Desktop GPU Performance Hierarchy Table (Archive)

[mellis]

Until 4K performance is achievable at a decent price, why upgrade? I’ll stick with 1080P until then. My current AMD 390X does the job for me still. When a graphics card can support 4K gamming at 60 fps for around $400, then I may buy it, but a compatible monitor will need to be even lower in price.

 

[tikal]

I dislike your partisanship, obviously you promote NVIDIA, otherwise you would never stoop to the level of trashing RX 570 in favour of GTX 1060 like that. The lack of RAM hits back at other games too and will only get worse and will prohibit you from using better textures packs. That is distinctive advantage of RX 570 and should have been praised.

 

[tojumikie]

Really helpful stuff when deciding which cards to get on eBay/Craigslist

 

[cryoburner]

I dislike your partisanship, obviously you promote NVIDIA, otherwise you would never stoop to the level of trashing RX 570 in favour of GTX 1060 like that. The lack of RAM hits back at other games too and will only get worse and will prohibit you from using better textures packs. That is distinctive advantage of RX 570 and should have been praised.

I can't say I see anywhere where they "trashed" the RX 570. In fact, they specifically point out that the 1060 3GB's limited VRAM might be a problem in some cases...

There’s a larger group of Radeon RX 580, GeForce GTX 1060 6GB, GeForce GTX 1060 3GB, and Radeon RX 570 boards that excel at 1920x1080. They can even be made serviceable at 2560x1440 if you dial down your detail settings far enough. But beware: insufficient GDDR5 memory on cards like the 3GB GeForce GTX 1060 may cause severe performance issues at higher resolutions. That model is usually faster than Radeon RX 570. However, because we benchmark Forza at 2560x1440 using High settings, its 3GB just aren’t enough, skewing the score lower.

In most existing games, when at 1080p resolution, which is what these cards are primarily designed for, the 1060 3GB will on average tend to be slightly faster than an RX 570. There are certainly games where the RX 570 is faster, but the 1060 3GB tends to still have a slight edge in more titles. It may experience more performance issues in the future as VRAM requirements rise, but based on games that can be tested today, it's reasonable to position it above the RX 570. And it's not like they're saying that one is significantly better than the other. The format of the new chart doesn't allow them to be positioned side-by-side though,which would probably be more reasonable, as both have their merits.

I must say though, having a score column is kind of silly if it serves no real purpose in the ranking. And like I said several months ago when the new format debuted, 3 games is simply not enough to accurately rank graphics cards by, particularly when they are running at settings that wouldn't likely be used with some of the cards. This results in some weird numbers, especially at the low-end, like the RX 550 scoring about 67% higher than the GT 1030, when in reality these cards are pretty close to one another in performance, and having more VRAM isn't likely to matter as much at the settings these low-end cards can comfortably run. What are they getting, less than 10fps in these 1440p tests? : P The numbers might be okay for giving a rough idea of how much faster a high-end card is compared to a lower-end one, but when comparing the performance of nearby cards in the chart, these numbers can often be a bit deceptive. And if anything, I'd say that the numbers might actually favor AMD's cards a little more than they probably should to be representative of typical gaming performance in existing games.

 

[shrapnel_indie]

I dislike your partisanship, obviously you promote NVIDIA, otherwise you would never stoop to the level of trashing RX 570 in favour of GTX 1060 like that. The lack of RAM hits back at other games too and will only get worse and will prohibit you from using better textures packs. That is distinctive advantage of RX 570 and should have been praised.

I can't say I see anywhere where they "trashed" the RX 570. In fact, they specifically point out that the 1060 3GB's limited VRAM might be a problem in some cases...

There’s a larger group of Radeon RX 580, GeForce GTX 1060 6GB, GeForce GTX 1060 3GB, and Radeon RX 570 boards that excel at 1920x1080. They can even be made serviceable at 2560x1440 if you dial down your detail settings far enough. But beware: insufficient GDDR5 memory on cards like the 3GB GeForce GTX 1060 may cause severe performance issues at higher resolutions. That model is usually faster than Radeon RX 570. However, because we benchmark Forza at 2560x1440 using High settings, its 3GB just aren’t enough, skewing the score lower.

In most existing games, when at 1080p resolution, which is what these cards are primarily designed for, the 1060 3GB will on average tend to be slightly faster than an RX 570. There are certainly games where the RX 570 is faster, but the 1060 3GB tends to still have a slight edge in more titles. It may experience more performance issues in the future as VRAM requirements rise, but based on games that can be tested today, it's reasonable to position it above the RX 570. And it's not like they're saying that one is significantly better than the other. The format of the new chart doesn't allow them to be positioned side-by-side though,which would probably be more reasonable, as both have their merits.

I must say though, having a score column is kind of silly if it serves no real purpose in the ranking. And like I said several months ago when the new format debuted, 3 games is simply not enough to accurately rank graphics cards by, particularly when they are running at settings that wouldn't likely be used with some of the cards. This results in some weird numbers, especially at the low-end, like the RX 550 scoring about 67% higher than the GT 1030, when in reality these cards are pretty close to one another in performance, and having more VRAM isn't likely to matter as much at the settings these low-end cards can comfortably run. What are they getting, less than 10fps in these 1440p tests? : P The numbers might be okay for giving a rough idea of how much faster a high-end card is compared to a lower-end one, but when comparing the performance of nearby cards in the chart, these numbers can often be a bit deceptive. And if anything, I'd say that the numbers might actually favor AMD's cards a little more than they probably should to be representative of typical gaming performance in existing games.

It was also mentioned before in an article here that with some games the 1050Ti with its 4GB of RAM would outperform the gimped 1060 3GB model because it ran out of RAM.

 

[climber]

As mentioned before in the comments thread, where's the Titan V? It was shown in the benchmarks for the review articles for the GTX 2080 FE and GTX 2080 Ti...
https://www.tomshardware.com/reviews/nvidia-geforce-rtx-2080-founders-edition,5809.html
https://www.tomshardware.com/reviews/nvidia-geforce-rtx-2080-ti-founders-edition,5805.html

 

[King_V]

I think I'd agree, and that it very likely gets slotted between the 2080 an 2080Ti

 

[cryoburner]

It was also mentioned before in an article here that with some games the 1050Ti with its 4GB of RAM would outperform the gimped 1060 3GB model because it ran out of RAM.

I think a lot of that came down to some questionable testing methodology, actually. I believe they mentioned that for at least some of those tests, they altered VRAM settings in such a way to make games attempt to utilize more VRAM than was available, but since no one would actually do that in practice, they ended up with results that were not all that representative of how the card would normally perform in those games.

They apparently didn't do this every time they tested the cards though. For example, let's look at their RX 570 review, and compare that to the results in their GTX 1060 Roundup written by the same reviewer a month prior. In the game Hitman, the performance of the 1060 3GB appeared to be abysmal even at 1080p ultra, averaging just 16fps, compared to 48fps for the 1050 Ti, or 83fps for the RX 570. However, if we look at the same game benchmarked in the 1060 roundup just weeks before, we see that the 1060 3GB averaged a very respectable 68fps at 1080p ultra, and even managed a relatively decent 48fps at 1440p. Below those abnormally low results in the RX 570 review, they mentioned "Hitman runs into big problems on the GeForce GTX 1060 3GB when you allow the game to override its memory safeguards." Supposedly that was done "to ensure a similar workload across the cards", but it doesn't at all accurately represent how the 1060 3GB will perform in that game. Now, I'm sure there are some cases where the limited VRAM does impact performance, and there will undoubtedly be more examples in the future, but as of now, in most games, the 1060 3GB tends to be slightly faster than an RX 570.

Of course, given current pricing, at least in the US, you can currently find a number of RX 580s, even some of the 8GB models, priced really close to the 1060 3GB, and those should perform better in just about every game. Unless one considers the higher power draw of the RX 580 to be a major concern, there isn't much reason to go with the 1060 3GB at this time.

 

[cryoburner]

As mentioned before in the comments thread, where's the Titan V?

I suspect it's not there because the Titan V was never really all that practical as a gaming card, seeing as it is priced upward of $3,000, more than double what the Titan cards have usually launched for. It's more intended as a card to be used for professional tasks, and is priced accordingly. So, it's probably a similar reason to why you don't see Quadro or Tesla cards in the list. Nvidia has shifted their product naming scheme this generation, so now the 2080 Ti fits the role of what a Titan previously would have been.