Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature currently requires accessing the site using the built-in Safari browser.
Hey everyone, staff have documented a list of banned content and subject matter that we feel are not consistent with site values, and don't make sense to host discussion of on Famiboards. This list (and the relevant reasoning per item) is viewable here.
Is it true that it is being rumored that Drake will not arrive in 2023?
But rather in 2023 an improved version of Switch that would arrive until PS4 Pro when it is in Dock, while the version with Drake would arrive in 2025?
Where do these rumors come from?
Is it true that it is being rumored that Drake will not arrive in 2023?
But rather in 2023 an improved version of Switch that would arrive until PS4 Pro when it is in Dock, while the version with Drake would arrive in 2025?
Where do these rumors come from?
Is it true that it is being rumored that Drake will not arrive in 2023?
But rather in 2023 an improved version of Switch that would arrive until PS4 Pro when it is in Dock, while the version with Drake would arrive in 2025?
Where do these rumors come from?
Is it true that it is being rumored that Drake will not arrive in 2023?
But rather in 2023 an improved version of Switch that would arrive until PS4 Pro when it is in Dock, while the version with Drake would arrive in 2025?
Where do these rumors come from?
disregard "pro vs successor" talk. it's a semantical argument. just focus on the specs, it's more powerful than a PS4 in docked mode if the clocks pan out
disregard "pro vs successor" talk. it's a semantical argument. just focus on the specs, it's more powerful than a PS4 in docked mode if the clocks pan out
disregard "pro vs successor" talk. it's a semantical argument. just focus on the specs, it's more powerful than a PS4 in docked mode if the clocks pan out
I would expect the Dock to be as powerful as PS4 Pro, even performance between PS4 Pro and Xbox Series S being optimistic. Anything other than that could be a disappointment, right?
I would expect the Dock to be as powerful as PS4 Pro, even performance between PS4 Pro and Xbox Series S being optimistic. Anything other than that could be a disappointment, right?
This is still portable hardware we are talking about, so I am not sure we can reasonably expect PS4 Pro and beyond levels of performance just yet. At a reasonable price point at least.
I would expect the Dock to be as powerful as PS4 Pro, even performance between PS4 Pro and Xbox Series S being optimistic. Anything other than that could be a disappointment, right?
"As powerful as Other Console" is so often an apples to oranges comparison. By what metric?
Drake could get games that look the same as their PS4 Pro Enhanced versions. Not all PS4 Pro games can come over without sacrifices. Good ports of PS4 games should have no problem. Series S|X games will be able to make a different set of sacrifices than they do for Series S - some may look better, some may look worse, and many will look "different." PS5 miracle ports will look as good/bad compared to PS5 as the Witcher III does compared to its PS4 version - notably cut down.
Some games will be better in handheld mode, with the drop in res suiting the art style and helping performance. Other games will be worse in handheld mode, looking muddy and performing poorly.
All of this depends on how adoption goes, and how much investment devs give to Drake exclusives.
I'm buying Drake no matter what, but this will be annoying since I'd be upgrading from a launch Switch dock which I don't think supports 4K 60, so I have to buy a new dock anyway.
Is it true that it is being rumored that Drake will not arrive in 2023?
But rather in 2023 an improved version of Switch that would arrive until PS4 Pro when it is in Dock, while the version with Drake would arrive in 2025?
Where do these rumors come from?
I think the only place we've heard anything like that is unlikely speculation from this very thread. But even then I think both of these were supposed to be Drake, just on different extremes.
If Nintendo really wanted to have a pro version...they could had done it when the original tegra x1 had a die shrink from 20nm to 16nm with Mariko...Could had higher clocks and added 2gb of extra memory and called it the switch pro...instead they match Mariko clocks to the original switch specifications for extended battery life.
Drake is most likely going to be a new generation for Nintendo.
Just as I did for every Nintendo system before it, yes. But that won’t stop it from being a horrible mistake. It’s just so improbable that it’s not worth humoring further.
I don’t think any of these possibilities are likely (but they might present a cheaper and/or slimmer way of docking the system). But answering you, I wouldn't. Nintendo Switch is my only videogames system since 2017, and I’m absolutely sold into both modes.
That's because they were still releasing games on hardware where SSD read speeds were not assured and had to duplicate assets in a game package to maneuver around seek times. Between that and hardware acceleration for decompression of things like textures and other things, this is why PS5 game packages are fairly routinely smaller than their PS4 equivalents, sometimes by more than half (Dying Light 2 is the most extreme example of that I've seen so far).
Nvidia is providing its own hardware acceleration for decompression, as I'm sure you're aware, and we have yet to learn how it stacks up against other implementations such as Oodle (what's used in PS5).
THis is refreshing to know. I know that ps5 games were around the same or lower, but didn't know how many were significantly lower in size. Somebody in reddit complied some PS5 vs PS4 mulitplatforms and it averaged about 38% smaller in file size for the ps5 game.
"As powerful as Other Console" is so often an apples to oranges comparison. By what metric?
Drake could get games that look the same as their PS4 Pro Enhanced versions. Not all PS4 Pro games can come over without sacrifices. Good ports of PS4 games should have no problem. Series S|X games will be able to make a different set of sacrifices than they do for Series S - some may look better, some may look worse, and many will look "different." PS5 miracle ports will look as good/bad compared to PS5 as the Witcher III does compared to its PS4 version - notably cut down.
Some games will be better in handheld mode, with the drop in res suiting the art style and helping performance. Other games will be worse in handheld mode, looking muddy and performing poorly.
All of this depends on how adoption goes, and how much investment devs give to Drake exclusives.
Making Drake handheld or docked only would be bad. Portable only is just a waste of performance potential. And of course docked only misses that fluidity of being on the go and it might just lose out sales to those that want an in the go device. Hybrid is here to stay and the best of both worlds.
Now if we do get handheld and docked versions down the road, I wouldn't mind, but hybrid should be the main model always.
To be clear, I personally wouldn't want a tv-only switch drake or a portable-only either.
But in case it happened, I would buy it anyways as long as a good enough prospect for the console existed.
Handheld only? does it last as much as a v2/OLED switch? no? does it last at least as long as the switch lite? yes? ok then. Are there enough launch titles to justify a purchase early on? no? do old titles at least run better? yes? fine.
TV-only? are they making proper use of the hardware and setting a decent enough clock speed for the CPU and GPU? no? is the performance at least considerably superior to the older models when running old titles? then fine.
In my opinion, it's believable to see a Switch Drake only home console at the beginning of this new generation, and a few years later the hybrid version of it. Or maybe both at the same time.
The thing I see way too far from happening is a Switch Drake hybrid at $450 or $500. And I don't see either Nintendo selling the next console at a loss.
I'd buy it if it was handheld-only. I'd buy it if it was TV-only. I'd buy it if it was touchscreen-only. But I'd frequently talk about what a stupid move it was.
I always hoped there would be three tier system to the Switch (now Drake): Handheld, hybrid, and TV. Buy whichever one best fits your needs. I also hoped the TV only one would push the hardware a bit more from the benefit of a better heat control system, but that’s just me.
I'd buy it if it was handheld-only. I'd buy it if it was TV-only. I'd buy it if it was touchscreen-only. But I'd frequently talk about what a stupid move it was.
Same. I rarely play handheld mode and secretly wish I could get a super powerful switch that’s tv only..but it’d be nonsensical for Nintendo to give me that
I have a question about the next generation of Nintendo hardware I haven't seen discussed before and presume this is the thread for it.
What colour scheme and surface finish do we think Nintendo could use for the console?
Black Joy-Con and bezel, brushed aluminium bare metal backplate. I mean they've already made the stand, hinges, bezel and rails metal, there's surprisingly little exterior metal on OLED Model. The speaker holes are literally machined into the metal. Black Nintendo Switch Dock with LAN Port (the existing one that comes with the OLED Model with Neon Red and Neon Blue Joy-Con).
Let's set some performance expectations*. Drake will not be a PS4 Pro, even in docked mode. It certainly will not be a Series S.
Series S and PS4 Pro are 4 TFLOPS. If you really pushed docked mode, bleeding 20W+ of power, with a giant fan in the dock, you could get there. But you wouldn't be able to match their 200+ GB/s of memory bandwidth**. You wouldn't be able to match Series S 3.6 GHz octo-core Zen 2 CPUs. You wouldn't be able to match PS4 Pro's pixel and texel fill rates.
The gap between Series S and Series X is massive, and devs are complaining. While you might be able to run enough power in docked mode to light the NuSwitch up like a Christmas Tree, handheld would still be constrained - pushing docked mode to 4 TFLOPS would create a gap just as big between handheld and docked mode as between Series S and Series X - which just means performance left on the table in one mode, or juddery frame rate garbage on the other.
DLSS can help cover that gap, somewhat. A good example would be Death Stranding.
Death Stranding runs at 1080p30fps on PS4, and 4k with checkerboard rendering on PS4 Pro. The Pro has custom checkerboard rendering hardware, and Sony says that it runs about 1.5ms per frame. Checkerboard rendering requires a 2k (1920 x 2160) image to run. If we do a little math...
1080p30fps = 2 million pixels every 33.3 ms
2k30fps, checkerboarded to 4k = 4.1 million pixels every 31.8 ms
In other words, PS4 Pro needs to be a little more twice as fast as the PS4 in order get the same game from 1080p to 4k. Which is where it is at!
Death Stranding has a PC port that uses DLSS 2.0. Digital Foundry used it for one of their first DLSS 2.0 comparison videos. In quality mode, it looks better than checkerboard rendering, while requiring a lower input resolution. It is slower however - some handwavy match suggests that DLSS 2.0 on NuSwitch in docked mode will likely run in 3ms.
2k30fps, DLSS to 4k = 3.6 million pixels every 30.3 ms
NuSwitch, running at something like 2.4 TFLOPS in docked mode, just can't get there. It's like 1.3x a PS4, not 2.2x like the PS4 Pro. But here is the magic trick. DLSS can get a decent looking 4k from only a 1080p input.
1080p30fps, DLSS to 4k = 2 million pixels every 30.3 ms
So, roughly a 10% increase in frame rate over the base PS4. NuSwitch's 2.4 TFLOPS, it's better-than-jaguar A78C CPUs, its decent but not great memory bandwidth can probably get there. Memory bandwidth might be an issue, so perhaps some texture compression. And yes, DLSS performance mode looks muddier than checkerboarding, but it is much sharper and more stable, so it's not an unambiguous loss.
Drake doesn't need to run anywhere near PS4 Pro's level of TFLOPS to get a product that is very close in quality with only slight compromises. Sure, more power is better, but as wild as our imaginations run on docked mode, the constraints on handheld mode are much much tighter.
Well, this is the argument for a 720p screen. 720p is half the pixels of 1080p, and handheld mode to docked mode on the current system is roughly half as powerful. But what about the tensor cores? Well, there is actually a use case for that, it's called DLAA - it's where you use the DLSS pipeline but you don't upscale. This gives you gorgeous MSAA quality anti aliasing, at a much lower cost. For the same of argument, let's say that DLAA takes half as long to run as DLSS, but on handheld you have half as much power, so it balanced.
720p@30fps, DLAA'd: 900,000 pixels, ever 30.3 ms.
About half the pixels, but generate them 10% faster than PS4. At 1.2 TFLOPS in handheld mode, this is as doable for handheld as for docked mode. It preserves a rough "10-20% better than PS4 per pixel" attribute that Docked mode has, and takes advantage of the tensor cores to offload some problematic operations (AA) to further improve performance.
Drake running at 2.5 TFLOPS in docked mode lets games gives games roughly identical performance characteristics per-pixel across the two modes and uses the tensor cores appropriately between the two modes, without any conditional coding on the dev's part. Handheld mode and TV mode shouldn't feel like wildly different devices.
* Actually, I think most of us understand all this, I'm bringing it up for the lurkers and the drive by folk
** In fact, I think memory bandwidth will be the one place that Drake won't get to the base PS4 level. Hopefully it will get close enough it won't much matter?
Memory bandwidth should be... okay-ish? Like if it ends up being 2.4 or 2.5 TFLOPS while docked, it seems reasonably comfy if we copy and pasted desktop Ampere's balance of bandwidth:flops* (~30 GB/s:1 TFLOP is roughly the upper end). And if we deduct ~75 GB/s from 102.4 (128-bit 6400 MT/s ram) to get ~27.4 GB/s leftover for the CPU... at 30 fps, that's a little over 0.9 GB/s per frame.
Hmm, what to use as comparison for bandwidth necessary for the CPU... you know what, let's take a quick trip to PC space. 'Dual channel' = 128-bit, and for DDR4 nowadays, 3200 MT/s is the baseline for decent, I think? (it's as high as JEDEC goes for DDR4, and last I've heard, it's about the highest** one can promise with great confidence that it'll be stable with Intel 12th gen non-K's locked voltage for the memory controller) That's 51.2 GB/s. Multiply that by (27.4/30) to get ~46.76 fps. I flubbed that. Ok, that should be divide to get ~56.05 fps
That is, ignoring OS and background stuff going on as well as fixed config to optimize for versus near-infinite-configs, having ~27.4 GB/s memory bandwidth for the CPU to work with and a target of 30 fps is analogous on some level to 51.2 GB/s and ~47 ~56 fps.
And we know that DDR4 CPUs are capable of higher than that in plenty of games. Which helps inform us a little bit about CPU-side memory bandwidth needs.
*of course, question here is: was desktop Ampere well balanced in the first place?
Memory bandwidth should be... okay-ish? Like if it ends up being 2.4 or 2.5 TFLOPS while docked, it seems reasonably comfy if we copy and pasted desktop Ampere's balance of bandwidth:flops* (~30 GB/s:1 TFLOP is roughly the upper end). And if we deduct ~75 GB/s from 102.4 (128-bit 6400 MT/s ram) to get ~27.4 GB/s leftover for the CPU... at 30 fps, that's a little over 0.9 GB/s per frame.
Hmm, what to use as comparison for bandwidth necessary for the CPU... you know what, let's take a quick trip to PC space. 'Dual channel' = 128-bit, and for DDR4 nowadays, 3200 MT/s is the baseline for decent, I think? (it's as high as JEDEC goes for DDR4, and last I've heard, it's about the highest** one can promise with great confidence that it'll be stable with Intel 12th gen non-K's locked voltage for the memory controller) That's 51.2 GB/s. Multiply that by (27.4/30) to get ~46.76 fps.
That is, ignoring OS and background stuff going on as well as fixed config to optimize for versus near-infinite-configs, having ~27.4 GB/s memory bandwidth for the CPU to work with and a target of 30 fps is analogous on some level to 51.2 GB/s and ~47 fps.
And we know that DDR4 CPUs are capable of higher than that in plenty of games. Which helps inform us a little bit about CPU-side memory bandwidth needs.
*of course, question here is: was desktop Ampere well balanced in the first place?
To be fair, the balancing for Ampere is noticeably tighter compared to Turing as well. (the tightest for desktop Turing is the Titan RTX at ~41 GB/s per TFLOP, I think)
So who knows how much of it is sheer improvement in bandwidth efficiency and how much of it is the economic side (ie screw it, that's enough GDDR6/6X chips for this die or sku) playing a part.
Nope. My heavy travel schedule is why I sold my PS4 and gave my Xbox series X to my brother. I exclusively game on SD and Switch, where Switch accompanies me on most business trips
Hi all
I haven’t looked here for a while and I don’t have time to read the thread atm cause I’m at work - but what was the latest update on a “switch pro” coming out?
Last I heard one is there was one in development (some guy on a Chinese forum was cross posting here about child or cases or something?)
I’m only asking because the internal fan on my current switch is on its way out and I need to get it repaired. Part of me wants to get it repaired, but I’m also considering just buying the OLED model, but if a new addition is on the way out I might just get it repaired and wait for that one
Hi all
I haven’t looked here for a while and I don’t have time to read the thread atm cause I’m at work - but what was the latest update on a “switch pro” coming out?
Last I heard one is there was one in development (some guy on a Chinese forum was cross posting here about child or cases or something?)
I’m only asking because the internal fan on my current switch is on its way out and I need to get it repaired. Part of me wants to get it repaired, but I’m also considering just buying the OLED model, but if a new addition is on the way out I might just get it repaired and wait for that one
GCN uses immediate rendering, so makes much less efficient use of L2 cache than newer tile-based rendering GPUs. Nvidia moved to tile-based rendering with the Maxwell architecture, and AMD I believe switched over on RDNA1. This is one of the reasons I'm hoping Drake has the larger 4MB cache that's indicated (but seemingly not confirmed) by the leak. For any tile-based renderer, increasing the cache size should be a much more power-efficient way to work around bandwidth limitations than just cranking up the memory clocks or bus width.
I don’t really agree with the conclusion about the PS4 Pro though, that was on 16nm and was GCN based. Not only are AMD architectures more memory bandwidth hungry, the GCN that the PS4 Pro used has a lot of bottlenecks in the minutia of it all that hampered its potential. It was inherent to the architecture.
PS4 Pro didn’t really increase the bandwidth that much though, it went from 176GB/s to 217.6GB/s.
One X is a lot better of a comparison if using it over its predecessor and looking at the bandwidth increase. Goes from 68GB/s on the One S to 326GB/s on the One X. That’s absolutely humongous of a bandwidth jump.
To be fair, the balancing for Ampere is noticeably tighter compared to Turing as well. (the tightest for desktop Turing is the Titan RTX at ~41 GB/s per TFLOP, I think)
So who knows how much of it is sheer improvement in bandwidth efficiency and how much of it is the economic side (ie screw it, that's enough GDDR6/6X chips for this die or sku) playing a part.
To be fair, the real comparison point the people should be looking at is the Ampere is actually similar to Pascal not to Turing. Comparisons would be better off in that direction.
Reading through this tweet thread, (you too @oldpuck ) can give an idea of what I mean:
Note this: 3060 has noticeably less memory bandwidth than the 1080TI, but is within spitting distance probably helped by the slightly higher TFLOP count.
Let's set some performance expectations*. Drake will not be a PS4 Pro, even in docked mode. It certainly will not be a Series S.
Series S and PS4 Pro are 4 TFLOPS. If you really pushed docked mode, bleeding 20W+ of power, with a giant fan in the dock, you could get there. But you wouldn't be able to match their 200+ GB/s of memory bandwidth**. You wouldn't be able to match Series S 3.6 GHz octo-core Zen 2 CPUs. You wouldn't be able to match PS4 Pro's pixel and texel fill rates.
The gap between Series S and Series X is massive, and devs are complaining. While you might be able to run enough power in docked mode to light the NuSwitch up like a Christmas Tree, handheld would still be constrained - pushing docked mode to 4 TFLOPS would create a gap just as big between handheld and docked mode as between Series S and Series X - which just means performance left on the table in one mode, or juddery frame rate garbage on the other.
DLSS can help cover that gap, somewhat. A good example would be Death Stranding.
Death Stranding runs at 1080p30fps on PS4, and 4k with checkerboard rendering on PS4 Pro. The Pro has custom checkerboard rendering hardware, and Sony says that it runs about 1.5ms per frame. Checkerboard rendering requires a 2k (1920 x 2160) image to run. If we do a little math...
1080p30fps = 2 million pixels every 33.3 ms
2k30fps, checkerboarded to 4k = 4.1 million pixels every 31.8 ms
In other words, PS4 Pro needs to be a little more twice as fast as the PS4 in order get the same game from 1080p to 4k. Which is where it is at!
Death Stranding has a PC port that uses DLSS 2.0. Digital Foundry used it for one of their first DLSS 2.0 comparison videos. In quality mode, it looks better than checkerboard rendering, while requiring a lower input resolution. It is slower however - some handwavy match suggests that DLSS 2.0 on NuSwitch in docked mode will likely run in 3ms.
2k30fps, DLSS to 4k = 3.6 million pixels every 30.3 ms
NuSwitch, running at something like 2.4 TFLOPS in docked mode, just can't get there. It's like 1.3x a PS4, not 2.2x like the PS4 Pro. But here is the magic trick. DLSS can get a decent looking 4k from only a 1080p input.
1080p30fps, DLSS to 4k = 2 million pixels every 30.3 ms
So, roughly a 10% increase in frame rate over the base PS4. NuSwitch's 2.4 TFLOPS, it's better-than-jaguar A78C CPUs, its decent but not great memory bandwidth can probably get there. Memory bandwidth might be an issue, so perhaps some texture compression. And yes, DLSS performance mode looks muddier than checkerboarding, but it is much sharper and more stable, so it's not an unambiguous loss.
Drake doesn't need to run anywhere near PS4 Pro's level of TFLOPS to get a product that is very close in quality with only slight compromises. Sure, more power is better, but as wild as our imaginations run on docked mode, the constraints on handheld mode are much much tighter.
Well, this is the argument for a 720p screen. 720p is half the pixels of 1080p, and handheld mode to docked mode on the current system is roughly half as powerful. But what about the tensor cores? Well, there is actually a use case for that, it's called DLAA - it's where you use the DLSS pipeline but you don't upscale. This gives you gorgeous MSAA quality anti aliasing, at a much lower cost. For the same of argument, let's say that DLAA takes half as long to run as DLSS, but on handheld you have half as much power, so it balanced.
720p@30fps, DLAA'd: 900,000 pixels, ever 30.3 ms.
About half the pixels, but generate them 10% faster than PS4. At 1.2 TFLOPS in handheld mode, this is as doable for handheld as for docked mode. It preserves a rough "10-20% better than PS4 per pixel" attribute that Docked mode has, and takes advantage of the tensor cores to offload some problematic operations (AA) to further improve performance.
Drake running at 2.5 TFLOPS in docked mode lets games gives games roughly identical performance characteristics per-pixel across the two modes and uses the tensor cores appropriately between the two modes, without any conditional coding on the dev's part. Handheld mode and TV mode shouldn't feel like wildly different devices.
* Actually, I think most of us understand all this, I'm bringing it up for the lurkers and the drive by folk
** In fact, I think memory bandwidth will be the one place that Drake won't get to the base PS4 level. Hopefully it will get close enough it won't much matter?
I don’t really agree with this premise because of what is present above with the architecture.
The highest I can see Drake is actually 3-3.2 TFLOPs while being very, very comfortable position in docked mode. In portable mode it will be 1.5-1.6 TFLOPs at best.
I know some developers have complained about the Xbox series S being a pain to work with, but I’m also going to hold judgment on that because some of those developers are the same ones that released a game that can’t even hold a steady frame rate even on the higher end consoles (or a high end PC), so I’m going to chalk it up as there’s something wrong with them and their game, not the hardware in those cases.
MVG spoke about the series here:
ultimately, I do not think the hardware is really entirely to blame here, I think it is part of the blame, but I also think a big elephant in the room is how developers are approaching the platform to develop for it. They seem to treat the Series S as an after thought rather than first setter, when it’s the lower end platform not the higher end.
Or probably the series S does not get the same amount of attention as a series x, ie they work with the PS5 and SX first, and then the S.
And finally, the muddier look of a game with DLSS performance mode, I’m assuming that muddier is referring to textures here in the comparison to checkerboard rendering, that’s more to do with the bias. A negative bias would have higher quality textures but one with lower textures would be due to the bias accommodated for the internal resolution.
The best example being the Nioh PC port video about it by DF.
GCN uses immediate rendering, so makes much less efficient use of L2 cache than newer tile-based rendering GPUs. Nvidia moved to tile-based rendering with the Maxwell architecture, and AMD I believe switched over on RDNA1. This is one of the reasons I'm hoping Drake has the larger 4MB cache that's indicated (but seemingly not confirmed) by the leak. For any tile-based renderer, increasing the cache size should be a much more power-efficient way to work around bandwidth limitations than just cranking up the memory clocks or bus width.
@LiC has confirmed that it does not have 4MB L2, but 1MB L2. The 4MB was from a simulation layer elsewhere and it’s only mentioned once(?), but 1MB is seemingly present all over the NVN2 API.
I don’t really agree with the conclusion about the PS4 Pro though, that was on 16nm and was GCN based. Not only are AMD architectures more memory bandwidth hungry, the GCN that the PS4 Pro used has a lot of bottlenecks in the minutia of it all that hampered its potential. It was inherent to the architecture.
PS4 Pro didn’t really increase the bandwidth that much though, it went from 176GB/s to 217.6GB/s.
One X is a lot better of a comparison if using it over its predecessor and looking at the bandwidth increase. Goes from 68GB/s on the One S to 326GB/s on the One X. That’s absolutely humongous of a bandwidth jump.
I wouldn't say that's quite a fair comparison, though, as the Xbox One (and S) has embedded SRAM that's intended to be used for bandwidth-heavy operations (ie buffers). This SRAM bandwidth on the One S is reportedly 218GB/s in addition to the 68GB/s to the DDR3. Of course, this is trickier to make use of than the PS4's RAM, particularly as the 32MB of SRAM isn't enough for full frame buffers, so developers have to implement tiling to use it effectively, but there's nonetheless way more memory bandwidth available than the commonly reported 68GB/s figure.
@LiC has confirmed that it does not have 4MB L2, but 1MB L2. The 4MB was from a simulation layer elsewhere and it’s only mentioned once(?), but 1MB is seemingly present all over the NVN2 API.
Thanks for the confirmation. A bit of a shame, but 1MB for Drake is pretty much in line with performance vs desktop Ampere GPUs, so not the end of the world.
I wouldn't say that's quite a fair comparison, though, as the Xbox One (and S) has embedded SRAM that's intended to be used for bandwidth-heavy operations (ie buffers). This SRAM bandwidth on the One S is reportedly 218GB/s in addition to the 68GB/s to the DDR3. Of course, this is trickier to make use of than the PS4's RAM, particularly as the 32MB of SRAM isn't enough for full frame buffers, so developers have to implement tiling to use it effectively, but there's nonetheless way more memory bandwidth available than the commonly reported 68GB/s figure.
While true, I forgot where I read it but devs opted to not really use it as it complicated multiplatform development. With the bandwidth being 109GB/s in one direction and 109GB/s in another direction in actuality rather than the stated 218GB/s . 102GB in and out on the VCR model.
perhaps they went with a hybrid approach, having some in the eSRAM, and the other using the slower DDR3 memory pool.
Thanks for the confirmation. A bit of a shame, but 1MB for Drake is pretty much in line with performance vs desktop Ampere GPUs, so not the end of the world.
I’ve been meaning to ask you this for a while, but what are your thoughts on a system level cache? I am aware that Drake is for video games and it is not a mobile processor, but in mobile processors it is pretty common to see a system level cache of 4 to 8 MB, excluding Apple who does a much larger SLC on the SOC.
ORIN also has a 4MB SysCache and I wonder if that remained. It would certainly reduce memory bandwidth constraints.
Nintendo isn’t necessarily a stranger to having extra cache (that I’m aware of)
I finally dipped my feet into the PC Handheld market with an Ayaneo GEEK Preorder (the cheaper version of the Ayaneo 2). $749 for great performance, 16GIG+512GB. Specs stay the same as the base 2, but with worse build quality basically. The next 2 available was up to $999, which I couldn’t justify.
I share this here bc it’ll be interesting to see how this changes my relationship with my Switch. I’ll no doubt be all in on the Drake regardless for exclusives, but I dont see myself buying ports (one such title already impacted, Portal, which I didn’t get the Switch port of knowing I’d be getting this down the line.)
I know this won’t impact the general market of the Drake since it’s a higher price point over 2x the Switch, but that price is sure coming down year by year. It’ll be interesting to see what happens to the dedicated handheld market over time!
This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
By continuing to use this site, you are consenting to our use of cookies.
twistedvoxel.com
