AMD Announces 7900XTX $999, 7900XT $899, Arriving December 13

[Shzzit]

1 minute ago, starsmine said:

Why are you apologizing to nobody? not a single person thought that.

where did they say 1.7x raytracing number? like honestly, how was that derived?, and how did you do that math and come to that conclusion?
the 6950 is usually pretty damn close to the 3090, a 1.7x will exceed the 3090ti in all games.

They said 1.7x ray tracing in there slides. So i just took 6900 ray tracing numbers and X them by 1.7 to get 7900 ray tracing performance. 

 

Did i do it wrong?  Not so good at math.

[Shzzit]

3 minutes ago, Imbadatnames said:

It should do because it is capable. You should be able to use FSR on AMD cards to bump up the frame rate too 

Ill see if i can try out FSR, thanks mate.  Ya sometimes i get bad fps drops, but most of the time its pretty fun.  Just started playing minecraft last month, been kind interesting to say the least lol. 

 

Brother upgraded and gave me his 6800xt, and it destorys my gtx 1080 in gaming for sure.  Pretty nice upgrade. 

[starsmine]

9 minutes ago, Shzzit said:

They said 1.7x ray tracing in there slides. So i just took 6900 ray tracing numbers and X them by 1.7 to get 7900 ray tracing performance. 

 

Did i do it wrong?  Not so good at math.

What slide said 1.7x ray tracing. 
They had slides that said up to 1.7x overall performance increase in some games. Not a single slide said that about ray tracing performance. And it was compared to the 6950, not 6900. AND, again, what sort of benchmark are you using to get that 1.7x of a 6900 is less then a 3090ti

Remember, Amdahl's law applies when you are talking about subset features (ray tracing). 
Double also, someone already posted some extrapolated performance
tipple also, so has LTT in their video on it
 

  

On 11/3/2022 at 7:44 PM, Delicieuxz said:

A rough estimate of how the 7900 XTX should perform, taking AMD's word at face-value:

The slides that talked about raytracing mentioned 50% (or 1.5x) a boost in performance per core, and since we have 50% more cores, thats 2x (ray tracing parallelizes very well). But that never would translate to a 2x increase in framerate because a frame isn't just ray tracing, its drawing the damn frame first and formost. 

 

https://en.wikipedia.org/wiki/Amdahl's_law#:~:text=In computer architecture%2C Amdahl's law,system whose resources are improved.

 

[Shzzit]

10 minutes ago, starsmine said:

What slide said 1.7x ray tracing. 
They had slides that said up to 1.7x overall performance increase in some games. Not a single slide said that about ray tracing performance. And it was compared to the 6950, not 6900. AND, again, what sort of benchmark are you using to get that 1.7x of a 6900 is less then a 3090ti

Remember, Amdahl's law applies when you are talking about subset features. 

Ya up to 1.7x but then under the games they show 3 games with (ray tracing) and 3 with out.  And ya i guess it says 1.5x 1.5x 1.6x for those.   So i guess me doing 1.7x gave it a little higher fps lol. 

 

 

So i looked up Doom eternal, one of the games they showed 1.5x.  And  AMD gets 194 AVG at 4k ultra settings.  So 194 X 1.5x = 291 fps.  Same game same settings 4090 gets 400 avg fps.   Thats not even with RT.

Dont see any RT benchmarks with 4090 for that game. 

 

with RT im guessing its even further apart. 

[Imbadatnames]

7 minutes ago, starsmine said:

What slide said 1.7x ray tracing. 
They had slides that said up to 1.7x overall performance increase in some games. Not a single slide said that about ray tracing performance. And it was compared to the 6950, not 6900. AND, again, what sort of benchmark are you using to get that 1.7x of a 6900 is less then a 3090ti

Remember, Amdahl's law applies when you are talking about subset features (ray tracing). 
Double also, someone already posted some extrapolated performance
tipple also, so has LTT in their video on it
 

  

The slides that talked about raytracing mentioned 50% (or 1.5x) a boost in performance per core, and since we have 50% more cores, thats 2x (ray tracing parallelizes very well). But that never would translate to a 2x increase in framerate because a frame isn't just ray tracing, its drawing the damn frame first and formost. 

 

https://en.wikipedia.org/wiki/Amdahl's_law#:~:text=In computer architecture%2C Amdahl's law,system whose resources are improved.

 

It depends on the games they used, the resolutions and if they uses FSR or not. Some games the 7900XTX will perform 1.7x better and other games it won’t. 

[starsmine]

10 minutes ago, Shzzit said:

Ya up to 1.7x but then under the games they show 3 games with (ray tracing) and 3 with out.  And ya i guess it says 1.5x 1.5x 1.6x for those.   So i guess me doing 1.7x gave it a little higher fps lol. 

 

So i looked up Doom eternal, one of the games they showed 1.5x.  And 6950 gets 194 AVG at 4k ultra settings.  So 194 X 1.5x = 291 fps.  Same game same settings 4090 gets 400 avg fps.   Thats not even with RT.

Are you trolling? I post (quoted) a graph, and you just pull out numbers out of your ass with zero source
Also LTT's video 

 

10 minutes ago, Imbadatnames said:

It depends on the games they used, the resolutions and if they uses FSR or not. Some games the 7900XTX will perform 1.7x better and other games it won’t. 

did you quote the wrong person?
I said, and I quote 

18 minutes ago, starsmine said:

They had slides that said up to 1.7x overall performance increase in some games. Not a single slide said that about ray tracing performance.

Also, that chart has no mention of FSR, they have other slides where the talk about FSR and improvements there.

[Shzzit]

15 minutes ago, starsmine said:

What slide said 1.7x ray tracing. 
They had slides that said up to 1.7x overall performance increase in some games. Not a single slide said that about ray tracing performance. And it was compared to the 6950, not 6900. AND, again, what sort of benchmark are you using to get that 1.7x of a 6900 is less then a 3090ti

Remember, Amdahl's law applies when you are talking about subset features (ray tracing). 
Double also, someone already posted some extrapolated performance
tipple also, so has LTT in their video on it
 

  

The slides that talked about raytracing mentioned 50% (or 1.5x) a boost in performance per core, and since we have 50% more cores, thats 2x (ray tracing parallelizes very well). But that never would translate to a 2x increase in framerate because a frame isn't just ray tracing, its drawing the damn frame first and formost. 

 

https://en.wikipedia.org/wiki/Amdahl's_law#:~:text=In computer architecture%2C Amdahl's law,system whose resources are improved.

 

Ya im wondering if those numbers are with FSR.   and they say UP TO. lol so is that MAX fps and not average?  Very hard to know, even NVIDIAS slides were bull shit.  UP TO, to me doesment mean average but max. 

 

And 3 of those games say RAY TRACING.  1.5 x 1.5x 1.6x  so i was off b y .1 lol

 

 

Numbers out of my ass?  Wow its clear were not debating and ur just angry lol.  Fanboy?

[porina]

On the raster performance, I would throw in a thought here. Higher doesn't always make a product better. I view performance in about 3 zones: 1, not enough. 2, usable, 3, excessive. For most things, people aim to get at least "good enough" at lowest cost. Beyond that isn't so optimal although a moderate headroom to grow into isn't a bad thing. I think this is AMD's GPU strategy weakness. They focus on raster, but are lacking in other areas like RT. Personally I don't think I need more raster perf than a 3080 as a native 4k60+ target. What I'd like is more RT perf. Much more. This is the same trick AMD did with early gen Ryzen. It was great at int, but collapsed if you gave them fp workloads. They may catch up eventually, but RDNA 3 is not that product.

 

 

Also following up on the PCWorld interview, where it was said the 7900 is going up against the 4080, the question then is, why not call them 7800? A bit of naming inflation?

[Shzzit]

@starsmine

So ill do what you did. And from same site.

 

 

 

6950xt gets 46  X 1.5 = 69  with 4090 getting 117 LOL

 

6950 gets 84,  x 1.5  = 126  vs 4090 at 175.  

 

 

6950xt gets 12.7 FPS,  x 1.7  = 21.6 FPS.  VS 4090 at 41 lol.  

[Imbadatnames]

E

15 minutes ago, starsmine said:

Are you trolling? I post (quoted) a graph, and you just pull out numbers out of your ass with zero source
Also LTT's video 

 

did you quote the wrong person?
I said, and I quote 

Also, that chart has no mention of FSR, they have other slides where the talk about FSR and improvements there.

No I didn’t. Just because the slide doesn’t say it doesn’t mean anything. Can’t remember who it was, might have been Intel, who were doing events and not telling people they were running the PC sub ambient. Don’t take a companies event keynote as gospel. 

[Shzzit]

ALSO, read the astriks on all these slides.   

 

Slide 2 "RX 7900 XT High performance 4K Raytracing" *benchmarked with FSR* (probably FSR2 on both cards), according to endnote RX-832

 

AMD says they used FSR for 4k benchmarks ,  /facepalm

[StephanTW]

47 minutes ago, Imbadatnames said:

But it’s Minecraft. Even at max with RT you should be comfortably higher than 60 plus

sad r9 290 noises, it can do a light shader with drops to 30 fps XD

[Imbadatnames]

8 minutes ago, StephanTW said:

sad r9 290 noises, it can do a light shader with drops to 30 fps XD

The GPU is like around 10 years old though tbf

[WereCat]

10 minutes ago, porina said:

On the raster performance, I would throw in a thought here. Higher doesn't always make a product better. I view performance in about 3 zones: 1, not enough. 2, usable, 3, excessive. For most things, people aim to get at least "good enough" at lowest cost. Beyond that isn't so optimal although a moderate headroom to grow into isn't a bad thing. I think this is AMD's GPU strategy weakness. They focus on raster, but are lacking in other areas like RT. Personally I don't think I need more raster perf than a 3080 as a native 4k60+ target. What I'd like is more RT perf. Much more. This is the same trick AMD did with early gen Ryzen. It was great at int, but collapsed if you gave them fp workloads. They may catch up eventually, but RDNA 3 is not that product.

 

 

Also following up on the PCWorld interview, where it was said the 7900 is going up against the 4080, the question then is, why not call them 7800? A bit of naming inflation?

Do people actually care about RT that much? From my personal experience nobody uses it. I know few people that have RT capable cards and most of them have not even tried it and the ones that did don't use it.

I found RT to be really nice in Minecraft and that's about the only game where I would use it if I was actually playing it.

I've tried Metro Exodus, DOOM, Quake 2, etc... I only really found it nice in Quake 2 and couldn't really justify it in the other games as the visual difference was not worth the performance hit imo.

 

I only like RT accelerators because it helps a lot in Blender but I'm quite a niche for that.

 

Don't get me wrong, I know there are people who like it but are these also the same people who buy low/mid range cards? Or just people who buy the best cards always so that they can always run max settings? Everybody I know is just a FPS snob playing games on LOW/Medium with high end cards and CPUs even if their FPS would barely budge if they increased settings. But then again, these same people mostly only play Multiplayer games so probably also not a great example.

 

The only thing that really bothers me in games which RT solves are the screenspace reflections if they are done poorly with traditional method of faking it. Most of games these days however do them quite well even without RT though.

 

I'm sure RT is a future but I still think it will take few years for it to become mainstream and widely adopted. From my point of view it's still in the category of "nice to have" unlike DLSS or FSR which are actually beneficial in a quite big way.

[Shzzit]

Just now, Imbadatnames said:

The GPU is like around 10 years old though tbf

HAHA good old r9 290.  Brings back memory's when i had a 290x lol.  Or way back with my hd5970. 

[starsmine]

41 minutes ago, Shzzit said:

@starsmine

So ill do what you did. And from same site.

 

 

 

6950xt gets 46  X 1.5 = 69  with 4090 getting 117 LOL

 

Correct, that chart should have used these numbers since ray tracing.
 

 

41 minutes ago, Shzzit said:

6950xt gets 12.7 FPS,  x 1.7  = 21.6 FPS.  VS 4090 at 41 lol.  

Though why you did that here, who fucking knows since it was not on in the given benchmark.


AMD addressed this game with RT on with the FSR2 slides. 


Also something people might find interesting

 

Foot note 

1. Testing done by AMD performance labs November 2022 on RX 7900 XTX, on 22.40.00.24 driver, AMD Ryzen 9 7900X processor, 32GB DDR5-6000MT, AM5 motherboard, Win11 Pro with AMD Smart Access Memory enabled. Tested at 4K in the following games: Call of Duty: Modern Warfare, God of War, Red Dead Redemption 2, Assassin’s Creed Valhalla, Resident Evil Village, Doom Eternal. Performance may vary. RX-842

 

36 minutes ago, Shzzit said:

ALSO, read the astriks on all these slides.   

 

Slide 2 "RX 7900 XT High performance 4K Raytracing" *benchmarked with FSR* (probably FSR2 on both cards), according to endnote RX-832

 

AMD says they used FSR for 4k benchmarks ,  /facepalm

Sir... in what world is 822, 832?


FIrst, it does not mater if FSR is used here or not, since it would be used on both, in fact, FSR might hide performance gains as this is a ratio. 1.7x boost in base fps may not equate to a 1.7x boost with FSR on when ratioed. Hard to say, but also, does not really mater.

Second yes its weird that the games are not all in the foot note. 
Third,  FSR was not enabled for these runs based off the foot note
 

for the slide with end not 832, yes, FSR used. NOT THIS ONE

Notice that they say "AND FSR performance mode enabled" when its on. 

Personal end note. NO ONE USES MAX FPS as a benchmark. I do not know why you brought that up. that is literally NOT a thing in the benchmarking space.
Another end note. 
The rest of us here are ok with ballpark numbers at this point. Getting this much into the weeds is just asking for trouble since to much is unknown, and driver optimization's are still happening constantly meaning all these numbers will be bunk by launch. 

 

 

 

 

[Shzzit]

3 minutes ago, starsmine said:

Correct, that chart should have used these numbers since ray tracing.
 

 

Though why you did that here, who fucking knows since it was not on in the given benchmark.


AMD addressed this game with RT on with the FSR2 slides. 


Also something people might find interesting

 

Foot note 

1. Testing done by AMD performance labs November 2022 on RX 7900 XTX, on 22.40.00.24 driver, AMD Ryzen 9 7900X processor, 32GB DDR5-6000MT, AM5 motherboard, Win11 Pro with AMD Smart Access Memory enabled. Tested at 4K in the following games: Call of Duty: Modern Warfare, God of War, Red Dead Redemption 2, Assassin’s Creed Valhalla, Resident Evil Village, Doom Eternal. Performance may vary. RX-842

 

Sir... in what world is 822, 832?


FIrst, it does not mater if FSR is used here or not, since it would be used on both, in fact, FSR might hide performance gains as this is a ratio. 1.7x boost in base fps may not equate to a 1.7x boost with FSR on when ratioed. Hard to say, but also, does not really mater.

Second yes its weird that the games are not all in the foot note. 
Third, no FSR was enabled for these runs based off the foot not
 

for the slide with end not 832, yes, FSR used. NOT THIS ONE

Notice that they say "AND FSR performance mode enabled" when its on. 

Personal end note. NO ONE USES MAX FPS as a benchmark. I do not know why you brought that up. that is literally NOT a thing in the benchmarking space.
Another end note. 
The rest of us here are ok with ballpark numbers at this point. Getting this much into the weeds is just asking for trouble since to much is unknown, and driver optimization's are still happening constantly meaning all these numbers will be bunk by launch. 

I agree on most parts, i just found it weird they would say UP TO and not average.   Thanks for all the additional data. 

[porina]

22 minutes ago, WereCat said:

Do people actually care about RT that much? From my personal experience nobody uses it. I know few people that have RT capable cards and most of them have not even tried it and the ones that did don't use it.

Because of the performance hit, people don't want to use it unless you have high end hardware. If RT came at no performance hit compared to raster, who wouldn't want it? So the question becomes at what point is it worth it?

 

I'm too lazy to look up exact number right now, but RT capable is in the tens of % on Steam Hardware Survey. Plus you also have the pool of current gen consoles, even if they have a low performance courtesy of AMD.

 

The be fair, there is no RT enabled game I play right now. But my gaming interests are a bit niche and I'm not into the interactive movies they call AAA gaming these days. There are many older games I'd really love to have the RT treatment, but it isn't going to happen. Edit: Portal RTX is supposed to be out this month, gives me an excuse to revisit it.

 

Still, my main argument is, on the higher end at least I don't need more raster performance. I certainly wont complain at getting it cheaper, but to move forwards we need better RT performance. RT to me adds far more value than 8k gaming.

[starsmine]

2 minutes ago, Shzzit said:

I agree on most parts, i just found it weird they would say UP TO and not average.   Thanks for all the additional data. 

"up to" is industry standard because people run AV in the background and will sue for not getting the exact FPS advertised.
Its been this way for decades. 
Its not  a cop out, its the realities of most users run shit in the background. 

[StephanTW]

21 minutes ago, Imbadatnames said:

The GPU is like around 10 years old though tbf

9 in like 9 days XP

 

 

21 minutes ago, Shzzit said:

Brings back memory's when i had a 290x lol.

yeah its still running, just quite noticeable lagging behind with new titles and features
 

[Shzzit]

2 minutes ago, starsmine said:

"up to" is industry standard because people run AV in the background and will sue for not getting the exact FPS advertised.
Its been this way for decades. 
Its not  a cop out, its the realities of most users run shit in the background. 

That makes sense.  Just like Nvidia's slides im taking it all with a grain of salt.  But the price for 999 does seem good vs Greedy ass Nvidia. 

[WereCat]

10 minutes ago, porina said:

 

 

The be fair, there is no RT enabled game I play right now. But my gaming interests are a bit niche and I'm not into the interactive movies they call AAA gaming these days. There are many older games I'd really love to have the RT treatment, but it isn't going to happen.

 

Still, my main argument is, on the higher end at least I don't need more raster performance. I certainly wont complain at getting it cheaper, but to move forwards we need better RT performance. RT to me adds far more value than 8k gaming.

Well, I can definitely agree with that.

 

10 minutes ago, porina said:

Because of the performance hit, people don't want to use it unless you have high end hardware. If RT came at no performance hit compared to raster, who wouldn't want it? So the question becomes at what point is it worth it?

Really depends. I often don't use higher settings if I can't see the difference vs lower settings even if it wouldn't affect my FPS. If I really had no performance hit with RT and there would be substantial visual difference like in Minecraft then I would definitely use it.

If the difference was like no RT vs RT in Metro Exodus then I would likely not even bother to enable it. I would rather games have better draw distance with no pop-in objects when looking in distance rather than good RT right now since that's the thing that bothers me most in RPG games... depending on game even mods don't fix it.

 

Most things that bother me are pop-in, shimmer/flicker (mostly solved by DLSS or good AA).

[xAcid9]

12 minutes ago, porina said:

There are many older games I'd really love to have the RT treatment, but it isn't going to happen. Edit: Portal RTX is supposed to be out this month, gives me an excuse to revisit it.

I wish AMD or Nvidia create something similar to Pascal's RTGI shader but able to utilize RT cores and implement them so you can toggle and adjust the parameters directly from Radeon/GE overlay. It's quite essentials for many games I play where the AO implementation is crap. 

[porina]

10 minutes ago, WereCat said:

I often don't use higher settings if I can't see the difference vs lower settings even if it wouldn't affect my FPS. 

Kinda get that. In most cases the difference between high and v.high/ultra is not really noticeable unless you do a side by side comparison. If you're on the limit on performance, that headroom might be more comfortable.

 

Note I said I don't think I need more raster perf than a 3080. I still don't have one, but a 3070. I have to compromise on settings at 4k to get enough framerate. Games modern enough to have RT will also have DLSS so that helps offset it sufficiently. It's slightly older games without DLSS I find the bigger problem.

 

10 minutes ago, WereCat said:

If the difference was like no RT vs RT in Metro Exodus then I would likely not even bother to enable it. 

I only have Metro Exodus EE which requires RT. That seems really well optimised. I don't play it, but I have it.

 

10 minutes ago, WereCat said:

Most things that bother me are pop-in, shimmer/flicker (mostly solved by DLSS or good AA).

I'd agree those things are undesirable, but add stutter to the list. However we're getting into game dev choices more so than GPUs at that point. More GPU power could overpower it I guess.

 

5 minutes ago, xAcid9 said:

I wish AMD or Nvidia create something similar to Pascal's RTGI shader but able to utilize RT cores and implement them so you can toggle and adjust the parameters directly from Radeon/GE overlay. It's quite essentials for many games I play where the AO implementation is crap. 

I've not looked in detail at RTGI but presuming it is giving RT-like effects using traditional means, it seems unlikely real RT hardware would be able to assist it.

[Imbadatnames]

39 minutes ago, starsmine said:

Correct, that chart should have used these numbers since ray tracing.
 

 

Though why you did that here, who fucking knows since it was not on in the given benchmark.


AMD addressed this game with RT on with the FSR2 slides. 


Also something people might find interesting

 

Foot note 

1. Testing done by AMD performance labs November 2022 on RX 7900 XTX, on 22.40.00.24 driver, AMD Ryzen 9 7900X processor, 32GB DDR5-6000MT, AM5 motherboard, Win11 Pro with AMD Smart Access Memory enabled. Tested at 4K in the following games: Call of Duty: Modern Warfare, God of War, Red Dead Redemption 2, Assassin’s Creed Valhalla, Resident Evil Village, Doom Eternal. Performance may vary. RX-842

 

Sir... in what world is 822, 832?


FIrst, it does not mater if FSR is used here or not, since it would be used on both, in fact, FSR might hide performance gains as this is a ratio. 1.7x boost in base fps may not equate to a 1.7x boost with FSR on when ratioed. Hard to say, but also, does not really mater.

Second yes its weird that the games are not all in the foot note. 
Third,  FSR was not enabled for these runs based off the foot note
 

for the slide with end not 832, yes, FSR used. NOT THIS ONE

Notice that they say "AND FSR performance mode enabled" when its on. 

Personal end note. NO ONE USES MAX FPS as a benchmark. I do not know why you brought that up. that is literally NOT a thing in the benchmarking space.
Another end note. 
The rest of us here are ok with ballpark numbers at this point. Getting this much into the weeds is just asking for trouble since to much is unknown, and driver optimization's are still happening constantly meaning all these numbers will be bunk by launch. 

 

 

 

 

Again what settings, what power targets, what method of average are they using? Was it a canned benchmark or a manual run, what methods were used to make a manual run as repeatable as possible? 
 

It is not in AMDs best interest to make the 6000 series look good therefore you shouldn’t assume they are making every test fair and representative. It is also literally on the slide “results may vary”. Take your fanboy hat off and stop reading what you want.