AMD Navi 23 ‘NVIDIA Killer’ GPU Rumored to Support Hardware Ray Tracing, Coming Next Year

[Howling]

You guys make my heart hurt

 

[LoadsaMoney]

Well 5700 series went EOL so they going to disappear that's why prices are going up too.

Nope.

https://videocardz.com/newz/amd-radeon-rx-5700-series-reach-end-of-life

 

[LoadsaMoney]

Well, I believe Variable Rate Shading alone can give very good performance increases if done right there to decrease the quality slightly but increase the performance more.
RX 5700 XT doesn't support Variable Rate Shading or any known performance optimisation via lowered image quality, maybe Radeon Boost but that doesn't work for static parts.
It only downscales the resolution in race simulators, or when the mouse moves.

 

[james.miller]

@james.miller what you're talking about is a "percentage points difference" not a "percentage gain" which is what you need to use when calculating how much performance GPU X needs to gain to compete with GPU Y.

Percentage gain = percentage increase. You said difference, difference is neither increase or decrease.

If GPU X is 50 FPS and GPU Y 75 FPS the difference is not 25%, its 50%, the 50 FPS GPU needs to be half as fast again (50% faster) to match the 75 FPS GPU.

The increase is 50%. The percentage difference would be the difference between the two divided by the average.

 

[elroberto]

Going from a 980ti to a 2080ti and from a 2600k to a 9900k was like time travel for me 2 years ago!

I'm going from a Piledriver 8320 to an 8-12 core Zen3 and whichever graphics card comes out the best at circa £500...my mind's going to be blown.

 

[humbug]

The increase is 50%. The percentage difference would be the difference between the two divided by the average.

Yes, 50%, you said that as if its not exactly what i wrote there. You're being pedantic because i said difference not increase.

 

[Lokken86]

What your missing @Lokken86 is as @humbug mentioned the HUB retest of all the games put the 5070XT better than the 2070s in some games and is not 11% slower anymore IIRC.

Yeah I know that it's different now I saw the hardware unboxed video it's like 1% difference now. I was saying "had" and not "is" This was just to help out to those who are misinterpreting those charts so the numbers were irrelevant really.

 

[james.miller]

I'm being accurate because you got it wrong, humbug. If you are trying to give people lessons then you need to be getting it right. Nothing pedantic about that.

 

[Zeeflyboy]

You guys make my heart hurt

You tried! Horse to water and all that jazz...

 

[gpuerrilla]

Yeah I know that it's different now I saw the hardware unboxed video it's like 1% difference now. I was saying "had" and not "is" This was just to help out to those who are misinterpreting those charts so the numbers were irrelevant really.

Sorry I came across poorly there, it wasn't a dig at you, was meant to add to your point about statistics %, with the added spin of its actually not that wide anymore.

 

[humbug]

In a few days we will be distracted away from all this.... thank ####

 

[Lokken86]

Sorry I came across poorly there, it wasn't a dig at you, was meant to add to your point about statistics %, with the added spin of its actually not that wide anymore.

Nah no worries it wasn't taken as a dig . Yes it does make a difference when we try to estimate the Big Navi performance. It's not quite as much as people are saying if we assume the 5700xt gained 10% performance then we need 81% on top to reach 3080@4k levels. Looking at the 3080 founders review performance summary.

 

[gpuerrilla]

Yep spot on. Its realistic for RDNA2 to hit that aswell. Would love a liquid cooler edition but then it would be expensive. Will be happy with something at £600 mark for 3080 performance.

 

[Lokken86]

Yep spot on. Its realistic for RDNA2 to hit that aswell. Would love a liquid cooler edition but then it would be expensive. Will be happy with something at £600 mark for 3080 performance.

I think so too. I get at least 10% more performance still then that with 2100mhz on my 5700xt. At that point it would be already be at 60.5% the performance of the 3080 on the chart or roughly similar to the regular 2080. Then we only need 65% more to catch a 3080. I'm sure with double the CUs, an increase in IPC and another 100mhz it could get that 65%.

 

[Soapysage]

I think so too. I get at least 10% more performance still then that with 2100mhz on my 5700xt. At that point it would be already be at 60.5% the performance of the 3080 on the chart or roughly similar to the regular 2080. Then we only need 65% more to catch a 3080. I'm sure with double the CUs, an increase in IPC and another 100mhz it could get that 65%.

Which isn't taking into account the at least 50% perf/watt gain from RDNA1 to RDNA2. Full die Navi 22 with all gains considered will be between the 3070 and 3080, leaving Navi 21 to take the fight to the 3080, 3090 and above.

 

[4K8KW10]

Red Dead Redemption II - UHD 2160p
RTX 2070S - 36 +6%
RX 5700 XT - 34

Resident Evil 3 - UHD 2160p
RTX 2070S - 53 +10.4%
RX 5700 XT - 48

Battlefield V - UHD 2160p
RTX 2070S - 61 +0%
RX 5700 XT - 61

Death Stranding - UHD 2160p
RTX 2070S - 61 +3.4%
RX 5700 XT - 59

Gears of War 5 - UHD 2160p
RTX 2070S - 42 +10.5%
RX 5700 XT - 38

Shadow of the Tomb Raider - UHD 2160p
RTX 2070S - 44 +10%
RX 5700 XT - 40

Metro Exodus - UHD 2160p
RTX 2070S - 35 -2.8%
RX 5700 XT - 36

Strange Brigade - UHD 2160p
RTX 2070S - 71 +7.6%
RX 5700 XT - 66

Codemasters Formula 1 - UHD 2160p
RTX 2070S - 67 -1.5%
RX 5700 XT - 68

Flight Simulator - UHD 2160p
RTX 2070S - 26 +30%
RX 5700 XT - 20

Borderlands - UHD 2160p
RTX 2070S - 34 +3%
RX 5700 XT - 33

Far Cry Dawn - UHD 2160p
RTX 2070S - 57 +3.6%
RX 5700 XT - 55

The Witcher Wild Hunt - UHD 2160p
RTX 2070S - 57 +29.5%
RX 5700 XT - 44

In games - on average the RTX 2070S is 8.4% faster than RX 5700 XT across the tested 13 games https://www.guru3d.com/articles_pages/geforce_rtx_3090_founder_review,24.html

And in Compute the RTX 2070S is 30.9% faster on average in 8 benchmarks https://www.guru3d.com/articles_pages/geforce_rtx_3090_founder_review,27.html

Unigine 2 - Superposition - Extreme quality
RTX 2070S - 50 (6722 points) +21.5%
RX 5700 XT - 41 (5532 points)

3DMark: Firestrike - Ultra Preset
RTX 2070S - 6028 +0%
RX 5700 XT - 6060

3DMark Time Spy - Graphics Score
RTX 2070S - 10088 +8%
RX 5700 XT - 9334

Basemark GPU v1.2 Vulkan Score
RTX 2070S - 10488 +23.8%
RX 5700 XT - 8475

Basemark GPU v1.2 OpenGL 4.5 Score
RTX 2070S - 8970 +76.9%
RX 5700 XT - 5071

Basemark GPU v1.2 DX12 Score
RTX 2070S - 9562 +9.4%
RX 5700 XT - 8736

Indigo v4.0.64 SuperCar MSamples/s
RTX 2070S - 21.572 +50.9%
RX 5700 XT - 14.291

Blender v2.82 Classroom in Seconds - lower is better
RTX 2070S - 100 +57%
RX 5700 XT - 175

RX 5700 XT is a joke of a card. It scores 0 also in all CUDA and Ray-tracing benchmarks.

 

[Twinz]

I don't want to start the argument up again so I'll just say this: you're missing the point. Yes, Titans exist as far back as Kepler and they were never billed as the flagship gaming card. That was always the Ti. The point here is Nvidia pulling a mind game with the Ampere's marketing which a lot of people are falling for and fawning over themselves and Uncle Leather Jacket. The narrative is suddenly after 5 generations, a non-Ti card is the flagship model with all of the intentional and distorted comparisons that brings. Of course Ampere is this "massive" leap in price and performance over Turing because the flagship is almost half the price and 50% faster. Who could ask for more, it's amazing! But it's just smoke and mirrors to yet again actually increase prices. Pushing the 3080 as the flagship has sent people scurrying to buy a £700 card that just doesn't exist in any practical terms, forcing them into paying at least £150 more for a version that they can get. Now granted that's still a solid move up from the 2080 Ti, but it's still a distortion and deflection intended to dupe the ill-informed, and fans the flames of idiot fanboyism and brand loyalists.

The simple fact that you've said to me "but the tier game is just that, tiers mean nothing" perfectly illustrates this. It doesn't matter if its a flagship or not, the 800 class card is more expensive yet again. But as long as people say "it's the flagship" it apparently doesn't matter and idiots will blindly open their wallets. The tier "game" is not a game, it's very real product segmentation.

The bottom line is that $700 Ampere dollars are faster than $1200 Turing dollars.

Class. Tier, "flagship"...none of it matters. Ampere money gets you more than Turing money.

 

[Lokken86]

Which isn't taking into account the at least 50% perf/watt gain from RDNA1 to RDNA2. Full die Navi 22 with all gains considered will be between the 3070 and 3080, leaving Navi 21 to take the fight to the 3080, 3090 and above.

Yeah well it's the performance/watt gains that's going to allow for these kind of specs. You could essentially double the die size of a 5700xt and have 80CUs and have 440w but you'd be straight back to 220w with 50% perf/watt improvement..which is good . The extra TDP headroom can be for the core clock and memory increases.

 

[Howling]

You could essentially double the die size of a 5700xt and have 80CUs and have 440w but you'd be straight back to 220w with 50% perf/watt improvement..which is good . The extra TDP headroom can be for the core clock and memory increases.

Jesus ******* wept.... This thread is actually making me die inside with the level of maths... You'd be at 300w, not 220w with a 50% improvement

Edit: changed value, I quoted too high. It's closer 295w but I rounded for ease to 300.

 

[Grim5]

Jesus ******* wept.... This thread is actually making me die inside with the level of maths... You'd be at 330w, not 220w with a 50% improvement

So many people are going to be scratching their heads when these cards release looking nothing like their failed 5th grade maths skills