-
Ever wanted an RSS feed of all your favorite gaming news sites? Go check out our new Gaming Headlines feed! Read more about it here.
-
We have made minor adjustments to how the search bar works on ResetEra. You can read about the changes here.
Next-gen PS5 and next Xbox speculation launch thread |OT5| - It's in RDNA
- Thread starter Mecha Meister
- Start date
- Status
That's quite big too then because I doubt we're gonna see the death of things popping up in games, especially these big open world games like from BGS and CDPR. I might be buying Starfield on PS5 then, who knows.
Believe me or not I don't have the speed but a friend working in the industry told when I show the patent, the technology for the SSD was discovered. He will not give the speed and broke his NDA. He just told if they could have something comparable on PS4 only 2.5 to 3 GB of RAM would have been ok for a game well-optimized*. He does not follow forum gaming internet and he was surprised and ask me if someone discovers something about the DualShock 5. On Sony side, they don't talk about SSD but parallel NAND technology.
For Dualshock 5, the technology is probably inside a Sony patent...
EDIT:*Less fast on PS4 because assets are smaller and less memory I think...
EDIT: I don't have insider information if I did not show Gofreak post he would not have talked about the SSD.
Last edited:
Still using 14gb's of GDDR6? that's disappointing. Also calling it the Ti means we can say goodbye to a $400 2070.
i kinda missed you here, 2.5 to 3 Mo of RAM? i am a bit confused.
it was a huge issue for insomniac as well, especially because the HDD was replaceable so they needed to design the game with the slowest HDD they could find.
You would imagine that at least one of Sony's developers will be working on a game that utilises the unique benefits of the superfast bandwidth. Imagine a game that is just insane and the tagline at the end is "Only possible on PlayStation 5". I'm excited to see how this benefits games and just generally excited for instant access, enough of loading screens in general.
Honestly, I'm not holding my breath. I have a 1080ti gaming rig which is around 11TF. Biggest difference regarding the visuals from the Xbox One X is the higher framerates. While faster CPUs in consoles will push the whole industry forward it's not likely that the difference will be even close to how the hype goes here. And I doubt that the framerates will get any higher either, we'll se skin pores and individual strands of hair with physics before we see 60fps become a standard on consoles.
I think the biggest change will be seen in ray tracing, if the consoles can do that.
i imagine god of war 2, where you can now use the portable bifrost to instantly teleport to a new realm. this will be so dope.
If we could have an SSD fast enough to have a ratio RAM/SSD speed as good than on PS5 2 to 3 GB of RAM would have been enough instead of 8GB of RAM for a game with good optimization on PS4.
EDIT: The SSD is fast enough for a game without big optimization to load a little slower than SSD on PC with RAMDISK.
EDIT: SSD is fast enough to have graphics nearly as good in Open world for an optimized game than a linear game.
Only things he told me...
Okay that is what i thought. This actually sounds ridiculous.
I'm watching that interview of Todd Howard now, it's quite good. Anyway, so as he says MS and Sony won't be screwing up at the starting line for next gen. But I'm sorry, if Sony ends up having a big advantage with their loading and streaming solution i definitely would call that MS screwing that part up. It's for sure gonna make a good amount of gamers change their console of choice, I sure will, especially when BGS games like TES and mainline Fallout games are very big reasons why I game.
And from what I'm reading the differences between the two could be quite significant and maybe even bigger than X VS Pro right now when it comes to how games look and perform.
And from what I'm reading the differences between the two could be quite significant and maybe even bigger than X VS Pro right now when it comes to how games look and perform.
The cost part is interesting, because it requires less but faster RAM than a normal drive. It very well could be cheaper than a regular SSD in production, the downside is the years of engineering and it not being terribly useful outside of game consoles. If Sony sends all the controller stuff to a Ryzen CPU core instead of an embedded ARM chip, the speed/cost potential is even higher, with a little extra anxiety about stability. The usual dedicated OS core is probably overkill with a CPU that fast, probably no problem giving it an extra task. Iirc, the reason there's a dedicated core is for I/O encryption/decryption, so it might make a lot of sense to do it that way. I just don't remember if that was a possibility in the patent.
To echo gofreak's reminder, just because there's a patent doesn't mean they're using it. They could've run into issues making the low-level file system code stable for all we know.
edit:
I wouldn't call it a screw up on MS's part if they've got a decent m.2 SSD in their system. They'd have longer loads and less streaming capability*, but it'll be in line with PC specs, so it'll be perfectly fine for everything running on the system. The only games it'd hold up is third-party console exclusives. The PS5 would be better in that regard for third-party games, but the Xbox might have other advantages.
*again, assuming the full potential of Sony's patent is reached.
Last edited:
Interesting chatter...although even taking it as true, I'm not sure it indicates a specific level of SSD bandwidth. I'd still urge caution and taking everything with a grain of salt :)
Taking Cerny's comments, I'd expect something north of 3.5GB/s (the typical PCi3 standard). Anything significantly north of 5GB/s seems to start straying into too-good-to-be-true territory IMO. Would love to be wrong, but I'll keep my expectations down there.
Whatever the bandwidth number, if it's high enough for devs to start tinkering about with traditional speed/space assumptions and tradeoffs, that would be exciting though.
I correct the typo, this is 2 to 3 GB of ram. Imo I think the speed of the SSD help them to use a different memory setup with HBCC.
Lets be honest though. Who on this forum is actually going to buy a Lockhart? Seems like a pointless console to me. At least in regards to this forum
I think a lot of people are putting too much faith in patents that get thrown away often & too little faith in Microsoft, i bet both of them just throw in a 1TB NVME SSD & call it a day. They could buy them in bulk super cheap from the likes of Samsung.
I can't see MS letting Sony one up them on loading times so easily.
I can't see MS letting Sony one up them on loading times so easily.
There are plenty of cheap bastards here, you'd be surprised.
It is someone I know for 14 years. I have full confidence in this guy. He is now in his forties and works for years in the videogame industry. And Lisa Su herself told it is a proprietary solution. My allegation will be easy to verify when the PS5 is revealed with the SSD and DualShock 5. I just hope he did not lie to me but I don't believe it one second...
That's fair enough then, still im sure Microsoft will have their own unique solution if that is the case. if MS doesn't however, i guess i can be right in another way of Lockhart holding back multiplats lol.
Turing isn't bandwidth hungry as GCN, the 2080RTX has 14Gb memory that gives 448GB/s of bandwidth and it's a 10TF GPU so the same 448GB/s will be more than enough for the 2070 TI. Regarding price, the rumor is that the 2070RTX will drop to 399$, the 2070TI (or Super) will be 499$ and the 2080RTX will drop to 599$. If true then it's a move to undermine Navi. The Sapphire guys said that Navi will be priced 399$ and 499$ and AMD showed us that the 5700 performance is just under the 2070RTX so basically NVIDIA will be making Navi DOA with a 399$ 2070RTX and 499$ 2070TI.
For HBCC it is just my own speculation. I don't think they use less RAM but a different setup of RAM in PS5...
wait, you have kind of inside information regarding the sony implementation with their ssd?
From my understanding he knows someone with a lot of experience in this field and in game development
I just say they use the technology in the patent Gofreak discover maybe they use a bomb to kill a little bird and the SSD speed is only 4 Gb/s but so much effort for so little speed. I just know the speed is between 1 GB/S and 20 GB/S like in the patent.
And the DS5 technology is inside a Sony patent somewhere...
EDIT: True professional don't break their NDA, he talks about it seeing the patent and the Gofreak explanation but did not give speed of the SSD for example.
Last edited:
what was that about DS5?
EDIT:
haha, fair enough
No. Did I give that impression? I'm expecting 12TFLOPS as the higher ceiling for Anaconda or PS5... Vega TFLOPS... current TFLOPS?... GCN TFLOPS LOL
Thank you for coming in and expanding on it!
I wonder if they'll be showing the improvements of an SSD like Sony's been doing with Spider-Man or perhaps commenting on ray tracing, although I understand how they'd rather keep quiet regarding these matters.
Hahaha... fuck. I'm just hoping a pricing battle is upon us rather than a quick defeat for AMD, but this is looking like a Vega 56 and Vega 64 scenario once again.
After seeing the SSD patent, he asked if someone discover DS5 technology. I suppose it is in a Sony patent somewhere but there is so much patent for the controller I will not search...
No, LOL, the plot I was pointing at was that NVIDIA will release a 2070TI to compete with NAVI. That is a good sign for NAVI imo.
It's still just a rumor :)
I'm not sure we should be promoting the idea of SSD instead of RAM volume.
1) Even a 5GB/s SSD (which qualifies as the fastest PC SSD ever in April 2019 when Cerny said that) will be ~100 times slower than GDDR6.
2) SSD random read speeds are ~15% of sequential read speeds. So even a 5GB/s SSD will drop to ~0.75GB/s if the data isn't perfectly sequential. So we might get back to data duplication and even that won't be able to prevent non-sequential reads.
The SSD is there to feed a big memory pool and feed it fast, not replace it. I do agree that the part of the memory that is used as a "just in case buffer" won't be needed anymore but I doubt if freeing that space will save much memory, definitely not in the volume of GigaBytes. We should see the SSD as a beautiful companion to a big RAM pool, not a replacement.
Last edited:
Game files are big and organized for sequential read even on HDD. If they reduce the size for the allocation table this is because games use big files. After it did not mean game developer does not need to organize the data for installation on SSD. It will help without needing duplication data and no problem like HDD fragmentation too but they will need to optimize the data with less constraint than on an HDD but the constraint exists.
Last edited:
A. The fact that Samsung is licensing some graphics IP from AMD doesn't mean that RDNA can scale down to smartphone chips.
B. Even if it can it's doesn't tell us anything on how "new" RDNA is compared to GCN.
Consider that smallest GCN chip (Cape Verde) is made of ~1.5B transistors and had a TDP of just 47W on 40nm process.
Apple's A12 SoC is made of 6.9B transistors (that's 4.6X more than Cedar) of which about 1/3 to 1/2 is the GPU - meaning that you can easily make a mobile GCN(1) GPU on modern day production process, no need for anything "new".
That's where the proprietary stuff (IO, sram, custom filesystem) does its additionnal magic. It's not only about SSD speed. From the patent they expect from 1MB/s to 20MB/s real bandwidth, not theoretical. Of course they wouldn't disclose the exact numbers in a patent...
Sony and MS engineers are still undecided on the exact TF they will deliver. They are still busy making their way through this thread to figure it out but every time they settle on a figure we change our minds.
And you never use full memory for display one frame, the faster the SSD is the faster you can have data fast enough to not preload the data long time advance in the system. It will never be the data on screen but maybe the data useful some few frame later.
Yup. On average, half to 2/3rds of the contents in memory get like 90% of the hits. That's how, through HBCC, AMD was able to halve memory required for a game with a really tiny hit to perf (modded version of Deus Ex, IIRC).
He said too using a SSD will be good too on Xbox. It is maybe one of the reasons GTA 6 did not release on PS4 and Xbox One, the HDD nightmare...
Ms 12 navi TF ...ps5 9.something navi TF ...lockhart 6 navi TF
Edit: there will be other differences in customization
Edit: there will be other differences in customization
That's how some games use VRAM for caching assets which HBCC may take advantage of spilling this cache into system RAM while having less VRAM on the GPU. The catch is that a) if such caching will in fact come into play - as in the cached data will be required for some frame - you'll get a frametime spike as HBCC will have to load the data from system RAM over PCIE. And b) it's essentially how modern engines work without any HBCC anyway. Hence why there never really was many benefits from HBCC in practice.
Saying that fast storage would allow to render the same assets with less V/RAM is a bit misleading. Streaming would work a lot faster and would thus need less V/RAM to precache assets from storage as you would need to keep less of them in V/RAM at any given moment. But you would trade some GBs of relatively cheap DRAM for a rather expensive high speed NAND storage and that would give you essentially the same visual result for a higher price of the hardware unit. The only reason I see for adding such storage is to provide user experience benefits which would be otherwise impossible to achieve, not save on system RAM size as such savings would likely not be savings at all.
Yes, Navi, but what about the scaling to consoles? Like this then, let's get rid of Nvidia. AMD's GPUs which released alongside the PS4 in the winter of 2013/2014 was the Rx 200 series. The top end card was at 5.6 TFLOPs (250W), whereas what went in to the PS4 was at 1.8 TFLOPs. So the PS4 had roughly a third of the TFLOPs from the best performing card in the range. With the same scaling this time, for a 12 TFLOP PS5, we'd need consumer AMD cards at 36 TFLOPs by next year.
I'm really just noting that the PS4 got a modest GPU compared to what was available for the desktop market at that time, whereas the expectations here are on a seriously high performing one. Could happen! But with higher prices now, higher power draws etc, I just don't know how realistic it is.
PS5:
9 TF
1 TB of RAM via Optane-like custom SSD
All PS4 games render at 8K automatically. Framerate boosts require patches.
Is this sufficiently uncomfortable? :)
You have to count production processes, wafer prices, relative die sizes and transistor complexities into such comparisons. What PS4 got in comparison to the GPU of a top end AMD card of its time says little about what PS5 may get in comparison to a mid range AMD card (RX5700 isn't top end) of a previous year (PS5 isn't launching in 2019).
- Status