AMD at 7nm is here, and at 300W for $700 it only starts to approach the performance of early 2017 PC tech (1080ti). So, what does this means for the PS5/XB3? Well, next gen consoles will reach the performance of ~1070/oc980ti if we are lucky, and the later is 2015 PC tech! Have next gen consoles been pwned already?
As long as the consoles are X86 APU based, they will never ever be able to compete or even compare to high-end PC builds. Only something wacky and exotic could faithfully compete in raw numbers with PC and then you have the issue with developers, largely from the west, not wanting to stray away from their bread and butter. This is just a reality since XBox is a competitor in consoles. You should temper expectations with this reality.
AMD at 7nm is here, and at 300W for $700 it only starts to approach the performance of early 2017 PC tech (1080ti). So, what does this means for the PS5/XB3? Well, next gen consoles will reach the performance of ~1070/oc980ti if we are lucky, and the later is 2015 PC tech! Have next gen consoles been pwned already?
There's another 7nm+ node improvement after the first gen 7nm. Xbox One X used 2nd gen 16 nm improvements after PS4 Pro's 16 nm and both APUs are similar in size i.e.
PS4 Pro's APU
8 CPU Jaguar
40 CU Polaris IP with selected Vega IP e.g. RPM. 36 CU active with 4.2 TFLOPS.
Xbox One X's APU
8 CPU Jaguar with lower latency
44 CU Polaris IP with near Vega ROPS cache design. 40 CU active with 6 TFLOPS. Can reach near GTX 1070 or GTX 980 Ti performance in some games.
VII's Vega 20 silicon has proper support for 64bits floating point while mid-range and game console GCNs has garbage support for 64bits floating point. AMD can reduce transistor usage/reduce power consumption by removing 64bits floating point support.
Yeah, AMD needs to build a 4K capable APU at about 200W. Things aren't looking too great right now. We aren't getting Navi and Zen 2 until later this year on the PC side, so things are looking even worse for a holiday 2020 set of consoles at a sub $400 price-point with capable hardware.
AMD at 7nm is here, and at 300W for $700 it only starts to approach the performance of early 2017 PC tech (1080ti). So, what does this means for the PS5/XB3? Well, next gen consoles will reach the performance of ~1070/oc980ti if we are lucky, and the later is 2015 PC tech! Have next gen consoles been pwned already?
There's another 7nm+ node improvement after the first gen 7nm. Xbox One X used 2nd gen 16 nm improvements after PS4 Pro's 16 nm and both APUs are similar in size i.e.
PS4 Pro's APU
8 CPU Jaguar
40 CU Polaris IP with selected Vega IP e.g. RPM. 36 CU active with 4.2 TFLOPS.
Xbox One X's APU
8 CPU Jaguar with lower latency
44 CU Polaris IP with near Vega ROPS cache design. 40 CU active with 6 TFLOPS. Can reach near GTX 1070 or GTX 980 Ti performance in some games.
VII's Vega 20 silicon has proper support for 64bits floating point while mid-range and game console GCNs has garbage support for 64bits floating point. AMD can reduce transistor usage/reduce power consumption by removing 64bits floating point support.
Hey ron! You were right about the 128 rops you told me a few months ago.
AMD at 7nm is here, and at 300W for $700 it only starts to approach the performance of early 2017 PC tech (1080ti). So, what does this means for the PS5/XB3? Well, next gen consoles will reach the performance of ~1070/oc980ti if we are lucky, and the later is 2015 PC tech! Have next gen consoles been pwned already?
I think the general output would be like those in traditional compute but I believe it'll have some of the later features of 2020/2021 GPUs.
Closer to 1070, 1070 Ti, Vega 64, definitely above Vega 56 or equivalent.
I seriously think the next gen systems will have emphasis on better upscaling (DLSS/CB) and maybe dedicated hardware for it that might not be on the RVII.
They might also consider shifting towards more CPU grunt the next round, at least possibly more impressive than this gen's GPU jump towards the next gen's GPUs.
Unless consumers are ready to shell out some serious moolah, consoles will never catch up to PC tech. Not that they need to, though. I never game on the PS4 expecting shit to look as good as what my PC can display.
AMD at 7nm is here, and at 300W for $700 it only starts to approach the performance of early 2017 PC tech (1080ti). So, what does this means for the PS5/XB3? Well, next gen consoles will reach the performance of ~1070/oc980ti if we are lucky, and the later is 2015 PC tech! Have next gen consoles been pwned already?
There's another 7nm+ node improvement after the first gen 7nm. Xbox One X used 2nd gen 16 nm improvements after PS4 Pro's 16 nm and both APUs are similar in size i.e.
PS4 Pro's APU
8 CPU Jaguar
40 CU Polaris IP with selected Vega IP e.g. RPM. 36 CU active with 4.2 TFLOPS.
Xbox One X's APU
8 CPU Jaguar with lower latency
44 CU Polaris IP with near Vega ROPS cache design. 40 CU active with 6 TFLOPS. Can reach near GTX 1070 or GTX 980 Ti performance in some games.
VII's Vega 20 silicon has proper support for 64bits floating point while mid-range and game console GCNs has garbage support for 64bits floating point. AMD can reduce transistor usage/reduce power consumption by removing 64bits floating point support.
Hey ron! You were right about the 128 rops you told me a few months ago.
128 ROPS is only part of the bottleneck fix. Rasterization (large scale floating point to integer conversion hardware) is another pipeline bottleneck.
RTX 2080 has 6 rasterization (from six GPC) units with 64 ROPS up to +1900 Mhz clock speed
VII has 128 ROPS and seems to be quad rasterization units up to 1800Mhz clock speed. For VII, AMD doubled HBM v2 stack count and doubled the ROPS units. I would be surprise if VII has 8 rasterization units with 128 ROPS which is doubled Vega 64's classic GPU hardware.
Mainstream NAVI 12 has 40 CU, 64 ROPS(?) with 256bit GDDR6 memory. At 1800 Mhz, yields 9.216 TFLOPS. 40 CU is 66 percent from VII's 60 CU.
Either the next gen will be a bit more powerful than the x1x with a pro version half way through or they might have tech in development that's not announced yet. Next gen will be more powerful the question is by how much
Yeah I’m not really seeing the point in next gen consoles coming out in less than 3 years from now. They will essentially just be refined versions of the Pro and X. No major leaps this coming gen.
AMD at 7nm is here, and at 300W for $700 it only starts to approach the performance of early 2017 PC tech (1080ti). So, what does this means for the PS5/XB3? Well, next gen consoles will reach the performance of ~1070/oc980ti if we are lucky, and the later is 2015 PC tech! Have next gen consoles been pwned already?
"Moore's Law used to grow at 10X every five years [and] 100X every 10 years," Huang said during a Q&A panel with a small group of reporters and analysts at CES 2019. "Right now Moore's Law is growing a few percent every year. Every 10 years maybe only 2X. ... So Moore's Law has finished."
Just like I said multiple times on this forum - 7nm will not fix AMD's power draw issues. Here we are: 7nm VII is higher TDP than it's 12nm competitor 2080.
I hope people won't be repeating the RX3080 price/performance/TDP hoax anymore after this announcment.
Funny how Sony devs get more out of 2013 console tech than most devs get out of 2018 PC tech.
Probabaly in a world where pigs fly but definitely not in this world at sub 30 FPS, 4xAF, almost no AA, almost totally baked lighting, baked physics, no GI, garbage Draw Distance, garbage textures, almost no dynamic shadows, extremely static worlds etc etc. It's 9 years and Sony have yet to make a game with fully dynamic lighting outside of a simple world indie game.
I said it before in one of the threads that next gen consoles will likely be around 1070/1070Ti levels if they were to release in 2020 and I was trolled to heck and back. Seems like I was right unless consolites are willing to shell $600 and even the problem of TDP is there.
AMD at 7nm is here, and at 300W for $700 it only starts to approach the performance of early 2017 PC tech (1080ti). So, what does this means for the PS5/XB3? Well, next gen consoles will reach the performance of ~1070/oc980ti if we are lucky, and the later is 2015 PC tech! Have next gen consoles been pwned already?
There's another 7nm+ node improvement after the first gen 7nm. Xbox One X used 2nd gen 16 nm improvements after PS4 Pro's 16 nm and both APUs are similar in size i.e.
PS4 Pro's APU
8 CPU Jaguar
40 CU Polaris IP with selected Vega IP e.g. RPM. 36 CU active with 4.2 TFLOPS.
Xbox One X's APU
8 CPU Jaguar with lower latency
44 CU Polaris IP with near Vega ROPS cache design. 40 CU active with 6 TFLOPS. Can reach near GTX 1070 or GTX 980 Ti performance in some games.
VII's Vega 20 silicon has proper support for 64bits floating point while mid-range and game console GCNs has garbage support for 64bits floating point. AMD can reduce transistor usage/reduce power consumption by removing 64bits floating point support.
Hey ron! You were right about the 128 rops you told me a few months ago.
128 ROPS is only part of the bottleneck fix. Rasterization (large scale floating point to integer conversion hardware) is another pipeline bottleneck.
RTX 2080 has 6 rasterization (from six GPC) units with 64 ROPS up to +1900 Mhz clock speed
VII has 128 ROPS and seems to be quad rasterization units up to 1800Mhz clock speed. For VII, AMD doubled HBM v2 stack count and doubled the ROPS units. I would be surprise if VII has 8 rasterization units with 128 ROPS which is doubled Vega 64's classic GPU hardware.
Mainstream NAVI 12 has 40 CU, 64 ROPS(?) with 256bit GDDR6 memory. At 1800 Mhz, yields 9.216 TFLOPS. 40 CU is 66 percent from VII's 60 CU.
Where can we get info of "Rasterization"?
Where did the Navi 12 info come from? Thanks again.
128 ROPS is only part of the bottleneck fix. Rasterization (large scale floating point to integer conversion hardware) is another pipeline bottleneck.
RTX 2080 has 6 rasterization (from six GPC) units with 64 ROPS up to +1900 Mhz clock speed
VII has 128 ROPS and seems to be quad rasterization units up to 1800Mhz clock speed. For VII, AMD doubled HBM v2 stack count and doubled the ROPS units. I would be surprise if VII has 8 rasterization units with 128 ROPS which is doubled Vega 64's classic GPU hardware.
Mainstream NAVI 12 has 40 CU, 64 ROPS(?) with 256bit GDDR6 memory. At 1800 Mhz, yields 9.216 TFLOPS. 40 CU is 66 percent from VII's 60 CU.
Where can we get info of "Rasterization"?
Where did the Navi 12 info come from? Thanks again.
The above diagram is for R9-290X "Hawaii".
Each Shader Engine has a rasterizer hardware unit.
RB includes color ROPS and z-ROPS and these are traditional graphics read/write units.
Hawaii's basic layout was recycled up to Vega series. Vega IP introduces ROPS being connected to L2 cache and triangle binning (early triangle culling).
--------
For NVIDIA... using GTX 1080 example
Each GPC includes Raster Engine.
NVIDIA's "Raster Engine" and AMD's rasterizer are major fix function hardware that separates GPUs from DSP (Digital Signal Processor).
Both Vega 64 and GTX 1080 has similar raster power (ie. quad raster, 64 ROPS read/write units) while GTX 1080 Ti's raster hardware is superior.
RTX 2080 has six GPC, hence six raster engines with 64 ROPS read-write units at +1900 Mhz stealth overclock mode.
VII has unknown rasterizer (I guess four rasterizer setup) with 64 ROPS read-write units at 1800 Mhz
TFLOPS are meaningless since it doesn't include graphics read/write units nor rasterizer hardware scale.
I'm expecting 2020 launch with whatever GPU is in AMD's $200-300 tier, but underclocked by 15-20%. Expecting it to be weaker than the RTX 2060, so maybe like a 1070/1070ti.
It depends on rasterziation hardware scaling for year 2020 and 7nm+ has 10 percent lower power consumption improvements.
On certain games, X1X's 40 CU GPU can nearly reach GTX 1070 or GTX 980 Ti level while RX-580 remains in the lower tier GPU e.g. Far Cry 5
X1X dev kit with 44 CU is 10 percent faster than X1X retail version. X1X GPU only has 32 ROPS with 2 MB render cache + TMU has 2MB L2 cache while GTX 1070 has TMUs and 64 ROPS with ~2 MB L2 cache unified. 64 ROPS enables any future Xbox One improve games with heavy ROPS read-write usage.
X1X GPU's ROPS improvements are half way house between Vega ROPS and Polaris ROPS.
RX-580's 32 ROPS is missing X1X's 2MB render cache unable to perform software cache tile render optimizations.
X1X's 44 CU GPU area size is ~280 mm2 at 16 nm process tech with near zero FP64 support.
RX-580 36 CU GPU area size is ~232 mm2 at 14 nm process tech with near zero FP64 support.
@ronvalencia: You spent a year+ telling us the Xbox One X GPU was equal to a GTX 1070. (and at one point you started talking 1080) What did we end up with???
@pc_rocks: You're delusional..1070 is barely better then a X1X.
Wrong.
I wouldn't if I were you. That cows makes his own facts and disregards anything that contradicts with his view of Sony.
On topic, the history shows us this that I already shared in another thread way back:
PS4 - the most powerful console in 2013 was an equivalent of 570 from 2010
PS4 Pro - released in 2016 was still weaker than 970 from 2014
X1X - released in 2017 was barely faster than 970 from 2014 and weaker than 1060
Any console coming in 2020 will match 1070Ti at best regardless of how many AMD's TFLOPS it has and the history will again repeat it self.
It'll be interesting to see how the next gen console plays out. Are they going to push 4k and lower settings or do the whole dynamic scaling thing, again? My money is stick with dynamic scaling and push higher settings, but PC will be so far ahead at that point. We might even get the 3080/2180 Ti by then. Ray tracing will be much more ubiquitous, and yeah, it's gonna be interesting.
^ See I told you. And this is not the first time, I have already presented these graphs as well as several DF articles/comparions/benchmarks but the cow has his (alternate reality) reasons...Cow logic: X1X stomps(8%) the 970 but 1070/Ti is barely(50-60%) faster than X1X, LOL.
@pc_rocks: You're delusional..1070 is barely better then a X1X.
Wrong.
I wouldn't if I were you. That cows makes his own facts and disregards anything that contradicts with his view of Sony.
On topic, the history shows us this that I already shared in another thread way back:
PS4 - the most powerful console in 2013 was an equivalent of 570 from 2010
PS4 Pro - released in 2016 was still weaker than 970 from 2014
X1X - released in 2017 was barely faster than 970 from 2014 and weaker than 1060
Any console coming in 2020 will match 1070Ti at best regardless of how many AMD's TFLOPS it has and the history will again repeat it self.
It'll be interesting to see how the next gen console plays out. Are they going to push 4k and lower settings or do the whole dynamic scaling thing, again? My money is stick with dynamic scaling and push higher settings, but PC will be so far ahead at that point. We might even get the 3080/2180 Ti by then. Ray tracing will be much more ubiquitous, and yeah, it's gonna be interesting.
Yup, I'm expecting the same although my money is on they will push 4K because resolution wars however I'm 100% certain quite a number of games miss the 4K mark just like this gen many games missed 1080p.
On that comparison, I would also like to add that X1X achieved the barely faster feat than 970 by adding $100 to the price a year later than Pro. So there's that too.
@ronvalencia: You spent a year+ telling us the Xbox One X GPU was equal to a GTX 1070. (and at one point you started talking 1080) What did we end up with???
Source
That's 1080p resolution
1. Gears of War 4 4K sustain 30 fps story mode, GTX 1060 and RX-580 fails to reach X1X's 4K results
2. Forza Motorsport 7 4K wet scenes sustain 60 fps, GTX 1060 and RX-580 fails to reach X1X's results
3. Far Cry 5 4K sustain 30 fps, GTX 1060 and RX-580 fails to reach X1X's results
4. Kill Instinct Season 3 4K nearly sustain 60 fps, GTX 980 Ti or GTX 1070 level. GTX 1060 fails to reach X1X's results.
At lower resolution like 1920x1080p, it's mostly CPU bound.
Log in to comment