NVIDIA GeForce RTX 5050 8 GB

[lexluthermiester]

I update it all the time. Let me check

Edit: The 5050 is listed at 1080p performance, the 5060 is listed at relative 4K performance, because it's faster than the 2080 Ti. So there's some area here where some cards are in a bit weird places. How to fix?

Edit2: reversed the sort order so that it's better aligned with tpu charts showing best at the top

Edit3: Maybe we could change 4K cutoff to RTX 3080, instead of 2080 Ti, to ensure all cards in the range have more than 8 GB VRAM?

That sounds reasonable. I have always looked at those numbers as relative to the stated specs rather than a perfectly hard and fast standard. Still, it might be a good idea to recompute the numbers from a data-base baseline, if that's possible.

 

[W1zzard]

All updated. Anything that still looks odd?

@Lew Zealand, @THU31, @lexluthermiester

 

[THU31]

The 5050-5060 difference seems good now. But overall I feel like the segmentation by 1080p and 4K creates some weird results.

Games get more demanding, so you can't really say that a certain level of performance is enough for 4K. Like the 3080 was amazing at 4K when it came out, but now it's not going to run any new game without heavy upscaling.
Also, new lower-end cards usually offer good 1080p and 1440p performance, but really struggle at 4K because of VRAM (bandwidth and capacity) and low ROP-counts. The 5060 is almost as fast as the 3070 at 1080p, but at 4K the 3070 is 27% faster (same bandwidth, but double the ROPs).

I'd go as far as to say that only top end cards are suitable for native 4K rendering in newest games, and to be honest, currently I'd only count the 5090, 4090 and barely the 5080 (and nothing from AMD).
If you look at your reviews, games like Wukong, Alan Wake 2, SW Outlaws or Stalker 2, you pretty much can't get 60 FPS at 4K on anything but those three cards. And those games are very representative of most big games that came out this year (Monster Hunter, AC Shadows, Oblivion, Clair Obscur and even Doom).

Wouldn't it make more sense if the entire database was based on native 1440p results (and maybe 1080p for very old cards)? That's the most balanced resolution that's not CPU-bottlenecked, but also not requiring the most powerful cards. I think it would be a more realistic and accurate representation.

Thanks for updating, though, and for taking input.

 

[Lew Zealand]

I update it all the time. Let me check

Edit: The 5050 is listed at 1080p performance, the 5060 is listed at relative 4K performance, because it's faster than the 2080 Ti. So there's some area here where some cards are in a bit weird places. How to fix?

I wish I knew as there's never a great answer that's also time-efficient.

Edit2: reversed the sort order so that it's better aligned with tpu charts showing best at the top

Edit3: Maybe we could change 4K cutoff to RTX 3080, instead of 2080 Ti, to ensure all cards in the range have more than 8 GB VRAM?

That sounds like a really good update! I'm not sure many people are considering a lower end GPU for 4K for new games.

All updated. Anything that still looks odd?

@Lew Zealand, @THU31, @lexluthermiester

The data for everything recent looks spot on now. Older things like RDNA1 and Pascal are still skewed a bit but not by enough to be worth any time spent, based on their age and relevance.

 

[W1zzard]

Older things like RDNA1 and Pascal are still skewed a bit but not by enough to be worth any time spent based on their age.

Happy to look at them, which ones?

 

[Lew Zealand]

Happy to look at them, which ones?

This is a placeholder as I'd really like to find where to make the best reference points for the purpose of being helpful instead of annoying.

Right now the RX 5700 and GTX 1080 Ti make the best comparison: In the last review they both appear in before the 1080 Ti is culled for age, the RTX 3080 AMP Holo review, the 1080 Ti is 13% faster than the 5700 at 1080p whereas in the database the 1080 Ti is listed as 31% faster.

However comparing their step down sibings, the 1080 and 5600 XT, in the AMP Holo review and the database their differences are only 3% vs 2%, noise-value so no difference.

My guess is whatever data is being used for the 1080 and 5600 XT (and likely almost all other cards) is spot-on while the data for the 5700 and 1080 Ti are skewed. Based on a cursory comparison, both appear skewed by about the same amount in opposite directions which means the data variation on each point is smaller, a good thing! The 5700 XT may also be skewed but not by as much.

I'll nose about this some more as I'm interested in good data representation but have a far less enjoyable assignment elsewhere right now...

 

[W1zzard]

This is a placeholder as I'd really like to find where to make the best reference points for the purpose of being helpful instead of annoying.

Right now the RX 5700 and GTX 1080 Ti make the best comparison: In the last review they both appear in before the 1080 Ti is culled for age, the RTX 3080 AMP Holo review, the 1080 Ti is 13% faster than the 5700 at 1080p whereas in the database the 1080 Ti is listed as 31% faster.

However comparing their step down sibings, the 1080 and 5600 XT, in the AMP Holo review and the database their differences are only 3% vs 2%, noise-value so no difference.

My guess is whatever data is being used for the 1080 and 5600 XT (and likely almost all other cards) is spot-on while the data for the 5700 and 1080 Ti are skewed. Based on a cursory comparison, both appear skewed by about the same amount in opposite directions which means the data variation on each point is smaller, a good thing! The 5700 XT may also be skewed but not by as much.

I'll nose about this some more as I'm interested in good data representation but have a far less enjoyable assignment elsewhere right now...

Are you aware that for cards not the in the comparison list (i.e. no manual data entry), the results are estimated? It says so in the footer of the table when estimated.

 

[lexluthermiester]

All updated. Anything that still looks odd?

@Lew Zealand, @THU31, @lexluthermiester

To be honest, I never saw the complaint. However, I do like the descending order the graphs are in. The numbers seem to look right as well.
When we compare TPU's numbers for the 5060vs3060vs1060vs960, they all correlate with performance measurements elsewhere, within a reasonable margin of error and difference.
Example;

GeForce RTX 5060 vs GeForce RTX 3060 vs GeForce GTX 1060 vs GeForce GTX 960 [videocardbenchmark.net] by PassMark Software

So it all looks good to me, nothing out of sorts and there's an improvement to the look of the graph. I say a job well done!

 

[Calicifer]

Games get more demanding, so you can't really say that a certain level of performance is enough for 4K. Like the 3080 was amazing at 4K when it came out, but now it's not going to run any new game without heavy upscaling.
Also, new lower-end cards usually offer good 1080p and 1440p performance, but really struggle at 4K because of VRAM (bandwidth and capacity) and low ROP-counts. The 5060 is almost as fast as the 3070 at 1080p, but at 4K the 3070 is 27% faster (same bandwidth, but double the ROPs).

I'd go as far as to say that only top end cards are suitable for native 4K rendering in newest games, and to be honest, currently I'd only count the 5090, 4090 and barely the 5080 (and nothing from AMD).
If you look at your reviews, games like Wukong, Alan Wake 2, SW Outlaws or Stalker 2, you pretty much can't get 60 FPS at 4K on anything but those three cards. And those games are very representative of most big games that came out this year (Monster Hunter, AC Shadows, Oblivion, Clair Obscur and even Doom).

I do count 50-60 series as 1080p cards. 70s series as 1440p cards and 80-90 series as 4k cards. With Ti models being able to be lite version of a card for a higher resolution. For many years now, GPUs were lagging behind gaming demands. I think it is the most healthy way to view the market through such segmentation. We never got lower cards to run higher resolutions as once dreamed. Instead, we needed to create an entire new tier (90s) cards for 4k gaming.

 

[OneSorcerer]

well, it is a side grade compared to my 6650xt, but has better temps and upscaller. I just sold my amd card in my country for 250. I might buy it if goes for 200 as a black friday deal and make 50 bucks. or break even if a 5060 costs 250 by then.

 

[THU31]

Daniel Owen used the TPU database in his 5050 video to compare to the 5060 (he always uses it to assess relative performance). It's a good thing we managed to tweak it, otherwise he'd probably be a bit confused if he only saw an 8% difference.

35:05