card and its cooling capabilities are important factors to consider.

Outlines

00:00

🚀 GeForce RTX 4070 Ti Super Review Overview

Adam and Will are reviewing NVIDIA's GeForce RTX 4070 Ti Super, comparing it with the original 4070 Ti and Radeon cards. They plan to conduct gaming benchmarks in 1440p and 4K resolutions, assess raytracing capabilities, content creation workloads, power consumption, thermal performance, and overall impressions. The 4070 Ti Super is positioned as a high-end 1440p gaming card with a starting price of $800, similar to the original 4070 Ti. The main differences include a move to an AD1 to 3 die, a core count bump from 7680 to 8448, and a significant memory upgrade from 12GB DDR6 to 16GB with a 256-bit memory bus, addressing previous concerns about future-proofing for 4K gaming.

05:01

📊 Synthetic Benchmarks and 1440p Gaming Performance

The hosts discuss synthetic benchmarks, highlighting the 3D Mark Speedway test which emphasizes CPU and future-looking feature workloads. They note that the 4070 Ti Super variants perform similarly to each other, with a slight advantage over the non-super variant and the AMD Radeon 7900 XT. The conversation shifts to actual gaming performance at 1440p, with tests on games like Cyberpunk 2077, Returnal, Shadow of the Tomb Raider, Rainbow Six Siege, and F1 2022. They observe that the NVIDIA cards show a consistent performance, with AMD cards falling slightly behind, especially in terms of 1% low frame rates indicating potential micro-stuttering.

10:05

🎮 4K Gaming and Raytracing Benchmarks

The discussion continues with 4K gaming results, where the hosts express surprise that the additional VRAM and memory bandwidth in the 4070 Ti Super do not show a significant uplift in performance. They test games like Cyberpunk 2077, Returnal, and others, noting that while there is a slight improvement, it's not as pronounced as expected. They also touch on the potential for shader limitations rather than memory bandwidth limitations in modern games. The conversation includes a look at raytracing performance, with the expectation that the increased specs of the 4070 Ti Super might make a difference, contrasting this with the raster performance which showed minimal differences.

15:06

🔌 Power Consumption and Thermals Analysis

The hosts analyze power consumption and thermal performance, reporting variations in power draw among the tested GPUs. They find that the rated spec for all cards should be 285W, but the actual draw varies, with some cards performing closer to the spec and others falling short. They also discuss thermals, noting that the smaller the card, the warmer it runs, which is expected due to less cooling capability. They mention that the non-super variant card runs cooler, possibly due to a quieter mode or less aggressive overclocking.

20:06

🎥 Video Editing Performance and Final Thoughts

Adam shares preliminary results from video editing tests in software like Premiere Pro and DaVinci Resolve, indicating that the additional VRAM in the 4070 Ti Super shows a more pronounced improvement in these applications compared to gaming. He suggests that for video editing tasks, the extra memory could be beneficial. The hosts discuss the trade-offs of card size, cooling, and performance, recommending larger cards for better cooling and quieter operation. They advise against spending extra on highly overclocked cards, suggesting that the base model offers good value. They also note the lack of a dual BIOS on the massive version of the card, which could be a consideration for some users.

25:07

📝 Wrapping Up the GPU Review

In the final paragraph, the hosts reflect on the process of GPU testing, acknowledging the time-consuming nature of running benchmarks. They express a desire to produce more content like this in the future, particularly regarding video editing performance. They encourage viewers to subscribe for more in-depth reviews and conclude the discussion on GPU testing, highlighting the effort involved in creating such comprehensive reviews.

Keywords

💡GeForce RTX 40 Super Series

The GeForce RTX 40 Super Series refers to a line of high-performance graphics processing units (GPUs) designed by NVIDIA for gaming and other graphics-intensive applications. In the context of the video, the presenters are reviewing the 4070 Ti Super model, comparing it with the original 4070 Ti and other Radeon cards. This series is significant as it represents the latest in GPU technology, offering enhanced gaming experiences through improved performance and features like ray tracing.

💡4070 Ti Super

The 4070 Ti Super is a specific model of GPU within the GeForce RTX 40 Super Series. It is positioned as a high-end GPU for premium 1440p gaming and some 4K gaming. The video discusses its performance in various benchmarks, comparing it with the original 4070 Ti and other GPUs. The 'Super' moniker indicates an upgraded version of the original 4070 Ti, suggesting improved specifications and performance.

💡Gaming Benchmarks

Gaming benchmarks are a measure of a GPU's performance in running games. These are typically conducted at various resolutions like 1440p and 4K to see how well the GPU can handle different graphics loads. In the video, the presenters run gaming benchmarks to evaluate the 4070 Ti Super against other GPUs, which helps viewers understand the card's real-world gaming performance.

💡Ray Tracing

Ray tracing is a rendering technique that simulates the physical behavior of light to create highly realistic lighting, shadows, and reflections in games and other graphics applications. It is a significant feature of modern GPUs like those in the GeForce RTX series. The video includes a discussion on ray tracing performance, which is crucial for gamers looking for realistic visuals in supported games.

💡VRAM

Video Random Access Memory (VRAM) refers to the memory used by a GPU to store image data. The amount of VRAM can impact the performance of a GPU, especially at higher resolutions and in memory-intensive applications. The video mentions the 4070 Ti Super's VRAM upgrade from 12GB to 16GB, which is significant for future-proofing the GPU for more demanding games and applications.

💡Memory Bandwidth

Memory bandwidth is the rate at which a GPU can read from or write to its memory. Higher memory bandwidth can lead to better performance, especially in scenarios where large amounts of data need to be processed quickly, such as in 4K gaming or certain professional applications. The video discusses the improvement in memory bandwidth from the original 4070 Ti to the 4070 Ti Super, which is a key selling point for the new model.

💡TGP

TGP stands for Total Graphics Power and refers to the maximum amount of power a GPU can draw. A higher TGP can indicate better performance potential but also higher power consumption. The script mentions that both the 4070 and 4070 Ti Super have the same 285-watt TGP, suggesting similar power consumption despite the performance improvements in the Super model.

💡1440p

1440p is a display resolution of 2560 x 1440 pixels, which is a step up from 1080p and offers a sharper image. In the video, 1440p is one of the resolutions used for gaming benchmarks to test how the GPUs perform at this increasingly popular resolution for gaming monitors.

💡4K

4K, or Ultra HD, refers to a display resolution of approximately 3840 x 2160 pixels. It is one of the highest resolutions used in consumer displays and represents a significant graphics challenge, especially for real-time rendering in games. The video includes 4K gaming benchmarks to assess the GPU's capability to handle high-resolution gaming.

💡Power Draw

Power draw refers to the amount of electrical power a device consumes. In the context of GPUs, power draw is an important factor as it can affect performance, efficiency, and thermals. The video discusses power draw measurements taken during testing, which can help users understand the real-world power consumption of the GPUs being reviewed.

💡Thermals

Thermals in the context of GPUs refers to the heat generated by the GPU and how effectively it is managed or dissipated. Good thermal performance is crucial for maintaining stable performance and avoiding thermal throttling. The video includes thermal measurements and discusses how different GPU models manage heat, which is important for users concerned about reliability and longevity.

Highlights

Reviewing the GeForce RTX 40 Super Series, specifically the 4070 Ti Super.

Comparison against the original 4070 Ti and Radeon cards in gaming benchmarks at 1440p and 4K resolutions.

Examining raytracing, content creation workloads, power draw, and thermals.

The 4070 Ti Super retains the same 285 watt TGP and $800 price point as the original 4070 Ti.

The 4070 Ti Super transitions from an AD1 to an AD104 die, aligning more with the 4080 than the original 4070.

A slight core count bump from 7680 to 8448 in the 4070 Ti Super.

Significant memory upgrade in the 4070 Ti Super from 12GB to 16GB of GDDR6X and a 256-bit memory bus.

Benchmarks show minimal differences in performance between the 4070 Ti and 4070 Ti Super in 1440p gaming.

AMD Radeon cards show a performance deficit in synthetic benchmarks.

The 4070 Ti Super shows a lead over the non-super variant in certain benchmarks.

In 4K gaming, the 4070 Ti Super does not show a significant uplift in performance due to increased VRAM.

Power draw varies among different card models, with some showing higher efficiency.

Thermals are influenced by card size and cooling solutions, with larger cards generally performing better.

DaVinci Resolve benchmarks show a preference for more VRAM.

The importance of considering card size, especially for those with smaller cases.

Recommendation against spending extra for highly overclocked cards, suggesting value in the base model.

The absence of a dual BIOS in the large version of the card.

Transcripts

Hey PCWorld fans.

Adam here with Will and we are currently reviewing

the suite of GeForce RTX 40 Super Series.

On tap today we are going to look at the 4070 Ti Super.

We are comparing it against the original 4070 Ti

as well as a couple Radeon cards.

We'll get to that.

We're going to do gaming benchmarks in 1440p and also 4K.

We also have a little bit of raytracing, a little bit of content

creation workloads, some power draw, some thermals,

you know, some some just general overall feelings on these cards.

We want to know how they stack up, how they stack up.

So stay tuned for that.

Let's dig in.

All right.

So to get started, let's go over the specs and it kind of falls in two buckets.

The ones where it's the same and ones where it's different

between the 47, 80 and the 4780 super.

So in terms of sameness, they're both targeting

high end 1440 gaming, some low end 4K gaming, but really the NVIDIA

kind of positions this as a 1440 super premium,

1440 super premium 1440p

they both have the same 285 watt TGP so there's no power increase here.

And then the last thing that's the same is the price.

It comes in or starts at $800.

The 4789 super is not going to be sold anymore in the trash.

Throw it in the trash, but it begins at $800.

So, you know, depending on the model these.

But what's different,

the main one is that it moves from an ad 1 to 3 die to an 81 to 4 die.

So really this is more like an imperfect 40, 80, rather than the perfect.

So it's the same die is a 48.

It's just they're they're Yes.

Ones that don't spec or ones

they're clocking down or whatever or what we're seeing here. Correct.

Another major change, it does get a little bit of core count

bump from 7680 to 8448.

So slight bump.

That's itsy bitsy bump, man.

That's like we're going to see that.

But the bars are going to be real

close to each other, I think is my prediction there.

But really, the big thing people complain about when the 4780 launched

is that it only came with 12 gigabytes of DDR six x vram

and the bus width was only 192 bit,

which meant that while it was okay for 1440p and that's more where in video

is aiming that looking forward to the future or trying to step up to 4K

but it just wasn't really well spent there.

So it moves from that now to 16 gigabytes of G or six X and a 256

bit memory, but so a substantial memory bandwidth improvement.

Yes, yes.

This is more just getting the 4780 and kind of future proofing a little better.

How that bears in benchmarks.

We will see.

We will see.

Let's talk about the actual cards I have here.

Testing against the vanilla 4780 that I have Here is a gigabyte

GeForce RTX, 4070 type Morosi comes in at a $900 MSRP,

but can be found for actually around 850 as of this recording right now.

But once again, it's going away so you won't be able to target it.

Big beefy card as a big beefy card.

While Infinite does not did not produce a founders edition of the 4780

for the initial review, I got sent this Asus

TUF Gaming GeForce RTX 4072 Super C edition.

This clocks in at $880.

So I'm not not that is not $800 out of nothing in dollars but closer to $800.

The other card that I have in to test

is the MSI GeForce RTX 4072, Super Ventus

two x ac, and this one clocks in at $820.

So close. Not quite there, but close.

So yeah.

So on the Arabian side, comparing it kind of we got two cards

that kind of sit at the different ranges compared to these TI variants.

This one is the as rock radian or x 7800 x t Challenger AC edition.

This comes at $500, so quite a bit less.

That's a much cheaper card out.

Yeah, about $300 cheaper.

But on the higher end we have the sapphire radial on our x 7900 XT.

This when it was released was $900.

But because of the the super refresh,

the new price on this GPU is $750.

Well, I should say I'm sorry the 7900 XT starts at 750.

I actually can't find this edition anymore.

Available for sale.

For sale anymore.

So we're talking about still a little bit cheaper than the the the ti.

I guess it's an interesting spot, right?

Like an $800 video card.

I expect to kind of light my world on fire. Yeah.

And I'm really into to see what the benchmarks play.

Yeah well let's let's dig into them and let's start with the synthetic

benchmark, the one synthetic benchmarks that I have for this.

This is 3D Mark Speedway.

This is a very intensive de X12 ultimate benchmark.

It is meant to really hammer a CPU, especially for arty

and future looking feature workloads and you know, you can see here

the two 2070 tie super cards are neck and neck with each other

and you get a healthy bump over the non super variant of the 4070

TI which already had a lead over the 7900 XT.

Yeah. So it's the same thing

we saw with the 4070 super reviews where AMD has a little bit of a deficit

when when you're doing like heavy, heavy compute load stuff like that.

This is a dated benchmark.

So I'm always a little skeptical.

It's a synthetic benchmark.

I want to throw it out there, you know, and it is good to measure it.

People love 3D, Mark 3D. Mark is cool.

It's it's score.

But let's switch over to actual gaming first.

We're going to go into 1440p, then we're going to do 4K,

then we're going to do Arti and upscaling.

Okay, so all of the 1440p and 4K gaming benchmarks that we're going to

look at first are all just pure raster.

Obviously there's a lot of cool arty features and upscaling features

that you can turn on, but you know, to get a baseline feel for what

these cards can do, pure registers is where we're going to focus on.

So first up is Cyberpunk 2077,

which definitely pushes all these GPUs hard.

But looking specifically at the these 47

TI versions, we're not too far off from each other, though.

We know the gap from 96 to 104 frames

a second is, is, you know, it's less than 10%.

We're not going to it's not a huge gap Bigger bar better though,

bigger bar better.

So technically the 4070 super is quicker.

But how does that stack up against AMD on these?

It actually falls in between the 570 800.

That's true.

And the $750.70 900 XT that bigger bar

is cheaper than the than the shorter bar though that bigger bar is cheaper.

So bigger bar better on the AMD side at least when it comes to no.

Yeah no RTX. So that's just one game.

Well it's breezed through a couple more here.

A returnal at 1440p

Wow that's once again the same kind of thing where

all the 7080 variants are, you know,

within spitting distance and it falls right between those two AMD cards.

Yeah.

But again, somebody on their axes looking pretty good right now to me

it is looking pretty good.

Well, let's switch over to Shadow, the Tomb Raider, which is actually

known to run a little bit better on NVIDIA systems compared to AMD systems.

And here the 4072 variants

all stack up right there with the 7900 XT.

Yeah.

Let's move over to Rainbow six Siege, which is a

really high refresh rate e-sports game.

So you want a lot of frames and guess what the the actual non super variant

turned in a little bit better of a score than one of the the Thai variant.

So that to me says you're getting

to the point where you're approaching CPU bound probably and it's maybe

what the what the benchmarks actually testing may be a little questionable.

Yeah, exactly.

But they do have a nice little lead over AMD in this one.

So yeah, let's move over to F1 2022 which also,

you know, you want really high frame rates for this

and once again, we're not seeing too much of a difference except for here.

That's 7800 XD is coming a little too close for comfort

you know $500 well and your 7900 XT is a guess off the charts.

Yeah, off the charts.

I will say that I actually did testing on this ASUS card.

This is the one that was set for review.

Later on I got this message card in and when I went to go back and run

this benchmark, there was an update for Call of Duty and it broke the numbers.

So I actually wasn't able to run the benchmark on here.

The perils of benchmarking live service games.

Yeah, but guess what?

It's the same kind of story.

It's it's right there with the non super version.

And once again here we go the that 70 $500.78

166 is very, very close

and that's that's an average frame rate if we look at the 1% lows though

AMD does pull you know, fall back a little further.

You know, we've kind of seen this before where amd the difference between average

frame rate and 1% lows is going to be a little tighter on NVIDIA compared to AMD.

Yeah, and the 1% lows again are measuring like your micro stutter.

So the the slight hitches that you feel as you're moving your mouse from side

to side, it's the average average frame

rate of the 1% slowest frames that are rendered.

And it's it's a really good indicator

of like I said, micro stutter and kind of Hetch Hetchy performance.

How smooth the game. Yeah.

The kind of stuff that really grabs you would have happened. Yeah.

I've got to say, I do worry when I look at a 7200 x with 186 on the

on the average and 104 on the 1% slo That's, that's a worrisome delta for me.

Yeah. Yeah.

Once again this is just the settings, It is a life service game.

It gets updated.

Obviously I have been having some problems with it.

So that's, that's why we test multiple games.

Multiple resolutions. Yes.

The last one for 1440p is Watch Dogs Legion.

And we're seeing the same kind of thing here.

Everywhere.

Everyone's within spitting distance and looking at the 1% lows on this one.

It's that same kind of deal where the the the delta between the AMD averages

and 1% lows is wider than on the NVIDIA GPUs.

It's a little bit

less pronounced on this one though, which is which is a good sign, definitely.

So that's 1440p and that's where these cards generally are meant to play.

But the big question is does adding that extra VRAM and adding

a bigger bus width actually help in 4K gaming

and that that was the thing everyone screamed about.

So now let's look at 4K results.

First up is Cyberpunk 2077.

These look exactly the same at par.

Okay, well, maybe.

Maybe we moved returnal same exact. Yeah.

I mean, hey, bigger bar better.

That's a you know, that is a bigger, I don't know if I care $800

worth of shadow of the tomb Raider.

You know, a little bit more pronounced. Yeah.

I will say that the Nvidia cards are definitely pulling away

from the AMD cards here, you know, a little bit more than when they were.

14:40 p.m..

Rainbow six siege once again, bigger bar better on the super variants,

but yeah, it's F1 2022.

Same kind of deal.

I'll be honest, I'm pretty surprised that we're seeing at the high resolution

especially we're not seeing an impact from the bigger memory bandwidth.

I would expect more RAM, maybe less impactful.

The bigger memory bandwidth.

I would expect to show a little bit more perfect game here, right? Yeah.

So we've got a couple more games, watch Dogs, Legion, same kind of thing.

I mean, technically it is a little bit better.

Call of Duty, Modern Warfare three,

you know, a slight step up in the averages and the 1% loss.

So nothing there.

It's it's really interesting to me.

I think it's a testament

to how much shader work is happening in games, even without ray tracing now

that we're actually seeing the impact of the cores.

Like maybe maybe we're shader limited on these games

more than memory bandwidth Limited.

Yeah.

So I mean, I got to say that's, that's all the, the benchmarks for the

the the rest rise gaming that I did you know and I think

one of the things to think about is I was the only one doing the benchmarks.

Right. It took,

you know, about three quarters to a day each of these GPUs to do the testing.

So I had to limit my testing and

and kind of, you know, see what I could get done in that time.

So I think there are situations I'm sure there are games that are

you are going to find better performance in very limited situations.

Maybe we're just not seeing that here.

Yeah, but also on the same token, in the future,

the demands on VRAM is probably just going to go up, right?

So I mean, it's not going to happen.

It never hurts to have more RAM and never has to have more memory bandwidth.

I think we've reached maybe the limit of what

we're going to expect to see with raster stuff, really, like just just

until we see way faster memory DDR seven or eight, I don't know.

Yeah, yeah.

So, you know, while these benchmarks are not showing

like the World on Fire 4k performance uplifts because of the

the VRAM changes, it doesn't mean there aren't any changes.

It just means it's not as pronounced as as we kind of expected, tell you the truth.

But yeah, in Raytracing, that's another story.

That's what I want to know about.

So now let's look at a couple of benchmarks with art,

because maybe we'll see some changes there.

I will say that the way that I ran the benchmarks here is with Raytracing

on, but also some sort of upscaling on, because I feel like for the most part

that if you're running art features, you're probably going to be want

using some upscaling right And while cyberpunk with path

tracing cranked all the way up is basically the bend

the high water benchmark for ray tracing right now on any platform

it's not representative.

It's like we playable on

this kind of tier card at the resolutions we're talking about.

So I mean, unless you really fiddle with the settings and kind of,

you know, it depends on how you want to kind of play the game.

But I mean, I guess I guess what I'm saying is you

you picked the ray tracing settings that are gave you a playable frame rate.

Exactly at the level of machines you're talking about.

So when it came to upscaling, I had to

look and say, okay, well, what what kind of upscaling we're going to use.

Obviously, NVIDIA has deal sets available to it.

You know, you can get some good gains, good quality out of it.

I needed kind of something to play a little bit more of an even playing field.

I didn't want to run deals

on the Nvidia cards, but only, you know, FSR on the AMD cards.

So I settled with FSR,

which is is still, you know, quite good, but it is vendor agnostic.

So all of these are set with FSR set to balanced

to keep it an even playing field.

Once again,

you might get more performance, you might get better quality out of deals,

but the way that I benchmarked, I just want to keep it.

You want to be able to compare apples to apples and oranges To oranges. Exactly.

Yeah, exactly.

So let's dive into Cyberpunk 2077.

This is with the RTM medium preset.

So not full path tracing, not not going all the way to the wall with it.

FSR set to balance a 1440p that means that renders a native

resolution of 1506 at by 847 everyone's favorite resolution, Yes.

Yeah. So about 720p but not quite 1083

And if we look at the results, you know,

the non super variant was actually not doing too bad to begin with.

It was I mean this is indicated indicative to me

of the relatively paltry increase in Q2 cores.

Right?

We saw a relatively small increase in Q2 cause

those are what does the Raytracing math, it turns out.

Yeah, but I thought Raytracing was a little bit more memory intensive as well.

So. Well, it depends. It's whichever is the slowest thing.

It's going to be the most. Okay. Yeah, right. Yeah.

So I will say though, if you compare it to the AMD cards, AMD and Raytracing

and it has been known for a bit drags behind Nvidia.

I mean, not bad.

It's still playable frame rates but it's Yeah,

but you know it's the Nvidia is a better choice on that.

But maybe we switch to 4K.

So at 4K with the same settings sort of medium and sort of balanced,

this is rendering and even resolution at 2259 by 1270.

So higher than 710 ADP but not quite 1440p

and here you know it's it's not quite 60

and you know technically the super cards are a little bit faster.

But you know, once again, we're in numbers that I'd be thrilled

to see on a steam deck and I'm kind of underwhelmed

for a desktop and see yeah I mean yeah, it's one of those situations

you kind of have to fiddle with the settings

and depending what you want, for the most part, I would say most people

probably are not willing to to do a first person shooter.

But I mean, if you wanted to, I mean you could turn on the TI features

and check it out. I mean, these are these are playable.

You can see what it looks like.

Yeah.

You can see what it looks like, you know, and maybe use that as a starting place to

to kind of fiddle with some settings, but that's just cyberpunk.

So the other game that I benchmarked that has art available in it is returnal

same kind of deal 1440p with FSR set to bounce,

but here the implementation is a little bit different.

I would say it's lighter.

Is that a better word?

Yeah, I mean this is, this is a launch window Ps5 title.

So this is I think the interesting thing about Returnal is it's probably more

indicative of what a console port from a Ps5 or Xbox Series

X is going to look like for the rest of this generation, right?

Like there's a limited there's a limited budget for raytracing on, on the consoles.

They're relatively low powered by today's PC standards.

So when you're bringing your your ps5 games, your Xbox series X games over,

this is a

this is about the level of raytracing and compute you're going to expect to see.

Okay.

So and because it's actually a little bit lighter, I turned on the

TI preset to Epic.

So this is the most rr t you can get out of this game.

But like all raytracing games, developers pick and choose the features.

They want to add something

some people do global illumination, which is really expensive.

Some some games just do like a little bit of shadows,

a little bit of colored lights and a little bit of reflections,

which is much less, much less impactful, which is why it's able to run,

which is why we're running on Epic and running on a PlayStation five. Yeah.

So looking at 1440p results here in Returnal, I mean, that's

that's nothing new on the on the 4780 results.

It's right right there with each other but it it

definitely is still faster than still faster than AMD but over 4K

with a native resolution that sits somewhere between ten XP

and 1440p, I would say that's still playable.

That's, that's all 100 frames per second.

But we got a couple more charts to look at.

So next up, let's look at the maximum power draw that I recorded here.

In this case, I ran F1 22 on a loop

so that it was able to kind of, you know, get a bunch of runs in.

And then in the background side H.W.

info recording the peak power that it saw during redraw.

Yeah so and here we're actually seeing some variation which is interesting

because the the rated spec for all of these cards should be 285.

The the tough version got pretty close to it that Ventas dropped a little bit.

But you know I would say it's

two to fan card you know it's smaller

the clocks aren't as boosting on there so that doesn't surprise me

but that arrow version the the non super card

was actually sitting well well below the the other versions.

I mean I'm looking at the three cards and I'm thinking is is it possible

first off there is a quiet mode over clock switch on this whose board you know

it's possible that they're in on this one as well on the gigabyte and the gigabyte.

So it's possible that they're saying, hey, to hell with thermals.

Let's let's get every every single water out of this thing.

Could be.

But now, when you look back at it, though, maybe, maybe that also helps,

you know, give context to some of the benchmarks

where it was just a little bit slower than the other ones.

It does explain that.

Yeah. Yeah.

So I'm not quite sure either way.

They definitely do.

All of them have have a healthy lead over that that top end Radian card,

which turned in kind of almost 70

watts more power draw in under the same load.

This is a known thing, you know Nvidia cards are more power efficient.

So that's what we're seeing here and how it relates to thermals,

not just the power draw but also the the card size we're hearing.

We're seeing that the smallest card,

the vent is is actually the warmest card here is surprise, surprise.

Yeah.

It doesn't have as much a cooling capability to it.

And once again, surprise, surprise, that arrow was less power draw.

And the beefy cooler that we have here

is actually 14 C cooler than than the good for that.

Yeah. So it's an older card too, right?

So like the thermal interface is older, all of all of the stuff on there,

that's impressive.

Yeah. Yeah.

So you know a lot of the thermals, it really depends on, you know,

what kind of card that you get.

I think the general rule of more cooling surface on your cooling apparatus

probably means you're going to get better temperatures.

Yeah, a couple more benchmarks.

I do want to tease a little bit of testing that I've been doing.

I've been working on this big grandiose GPU testing

suite for creative, for video editing specifically.

And so I do have a couple of charts here.

These are just the the overall scores.

I'm not going to get into sub scores here,

but I kind of wanted to see how these cards fare in video editing.

Obviously, a lot of people build PCs to do a lot more things than just gaming.

So I want to start getting into more and more of that kind of testing.

This is just kind of my my little, my, my little drop in the bucket.

So let's, let's check out Premiere Pro.

So actually, I mean, even though the numbers are fairly close,

we are seeing a good jump here in Premiere Pro.

And I do think that comes down to VRAM like maybe

a little bit of the cooling increase.

But yeah, you're going to have to tune in later for the full video

where we go into the sub scores to really dig into it.

I don't know. We'll have to see.

You have to watch the full video.

But I also do have Da Vinci results numbers here as well,

and it's a lot more pronounced.

But I can I can definitely tell you that DaVinci Resolve

definitely loves more VRAM when it has it.

So we're seeing that here.

It's interesting here because we're looking

our sheet has the 40, 70, super and 40, 70 numbers as well.

And, and those memory configurations, the lines are exactly it's

just a straight line where the memory configurations.

Interesting, interesting, weird. Yeah.

So once again, maybe tune in for the full video to see that.

But we're we're finishing this video up here

in the last kind of final thoughts over the card designs.

How sad are you that they didn't

actually make a founders edition low Look, I like the Founders edition.

They're almost always harder to get than they're worth chasing down.

I'm I'm kind of more interested these days

in a bigger quieter card than than

your founders, with a few exceptions, are usually pretty svelte.

So, you know, I'll take a big chunky boy if it's quiet and fast.

Yeah, yeah.

And I mean, we definitely saw that performance while it was even

the actual thermals definitely did range depending on the size of the card.

Yeah. I mean, if you have a smaller case,

then obviously, you know, you're only limited by, by such things.

But yeah, there is a reason to pony up a little bit more money

for the bigger cooler because that's the other thing, $820, $880.

Mostly you're talking about the actual esthetics

and you know so yeah, the cooler and capability and stuff.

So yeah, the, the one thing I think I'd probably take I'm taking away from

this is that I wouldn't

spend a lot of extra money above the base price on a highly overclocked card.

I think I would just if I and and honestly

I would look really hard at the 70/900, right?

Yeah.

I mean you're getting a lot of performance and if if at least keeping up

or surpassing in a lot of ways depending on on what you're playing

not in raytracing if you're playing non ray tracing

and you're playing a lot of multiplayer games,

if you're playing,

God help you dice 11 multiplayer games, I think the 7900 is probably

going to be the winner for you. Yeah, not bad.

I do want to point out, last thing to note of these cards

is that this massive version does not have a dual BIOS

version where both the gigabyte and the AC stuff.

So, you know, I think that's another thing you know,

depending on how much you're willing to spend that you're going to get out of it.

So at the same time,

you could fit that missile into an ATX case

and I don't think you could on either of these others.

You can try you can try.

It depends on the case but out the Dremel. Yeah.

Yeah. I guess we'll try it. I'll report back.

Thank you for sitting down and running through these numbers with me.

I appreciate.

That's kind of the first time I'm doing GPU testing like this.

So it's been been a little harrowing to see.

It's, you know, the thing that people don't remember

about having lived this life in the past, but in pre video days, the thing

people don't really understand

is that these benchmarks take a really, really long time to run.

You know, it's like put the cart in the morning

and maybe by the end of the day you finish the entire lap

and you're talking about what, 12, 15 benchmarks over

over 20 runs for the different for three runs each.

It's it's just it's a ton of sitting there and watching 3D Mach and Siege

and Call of Duty

and and so it's been it's been fun I've got an envelope out at home.

Yeah and hopefully it's been enlightening You know and it helps people

decide on on which GPUs they want to buy in the future.

If you want to see more GPU testing videos like this, and especially that

that video editing one that I keep teasing,

be sure to subscribe here on PC World until the next one.

We will catch you later.

What we should do is just

speed it up to X in the edit and then it's going to be a 15 minutes.

I only watch YouTube videos at two x.

I just I just put them on to X and turn on close captions.

Use them so you do a 0.5 x

just really like y on the

only when I'm watching slime videos and.