The thread which sometimes talks about RDNA2

[Purgatory]

I want to see Flight Simulator benchmarks. Is anyone aware of any?

One of the main reasons I switched from AMD to NVIDIA was flight simulator (FSX) as AMD broke it in the past when I went to a GTX 580 to fix it again I was on a ATI Radeon HD 5870 and when I complained on the AMD forums with the issues they banned me there and I only highlighted the issues and they went on a banning spree then to silence people with known issues. Also FS2020 and FSX are very similar in the code so guessing Nvidia is still going to work best on it..

I'm waiting to see the same thing but have my doubts AMD and their drivers will make it work 100% right .. but will see, but I'm kinda worried to go out blind to buy a 6900xt and find the same issues as in the past with their drivers for FS2020 and older sims that I like to use and so far Nvidia is proven to work right on FSX,X-Plane,Prepar3D, etc .. AMD had problems with Prepar3D too and that's basically FSX tuned by Lockheed Martin.

That's actually the only thing that is still swinging me to a 3090 and I have a new 49" FreeSync Premium Pro monitor that would work better with a AMD card for sure but again Nvidia allows FreeSync to be used now too, but... who knows what they will do in the future with drivers and may turn all that off for FreeSync users ..

 

[FoxEye]

Soo... do we think these cards will actually exist or will they be mythical creatures like the 3000 series?

 

[Chuk_Chuk]

Soo... do we think these cards will actually exist or will they be mythical creatures like the 3000 series?

AMD didn't rush their launch so I expect decent stock. Should last about 5-10 mins before it's sold out till 2021

 

[Nablaoperator]

Some more interesting stuff on Infinity Cache from AMD's press release:

Seems like the effective bandwidth is the sum total of infinity cache and GDDR6 bandwidth as if they are operating in parallel. The cache hit is expected to stabilise at 58% in infinite time.

"AMD Infinity Cache technology that offers up to 2.4X greater bandwidth-per-watt compared to GDDR6-only AMD RDNA™. Measurement calculated by AMD engineering, on a Radeon RX 6000 series card with 128 MB AMD Infinity Cache and 256-bit GDDR6. Measuring 4k gaming average AMD Infinity Cache hit rates of 58% across top gaming titles, multiplied by theoretical peak bandwidth from the 16 64B AMD Infinity Fabric channels connecting the Cache to the Graphics Engine at boost frequency of up to 1.94 GHz. RX-547"

https://www.amd.com/en/press-releas...-pc-gaming-amd-radeon-rx-6000-series-bringing

 

[Katalyst]

Howcome the 5000 cpu's popped up instantly on reseller sites, but nothing for the 6000 gpus? id assume this coming week we'll see?

 

[BogEyes]

When do we think actually pricing in £ will be?

I feel like I may have the urge to buy a 6800 if it's closer to £500 than £600.

 

[FoxEye]

When do we think actually pricing in £ will be?

I feel like I may have the urge to buy a 6800 if it's closer to £500 than £600.

Depends on stock levels and where you choose to buy it from.

I'd expect to have to budget close to £600.

 

[kitfit1]

One of the main reasons I switched from AMD to NVIDIA was flight simulator (FSX) as AMD broke it in the past when I went to a GTX 580 to fix it again I was on a ATI Radeon HD 5870 and when I complained on the AMD forums with the issues they banned me there and I only highlighted the issues and they went on a banning spree then to silence people with known issues. Also FS2020 and FSX are very similar in the code so guessing Nvidia is still going to work best on it..

I'm waiting to see the same thing but have my doubts AMD and their drivers will make it work 100% right .. but will see, but I'm kinda worried to go out blind to buy a 6900xt and find the same issues as in the past with their drivers for FS2020 and older sims that I like to use and so far Nvidia is proven to work right on FSX,X-Plane,Prepar3D, etc .. AMD had problems with Prepar3D too and that's basically FSX tuned by Lockheed Martin.

That's actually the only thing that is still swinging me to a 3090 and I have a new 49" FreeSync Premium Pro monitor that would work better with a AMD card for sure but again Nvidia allows FreeSync to be used now too, but... who knows what they will do in the future with drivers and may turn all that off for FreeSync users ..

To be honest, the only thing that annoys me about FS2020 is the lack of DX12. Considering FS2020 was a collaboration between AMD and MS in the first place, i simply do not understand why DX12 has not been enabled. DX12 on it's own would totally transform things regardless of what GPU was being used.

 

[Chuk_Chuk]

Some more interesting stuff on Infinity Cache from AMD's press release:

Seems like the effective bandwidth is the sum total of infinity cache and GDDR6 bandwidth as if they are operating in parallel. The cache hit is expected to stabilise at 58% in infinite time.

"AMD Infinity Cache technology that offers up to 2.4X greater bandwidth-per-watt compared to GDDR6-only AMD RDNA™. Measurement calculated by AMD engineering, on a Radeon RX 6000 series card with 128 MB AMD Infinity Cache and 256-bit GDDR6. Measuring 4k gaming average AMD Infinity Cache hit rates of 58% across top gaming titles, multiplied by theoretical peak bandwidth from the 16 64B AMD Infinity Fabric channels connecting the Cache to the Graphics Engine at boost frequency of up to 1.94 GHz. RX-547"

https://www.amd.com/en/press-releas...-pc-gaming-amd-radeon-rx-6000-series-bringing

Any idea what the standard cache miss rate is?

 

[TheRealDeal]

If you actually check the twitter pics, you'd realize yourself that the AMD results aren't with ULTRA preset (for the other settings) like wcctech implies. The ultra symbol (which you can see at the first texture setting ) cannot be found in any of the other settings, it's also running no AA, no screen effects, no Lens flares and High not ULTRA LOD and a few other different settings, unlike the NV which is maxed out.

In other words, 6800 will be close to 2080ti RT perf WITHOUT DLSS (which means it's good, but not great as the article implies) , but hey, keep hyping yourself up, this is going into the old 480 beating 980ti territory.



Also

My 2080 @ 1440p comparable to the twitter pics settings - NO DLSS (menu like i mentioned before is full 3440x1440 native, game was run at 2560x1440 and it stretched). A 2080ti will easily get to the same 80 fps with RT on.

Did you miss the part in the article where they stated this.

In terms of performance, the GPU scored 46 FPS at 4K and 80 FPS at WQHD (1440p) resolution in the title. For comparison, we tested the GeForce RTX 3070 in the same benchmark with a similar CPU configuration and the same settings and got 54 FPS at 4K and 85 FPS at WQHD. One thing to note is that the results were with DLSS enabled so you can expect lower performance for the card with the feature enabled. You can see our results below.

So they ran the same settings with a 3070 which is pretty much the same as a 2080ti in performance.

So they have 3070 and 2080ti results, the only thing in dispute is the 6800 results.

 

[shankly1985]

I still wonder what happened to amd using its SP cores for ray tracing, they claimed they had some way of accelerating ray tracing without needing extra hardware - then did a quick bait and switch by giving us cards with the same design as Nvidia and acting like they never said they didn't need fixed function hardware

?What? It was always known by the blueprints that AMD was going to be using the CU for Ray tracing not the SP this is how AMD will have the full GPU performance no matter the task.

 

[4K8KW10]

AMD's RA (Ray Accelerator) takes very small part of the transistors budget, anyways:


https://videocardz.com/newz/amd-ray...dia-rt-core-in-this-dxr-ray-tracing-benchmark

 

[Purgatory]

To be honest, the only thing that annoys me about FS2020 is the lack of DX12. Considering FS2020 was a collaboration between AMD and MS in the first place, i simply do not understand why DX12 has not been enabled. DX12 on it's own would totally transform things regardless of what GPU was being used.

If you read the FS2020 forums the developers are working on the DX12 and will be about soon, but they have also stated anyone expecting performance boosts not to expect them as DX12 will not help in that department but will add ray tracing and some more DX12 features but no performance boost to frame rates, so basically I read that to mean it will be prettier to look at and probably worse performance in DX12 mode. FSX did same thing too in past with it's broken DX10 "update" that actually made things worse and most people stuck to DX9 on it even to this day.


The FS2020 forums the devs there are stating a lot of things if you go read it and honestly there is a lot broken is FS2020 that needs fixing before the graphics updates and DX12, the planes have a lot of broken systems and other issues that didn't exist in FSX and previous MS flight sims. Still a lot of work needs to be done on it and I think only the true flight sim users have noticed the real issues with it, a lot of people are just using it to fly around and enjoy the scenery and not realising there are many broken things in it to make it a true flight simulator. Anyways I hope they get all the important issues fixed as so far anyone that uses the flight systems correctly will realise all the faults on the planes.

 

[4K8KW10]

One of the main reasons I switched from AMD to NVIDIA was flight simulator (FSX) as AMD broke it in the past when I went to a GTX 580 to fix it again I was on a ATI Radeon HD 5870 and when I complained on the AMD forums with the issues they banned me there and I only highlighted the issues and they went on a banning spree then to silence people with known issues. Also FS2020 and FSX are very similar in the code so guessing Nvidia is still going to work best on it..

I'm waiting to see the same thing but have my doubts AMD and their drivers will make it work 100% right .. but will see, but I'm kinda worried to go out blind to buy a 6900xt and find the same issues as in the past with their drivers for FS2020 and older sims that I like to use and so far Nvidia is proven to work right on FSX,X-Plane,Prepar3D, etc .. AMD had problems with Prepar3D too and that's basically FSX tuned by Lockheed Martin.

That's actually the only thing that is still swinging me to a 3090 and I have a new 49" FreeSync Premium Pro monitor that would work better with a AMD card for sure but again Nvidia allows FreeSync to be used now too, but... who knows what they will do in the future with drivers and may turn all that off for FreeSync users ..

If you read the FS2020 forums the developers are working on the DX12 and will be about soon, but they have also stated anyone expecting performance boosts not to expect them as DX12 will not help in that department but will add ray tracing and some more DX12 features but no performance boost to frame rates, so basically I read that to mean it will be prettier to look at and probably worse performance in DX12 mode. FSX did same thing too in past with it's broken DX10 "update" that actually made things worse and most people stuck to DX9 on it even to this day.


The FS2020 forums the devs there are stating a lot of things if you go read it and honestly there is a lot broken is FS2020 that needs fixing before the graphics updates and DX12, the planes have a lot of broken systems and other issues that didn't exist in FSX and previous MS flight sims. Still a lot of work needs to be done on it and I think only the true flight sim users have noticed the real issues with it, a lot of people are just using it to fly around and enjoy the scenery and not realising there are many broken things in it to make it a true flight simulator. Anyways I hope they get all the important issues fixed as so far anyone that uses the flight systems correctly will realise all the faults on the planes.

Are these flight simulators intended for professional level training or substitution of such? Or are they just fun for gaming purposes?

 

[Nablaoperator]

Any idea what the standard cache miss rate is?

Nope, but L0/L1/L2 cache miss should be higher (theoretically speaking) because the higher up you'd move access frequency would reduce.

However, the measurements assume that a non zero steady state is reached (58%) through observational means only. I will update if I stumble upon a mathematical proof.

 

[chris85oc]

Did you miss the part in the article where they stated this.




So they ran the same settings with a 3070 which is pretty much the same as a 2080ti in performance.

So they have 3070 and 2080ti results, the only thing in dispute is the 6800 results.

They ran 3070 benches with ultra settings (and RT on high) AA etc. The leaked twitter pic has multiple settings disabled, including AA ( same symbol as stereoscopic 3d which is disabled ) some settings on high not ultra etc. You can compare yourself by checking the symbols.

So they compared max settings on 3070 with AA ( high RT) with lower settings no AA on 6800. ‘Journalism’ clickbait.

Either way, people can do the math themselves if they want to, but easier to just copy paste wcc and spread misinformation.

 

[Purgatory]

Are these flight simulators intended for professional level training or substitution of such? Or are they just fun for gaming purposes?

Prepar3D is training and built around FSX but Lockheed Martin has tuned it. It's not sold as entertainment and the license doesn't allow that. Have a read here :- https://www.prepar3d.com/product-overview/prepar3d-license-comparison/

FSX , FS2020 , X-Plane etc are sold as entertainment products really but they are built as a true working simulator and people were even trained to fly on Microsoft Flight simulator in past in flight schools and other "entertainment" Flight Sims...

They are not really games but like anything can be used as an entertainment product or a training product too or just for someone interested in flight to have a play with a plane and learn how to fly it or you can allow the computer to do most things and just point the plane in the direction you want, but that's no fun really, it's fun to learn how to use the flight systems and learn to land correctly and take off correctly and guide yourself to other airports or countries correctly, follow air traffic control etc etc .. There is a lot to it if it really interests you and lots of planes to try and learn all the different things about them and their systems.

 

[Lokken86]

Same pal. Although it should last at least three years if I did splash out.

Yeah I'd be keeping it for quite some time. One thing I have learned over the years is to stand by my purchases and not get tempted by the upgrade bug. I've had this CPU for going on 3 years I think and the 5700xt was a gift. Before that my vega 64 was going strong near 3 years.

 

[Joxeon]

I can not see that RT performance for the AMD cards being right ..as AMD would have most certainly been on about this on there announcement surly.. I hope it is right guess not long to wait

I think it's more likely down to the 6800 being faster than the RTX 3070 to begin with and not so much faster RT performance.

If it's 15% faster in rasterisation and 15% slower in RT then it will be on par etc.

 

[Joxeon]

https://www.amd.com/en/gaming/graphics-gaming-benchmarks enjoy


Something fishy going on with AMDs benchmarks for the 3080 and 3090 if you use Techpowerups benchmarks and match them up the same with same settings and API.

They say they were all run on the x570 motherboard and 5900x CPU, but the figures for the 3080 and 3090 should be better as it's a better cpu than they used in their test setup which is a 9900k..

So the 3080 and 3090 results should be better than techpowerups but they read a lot worse and the 6900 xt, 6800 xt and 6800 humm if you match up the numbers to known benchmarks of the 3080 and 3090. I hope AMD are not playing us for fools.. Anyways link there from their benchmarks and use your favourite site with benchmarks to compare.

They likely handicapped the nvidia cards by using slow ram whereas the AMD cards are not effected as they had SAM.