6600xt vs 3060 , 1 month in!

[Rroff]

Hardware Unboxed appear to be creating a whole new subculture of trolls recently.. or is this something i've only just noticed.

As far as any evaluation of RT went that video was pretty much meaningless - no testing with RT on in several of the titles which were included where it would have shown more of a difference, using settings in stuff like Metro which would reduce the strain on RT capabilities at the expense of image quality. (With F1 just left the ultra quality setting which enables the lowest level of RT which minimises any difference between the capabilities of different GPUs and largely isn't much to write home about anyhow).

They might as well have just skipped anything relevant to RT in the video because the selective bits they did do were meaningless in terms of the bigger picture with RT.

 

[nvidiamd]

But with RT ultra on CP they are both under 30FPS, even at 1080p. Sure one may get 20 fps and the other 10fps but they are both useless...

 

[Grim5]

But with RT ultra on CP they are both under 30FPS, even at 1080p. Sure one may get 20 fps and the other 10fps but they are both useless...

by not reporting it, it's misleading

 

[Rroff]

But with RT ultra on CP they are both under 30FPS, even at 1080p. Sure one may get 20 fps and the other 10fps but they are both useless...

There are a lot more options than CP2077 at ultra RT though.

I completely understand tailoring settings to what gets playable framerates for a given set of cards but that isn't what has gone on with the RT testing (or lack of) here. So to use it as any meaningful commentary on RT performance/suitability is lacking.

 

[Harlequin]

Oh dear, the same ones who complained about tessellation not being on ultra max are now complaining about RT not being on ultra max , because it shows `their team` isnt as uber 1337 as they want to believe. /Facepalm.


Again, for the ones at the back who are shouting; the games chosen are each a representation of different DXR techniques ; just because YOU do not like them doesn't make them any less valid. Jensen Huang doesn't care, he sells shed loads to miners anyway

 

[Wrinkly]

Oh dear, the same ones who complained about tessellation not being on ultra max are now complaining about RT not being on ultra max , because it shows `their team` isnt as uber 1337 as they want to believe. /Facepalm.

Tessellation is just that, tessellation. Raytracing is a collective term for a number of effects, e.g. shadows, gi, reflection, caustics, ao, etc., that use raytracing tech. You are missleading the viewer when you cut down the number of effects to a level, e.g. just shadows, that shows cards to be equally good at RT. What is more worrying is that it gives the impression that RDNA2 in general is now far better at RT than it started out, building on the fine wine nonesense.

 

[Harlequin]

Oh dear, you must have been too young to not be aware of the entire over tessalated benchmarks and initial use in games, to show off the effects that it brought. Just as now with DXR, it was touted as the next bug thing. Its ok lad, we understand. DXR is the same but in 2021.

 

[Wrinkly]

Oh dear, you must have been too young to not be aware of the entire over tessalated benchmarks and initial use in games, to show off the effects that it brought. Just as now with DXR, it was touted as the next bug thing. Its ok lad, we understand. DXR is the same but in 2021.

I remember it quite clearly. I was burning EEPROMs for BBC micros, while swapping boot, wordstar, cobol and assembler floppies in an Advance A86, while you weren't even a naughty thought

 

[Phixsator]

Tessellation is just that, tessellation. Raytracing is a collective term for a number of effects, e.g. shadows, gi, reflection, caustics, ao, etc., that use raytracing tech. You are missleading the viewer when you cut down the number of effects to a level, e.g. just shadows, that shows cards to be equally good at RT. What is more worrying is that it gives the impression that RDNA2 in general is now far better at RT than it started out, building on the fine wine nonesense.

There is nothing misleading about it as long as you state the settings used and both GPUs are tested equally. It simply shows the performance at a certain image quality. RDNA is showing better results regarding RT in later/newer games. Why? because some of the newer titles optimize for it now that AMD has a product on the market that supports it. It has nothing to do with fine wine but I guess some people have a tough time comprehending even the simplest of things. If it was "fine wine" Cyberpunk with RT would run much better, it doesn't, RT performance is still poor. Then again it is a nvidia sponsered title that went pointies up on so many fronts.

 

[gpuerrilla]

There is nothing misleading about it as long as you state the settings used and both GPUs are tested equally. It simply shows the performance at a certain image quality. RDNA is showing better results regarding RT in later/newer games. Why? because some of the newer titles optimize for it now that AMD has a product on the market that supports it.

Pretty spot on. I generally see HU as a good source compared to the sea of **** out there. Whilst it was not a perfect comparison, it certainly isnt as bad as people are making out. If your buying the entry dGPU's I certainly wouldn't want to be dumping performance on a hog like RT anyway.. different story for the cards around 3080 level or above.

 

[Alan McFall]

There is nothing misleading about it as long as you state the settings used and both GPUs are tested equally. It simply shows the performance at a certain image quality. RDNA is showing better results regarding RT in later/newer games. Why? because some of the newer titles optimize for it now that AMD has a product on the market that supports it. It has nothing to do with fine wine but I guess some people have a tough time comprehending even the simplest of things. If it was "fine wine" Cyberpunk with RT would run much better, it doesn't, RT performance is still poor. Then again it is a nvidia sponsered title that went pointies up on some many fronts.

That's the problem.. it isn't, RDNA2 ray traces at RTX2000 levels which was discounted by HUB along with RT on the 3000 series (remember the Nvidia debacle ?). Now AMD have the hardware they are limiting the cores of the 3000 series and suggesting that RDNA2 is equivalent.. terribly misleading.

When RDNA3 and rtx 4000 series release and RT gets taken to the next level the people who bought an RDNA2 card based on the above premise are not going to be happy.

And don't think that graphics are now limited to what a console can produce, that's a joke.

 

[Wrinkly]

There is nothing misleading about it as long as you state the settings used and both GPUs are tested equally. It simply shows the performance at a certain image quality.

Which just happens to be much lower than the 3060 is capable of in certain cases.

RDNA is showing better results regarding RT in later/newer games. Why? because some of the newer titles optimize for it now that AMD has a product on the market that supports it.

I feel my original point has been proven here.

It has nothing to do with fine wine but I guess some people have a tough time comprehending even the simplest of things. If it was "fine wine" Cyberpunk with RT would run much better, it doesn't, RT performance is still poor. Then again it is a nvidia sponsered title that went **** up on some many fronts.

I didn't say it was finewine. I said they gave the impression of. CP2077 is a good example of RDNA2's limitations as it goes a bit further with RT use than MEE(console refresh) or REVillage(console targetted). Remember AMD took 6months(?) to work with CDPR before allowing RDNA2 users to turn on RT. Wonder why...

 

[Phixsator]

That's the problem.. it isn't, RDNA2 ray traces at RTX2000 levels which was discounted by HUB along with RT on the 3000 series (remember the Nvidia debacle ?). Now AMD have the hardware they are limiting the cores of the 3000 series and suggesting that RDNA2 is equivalent.. terribly misleading.

It is equivalent though at those specific settings... nothing misleading about it. By your logic, testing using only maximum settings including RT would also be misleading as it may distort the results due to settings people would/may never run cause of the performance impact on both brands even if one brand is beating the other. It could give an impression that if you just lowered your settings a bit you would get MUCH MUCH more performance, yet in reality not really being able to overtake the competition in any meaningful way due to the **** poor scaling at lower res/settings. And there you just spent 20, 30, 50 dollars for the privilege to find that out the hard way.

Which just happens to be much lower than the 3060 is capable of in certain cases.

If the 3060 is capable of more then why isn't it showing it? It clearly isn't.

I didn't say it was finewine. I said they gave the impression of. CP2077 is a good example of RDNA2's limitations as it goes a bit further with RT use than MEE(console refresh) or REVillage(console targetted). Remember AMD took 6months(?) to work with CDPR before allowing RDNA2 users to turn on RT. Wonder why...

CPDR has a histori of working with Nvidia and using their black box bs. Witcher is a fine example of this, with physx completely gimping Kepler's performance in certain places not to mention the Tesselation used for Hairworks also completely destroying performance for no reason. So I tell you why it took 6 months, because the game was and still is a cluster* of a game performance wise. A 3070 can't even run the game maxed at 1440p with normal raytracing on without dipping below 60. DLSS had to be used and that came with it's own problems such as ghosting. It's typical Nvidia MO. Overdo the effects for no reason come hell or high water. Reminds me of when you have to clean up a server rack and you see patch cables all over the place, all spaghetti because the previous guy couldn't be arsed to take the time and do it right the first time. Who gets blamed? you! for taking too long trying to make something of a complete mess

 

[Wrinkly]

If the 3060 is capable of more then why isn't it showing it? It clearly isn't.

Again I feel my point is proven.

CPDR has a histori of working with Nvidia and using their black box bs. Witcher is a fine example of this, with physx completely gimping Kepler's performance in certain places not to mention the Tesselation used for Hairworks also completely destroying performance for no reason. So I tell you why it took 6 months, because the game was and still is a cluster* of a game performance wise. A 3070 can't even run the game maxed at 1440p with normal raytracing on without dipping below 60. DLSS had to be used and that came with it's own problems such as ghosting. It's typical Nvidia MO. Overdo the effects for no reason come hell or high water. Reminds me of when you have to clean up a server rack and you see patch cables all over the place, all spaghetti because the previous guy couldn't be arsed to take the time and do it right the first time. Who gets blamed? you! for taking too long trying to make something of a complete mess

I wouldn't expect a 3070 to run it well at 1440p at maxed settings. That's why I purchased a 3080. I almost went 3090 but sanity told me that I wouldn't notice a difference at said resolution. Looking at the heat/power draw issues I'd say I dodged a bullet.

BTW. I'm fine with spaghetti...

 

[Grim5]

If the 3060 is capable of more then why isn't it showing it? It clearly isn't.

 

[Rroff]

Oh dear, the same ones who complained about tessellation not being on ultra max are now complaining about RT not being on ultra max , because it shows `their team` isnt as uber 1337 as they want to believe. /Facepalm.


Again, for the ones at the back who are shouting; the games chosen are each a representation of different DXR techniques ; just because YOU do not like them doesn't make them any less valid. Jensen Huang doesn't care, he sells shed loads to miners anyway

It is nothing like that - if you want to compare to tessellation it is like choosing games and settings where you are only using light levels of tessellation which fit inside the capabilities of the dedicated tessellation engine on early AMD GPUs so as to not expose their limits, or let the more capable hardware stretch its legs and pretend the rest doesn't exist. So the balance of performance is largely shifted away from RT capabilities.

There is a long way from medium RT in F1 and normal profile w/ 4x VRS in Metro and a situation where you are throwing absurd levels of it beyond makes any visual difference with a lot of games and settings in the middle ground.

The RT testing they did is in no way a broad representation of the situation with RT. It doesn't matter if you or I like or dislike it the testing does not reflect an accurate broad portrayal of the RT capabilities of either card vs existing games within their performance envelope let alone future titles.

That isn't to say the result will necessarily be any different in a broader range of titles and settings but those used in no way reflect the full extent of the situation.

 

[Rroff]

That isn't really a good comparison as they are using DLSS on w/ RT as it is more a DLSS benchmark. What would be more interesting is seeing both without upscaling and/or both with upscaling at the same quality level.

 

[oguzsoso]

looking for baal?

 

[D1craig]

wouldnt this then cause the same argument people are having now but the other way around?

nobody will be buying either of these cards for any raytracing apart from maybe benchmarks...

 

[Rroff]

i want to make a clarification, this is a bit overstatement (im supporting your opinions, just wanted to clarify some stuff). 3070 can't even reach 60 fps most of the time at native 1080p lmao at 1440p native with normal rt settings it wont punch anything above 40 fps unless you look at sky or be in nomad area

its hilarious tbh. "good" rt performance yet its not. even a 3080 wont push a 60 fps at 1440p native with rt in cyberpunk. yeah game bad performance etc etc. dlss come to the rescue.... it looks somewhat decent but it is what it is. supposedly dlss was to be the savior at 4k. now it is savior of a "rtx 3080" at "1440p" to render the game at "960p" so that it can push close to 60-70 fps lmao

or at 4k and 1080p render u might get 55-63 fps with 3080 probs. dlss works when u give it 4 mil pixels (dlss quality at 4k) but when u give it 2 mil pixels at 4k (dlss perf at 4k), its not that of a good stuff.

The game doesn't run that great with RT features off either. Not sure what is up with the performance of the game for some people though - it runs perfectly fine at ~60FPS on my 3070 at 1440p with all ultra settings + RT ultra with DLSS Quality - but I see some people who struggle to get that even with a 3080.

Only place I see it drop really from ~60FPS is the park area in the middle of the city.

Sadly the game rarely uses ray tracing to that great effect aside from isolated areas anyhow - it is mostly used to touch up the edges of where traditional features fall down at a significant performance cost - using an approach like Metro you could probably path trace the whole game at a less performance cost at the expense of RT hardware being a minimum requirement.