D
Deleted member 887
Guest
They are unlikely to go with UFS for GameCards. XtraROM allows Nintendo to embed a DRM scheme in the cards themselvesBut especially considering a modded UFS for the new Switchcards could be an option.
They are unlikely to go with UFS for GameCards. XtraROM allows Nintendo to embed a DRM scheme in the cards themselvesBut especially considering a modded UFS for the new Switchcards could be an option.
100MB of max speeds would absolutely limit novel engine features designed to take advantage of faster storage. We have devs in this very thread saying as much.Because of XtraROM, not wanting to force installs, and the availability of UHS-1 Vs faster SD cards, I remain in the camp that the new device will settle for 100MB/s read speeds.
What I'd like to see them do is embrace larger internal storage since they'll probably stick with eMMC, maybe 256GB would be nice.
My actual expectations remain really straightforward. 128GB to differentiate it. No difference in speed. Significantly faster loading thanks to the faster CPU and FDE. 100MB/S + decompression, and likely smaller assets than home consoles to begin with, should be more than enough. Remember, most if not all games still run on spinning rust in a desktop or a Micro SD card in a laptop or Steam Deck.
I sincerely don't think this speed limit will impact the kind of gameplay possible on the device. Asset streaming was a thing back in a disk based world, after all.
100MB+ decompression would be a lot more than 100MB, and smaller assets on top of that.100MB of max speeds would absolutely limit novel engine features designed to take advantage of faster storage. We have devs in this very thread saying as much.
I don't know how they would solve it, maybe select games will require installs.
Do note that 100MB/s is the number for uninterrupted sequential transfers, not including random access delays. When a game loads, it isn't reading one continuous strand of data from one end to the other. It's jumping around all over the place to get each piece it needs, which can be even as little as a handful of KB, tacking on random access delays each jump and dropping that overall transfer rate harshly. It's better than what PS4/XB1 had to deal with as those used mechanical drives with moving parts, but still.Because of XtraROM, not wanting to force installs, and the availability of UHS-1 Vs faster SD cards, I remain in the camp that the new device will settle for 100MB/s read speeds.
What I'd like to see them do is embrace larger internal storage since they'll probably stick with eMMC, maybe 256GB would be nice.
My actual expectations remain really straightforward. 128GB to differentiate it. No difference in speed. Significantly faster loading thanks to the faster CPU and FDE. 100MB/S + decompression, and likely smaller assets than home consoles to begin with, should be more than enough. Remember, most if not all games still run on spinning rust in a desktop or a Micro SD card in a laptop or Steam Deck.
I sincerely don't think this speed limit will impact the kind of gameplay possible on the device. Asset streaming was a thing back in a disk based world, after all.
An octacore a78 can probably easily decompress 100MB/s without sweating. Additional hardware seems overkill for that.Do note that 100MB/s is the number for uninterrupted sequential transfers, not including random access delays. When a game loads, it isn't reading one continuous strand of data from one end to the other. It's jumping around all over the place to get each piece it needs, which can be even as little as a handful of KB, tacking on random access delays each jump and dropping that overall transfer rate harshly. It's better than what PS4/XB1 had to deal with as those used mechanical drives with moving parts, but still.
Here's the thing. If Nintendo recommended a microSD based on having a sequential read speed between 60MB/s and 95MB/s for the Switch that had the CPU decompressing assets, then what does that mean for the goal of Drake's FDE with a 100MB/s microSD? To simply take the pressure off the CPU and not really provide faster loading?
Here's the thing. If Nintendo recommended a microSD based on having a sequential read speed between 60MB/s and 95MB/s for the Switch that had the CPU decompressing assets, then what does that mean for the goal of Drake's FDE with a 100MB/s microSD? To simply take the pressure off the CPU and not really provide faster loading?
The Switch has an extra mode where it turns the GPU way down and the CPU speed way up, explicitly designed for loading screens. You get just enough GPU power to run a little animation, and all the power goes to the CPU so decompression can be done as quickly as possible.An octacore a78 can probably easily decompress 100MB/s without sweating. Additional hardware seems overkill for that.
For better or worse, I don't think it'll be UFS storage. SD card baybeee.Gonna be really interesting to see how much faster loading games on Drake will be vs Switch.
Speaking of storage though... Would storage on Drake be cheaper than what's being offered on current gen consoles? Especially if its UFS? Microsoft is making permanent price drops on their Seagate expansio storage cards for their current gen consoles.. $90 for 512GB, 150 for 1 TB, and $280 for 2 TB.
Trying to gauge how much 128, 256, 512GB, and even 1TB would cost on Drake. 1TB seems unlikely imo. 128GB would be disgustingly low, but considering how Nintendo has barely moved the storage size since Wii U days, it's not impossible. 256GB and maybe 512GB might seem the most likeliest, but the latter would cost a bundle. I wonder how much would it cost them total . Can't find any prices on UFS storage.
Having just checked BotW using the CPU boost mode, loading a save from a fresh start on the main menu showed only ONE core doing most of the work, hitting a spike close to 100% for a split second, but remained roughly around 45% the remaining time. The other game cores would spike around 20%, but stuck at less than 5% most of the time. And this is with loading from my microSD that benchmarked at ~92MB/s.The Switch has an extra mode where it turns the GPU way down and the CPU speed way up, explicitly designed for loading screens. You get just enough GPU power to run a little animation, and all the power goes to the CPU so decompression can be done as quickly as possible.
Of course, increasingly, open world games don't have loading screens. Meaning that those CPU cores are running your physics engine while streaming in data for the next chunk of the game world. Dedicated decompression hardware effectively doubles the amount of CPU power in these scenarios, without having to add CPU cores or run at battery draining clocks.
It's the same way almost all AAA games are running some kind TAA/TU, so REDACTED has a win for farming that out to the tensor cores.
easily. it's still probably gonna be mSD cardsGonna be really interesting to see how much faster loading games on Drake will be vs Switch.
Speaking of storage though... Would storage on Drake be cheaper than what's being offered on current gen consoles? Especially if its UFS? Microsoft is making permanent price drops on their Seagate expansio storage cards for their current gen consoles.. $90 for 512GB, 150 for 1 TB, and $280 for 2 TB.
Trying to gauge how much 128, 256, 512GB, and even 1TB would cost on Drake. 1TB seems unlikely imo. 128GB would be disgustingly low, but considering how Nintendo has barely moved the storage size since Wii U days, it's not impossible. 256GB and maybe 512GB might seem the most likeliest, but the latter would cost a bundle. I wonder how much would it cost them total . Can't find any prices on UFS storage.
Speaking of Lego City, I do have that digitally on Switch. It may not use boost mode, but something tells me it would not have benefitted from it, as having just checked it now, it never exceeded 70% on a single core @ 1Ghz while the rest did practically nothing, and most of the time that core sat around 40%. There were times where during the loading, that core was at 0% while another core was around 24% before something was loaded onto it.I don't think loading times are that much of an issue on the Switch. We're not talking lego city on wiiu levels of slow. Not even persona 3 FES on the ps2 levels of annoying.
I'm kinda late getting to this but holy cow this video is amazing, thanks for sharing!If you wanna get really deep, but still accessible, this is a good video on how Breath of the Wild's lighting system works
The fact it even RAN on Wii U is still amazing to me.I'm kinda late getting to this but holy cow this video is amazing, thanks for sharing!
It and Xenoblade X both. Two of my favorite-looking games ever and both on "failed" hardware. WiiU was the little engine that could, I swear.The fact it even RAN on Wii U is still amazing to me.
LEGO City on Switch was better, but not nearly as much as I expected it to be.I don't think loading times are that much of an issue on the Switch. We're not talking lego city on wiiu levels of slow. Not even persona 3 FES on the ps2 levels of annoying.
LEGO City on Switch was better, but not nearly as much as I expected it to be.
Big B, so Gigabyte.1Gb, as in Gigabit or 1GB, as in 1 GigaBYTE?
As for where it'll land. I don't think it'll hit 1GB, even after decompression, personally. There's too many hurdles and not enough to gain.
Aye, I know that now, I was asking because they initially used "gb" (small b).Big B, so Gigabyte.
Good analysis, but not sure what your ellipses is implying? (genuinely, hope that doesn't read as passive aggressive)Having just checked BotW using the CPU boost mode, loading a save from a fresh start on the main menu showed only ONE core doing most of the work, hitting a spike close to 100% for a split second, but remained roughly around 45% the remaining time. The other game cores would spike around 20%, but stuck at less than 5% most of the time. And this is with loading from my microSD that benchmarked at ~92MB/s.
To me, this makes it seem like the bottleneck is now the microSD, mainly because even the one CPU core doing most of the work isn't being pushed to its max for the majority of the time. If it was the CPU that was the bottleneck before the boost mode update, it doesn't seem to be it now. Granted, this is just one game. I need to check others that use boost mode for loading, especially those that released AFTER the boost mode, because maybe they utilize the CPU cores better. But if they also follow what BotW was doing....
The fact it even RAN on Wii U is still amazing to me.
Absolutely. Breath of the Wild's renderer isn't special, exactly - I would say it's very modern, but standard, PS4 era engine. That is ran on a machine that was easily outclassed by the PS3 is wild.It and Xenoblade X both. Two of my favorite-looking games ever and both on "failed" hardware. WiiU was the little engine that could, I swear.
Have you considered the FY earnings report in tuesday in your schedule?I have a mental schedule of the thread annd the cycle it goes through and I expect next week( edit: this week) to be utter chaos and doom by 2AM.
If you’re late that’s detention.
Following that, which ends at 12PM, we will have lunch out. A BBQ, vegan options are welcome as well.
Then at 2PM, we will discuss launch titles potentially.
That ends until the 10th at 5pm.
And then after that we have a special guest in @hologram to give us a show on why TOTK will be great, then back to the regularly scheduled program of Node discussion until Wednesday morning, 9AM.
Have you all wrote that down? The normal cycle of the thread?
We also have a “we need more leaks” hour and “answering your questions” hour from Thursday and Friday, every hour on the hour for the other visitors.
My dear Tony, that’s where the chaos comes fromHave you considered the FY earnings report in tuesday in your schedule?
The chaos starts Tuesday AM. Japan is 13 hours ahead of where I live, so as early as 6-8am is when we’ll start seeing reportsMy dear Tony, that’s where the chaos comes from
Where I live it would be around 2AMThe chaos starts Tuesday AM. Japan is 13 hours ahead of where I live, so as early as 6-8am is when we’ll start seeing reports
Perfect.The chaos starts Tuesday AM. Japan is 13 hours ahead of where I live, so as early as 6-8am is when we’ll start seeing reports
There's fairly little public information to go on about XtraROM in general, and some of the wording on Macronix's site suggests that the from of it used by Nintendo is probably "ASIC XtraROM" (it's the one they say is used by "handheld gaming consoles"), meaning that it's probably semi-custom on top of that. If there are improvements available in the underlying technology (and you'd figure there probably would be after 6 years), I don't think we'd necessarily have a way of knowing before Nintendo started using it.Do we have any clues on if Nintendo might be using something more akin to an XtraROM successor? Maybe a straight XtraROM2 with higher read speeds if such a thing even exists or is being worked on for the succ?
I think you misunderstood when I spoke about the microSD hitting 92MB/s. That was with the benchmark tool via Hekate I presented some time ago for sequential reads. When random access was taken into account, that number dropped like a brick. Even SSDs aren't immune to it. The NVMe in my gaming laptop, for instance, can hit upwards of 7GB/s, but can drop all the way down to 85MB/s if it is pushing a lot of random access. I recently tested my laptop's drive with 3DMark's Storage benchmark that does a synthetic test of 3 games loading into their main menus from launch, and this is their result.Good analysis, but not sure what your ellipses is implying? (genuinely, hope that doesn't read as passive aggressive)
DEFLATE compression is really hard to make multi-threaded, so I wouldn't expect to see solid multi-core utilization. Texture decompression happens potentially multiple times per frame, so those operations might be split over multiple threads, but you would also need sub-frame sampling rates on your utilization tool to see actual usage. Not sure what the sampling rate is on the various homebrew tools out there.
You're probably right that the read speeds are the limiting factor here, but considering that you're getting a microSD up to speeds that near the sequential read max, then that implies that Nintendo has made some special optimization for a bulk loading operation, which makes sense. The use case for an FDE is likely to be more focused on highly bursty and unpredictable IO that happens during gameplay, so a bit of an apples to oranges situation.
Eh, in the home console space, games actually have been somewhat designed around trying to do as much sequential reads as possible up until pretty recently. Both optical discs and HDDs have much higher penalties for random access than solid state storage, and games have been using various techniques to mitigate that for decades. It's only with the Switch, PS5, and Xbox Series going full solid state that we're starting to see some of these optimizations go away, such as notably smaller files sizes in some cases from no longer duplicating a lot of data so it can be read more sequentially. Obviously the level of optimization varies per game, but it's definitely been a thing.I think you misunderstood when I spoke about the microSD hitting 92MB/s. That was with the benchmark tool via Hekate I presented some time ago for sequential reads. When random access was taken into account, that number dropped like a brick. Even SSDs aren't immune to it. The NVMe in my gaming laptop, for instance, can hit upwards of 7GB/s, but can drop all the way down to 85MB/s if it is pushing a lot of random access. I recently tested my laptop's drive with 3DMark's Storage benchmark that does a synthetic test of 3 games loading into their main menus from launch, and this is their result.
Battlefield V - 1130.83MB/s
CoD: BO4 - 932.36MB/s
Overwatch - 465.69MB/s
I realized just now that I can choose the drive to test it, so what I'm doing now is testing it with my microSD. I'll post those results later as it's going to take a long time, going from 7GB/s sequential to 92MB/s sequential.
My point is, games are not really designed with sequential reads in mind, because they are made up of thousands upon thousands of unique pieces of data scattered across the medium. Some large (like FMVs), some small, and a lot that are really small. Even the really big files are likely to be just archives that contain a lot of smaller pieces that must be traversed via random access.
In the end though, they will deal the cards (no pun intended), and we simply have to use what we are given. It may be proprietary. It may be common microSDs. It could even that REDACTED games must be installed to internal storage, and microSD is just a holding station like external HDDs for PS5 and Series. You can play the games from older devices on them, but not the games for the newer device.
Whoever brings potato salad with raisins in it; we fighting!I have a mental schedule of the thread annd the cycle it goes through and I expect next week( edit: this week) to be utter chaos and doom by 2AM.
If you’re late that’s detention.
Following that, which ends at 12PM, we will have lunch out. A BBQ, vegan options are welcome as well.
Then at 2PM, we will discuss launch titles potentially.
That ends until the 10th at 5pm.
And then after that we have a special guest in @hologram to give us a show on why TOTK will be great, then back to the regularly scheduled program of Node discussion until Wednesday morning, 9AM.
Have you all wrote that down? The normal cycle of the thread?
We also have a “we need more leaks” hour and “answering your questions” hour from Thursday and Friday, every hour on the hour for the other visitors.
I have a mental schedule of the thread annd the cycle it goes through and I expect next week( edit: this week) to be utter chaos and doom by 2AM.
If you’re late that’s detention.
Following that, which ends at 12PM, we will have lunch out. A BBQ, vegan options are welcome as well.
Then at 2PM, we will discuss launch titles potentially.
That ends until the 10th at 5pm.
And then after that we have a special guest in @hologram to give us a show on why TOTK will be great, then back to the regularly scheduled program of Node discussion until Wednesday morning, 9AM.
Have you all wrote that down? The normal cycle of the thread?
We also have a “we need more leaks” hour and “answering your questions” hour from Thursday and Friday, every hour on the hour for the other visitors.
Isnt he the same who backpedaled when reporting the very same thing?
Isnt he the same who backpedaled when reporting the very same thing?
lmaooPerfect.
* Hidden text: cannot be quoted. *
i have 12hs so even if its earlier for me, i expect this to blow up by the time i reach work so i have something to pass the time lolThe chaos starts Tuesday AM. Japan is 13 hours ahead of where I live, so as early as 6-8am is when we’ll start seeing reports
I know the gaming community LOVES to overshoot the capacities of PS hardware in general, especially while lowballing Nintendo hardware in the space of a single breath... But the Wii U was not "easily outclassed by the PS3". Not in any real life timeline. Perhaps in a bizarro one, but under proper scrutiny, that doesn't hold up at all.Absolutely. Breath of the Wild's renderer isn't special, exactly - I would say it's very modern, but standard, PS4 era engine. That is ran on a machine that was easily outclassed by the PS3 is wild.
Then you consider Metroid Prime: Remastered, which is doing something similar* and MSAA - and it's doing it twice as fast as Zelda, to get 60fps.
I believe MP:R is using forward rendering instead of Zelda's deferred rendering.
Forward rendering uses all the same buffers as deferred rendering you see in the video, but where deferred rendering flattens the image before lighting it, forward rendering lights the image first, then flattens it.
Forward Rendering would be a performance nightmare in an open world game, or in a game with lots of moving lights (like the sun), but modernizing a GameCube game is like an ideal use case, where forward lighting is simpler and faster. That's likely how MP:R manages to look so good and hit 60fps.
Indeed, you could argue the PS3 had greater CPU capabilities but Wii U had a clearly more modern, capable GPU so nothing is easily outclassed.I know the gaming community LOVES to overshoot the capacities of PS hardware in general, especially while lowballing Nintendo hardware in the space of a single breath... But the Wii U was not "easily outclassed by the PS3". Not in any real life timeline. Perhaps in a bizarro one, but under proper scrutiny, that doesn't hold up at all.
I'd like to underline and embolden what others said: Great video!If you wanna get really deep, but still accessible, this is a good video on how Breath of the Wild's lighting system works