The PS4 also had more fillrate, more (I can't recall the technical term) scheduling units to feed its GPU, etc. All in all a very similar approach to the PS5. Also, the DDR 3 may have had a slightly lower latency, it also had a much lower bandwidth to feed its CPU and GPU.
The interesting thing is PS4 outperformed XboxOne in quite few (supposed) CPU bound games/scenarios too. Perhaps there was a customization beyond the clock frequency or this was due to API differences.
Did they really say that? LOL by that logic, backwards compatible games should run worse on the next gen hardware, as they are rendering the last gen graphics with powerful hardware.
They don't say anything about engine rewrite to take advantage of anything. They said it very specifically this time. This at least implies this wasn't done for xs.It doesn't say anywhere that it's backwards compatibility, the SX also benefited from its native version at launch.
The devs explain well that it's thanks to the higher GPU frequency and unified RAM of the PS5 that they were able to push the resolution higher than it was on XSX
Maybe the developers can release an XSX patch to put it in 8K as well of course if they manage to do so with further optimization, but for now it remains in 6K.
It's him yeahIs that Jesús Quintero smoking a joint? lol
The thing I least expected to see on GAF.
I've always wondered why GAF is so full of Spanish freaks.
Considering the timing of the release, the devs likely would have used the updated GDK. They don't need to mention every little thing that they have done. No one does.
That's still a straight port vs engine rewrite.
What is a straight port? You're assuming it was a "straight port".
Like we've learnt that means fuck all, especially for larger companies. ME: Anniversary edition is BC game, even game file info on console says "Gen9aware" (so like Cyberpunk) and still has badge X|S....
Dev told them.
Nope.
MS themselves said they got more performance with 7% upclock than 17% increase in CUs with Xbox One.
Well if we forget the how the big advantage in the past years in GPU market was the high clocks of nVidia cards giving the edge in performance per TF even with way lower number of units.
Yes (partially) because the upclock was also impacting esram bandwidth.
Upclock affect all parts of the GPU including cache.
A GPU with the same memory bandwidth with 20CUs at 1.5Ghz will run faster than one with 30CUs at 1Ghz.
no if you have an 8k tv you wont benefit from it because the PS5 doesnt support 8k output.
Supersample is a benefit in render higher than your screen resolution.
Yes..
GPU | 5700 | 5700XT | 5700 OC |
|---|---|---|---|
CU's | 36 | 40 | 36 |
Clock | 1725 Mhz | 1905 Mhz | 2005 Mhz |
TFLOP | 7.95 | 9.75 | 9.24 |
TFLOP Diff. | 100% | 123% | 116% |
Assassin's Creed Odyssey | 50 fps | 56 fps | 56 fps |
F1 2019 | 95 fps | 112 fps | 121 fps |
Far Cry: New Dawn | 89 fps | 94 fps | 98 fps |
Metro Exodus | 51 fps | 58 fps | 57 fps |
Shadow of the Tomb Raider | 70 fps | 79 fps | 77 fps |
Performance Difference | 100% | 112% | 115% |
www.pcgamesn.com
I used to own an 8k TV which I used as a PC monitor with a RTX3090.
What's the point to rewrite the engine on XSX? Isn't it virtual coded with a multiplat development kit? Some you are something else sometimes. You don't have access to a low level of API on XSX, thanks the direct X Lord and the logic of coding by MS, why you blame the developers, they did what they can. Furthermore if I'm not wrong, they also said higher res was possible thanks to the faster frequency and the memory setup, not only because they "just" rewritten the engine.
That is the whole point of GDK… same development and code across all supported platforms.
How do you cope with the fact that the PS5 has less units and an higher fill rate?
Lysandros
I fixed PS4 PRO ROPs number.So are many other games that they just run thru the SDK to gdk and call it a day. Again Borderlands 3 is a native ps5 game. 1070 performance is the best we can expect from that engine on ps5, or do you think if they rewrite their engine specifically for the ps5 it could do better?
No, Xbox gives the option of making it simple, not that it has to be. You can do the same pretty much for ps4-ps5 and it's been done quite often. Just not this time.
I mean, at this point I think you are free to believe whatever you want. Developers said also ps5 specs give them some advantages for the resolution but eh.
Yes..
All GPU's are all based on AMD Navi 10, have GDDR6 memory at 448GB/s. Game benchmarks were done at 1440p.
AMD hid the true power of the Radeon RX 5700… here’s how to unlock it
Unlocking the frequency and power shackles on the second-string Navi GPU makes it a fantastic card
PS. Oc course there is a silicon limit for clocks speeds and so after you cross that limit in a chip the performance won’t scale anymore… RDNA it was around 2000-2100Mhz depending of the chip (bin lottery).
did you ear something? I heard nothing, maybe you should have yourself checked?
You're making a huge left-field assumption that devs did not create the Xbox version with the same efforts as they did the PS5 version, while there is absolutely no reason/evidence to do so.
The version of DX12 on the XSX is a low level API though. Heck, that was the entire point of DX12 in the first place, to compete with GNM on the PS4.
developer.nvidia.com
API has no place in that type of test… after all the game is using the same API across the two cards.
The point that units keep being not used in a render time already tells you clock increase us advantage over unit increase.
It's not exactly as to have an effective low API hardware access however. It's the same argument used by MS for the velocity architecture.The version of DX12 on the XSX is a low level API though. Heck, that was the entire point of DX12 in the first place, to compete with GNM on the PS4.
Encoding for DirectX 12 with NVIDIA Video Codec SDK 11.1 | NVIDIA Technical Blog
Read about DirectX 12, a low-level programming API from Microsoft that reduces driver overhead in comparison to its predecessors. DirectX 12 provides more flexibility and fine-grained control on the…
It's also the reason DX12 has been such a hit and miss on the PC side of things, your game engine has to specifically coded to work well with it. Hence some games saw massive games (on AMD hardware) and others saw very little or actually regressed in performance.
Now it may very well be the PS5 API has a lower overhead then the DX12 version on the XSX, but it likely isn't anything huge as both are effectively using low-level APIs. Both consoles will continue to get performance enhancements as the lifecycle of both consoles go on, we even have a recent example of this, Control was improved on both consoles with zero input from the developer.
Thanks for the reply. How do you explain the same amount of ROPs, GPU L1 cache, ACEs and HWS along with other units like rasterizers, prim units etc. across PS5/XSX despite the CU number difference then? As far as i know those scale with shader engine/array not CU count.A CU has TMUs, ROP, vector units, scheduler, cacche, etc.
Add a CU and you add a proportional amount of other units.
I don't think that it's necessary to constantly repeat that the X1 had fewer of all of these units, when just saying that it had fewer CUs, does the same.
Yes some, but double?
No, I'm not. One is a launch game other came out 1 year later. One they said they rewrite their engine to better use the hardware and the other they said no such thing. I think it's more left field to assume the small differences in hardware will ever give double resolution.
You're believing that has no evidence and outright discarding what the developer is literally telling us.
But that is what it is? Both GNM and DX12 are low level APIs. There is no real way to prove which one is more 'low level' than the other. We just know that both are. Unless you think Microsoft, Nvidia, and AMD are all lying?
www.eurogamer.net
COD MW was ran through the newer XDK and then reverted as it done bugger all bar force the game on the internal storage
No, not at all. You are disregarding the next line in the same quote. Unless writing engines to specific hardware doesn't give better performance?
