CyberPanda
Banned
If I were MS, I would scrap Lockhart, and only focus on Series X IMO.
[DF] In Theory: Can a 4TF Navi GPU Deliver a Next-Gen Console Experience?
- Thread starter Bullet Club
- Start date
01011001
Banned
it wouldn't... jesus christ these people...
cryptoadam
Banned
Thats what I am thinking. Remove the RT for lockhart and you are giving yourself more headroom.
All though I kinda of agree with the end of the video where this might just be MS's silver bullet and are holding it based on what Sony does. If Sony manages to get a cheap system out then we see lockheart, if they are at price parity with MS then lockhart might stay in the deck until MS needs to use it. We may very well see a reverse X/Pro situation this gen, where the cheaper less capable machine launches mid gen to bring in more customers.
I think it comes down to price point for MS.
Sosokrates
Report me if I continue to console war
Also to put things into perspective Lockhart is actually a bigger "leap" then series X at there targeted resolutions.
this gens 1080p console (the PS4) is 1.84 gcn tflops , Lockhart is roughly the equivalent of 6 gcn tflops so 6 ÷ 1.84 = 3.26 , so Lockhart is 3.26x more powerful then a PS4.
Now the series X is about 2.7x more powerful then this gens 4k console (the 1X) , the series X is the equivalent to about 16.2gcn tflops.
So at 1080p Lockhart is a 3.26x leap
And at 4k the series X is a 2.7x leap.
So those who are fearful of Lockhart holding games back next gen should be more concerned with 4k holding games back on the series X
this gens 1080p console (the PS4) is 1.84 gcn tflops , Lockhart is roughly the equivalent of 6 gcn tflops so 6 ÷ 1.84 = 3.26 , so Lockhart is 3.26x more powerful then a PS4.
Now the series X is about 2.7x more powerful then this gens 4k console (the 1X) , the series X is the equivalent to about 16.2gcn tflops.
So at 1080p Lockhart is a 3.26x leap
And at 4k the series X is a 2.7x leap.
So those who are fearful of Lockhart holding games back next gen should be more concerned with 4k holding games back on the series X
ResilientBanana
Member
GPU output does not affect the way games play.
dangopee
Neo Member
Obviously in the case of a "balls to wall" 12TF game that can't even maintain 2160p 100% of the time, the developers will have to lower settings other than just resolution for the 4TF console to have a playable frame rate at 1080p. Resolution isn't the only thing you can lower to improve performance. One big thing is probably this Lockhart thing won't even have raytracing at all. Just the basic screenspace reflections we have in current gen.
Hobbygaming
has been asked to post in 'Grounded' mode.
I'm not talking about the way they play I'm talking about those 12 max theoretical teraflops being pushed in different ways graphically besides higher resolution
Portugeezer
Member
Video paid by M$.
But it was interesting to see how much is wasted on 4k. I hope next gen comes with more scalability for those who do not have/want 4k.
But it was interesting to see how much is wasted on 4k. I hope next gen comes with more scalability for those who do not have/want 4k.
Tiamat2san
Member
I will buy a Lockhart (if it really exists) as my second console for the bedroom.
It's a 1080p tv.
I think it would be more interesting than keeping the one X as a second console because of the CPU, games should be more fluid.
Séries X for the living room though (4K tv)
It's a 1080p tv.
I think it would be more interesting than keeping the one X as a second console because of the CPU, games should be more fluid.
Séries X for the living room though (4K tv)
Sosokrates
Report me if I continue to console war
Another thing to take into account is that a Dev might actually be able to make a more graphically sophisticated @ 1080p on Lockhart then a 4k game on xsx.
People really underestimate how much resources resolution uses.
People really underestimate how much resources resolution uses.
Kenpachii
Member
That's the issue people here don't seem to understand 4k isn't a joke to render stuff at. Witcher 3 runs on a 5700xt at 44 fps on average at 4k that's equalivant towards a underclocked 290 radeon which is 4,8 tflops, get a new architecture and a 4tflop rdna2 will be by far enough to run those games at it.
Honestly 290 was a card that released i think even before the PS4.
Potato Chip PRO
Member
This makes alot of sense especially since kids only want to play games like Fortnite, Minecraft and Roblox.
From the video it sounds like 4k is an enormous resource hog. Is it really worth it and could they be doing something better with the power of XSX at 1440P? Are all games for that system going to be 4k or will there be a mix as it has been with the current gen and 1080p and the previous gen and 720p?
CrustyBritches
Gold Member
I was curious about performance scaling between 4k and 1080p at ~12TF and ~4TF, respectively. I have a 2060 Super that has 2176 CUDA cores and runs 1650MHz stock(7.2TF). Using MSI Afterburner I was able to create a custom frequency curve and limit the frequency and voltage to set values to prevent clock boosting. I verified the set frequency values using MSI Afterburner monitor set to Core Clock. I then verified the frequency scaling and stability using 3DMark TimeSpy, as seen in the following graph:
Ultra Settings:
2080 Super(12TF) @ 4K = 72fps avg
2060 Super(4.4TF) @ 1080p = 75fps avg
High Settings:
2060 Super(4.4TF) @ 1080p = 104fps avg
After each benchmark run, the frequency was verified in Afterburner monitor to confirm it was set and stable at 1010MHz.
Results indicate that, at least on Turing, ~12TF @ 4K perfectly scales to ~4TF @ 1080p using 'Ultra' Settings. In the event that required scaling can't be achieved, reverting to flat 'High' settings provides a 37.7% boost to frame rate.
Afterwards, I used Guru3D reviews and benchmark for The Witcher 3 to test scaling from 4K on 2080 Super(~12TF w/OC) to 1080p on 2060 Super(~4.4TF w/Underclock, Undervolt). I used the settings from Guru3D(Ultra, AA 'ON', SSAO, Hairworks 'OFF') to compare the GPUs using FRAPS benchmark with a ~30sec run through White Orchard.
Ultra Settings:
2080 Super(12TF) @ 4K = 72fps avg
2060 Super(4.4TF) @ 1080p = 75fps avg
High Settings:
2060 Super(4.4TF) @ 1080p = 104fps avg
After each benchmark run, the frequency was verified in Afterburner monitor to confirm it was set and stable at 1010MHz.
Results indicate that, at least on Turing, ~12TF @ 4K perfectly scales to ~4TF @ 1080p using 'Ultra' Settings. In the event that required scaling can't be achieved, reverting to flat 'High' settings provides a 37.7% boost to frame rate.
Last edited:
SportsFan581
Member
VRS changes things quite a bit. You could have a game at "native" 4k that is using considerably fewer resources to render than an equivalent 4k game this gen. Lockhart would be in the same position, depending how demanding the title was, they could target 1440p with VRS being heavily utilized. At least that's the way it appears from what I've read, not an expert on VRS by any stretch of the imagination.
DeepEnigma
Gold Member
Imagine a balls to the wall 1080-1440p game on the 12TF box tho!
Chairman Tickles
Banned
So what they are asking, can what is effectively a Geforce 1060 paired with a ryzen cpu run games coming out in 2020-2021? Sure, but probably only at 1080fps on medium settings. Or 1440p if you cut the fps down to 30.
Last edited:
StreetsofBeige
Member
So what's the summary of the video?
Yay? Or avoid?
Yay? Or avoid?
Leonidas
AMD's Dogma: ARyzen (No Intel inside)
4 Tflop machine will run games better at 1080p than 9.2 Tflop would at 4K.
4 Tflop machine at 1440p is close to the performance of 9.2 Tflop at 4K.
Last edited:
CrustyBritches
Gold Member
Depends on the implementation. The only hands-on I've had is with 3DMark's VRS test suite and Wolfenstein Youngblood. In the 3DMark VRS Tier 1 and Tier 2, it can provide up to around 40% performance improvement. In Wolfenstein Youngblood it's around 5%. VRS should be a big deal with VR, since they can do eye tracking and extreme VRS on the peripheral of your vision. AI-powered upscaling like DLSS will be boon for performance as well. I'm sure MS and Sony will have it. Can give you anywhere from 30-60% based on the implementation. If you're interested, I did a vid on 3DMark VRS Tier 1 and also Wolfenstein Native vs VRS vs DLSS.
---
In the case of XSX target of less than 4k, Lockhart can revert to AI upscaling to 1080p, VRS, slightly lowered settings, or a mix of these different features. I did a screenshot comparison in the game Control to demonstrate the variety of approaches they can take. *Note* This performance on RTX 2060 Super, something in between the XSX and XSS. The frame rate is included to show performance scaling. Also, these are stills for comparison, in motion artifacting will vary:
1. 1080p-Native/High Settings/High RT = 54fps
2. 1080p-DLSS/High Settings/High RT = 89fps
3. 1080p-Native/High Settings/OFF RT = 90fps
4. 1080p-DLSS/High Settings/Medium RT = 103fps
5. 1080p-DLSS/High Settings/Medium RT + Indirect Diffuse Lighting[RT] = 94fps
---
In my opinion, RT 'OFF' is unacceptable in this game. You lose the detailed reflections on the ground, detailed transparent reflections in the office windows, and quality indirect diffuse lighting in the scene. There are 5 RT settings, and imo Contact Shadows are too noisy for use at 1080p resolution.
The best performance settings for Lockhart in the case of sub-4k target on XSX would be "1080p-DLSS/High Settings/Medium RT + Indirect Diffuse Lighting[RT]". The Indirect Diffuse RT Lighting only has ~8% performance penalty, and the extra moodiness it provides is worth it. This mix offers a 74% performance increase over 1080p-Native/High Settings/High RT, and 5% increase over 1080p-DLSS/High Settings/High RT, without the noisy Contact Shadows.
Image on the right has a 74% increase in performance over image on the left.
Last edited:
Sosokrates
Report me if I continue to console war
Yeah, I think we might get some 1440p, 4k reconstructed games on xsx. Weather 4tf will be enough on that scenario depends on how powerful the ps5 is, is the ps5 is also around 12tflop, the 4tf Lockhart version will need cut backs in other settings and lots of VRS in order to reach 1080p.
Kenpachii
Member
Depends on the implementation. The only hands-on I've had is with 3DMark's VRS test suite and Wolfenstein Youngblood. In the 3DMark VRS Tier 1 and Tier 2, it can provide up to around 40% performance improvement. In Wolfenstein Youngblood it's around 5%. VRS should be a big deal with VR, since they can do eye tracking and extreme VRS on the peripheral of your vision. AI-powered upscaling like DLSS will be boon for performance as well. I'm sure MS and Sony will have it. Can give you anywhere from 30-60% based on the implementation. If you're interested, I did a vid on 3DMark VRS Tier 1 and also Wolfenstein Native vs VRS vs DLSS.
---
In the case of XSX target of less than 4k, Lockhart can revert to AI upscaling to 1080p, VRS, slightly lowered settings, or a mix of these different features. I did a screenshot comparison in the game Control to demonstrate the variety of approaches they can take. *Note* This performance on RTX 2060 Super, something in between the XSX and XSS. The frame rate is included to show performance scaling. Also, these are stills for comparison, in motion artifacting will vary:
1. 1080p-Native/High Settings/High RT = 54fps
2. 1080p-DLSS/High Settings/High RT = 89fps
3. 1080p-Native/High Settings/OFF RT = 90fps
4. 1080p-DLSS/High Settings/Medium RT = 103fps
5. 1080p-DLSS/High Settings/Medium RT + Indirect Diffuse Lighting[RT] = 94fps
---
In my opinion, RT 'OFF' is unacceptable in this game. You lose the detailed reflections on the ground, detailed transparent reflections in the office windows, and quality indirect diffuse lighting in the scene. There are 5 RT settings, and imo Contact Shadows are too noisy for use at 1080p resolution.
The best performance settings for Lockhart in the case of sub-4k target on XSX would be "1080p-DLSS/High Settings/Medium RT + Indirect Diffuse Lighting[RT]". The Indirect Diffuse RT Lighting only has ~8% performance penalty, and the extra moodiness it provides is worth it. This mix offers a 74% performance increase over 1080p-Native/High Settings/High RT, and 5% increase over 1080p-DLSS/High Settings/High RT, without the noisy Contact Shadows.
Image on the right has a 74% increase in performance over image on the left.
No raytracing is acceptable if the devs didn't do a horrid job of nuking the non raytracing visuals for gigs and lulz.
The game looks butt ugly without ray tracing because they wanted it to look terrible without raytracing.
Last edited: