Faked NVIDIA GeForce RTX 3090 Performance Slides are making Rounds on the WWW

kakiharaFRS

2020-08-31 18:10

commenting for a laugh because I'm currently benchmarking Port Royal on a 1080Ti : 2074points 10.07fps 😳 I don't know if it'll be 2x faster in ray-tracing than a 2080Ti but I sure hope it'll be 6x faster than my 1080Ti 😉

#5823699

TheDeeGee

2020-08-31 18:53

Saw them earlier, looked way too fishy.

#5823703

Noisiv

2020-08-31 19:16

Very weak fake having all kinds of problems. Surprised it lasted this long. And that it got confirmed by a "reliable" leaker. WTH is going on; only 23 more hours.

#5823710

Legacy-ZA

2020-08-31 19:37

Could be that the graphs were rendered with DLSS. 😛

#5823723

Martin5000

2020-08-31 20:23

Looks like AMD is doing the same to Nvidia than Nvidia did to AMD. Raise expectations to get disappointment. Its an old marketing tactic.

#5823726

Denial

2020-08-31 20:27

Martin5000:

Looks like AMD is doing the same to Nvidia than Nvidia did to AMD. Raise expectations to get disappointment. Its an old marketing tactic.

Or its just some random discord guy making a slide.. which is what it was. Why does everything have to be weird 4D backgammon corporate espionage, when it's much more believable that some forum nerd will make stuff up and a bunch of sites will run with it for $$?

#5823735

Exodite

2020-08-31 21:15

Martin5000:

Raise expectations to get disappointment.

Uh, maybe I'm an outlier here but a 2X performance increase at the top end using DLSS + RTX is hardly a high expectation. I'd probably be disappointed if it isn't more than that, given than RTX with Turing has been a huge drag on performance and DLSS being improved with Ampere is definitely the expectation. Now if we're talking 2X in pure raster performance, without DLSS, then yes - that would be a pretty high expectation!

#5823736

neikosr0x

2020-08-31 21:15

Fox2232:

I love when people make jokes... Like measured at 4K while DLSS enabled. (Means 1440p with DLSS upscaling to 4K.)

Doesn't matter if it looks equal or better.

#5823738

Denial

2020-08-31 21:18

Exodite:

Uh, maybe I'm an outlier here but a 2X performance increase at the top end using DLSS + RTX is hardly a high expectation. I'd probably be disappointed if it isn't more than that, given than RTX with Turing has been a huge drag on performance and DLSS being improved with Ampere is definitely the expectation. Now if we're talking 2X in pure raster performance, without DLSS, then yes - that would be a pretty high expectation!

If it's DLSS to DLSS then obviously some part of the increase is raster performance? You're basically comparing 1440p + DLSS impact (Turing) to 1440p + DLSS impact (Ampere) across two cards (if the slide was real). I don't think you can make DLSS much faster.. I'd imagine the biggest portion of the increase in performance is just coming from the base resolution decrease, DLSS adds some latency on top of rendering the frame, but I'm sure it's really small. I doubt you could decrease it much. The only other thing would be increase the quality that DLSS can reconstruct, allowing you to drop the base resolution even further.

#5823750

Denial

2020-08-31 21:55

Fox2232:

Does not look equal, nor better as even newest implementations stand. For now, it is always downgrade in fidelity. In some cases visual downgrade is so big, that when applied to 1440p and then down-scaled to 1080p, you can still see that it is much worse. But hey, you can always blame game developers... all of them. I would not touch DLSS for now, unless my HW was so poor, I could not play game without it.

Compared to TAA'd variant? Disagree.

#5823753

sbacchetta

2020-08-31 22:04

I am not saying those slide are legit, I personally don't care much (in less than 30 hours we will get real answers), but why some dirty artifacts due to bad jpeg compression are a proof of them being fake ? Unless there is other oddities not show in the article ?

#5823754

sbacchetta

2020-08-31 22:06

Denial:

Compared to TAA'd variant? Disagree.

I would say the latest implementation of DLSS is superior 90-95% of the time, for really small detail (far away fence, inside of bicycle wheel in the distance) DLSS fail, but for anything else is better.

#5823755

Astyanax

2020-08-31 22:07

sbacchetta:

I am not saying those slide are legit, I personally don't care much (in less than 30 hours we will get real answers), but why some dirty artifacts due to bad jpeg compression are a proof of them being fake ? Unless there is other oddities not show in the article ?

Because those aren't compression artifacts, they are incorrectly placed pixels artifacts.

#5823757

kakiharaFRS

2020-08-31 22:11

for anyone still reading comments here fun fact : I managed to get my 1080Ti to generate 384Watts running 3DMark Timespy Extreme...a vision of the future ? 😛

#5823758

devastator

2020-08-31 22:15

Legacy-ZA:

Could be that the graphs were rendered with DLSS. 😛

it be fine when its clean up with dlss 2.0 😀:D

#5823761

sbacchetta

2020-08-31 22:21

Astyanax:

Because those aren't compression artifacts, they are incorrectly placed pixels artifacts.

Sorry but I have seen jpg image with that kind of artifacts in the past. The one on the upper part of the zoomed picture look a tad strange, but not impossible either. PS I am a professional retoucher/photographer, you don't have to trust me, but I am not speaking out of my a##

#5823763

Exodite

2020-08-31 22:23

Denial:

If it's DLSS to DLSS then obviously some part of the increase is raster performance?

I suppose it's a bit silly to discuss made-up slides but my reasoning is that RT is by far the most significant limiting factor in certain situations, like Control. And we're all expecting significantly higher RT performance with Ampere. Hence, Ampere wouldn't need much more raster performance do significantly outdo Turing in the examples used - because what performance it does have won't be constrained by RT (or at least not to the same extent). The DLSS aspect is less important, generally speaking it just means the comparison is made at a lower resolution. I would imagine that any performance hit from the technology to be lower with Ampere though.

#5823766

Astyanax

2020-08-31 22:26

sbacchetta:

Sorry but I have seen jpg image with that kind of artifacts in the past. The one on the upper part of the zoomed picture look a tad strange, but not impossible either. PS I am a professional retoucher/photographer, you don't have to trust me, but I am not speaking out of my a##

well theres also the fact that coreteks came out and said it was a troll pic.

#5823771

sbacchetta

2020-08-31 22:40

Astyanax:

well theres also the fact that coreteks came out and said it was a troll pic.

Ok, sorry I didn't read that part. Any ways, in a few dozen of hours we will get actual performance number AND price.

#5823802

Fediuld

2020-09-01 00:32

ViperAnaf:

those graphs probably represent the price not the performance hahahaha

Muahahahahah 😀