Comparison objectivity.

[Eighjan]

... between AMD & nVidia graphics cards - is it truly possible?

 

Given that the architecture of the two companies products is totally different, how is there a way to look at two cards and say that one is better than the other... IF there's no benchmarking tools for the application you want to use the card for?  This means spec's like CUDA™ cores - or the AMD equivalent - have little strength in comparisons.

Core & boost clock speeds are immaterial, except for comparing cards of the same chipset/GPU.

Even differences in VRAM technologies can cause discrepancies in comparisons.

 

Is there realistically any way to say that exactly any card is a direct competitor to another... or is it literally down to price?  If so, how can older cards be 'written off' if - for example - they are no longer manufactured, let alone sold?

 

The easiest cuprit would be lack of in-application (games are applications, too) benchmarking, but if that's all a particular user "uses", is there really any way to get concrete, irrefutable advice?

[Tan3l6]

I do believe in the benchmark business.

[Eighjan]

1 minute ago, Tan3l6 said:

I do believe in the benchmark business.

OK... so a game with no benchmark (for example) is entirely down to user satisfaction/subjective opinion for whether any component is worth using?

[boggy77]

4 minutes ago, Eighjan said:

IF there's no benchmarking tools for the application you want to use the card for?

those applications must be pretty rare and still are probably similar in the way they use the gpu to other applications for which benchmarks exist.

 

do you have a particular application in mind for which you can't find any reliable benchmarks for gpus?

[Tan3l6]

1 minute ago, Eighjan said:

OK... so a game with no benchmark (for example) is entirely down to user satisfaction/subjective opinion for whether any component is worth using?

Benchmarks in other similar conditions might give the approximation of the program/game performance.

[Fasauceome]

5 minutes ago, Eighjan said:

Is there realistically any way to say that exactly any card is a direct competitor to another... or is it literally down to price?

Cards can only be described as a competitor with another based on market segment. That includes price, and also naming scheme/advertising.

 

As for comparison between the two cards to determine which is "better," the one and only way is benchmark data.

[Eighjan]

1 minute ago, boggy77 said:

those applications must be pretty rare and still are probably similar in the way they use the gpu to other applications for which benchmarks exist.

 

do you have a particular application in mind for which you can't find any reliable benchmarks for gpus?

Yes... a rather old & highly popular game from a relatively cynically appraised publisher; but naming and shaming is surely poor form.

[boggy77]

2 minutes ago, Eighjan said:

Yes... a rather old & highly popular game from a relatively cynically appraised publisher; but naming and shaming is surely poor form.

in games, before the rtx era, it's pretty simple, as all the games depend on the card's rasterisation power. so you can compare cards based on almost any game and that should still hold true for your guilty pleasure of a game.

[Eighjan]

13 minutes ago, Tan3l6 said:

Benchmarks in other similar conditions might give the approximation of the program/game performance.

Fair point... but how long does an individual component become viable for inclusion is such schemes?

 

7 minutes ago, boggy77 said:

in games, before the rtx era, it's pretty simple, as all the games depend on the card's rasterisation power. so you can compare cards based on almost any game and that should still hold true for your guilty pleasure of a game.

Equally fair, but is there a solid comparison of rasterisation that works & catalogues historical records for games as old as 2004, that is still used for current cards?

 

My 'guilty pleasure' has implemented some level of ray-tracing, though I doubt it would be of any significance to products released in the last 3-5 years.

[boggy77]

8 minutes ago, Eighjan said:

Equally fair, but is there a solid comparison of rasterisation that works & catalogues historical records for games as old as 2004, that is still used for current cards?

probably not, but what I meant is if gpu A is better than gpu B in a 2018 game, it's very likely this would hold true for a game from 2004.

[Eighjan]

4 minutes ago, boggy77 said:

probably not, but what I meant is if gpu A is better than gpu B in a 2018 game, it's very likely this would hold true for a game from 2004.

My fear there is that the GPU I use may be unlikely to be a 'serious contender' for inclusion for 2018 title comparisons; even considered too niche in the first place.

Edited January 20, 2021 by Eighjan

[boggy77]

2 minutes ago, Eighjan said:

My fear there is that the GPU I use may be unlikely to be a 'serious contender' for inclusion for 2018 title comparisons; even considered too niche in the first place.

you can use passmark for a first pass

sorry for the pun.

not sorry.

 

now I'm intrigued, what is this rare GPU you are using and what's the equally intriguing game? what other gpu are you trying to compare with your current gpu?

[Eighjan]

9 minutes ago, boggy77 said:

you can use passmark for a first pass

sorry for the pun.

not sorry.

 

now I'm intrigued, what is this rare GPU you are using and what's the equally intriguing game? what other gpu are you trying to compare with your current gpu?

Ugh... dammit...

 

Vega56 (got it to replace a 1050ti that wasn't up to more than 1080p; the monitor I ended up using now is 1440p/144)

World of Warcraft

RTX 2070 (or better)

[boggy77]

10 minutes ago, Eighjan said:

Ugh... dammit...

 

Vega56 (got it to replace a 1050ti that wasn't up to more than 1080p; the monitor I ended up using now is 1440p/144)

World of Warcraft

RTX 2070 (or better)

i mean, the vega 56 is not that niche, and it's a pretty fun card to try to overclock. at stock it's around 1070 performance, so below the rtx2060.

if you overclock it, it can in some instances reach beyond 2060 performance (but with huge power draw).

i'd say an upgrade to a 2070 would make sense, and yo should see a decent performance increase, but not with current prices on the market

[Eighjan]

8 minutes ago, boggy77 said:

but not with current prices on the market.

Oh, hekk no... not even considering buying 'now' (bad enough trying to look for summat to replace my wife's aging 2GB HD6670 with).

That said, by the time stability in price is achieved, we may have had another generation launched to eclipse RTX3xxx/RX6xxx...

8 minutes ago, boggy77 said:

if you overclock it, it can in some instances reach beyond 2060 performance (but with huge power draw).

Yeah; that doesn't sound like a fun thought... 

 

I'll keep my eye on 2070/Super prices on PCPartPicker...

[boggy77]

6 minutes ago, Eighjan said:

I'll keep my eye on 2070/Super prices on PCPartPicker...

doubt you'll find them new anywhere, they are most likely discontinued by nvidia. best bet is ebay, cex or wait for prices on 3060/ti to come down.

[Eighjan]

1 minute ago, boggy77 said:

doubt you'll find them new anywhere, they are most likely discontinued by nvidia.

Ohh...

1 minute ago, boggy77 said:

or wait for prices on 3060/ti to come down.

... preferable.  Thx.