Question Should I include 4070 Ti (non-Super) or 7900 XT in initial testing results for 5070 Ti?

[JarredWaltonGPU]

Hey forumites!

Quick question for you. Would you rather I include the 7900 XT or the 4070 Ti in the upcoming 5070 Ti initial test results? The problem right now is that I'm out of time and can only test one more GPU before the deadline. I can add the other(s) later, but for tomorrow, what would you prefer?

Video on the subject with a bit more detail, maybe:

View: https://www.youtube.com/shorts/9l2IiqUYlmo

 

[Lutfij]

RTX vs RTX, so include the RTX4070Ti. Hopefully you can followup with the RX 7900 XT not long after

 

[DRagor]

Who cares when they will cost 1K. You can TLDR your review without even testing anything with "Great GPU but not worth what you have to pay for it".

 

[JayGau]

I mean, that sounds obvious to me. You should compare the new 70 ti with the previous 70 ti. Did you really consider not including the previous gen model in your test? That's the most important metric.

 

[JarredWaltonGPU]

I mean, that sounds obvious to me. You should compare the new 70 ti with the previous 70 ti. Did you really consider not including the previous gen model in your test? That's the most important metric.

The intention was to include it and the 7900 XT, but time is running out. As for why I wouldn't include it initially, it was superceded by the 4070 Ti Super, which is a more direct precursor. The 4070 Ti was never one of the favored GPUs, due to having 12GB and a 192-bit interface. 4070 Super is nearly the same performance, 4080 was a big step up, and let's not forget it was the canceled RTX 4080 12GB before internet outcry forced Nvidia to rename it.

The 4070 and 4070 Super are more popular, and even the one year later 4070 Ti Super has nearly caught up to the discontinued 4070 Ti in market share (caveat: Steam HW survey numbers).

But I'm now testing the 4070 Ti, and will inevitably face the wrath of angry AMD fans wondering why the 7900 XT wasn't included. 🤷‍♂️

 

[JayGau]

The intention was to include it and the 7900 XT, but time is running out. As for why I wouldn't include it initially, it was superceded by the 4070 Ti Super, which is a more direct precursor. The 4070 Ti was never one of the favored GPUs, due to having 12GB and a 192-bit interface. 4070 Super is nearly the same performance, 4080 was a big step up, and let's not forget it was the canceled RTX 4080 12GB before internet outcry forced Nvidia to rename it.

The 4070 and 4070 Super are more popular, and even the one year later 4070 Ti Super has nearly caught up to the discontinued 4070 Ti in market share (caveat: Steam HW survey numbers).

But I'm now testing the 4070 Ti, and will inevitably face the wrath of angry AMD fans wondering why the 7900 XT wasn't included. 🤷‍♂️

Yes I see your point. But I personally think that it's better to compare the previous non super with the new non super. There will likely be a 5070 Ti Super at some point, and then the comparison with the 4070 Ti Super will be more relevant. But that's just my opinion.

 

[JarredWaltonGPU]

Yes I see your point. But I personally think that it's better to compare the previous non super with the new non super. There will likely be a 5070 Ti Super at some point, and then the comparison with the 4070 Ti Super will be more relevant. But that's just my opinion.

Possibly, but I don’t know what Nvidia will do with a 5070 Ti Super. I guess it could enable eight more SMs or something. Actually, I expect all of the “Super“ 50-series cards will use 3GB GDDR7 chips. Which will actually make them worthwhile! Maybe. We’ll see next year.

 

[artk2219]

Personally, id say the 7900 XT to have two non Nvidia results from around the same price bracket, but I get your dilemma. I think since you already have a 4070 ti super in there, the XT would be the more valuable data point of the two. Honestly, you're going to be facing angry fan boys either way, you could flip a coin 3 times and leave it up to chance? Regardless, you could make a point to address it with the 9000 series review, if you think you may have some more time for a few more data points on that one. Thank you for these reviews, i know the past month or so has definitely been keeping you busy.

 

[alceryes]

Hey forumites!

Quick question for you. Would you rather I include the 7900 XT or the 4070 Ti in the upcoming 5070 Ti initial test results? The problem right now is that I'm out of time and can only test one more GPU before the deadline. I can add the other(s) later, but for tomorrow, what would you prefer?

Video on the subject with a bit more detail, maybe:

View: https://www.youtube.com/shorts/9l2IiqUYlmo

Ideally, you would have both.
However, I think it depends on what your current comparison chart looks like. If you have no AMD cards currently in the comparison then definitely have the 7900 XT.

I also think the 4070 Ti and 4070 Ti S are close enough for people to extrapolate how the missing one would perform. You could note this a couple times in the review.

 

[JarredWaltonGPU]

Ideally, you would have both.
However, I think it depends on what your current comparison chart looks like. If you have no AMD cards currently in the comparison then definitely have the 7900 XT.

I also think the 4070 Ti and 4070 Ti S are close enough for people to extrapolate how the missing one would perform. You could note this a couple times in the review.

So, right now I have:

5090
5080
5070 Ti
4090
4080 Super
4070 Ti Super
4070 Ti
7900 XTX

I'll add 7900 XT in the next two days is the plan, after the initial write-up.

 

[artk2219]

So, right now I have:

5090
5080
5070 Ti
4090
4080 Super
4070 Ti Super
4070 Ti
7900 XTX

I'll add 7900 XT in the next two days is the plan, after the initial write-up.

Yeesh, talk about a tough crowd for that review. For what it's worth, thank you for putting that review together. I know that must have taken days, and it was a good read.

 

[alceryes]

So, right now I have:

5090
5080
5070 Ti
4090
4080 Super
4070 Ti Super
4070 Ti
7900 XTX

I'll add 7900 XT in the next two days is the plan, after the initial write-up.

Good review.
One thing I don't like is your mixing of DLSS 2/3 results and straight raster on a single chart. This could easily confuse people into thinking that the 5070Ti is way better than the 4090. Yes, you can explain it as much as you want but many, many people just look at the pretty bar graphs and ignore the useless words.

Upscaling/DLSS/FSR/FG on one (or more) charts - Straight raster on another - This is the way.

 

[JarredWaltonGPU]

Good review.
One thing I don't like is your mixing of DLSS 2/3 results and straight raster on a single chart. This could easily confuse people into thinking that the 5070Ti is way better than the 4090. Yes, you can explain it as much as you want but many, many people just look at the pretty bar graphs and ignore the useless words.

Upscaling/DLSS/FSR/FG on one (or more) charts - Straight raster on another - This is the way.

There is no DLSS/FSR/FG testing on any of the charts in this review. If that's not clear, please let me know which ones are causing the confusion. If you're talking about DXR + Rasterization, the only charts those are mixed on are the second set of charts on the ray tracing page. In other words, they're slightly buried by intent. By the time anyone gets to that set of charts, it should be clear what they represent.

And really, you do have to look at the big picture sometimes. Yes, rasterization remains more important and plenty of RT games run so poorly that it's not worth using. But besides having faster RT performance, Nvidia also offers provably superior upscaling quality, as well as frame generation and MFG. All of that gets proxied into the overall geomean via the RT performance getting blended in.

What's the perfect weighting for everything? That's a huge debate people will want to have, and it's why I have separate rasterization and RT charts and averages. But for those that want a "big picture" chart as well, I provide that. Take it as you will.

 

[Gururu]

When you speak about how the games feel with MFG or upscaling do you actually play or are you basing it on observations on automatic benchmarks? Do you use game sequences specific to Tom's if you do play? Thanks!

 

[alceryes]

There is no DLSS/FSR/FG testing on any of the charts in this review. If that's not clear, please let me know which ones are causing the confusion. If you're talking about DXR + Rasterization, the only charts those are mixed on are the second set of charts on the ray tracing page. In other words, they're slightly buried by intent. By the time anyone gets to that set of charts, it should be clear what they represent.

And really, you do have to look at the big picture sometimes. Yes, rasterization remains more important and plenty of RT games run so poorly that it's not worth using. But besides having faster RT performance, Nvidia also offers provably superior upscaling quality, as well as frame generation and MFG. All of that gets proxied into the overall geomean via the RT performance getting blended in.

What's the perfect weighting for everything? That's a huge debate people will want to have, and it's why I have separate rasterization and RT charts and averages. But for those that want a "big picture" chart as well, I provide that. Take it as you will.

Hmmm, I swear I saw some DLSS results in the bar charts. Not there now though. As far as seeing the big picture, I agree - as long as it's not misleading or just plain false.

The perfect weighting is having enough separate bar graphs so the reader can weigh for themselves. 😉
Go ahead and have your DLSS/FSR/FG graphs, but also have a baseline no upscaling, plain raster graph that NEVER has any examples of upscaling/DLSS/FSR/FG.

 

[JarredWaltonGPU]

When you speak about how the games feel with MFG or upscaling do you actually play or are you basing it on observations on automatic benchmarks? Do you use game sequences specific to Tom's if you do play? Thanks!

Of the 22 games tested for the review, only four use a "canned" benchmark where I don't really play (Black Myth Wukong, F1 24, and both Flight Simulator games). The rest are manually run benchmarks, so yes, I get to feel what the game is like. It's a pain, but it's also more valid as an approach IMO.

For MFG, I used manual benchmarks and also played the games outside of benchmarking. I tried to get a good idea of how it actually looks and feels rather than just looking at a performance number. And it ends up as a very nebulous thing. Sometimes it's good, sometimes it's okay, sometimes it sucks.