btw if 12tf rdna2 xsx is around 2080 level in gears performance increase per teraflop on rdna2 is probably marginal and close to rdna1
It's on cry engine, not best for massive open worlds
How is that very obvious ?
The results from Anandtech are for the reference design. Those are for the Asus Strix OC. So yes, it is maintaining over 2ghz. It is the actual game clock.
Because it would vary. It would not be at a static level as your graph shows. I honestly don't know if you are trolling me or not, but I don't think we can get any further with this. Anyway, thanks for the link to the Anandtech review, it contained very useful data.
It's an overclocked model. It's meant to run at that high clock. No, I'm not trolling you but I feel like you are. I keep bringing concrete evidences that you either don't read or only keep what you want to run some weird calculations that wont translate to gain in performances. You want your 10.3 Tflops RX 5700 XT ? You have it. It's 18% away from a 2080 during gaming. If facts dont suit you, well, you can still push them aside and go back to making crazy theories.
btw if 12tf rdna2 xsx is around 2080 level in gears performance increase per teraflop on rdna2 is probably marginal and close to rdna1
Absolutely, I love questions :)
So yes, it's possible the 2080 will be a "mid-range"-level card at the end of this year, just unlikely given the history of the prices of GPU power brackets going up over time. If the 3060 is equal to 2080, but at $400, that's not mid-range. But even when 2080 is midrange eventually, which it will be, it'll take a long time for the average gaming computer to actually be at that level.
- I'm going to say a "mid-range" graphics card is something in the $250-300 range, as whatever is released there seems to end up the most popular. So, if a 2080-level card is released for $300, and there are more powerful cards above that, would that mean console GPUs are "mid-range"? Yes. I'd be surprised at that though. I think that level will stay at $400-500.
- Absolutely, and I think this is actually the more relevant metric than the "current product range." The 1060 came out in 2016, and is still the most popular card. Using the wayback machine to review previous years, the percentage hasn't even changed much. The 970 was the most popular before that, and it was released in 2014. With the economy the way it is, I think it's going to take even longer than usual for the average person to upgrade if they've got a functional computer at the moment. I'd bet the 1060 will still be the most popular card a year from now, though it will go down in percentage a bit.
NX Gamer: PS5 Full Spec Analysis | A new generation is Born
It's unlikely that we'll see huge strides in lossless compression. Just a couple of percent more would be considered pretty revolutionary. Kraken is absurdly good already, and it sounds like it's largely baked into the hardware design, so that's what it's likely to remain as. The sorts of...www.resetera.com
for sure 2080 level in nextgen console is very impressive, sad we have to wait long time for proper xsx exclusive without xone burdenThe crazy thing is that's with a 2 week simple port not using any new features on the XSX it's performing like a 2080 so imagine what devs can do years going further into the features and whole teams coding a engine to it
Yup... I think what is really impressive gong from RDNA1 to RDNA2 is the power consumption. Its a more efficient chip.. oh and has RT stuff.btw if 12tf rdna2 xsx is around 2080 level in gears performance increase per teraflop on rdna2 is probably marginal and close to rdna1
NX Gamer: PS5 Full Spec Analysis | A new generation is Born
It's unlikely that we'll see huge strides in lossless compression. Just a couple of percent more would be considered pretty revolutionary. Kraken is absurdly good already, and it sounds like it's largely baked into the hardware design, so that's what it's likely to remain as. The sorts of...www.resetera.com
IndeedYup... I think what is really impressive gong from RDNA1 to RDNA2 is the power consumption. Its a more efficient chip.. oh and has RT stuff.
The efficiency is 50% better than RDNA 1. This is more than a marginal increase in Tflops over Previous gen. 50% improvement per watt Vs RDNA 1 will yield big gains. Nvidia will have competition finally. Just wrap your head around the fact AMD has APU's that compete with top END PC's. 2.23GHz on the GPU is impressive for any GPU much less one in small form factor(Console) 12Tflops performing like 2080(Gears 5) and unoptimized(not using any RDNA 2 features) Ya there is a big jump from first Gen RDNA.Yup... I think what is really impressive gong from RDNA1 to RDNA2 is the power consumption. Its a more efficient chip.. oh and has RT stuff.
sometimes I wonder what the point of RDNA1, but it did bring AMD's name back and make people excited for RDNA2Know what I find amazing?
These consoles just skipped right over RDNA 1 and are using 2.
That in itself is something to be excited for.
In order for games to have higher quality textures And other assets on PS5, games would have to saturate the SSD bus at (conservatively) 2.4GB/s. That's roughly 25 to 50 times faster than current gen games stream assets and current gen games look fantastic. Which means that games will have 20 to 50 times larger assets (on a 825GB SSD). And these assets will fit into RAM that is slightly larger than current gen. I think you guys have unreasonable expectations for next gen games. Compute and RAM is still going to be the limiter on games, not the SSD.
No, that's not typical, that's theoretical max without BCpack (which is used for texture AFAIK).
Yes, if the ssd was 22GB/s raw speed the io chip could manage all that data.
False. The 22GB/s figure assumes a perfect compression rate of 4:1 for the assets being read at 5.5GB/s expanding to 22GB/s.
Yeah, no mention of it at the AMD FAD. I assume <10% if any.btw if 12tf rdna2 xsx is around 2080 level in gears performance increase per teraflop on rdna2 is probably marginal and close to rdna1
sometimes I wonder what the point of RDNA1, but it did bring AMD's name back and make people excited for RDNA2
Why do people keep misquoting what BCPack is?No, that's not typical, that's theoretical max without BCpack (which is used for texture AFAIK).
I don't think the discrete cards coming later this year will be on 7nm+. Expecting N7P. There could and likely will be other differences though.It will be interesting to see the IPC delta between RDA1 and 2. Of course, it's going to be somewhat muddy if commercial RDNA2 cards end up being fabricated on N7+ as opposed PS5 and XSX's N7P platform.
both Sony and MS say they're using "custom" RDNA2 gpus, so there are probably some rather large differences, but who knows whatI don't think the discrete cards coming later this year will be on 7nm+. Expecting N7P. There could and likely will be other differences though.
yep. if the 12.1 tflops console is offering performance thats equivalent to 11.4 turing tflops, we can assume very little gains.
Disregard, I misread the post.Why do people keep misquoting what BCPack is?
BCPack is a texture compression format. It does not make the IO any faster than it is, it makes texture compression better so you can fit more texture per second through the IO when compressed, the IO speed does not change. The 6GB/s theoretical max does not change, what you can fit through it changes when you have a good compression ratio.
And BCPack is totally taken into consideration when giving the theoretical max throughput otherwise the number would be higher. That is why they said 4.8GB compressed even though the IO has a max theoretical of 6GB/s.
btw if 12tf rdna2 xsx is around 2080 level in gears performance increase per teraflop on rdna2 is probably marginal and close to rdna1
Source for this ?False. The 4.8GB/s figure includes BCPack as part of the typical rate.
False. The 22GB/s figure assumes a perfect compression rate of 4:1 for the assets being read at 5.5GB/s expanding to 22GB/s.
How would he know?What is the likelihood of this actually happening? Would it be possible for developers to use smart programming or game design to try and maximise it?
All this tech talk and as much as I love it, it'd be nice to see actual projects running on the machine.
Expecting DeS remake to be like:
I would say highly unlikely because not all data type compresses the same. And even at that, there is a mixture of data types in any transfer so there would be some stuff that could compress really well and others that can't. Sony and MS gave that 8GB/s-9GB/s and 4.8GB/s respectively numbers for a reason. The 6GB/s and 22GB/s thingy is just a PR theoretical max as far as I am concerned. Depicting levels of compression that isn't quite possible just yet.What is the likelihood of this actually happening? Would it be possible for developers to use smart programming or game design to try and maximise it?
I would say a good indication would be game sizes.
I would say highly unlikely because not all data type compresses the same. And even at that, there is a mixture of data types in any transfer so there would be some stuff that could compress really well and others that can't. Sony and MS gave that 8GB/s-9GB/s and 4.8GB/s respectively numbers for a reason. The 6GB/s and 22GB/s thingy is just a PR theoretical max as far as I am concerned. Depicting levels of compression that isn't quite possible just yet.
I would say a good indication would be game sizes.
Speaking of sizes though, I think we would benefit from total removal of "in-engine" cinematics stored as video files and usually used to hide loading since everything would likely be done in-engine and real-time on the hardware this time around.
You will obviously not unload the whole OS, just the memory hungry parts of it, aka the UI.Just as an aside 5 GB/s is not enough for an OS. You'd be getting hitched every time an OS function swaps into main memory.
DDR3's slowest speed is about 1 GB/s faster.
As it stands they went from making videos using full-on CGI, to making "in-engine videos" that they use as cutscenes. either way, they are data hogs. Now we would see more of in-engine cinematic running in real-time, no reason to still be putting video files on the disc unless their cutscene is totally unrelated to the actual in-game scenario... like doing a back story of sorts or something.Oh god, yes. CGI cutscenes might still have something of a place (though I still dislike their severe limitations like not being able to change character and weapon models and such), but 1080p30 in-engine videos that look like compressed shit, take up disc space and are made to hide loading? Burn them all, and salt the earth where they stood.
Yup... 5GB/s second isn't enough to run the OS, but its definitely enough to load it in and out quickly. If XSX can make do with a 2.5GB of RAM OS reserve, then the PS5 can make do with a 1GB reserve. And be able to move like 4GB of data into RAM in like 0.50 seconds.You will obviously not unload the whole OS, just the memory hungry parts of it, aka the UI.
I'm hoping Monster Hunter world looks like that. That would be amazing..
For me, it's not the fidelity that is striking. Rather, it is the animation. Look at the hit reaction as well as how locomotion and movement is determined by the footing of the creature and mass- it simply does slide from place to place.
This generation did not see a commensurate increase for animation quality like it did with fidelity (barring exceptions, like Death Stranding where human foes had some of the best animations to day). It is both sad and interesting that I still remember KZ2, a PS3 game, that featured great foe animations and esp. hit reactions which still have not been surpassed to this day.
Here's hoping next gen delivers on that front.
btw if 12tf rdna2 xsx is around 2080 level in gears performance increase per teraflop on rdna2 is probably marginal and close to rdna1
The only console optimization bonus this generation we saw on cpu, gpu behaved similarly to their desktop's equivalent in same/close graphic setup.In practice, I expect PS5 to perform around 2080TI level and SeX to slightly exceed that , thanks to the usual console optimization bonus.
NoIn practice, I expect PS5 to perform around 2080TI level and SeX to slightly exceed that , thanks to the usual console optimization bonus.
The introduction of new consoles will basically be: "This is what a game looks like when built with high end PC specs as a baseline". It will be dope.
I have generally seen a significant lack of imagination here regarding how the new fast and dream-level SSDs will be used to revolutionize game design.
This leads me to believe a similar phenomenon will happen among the dev community and the SSD will go under-used for many years to come. I don't remember seeing UE4 or Unity demos showing new awesome PCIE3 or PCIE4 NVME-exclusive techniques.
I expect big budget next gen games to look way better than that. WAY BETTERAll this tech talk and as much as I love it, it'd be nice to see actual projects running on the machine.
Expecting DeS remake to be like:
You should check out this video: http://www.youtube.com/watch?v=IDO-UHZebV4&t=4m40sI doubt that. The game clock varies between different games. If it was recorded for all games it would look different.
Since I don't have a life ;-), I took the performance data from Anandtech's review (https://www.anandtech.com/show/14618/the-amd-radeon-rx-5700-xt-rx-5700-review/12), since they write out the measured game clock and recomputed the benchmark FPS if the 5700XT was scaled up to 10.3TF. Then it looks like this,
Game Game Clock 5700XT/10.3TF 5700XT/2070 Super Performance ratio
Tomb Raider 1780MHz 39.8/44.98/42.6 1.0559
F1 2019 1800MHz 54.8/61.25/59.9 1.0225
Assassin's Creed 1900MHz 38.3/40.55/46.2 0.8777
Metro Exodus 1780MHz 34.7/39.2/35.0 1.12
Strange Brigade 1780MHz 69.6/78.66/75.0 1.0488
Total War: TK 1830MHz 26.2/28.8/30.0 0.96
The Division 2 1760MHz 34.7/39.66/40.2 0.9866
Grand Theft Auto V 1910MHz 41.0/43.1/47.5 0.9073
Forza Horizon 4 1870MHz 60.1/64.65/58.0 1.1147
avg: 1.0104
Unfortunately, the review doesn't include the 2080, so I used the 2070 super instead. For the 9 benchmarks a 10.3TF 5700XT would be about 1.0104 times faster than a 2070 super.
Then I used the relative performance from www.techpowerup.com to compare the 2070 super with the 2080. It looks like,
RTX 2070 Super: 114%
RTX 2080: 123%
Hence a 2080 would have about 1.23 / (1.14 * 1.0104) = 1.0678 higher performance than a 10.3TF 5700XT and a 10.3TF 5700XT would have about 1.14 * 1.0104 = 1.152 higher performance than a 5700XT.
We still don't take the RDNA 2 architectural improvements into account, nor the BW deficit caused by the CPU/SSD/Audio. I still believe there are good reasons to expect the PS5 GPU to perform on a comparable level to the 2080 and about ~15% above the 5700XT.
You're going to be disappointed if you think either are going to perform better or equal to a 2080ti.In practice, I expect PS5 to perform around 2080TI level and SeX to slightly exceed that , thanks to the usual console optimization bonus.
The introduction of new consoles will basically be: "This is what a game looks like when built with high end PC specs as a baseline". It will be dope.
I have generally seen a significant lack of imagination here regarding how the new fast and dream-level SSDs will be used to revolutionize game design.
This leads me to believe a similar phenomenon will happen among the dev community and the SSD will go under-used for many years to come. I don't remember seeing UE4 or Unity demos showing new awesome PCIE3 or PCIE4 NVME-exclusive techniques.
This as a baseline for animation, interactivity, physics, and LOD would be amazing.All this tech talk and as much as I love it, it'd be nice to see actual projects running on the machine.
Expecting DeS remake to be like:
His point is not that the PS5 would match the 2080 TF wise, but that with the usual slew of console optimizations and consoles having a history of punching well above their weight. The PS5 with its 10.3TF GPU would end up eeking out performance equivalent to a 2080. So basically a 10.3TF console GPU would perform like a 12.5TF CPU GPU. And a 12TF console GPU would perform like a 13.5/14TF CPU GPU.You should check out this video: http://www.youtube.com/watch?v=IDO-UHZebV4&t=4m40s
You need a 2450mhz 5700xt, which means it's a 12.5 TF GPU, to match a 2080. Probably heavily bandwidth limited but the PS5 will have even less bandwidth available as it needs to share.
While I love how it looks, I am not gonna get my hopes up. I just don't trust devs anymore when it comes to stuff like this. I would believe it when you show me someone playing it or when its coming from a studio that I know I can trust.This as a baseline for animation, interactivity, physics, and LOD would be amazing.
His point is not that the PS5 would match the 2080 TF wise, but that with the usual slew of console optimizations and consoles having a history of punching well above their weight. The PS5 with its 10.3TF GPU would end up eeking out performance equivalent to a 2080. So basically a 10.3TF console GPU would perform like a 12.5TF CPU GPU. And a 12TF console GPU would perform like a 13.5/14TF CPU GPU.
Obviously, my numbers are just estimates, but I am at least certain that consoles have (and likely will) always punched above their weight. You simply can't take a 10.TF console GPU and pair it with a 10TF CPU GPU and expect the same results.