GPU Tier List?

[Guest]

When I got into PC Building it was quite simple for me to understand what GPUs were better than others, but now with Nvidia doing their wacky thing with RTX, GTX & 16 series and then AMD having 2 series of GPU and APUs sporting things from their VEGA line I'm quite confused & can't recommend parts to friends nor judge for myself.

Also I don't understand if RTX is 2 graphics cards & if you use raytracing you don't use 50% of your GPU but if you don't user RayTracing then you don't use the other 50% but some graphics cards can enable RTX anyway.

 

Is there a chart that lines things up?

IE

 

Best

GTX 1080

GTX 1070

GTX 980TI = GTX 1060 = R9 Fury X = RX 580

AMD R9 390 = GTX 980 = RX 480 = RX 570

GTX 970 = AMD R9 380 = GTX 1050 TI = RX 470 = RX 560

GTX 960 = GTX 1050 = RX 460

 

Worst

 

I also don't know if my understanding of GPU performance is accurate in what's about the same performance quality. I am looking at performance only. Power consumption & cost are irrelevant to this perspective. I can look up prices on my own.

[gloop]

Modern cards go:

 

Titan RTX

2080 ti

2080 super

2080

2070 super

5700xt

2070

5700

2060 super

Thats as far as I'll go, will leave the lower ends for others to do.

[Chen G]

more expensive=better.

Quote

Also I don't understand if RTX is 2 graphics cards

It's not

Quote

if you use raytracing you don't use 50% of your GPU but if you don't user RayTracing then you don't use the other 50% but some graphics cards can enable RTX anyway.

It's not 50%

And it's common to not be able to use all the modules in a chip, you don't use 100% of the CPU either, even when it says 100%, that's just 100% of the execution unit.

[CommanderAlex]

RTX video cards have something called Tensor cores that are specifically engineered into the GPU die itself that handles the raytracing capabilities of the graphics card. If you turn the raytracing features off in the supported game's video settings, it simply disables those cores. Raytracing does have an impact on FPS because the graphics card now has to render those reflections if I understand it correctly. Think of raytracing as adding sprinkles to ice cream, where ice cream is your graphics. 

 

A chart would change constantly because there are a lot of graphics cards out on the market and there are new graphics cards coming out soon (Big Navi and RTX 3000 series). A RTX series card has some performance improvements over a GTX 10 series card but these cards do not have Tensor cores. 

[GoldenLag]

11 minutes ago, fpo said:

Is there a chart that lines things up?

If you ignore the drivers of cards that are maxwell and older. 

 

Techpowerup has a very rough lineup listing GPUs in relative performance. 

 

Edit: offcourse with a big disclaimer that unless you compare usage scenarios, you shouldnt compare GPUs using such lists outside of their own architecture. 

[Guest]

Gloop

 

 

 

 

Spoiler

Just now, gloop said:

Modern cards go:

 

Titan RTX

2080 ti

2080 super

2080

2070 super

5700xt

2070

5700

2060 super

Thats as far as I'll go, will leave the lower ends for others to do.

How do they relate to like 1060? My baseline is the GTX 900 series.

Is there a considerable step between each of these cards?
I put some equals like R9 390 = GTX 980. It's a little less than a 980 but it's close enough that they're basically equivalents.

 

Commander Alex

Spoiler

Just now, CommanderAlex said:

RTX video cards have something called Tensor cores that are specifically engineered into the GPU die itself that handles the raytracing capabilities of the graphics card. If you turn the raytracing features off in the supported game's video settings, it simply disables those cores. Raytracing does have an impact on FPS because the graphics card now has to render those reflections if I understand it correctly. Think of raytracing as adding sprinkles to ice cream, where ice cream is your graphics. 

So if I never use Raytracting, am I wasting my money because I'll never use those cores?

 

 

Chen G

Spoiler

Just now, Chen G said:

more expensive=better.

General consensus, but I don't need the best, just to know what should be a good GPU.

Just now, Chen G said:

It's not

It's not 50%

And it's common to not be able to use all the modules in a chip, you don't use 100% of the CPU either, even when it says 100%, that's just 100% of the execution unit.

With a CPU it won't reach 100% but I know 100% of the CPU is capable of executing code.
If I get RTX & never turn on raytracing, will I just never ever use part of my card? Why not just make a card that cuts the RTX part off & make it cheaper?

 

 

Golden Lag

 

 

 

1

Spoiler

Just now, GoldenLag said:

If you ignore the drivers of cards that are maxwell and older. 

 

Techpowerup has a very rough lineup listing GPUs in relative performance. 

Do you have a link?

I know there's a website that does some scores for GPUs but that doesn't mean anything because I don't understand how points are calculated. Additionally those are synthetic & not reflective of real-world performance so I've been told.

 

[CommanderAlex]

3 minutes ago, fpo said:

Commander Alex

  Reveal hidden contents

 

 

So if I never use Raytracting, am I wasting my money because I'll never use those cores?

 

 

Well in a way yes, but you're gaining graphical performance gains over Pascal (GTX 10 Series) but its not that much if you already own a GTX 10 Series card. 

[gloop]

6 minutes ago, fpo said:

Do you have a link?

https://www.tomshardware.com/reviews/gpu-hierarchy,4388.html

[Guest]

Just now, CommanderAlex said:

Well in a way yes, but you're gaining graphical performance gains over Pascal (GTX 10 Series) but its not that much if you already own a GTX 10 Series card. 

What's the difference between RTX without RayTracing, GTX & The Nvidia 16 series. 1660 and so on.

 

Personally I have the AMD R9 390 & my friend has a laptop with a GTX 1050 (Non-TI) that may want to build a desktop.

[Eighjan]

4 minutes ago, gloop said:

https://www.tomshardware.com/reviews/gpu-hierarchy,4388.html

Based on that link, would I be looking at a minimum of a 2060 Super as an appreciable (real world) upgrade to my Vega 56?

 

/hijack -- sorry, @fpo

[GoldenLag]

11 minutes ago, fpo said:

know there's a website that does some scores for GPUs but that doesn't mean anything because I don't understand how points are calculated. Additionally those are synthetic & not reflective of real-world performance so I've been told.

Hence the use of the words "very rough" and the edition of a disclaimer in the edit.

 

Link: https://www.techpowerup.com/gpu-specs/radeon-rx-5700-xt.c3339

 

Check the relative performance window on the page. 

 

Again, very rough, not for comparing GPUs directly, as that can only be done on a per task basis. 

[Bartholomew]

23 minutes ago, CommanderAlex said:

RTX video cards have something called Tensor cores that are specifically engineered into the GPU die itself that handles the raytracing capabilities of the graphics card.

Well yes, but no

 

Tensor cores = machine learning, DLSS

RT cores = ray tracing cores

 

(And no, they are not the same thing as some people insist on believing).

Both can be used simultainously or seperate, depending on the application.

[jones177]

I go buy what I used them for.

 

GTX 980     = 1080p 60hz

GTX 980 ti  = 1440p 60hz

GTX 1080   = 3440 X 1440 60hz

GTX 1080 ti = 3840 X 2160 60hz

RTX 2080 ti = 1440p 120hz = 3440 X 1440 120hz

[Guest]

1 hour ago, jones177 said:

I go buy what I used them for.

 

GTX 980     = 1080p 60hz

GTX 980 ti  = 1440p 60hz

GTX 1080   = 3440 X 1440 60hz

GTX 1080 ti = 3840 X 2160 60hz

RTX 2080 ti = 1440p 120hz = 3440 X 1440 120hz

Where do the 70, 60 & 50s line up? 

[jones177]

8 minutes ago, fpo said:

Where do the 70, 60 & 50s line up? 

The last 70 I bought was a GTX 970. All it did was stutter so I put it in my Son's computer and it ran perfectly. It was a good excuse to get the 980. 

Never used anything lower but I did have four 70s.

They all ran 1080p monitors. 

 

GTX 970

GTX 680 w/ Backplate 4GB 

EVGA GeForce GTX 670                       =   1080p

EVGA GeForce GTX 570 

EVGA GeForce GTX 470 

 

 

 

 

[Guest]

26 minutes ago, jones177 said:

The last 70 I bought was a GTX 970. All it did was stutter so I put it in my Son's computer and it ran perfectly. It was a good excuse to get the 980. 

Never used anything lower but I did have four 70s.

They all ran 1080p monitors. 

 

GTX 970

GTX 680 w/ Backplate 4GB 

EVGA GeForce GTX 670                       =   1080p

EVGA GeForce GTX 570 

EVGA GeForce GTX 470 

 

 

 

 

I’m looking for chart comparison to see what performs as good as what 

[xAcid9]

2 hours ago, fpo said:

Where do the 70, 60 & 50s line up? 

2080 Ti

2080S

2080, 1080 Ti, 2070S, 5700XT, Radeon VII

2070, 5700, 2060S, 1080

5600XT, 2060, 1070 Ti, Vega 64

Vega 56, 1070, 1660 Ti, 980 Ti, 1660S

1660, 590, 980, 580, 1060 6GB, 1650S

970, 570, 1650, 1060 3GB, 5500 XT

5500, 1050 Ti, what else?

[jones177]

7 hours ago, fpo said:

I’m looking for chart comparison to see what performs as good as what 

For comparing performance I use techpowerup.com relative performance chart that can be found on their GPU pages. https://www.techpowerup.com/gpu-specs/geforce-gtx-470.c267

 

[Propanepapi]

Surprised nobody has posted this:

 

https://www.videocardbenchmark.net/high_end_gpus.html

 

It's a list of graphics performance scores based on how the card did in Passmark's GPU benchmarking software.

 

There is a comparison tool for crosschecking GPU models, but the score typically shows you were they land near other models.

 

They also do CPUs:

https://www.cpubenchmark.net/

 

-pp