Support NeoGAF

[Liabe Brave]

This is the post from the thread I was referring to (again, apologies for not providing the link originally): http://www.neogaf.com/forum/showpost.php?p=231418033&postcount=447

Ah okay, this is a different thread than I was looking at. I thought you said it was the first Digital Foundry thread about Switch.

Ok, so is this what we are looking at?:

first column is A57 @ 1.9GHz, second column - Jaguar @1.6GHz (PS4), third column - first column / 1.9 (Switch).

If that the case, the Switch's performance may be around 70% of PS4's in single core performance.

So in efficiently multithreaded applications the Switch CPU would be around 35% as powerful as PS4's?

 

[Donnie]

Used Shield TV as the comparison instead (from the same site). First line is Jaguar at 1.6ghz, second is Shield normalized to 1Ghz (Switch).

Single-Core Score       967     779
AES                     916     546
LZMA                    1191    861
JPEG                    1177    897
Canny                   988     927
Lua                     799     660
Dijkstra                1366    828
SQLite                  751     720
HTML5 Parse             902     738
HTML5 DOM               819     891
Histogram Equalization  1022    872
PDF Rendering           926     811
LLVM                    825     1065
Camera                  1003    952
SGEMM                   300     300
SFFT                    472     564
N-Body Physics          650     447
Ray Tracing             744     485
Rigid Body Physics      977     825
HDR                     1333    1004
Gaussian Blur           813     725
Speech Recognition      1012    657
Face Detection          861     789
Memory Copy             1056    822
Memory Latency          3459    1525
Memory Bandwidth        647     680

http://browser.primatelabs.com/v4/cpu/751107

 

[Liabe Brave]

Used Shield TV as the comparison instead (from the same site). First line is Shield, second is Jaguar at 1.6ghz, third is Switch (Shield normalized to 1Ghz, all scores rounded up).

Single-Core Score       1558        967     779
AES                     1091        916     546
LZMA                    1721        1191    861
JPEG                    1793        1177    897
Canny                   1854        988     927
Lua                     1319        799     660
Dijkstra                1656        1366    828
SQLite                  1439        751     720
HTML5 Parse             1476        902     738
HTML5 DOM               1781        819     891
Histogram Equalization  1743        1022    872
PDF Rendering           1621        926     811
LLVM                    2130        825     1065
Camera                  1904        1003    952
SGEMM                   599         300     300
SFFT                    1128        472     564
N-Body Physics          894         650     447
Ray Tracing             969         744     485
Rigid Body Physics      1650        977     825
HDR                     2007        1333    1004
Gaussian Blur           1450        813     725
Speech Recognition      1314        1012    657
Face Detection          1577        861     789
Memory Copy             1643        1056    822
Memory Latency          3050        3459    1525
Memory Bandwidth        1359        647     680

So that'd put Switch single-threaded performance at about 80% of PS4, and total performance at about 40%, if I'm reading correctly.

 

[Donnie]

Yep those numbers suggest that if a game manages to saturate 6 PS4 cores Switch running on 3 cores should be about 40% of that.

 

[lwilliams3]

Used Shield TV as the comparison instead (from the same site). First line is Jaguar at 1.6ghz, second is Shield normalized to 1Ghz (Switch).

Single-Core Score       967     779
AES                     916     546
LZMA                    1191    861
JPEG                    1177    897
Canny                   988     927
Lua                     799     660
Dijkstra                1366    828
SQLite                  751     720
HTML5 Parse             902     738
HTML5 DOM               819     891
Histogram Equalization  1022    872
PDF Rendering           926     811
LLVM                    825     1065
Camera                  1003    952
SGEMM                   300     300
SFFT                    472     564
N-Body Physics          650     447
Ray Tracing             744     485
Rigid Body Physics      977     825
HDR                     1333    1004
Gaussian Blur           813     725
Speech Recognition      1012    657
Face Detection          861     789
Memory Copy             1056    822
Memory Latency          3459    1525
Memory Bandwidth        647     680

http://browser.primatelabs.com/v4/cpu/751107

Ah. Thanks for posting that.

Is there any way we can get a benchmark to the old G3 750CL that is in the GCN, Wii, and Wii U?

 

[plagiarize]

Yep those numbers suggest that if a game manages to saturate 6 PS4 cores Switch running on 3 cores should be about 40% of that.

That lines up pretty well with what we're seeing with Snake Pass too, interestingly enough.

 

[ShadowFox08]

That lines up pretty well with what we're seeing with Snake Pass too, interestingly enough.

So Swith's CPU is 40% of the PS4 base?

 

[Inuhanyou]

So Swith's CPU is 40% of the PS4 base?

Not anymore i don't think...I mean at launch with 6 cores sure, but they did unlock the 7th core for gaming right?

 

[Donnie]

Not anymore i don't think...I mean at launch with 6 cores sure, but they did unlock the 7th core for gaming right?

I think the 7th core was freed for developers, not sure if it's the entire core though. Either way the more cores you have to spread your code over the more difficult it becomes to get the most out of each core. So it's already a bit theoretical to talk about 40% for Switch vs PS4 CPU purely by multiplying cores.

 

[Mr Swine]

Has Nvidia said anything about Tegra X1 polygon output? It should be higher than what PS3 and 360 can do and that it has a tesselation engine right?

 

[Donnie]

Ah. Thanks for posting that.

Is there any way we can get a benchmark to the old G3 750CL that is in the GCN, Wii, and Wii U?

Their's this benchmark by Blu:

http://www.neogaf.com/forum/showpost.php?p=192321356&postcount=99

| CPU                   | N-way SIMD ALUs  | flops/clock | remarks                                        |
|-----------------------|------------------|-------------|------------------------------------------------|
| IBM PowerPC 750CL     | 2-way            | 1.51        | g++ 4.6, paired-singles via autovectorization  |
| AMD Bobcat            | 2-way            | 1.47        | clang++ 3.4, SSE2 via intrinsics               |
| Intel Sandy Bridge    | 8-way            | 9.04        | clang++ 3.6, AVX256 via generic vectors        |
| Intel Ivy Bridge      | 8-way            | 9.09        | clang++ 3.6, AVX256 via generic vectors        |
| Intel Haswell         | 8-way            | 9.56        | clang++ 3.6, AVX256 + FMA3 via generic vectors |
| Intel Xeon Phi (KNC)  | 16-way           | 6.62        | icpc 14.0.4, MIC via intrinsics                |
| iMX53 Cortex-A8       | 2-way            | 2.23        | clang++ 3.5, NEON via inline asm               |
| RK3368 Cortex-A53     | 2-way            | 2.40        | clang++ 3.5, A32* NEON via inline asm          |
| AppliedMicro X-Gene 1 | 2-way            | 2.71        | clang++ 3.5, A64 NEON via generic vectors      |
| Apple A7              | 4-way            | 11.07       | apple clang++ 7.0.0, A64 NEON via intrinsics   |
| Apple A8              | 4-way            | 12.19       | apple clang++ 7.0.0, A64 NEON via intrinsics   |
| Apple A9              | 4-way            | 16.79       | apple clang++ 7.x.x, A64 NEON via intrinsics   |

No A57 there, but does show a A53 beating the 750cl quite handily.

EDIT: Actually it appears to be a Cortex A32, which I assume is a bit lower performance than A53? (and obviously well below A57).

 

[Graphics Horse]

Has Nvidia said anything about Tehran X1 polygon output? It should be higher than what PS3 and 360 can do and that it has a tesselation engine right?

I remember someone saying it was twice the GPU freq, so ~1.5 billion or half for portable.
Not sure if that is correct but I'm not from Tehran.

 

[Mokujin]

Those Bandwidth/Pixel ratios thrown earlier are not a good way to look at the Bandwidth subject, that ignores differences between AMD and Nvidia bandwidth efficiency or different architecture generations.

Back in earlier threads I did a couple of tables showing cuda cores / bandwidth ratio and gigaflops / bandwidth ratio and in comparison to Nvidia Gpus and even at 25.6 Gflops Switch ratios were more than healthy.

Being on mobile now can't repost them, but will try later.

 

[blu]

EDIT: Actually it appears to be a Cortex A32, which I assume is a bit lower performance than A53? (and obviously well below A57).

A32 - ARM 32-bit ISA (circa ARMv8)
A64 - ARM 64-bit ISA
(yeah, I know - blame ARM Holdings for that nomenclature hotchpotch)

 

[Donnie]

Ah right, any guess on what kind of difference in performance A57 would show in that test? A53 is 2-way SIMD and A57 4-way right?

Would be interesting BTW to see that bench ran on a Jaguar based system and also on Shield TV

 

[blu]

Ah right, any guess on what kind of difference in performance A57 would show in that test? A53 is 2-way SIMD and A57 4-way right?

Right. A15, which in theory A57 derives from, does 3.82 flops/clock (which, for reference, is better than a Core2-era 4-way intel).

Would be interesting BTW to see that bench ran on a Jaguar based system and also on Shield TV

Don't have any A57s or Jaguars : /

Getting an A72 mini-ITX very soon, though, so we'll see what it could have been had nintendo been more, erm, ambitious.

A sidenote: the difference between this matrix mult test and, say, Geekbench's is that the former is entirely about 4x4 multiplications, whereas Geekbench's is an SGEMM (general matrix multiplication, an operation of the form A = B * C + D) of large dense matrices. Now, the two tests don't necessarily produce similar flop results on the same uarch, due to the nature of the dataset size, which affects the op stream composition and op grouping/densities. That said, I'd say that 4x4 is slightly more relevant to game scenarios (some would say it's actually quite common), whereas SGEMM is more about scientific applications (technically, SGEMM is an exercise in cache prefetching)

 

[lwilliams3]

Has Nvidia said anything about Tegra X1 polygon output? It should be higher than what PS3 and 360 can do and that it has a tesselation engine right?

I asked a similar question a few months ago, and it was answered by Alstrong of Beyond3D:

On paper, GCN is 1 tri/clk per geometry engine. Tahiti is 2/clk because it has 2, Hawaii/Tonga/Polaris 10 are 4 because they have 4 engines.

Anyways, a paper comparison doesn't help because of culling rates. nV has generally done a better job there, so the rates are somewhat higher per raster engine on Maxwell/Pascal.

RSX was particularly bad there (1tri/2clks on paper) hence all the work done via SPEs for culling, but WiiU also had a slightly faster core clock than PS360.

Switch likely only has one raster engine (it's a relatively big area cost), and WiiU's probably isn't that great coming from an earlier generation than GCN, which itself isn't particularly great (pre-GCN4/Polaris), so it wouldn't surprise me if the mobile mode was fine there.

It doesn't seem like the system has a geometry bottleneck.

The system also should tessellation units that are as good or better than the PS4XB1

 

[dr_rus]

I asked a similar question a few months ago, and it was answered by Alstrong of Beyond3D:






It doesn't seem like the system has a geometry bottleneck.

The system also should tessellation units that are as good or better than the PS4XB1

Tegra X1 has two SMs so it has two geometry units and should be able to push out 2 triangles per clock but since it has only one rasterizer this is true only for peak optimized situations.

 

[lwilliams3]

Tegra X1 has two SMs so it has two geometry units and should be able to push out 2 triangles per clock but since it has only one rasterizer this is true only for peak optimized situations.

I see. Should we expect models and background geometry to be within the similar range as games for the XB1/PS4?

 

[Polygonal_Sprite]

I'm plying Snake Pass on the Switch just now in handheld mode. It's a little blurry at first but it's amazing how fast your eyes get used to it.

It's a very nice looking game, the grass especially is technicallly impressive.

There are quite a lot of framerate drops through especially when moving the camera around and especially when the engine has to draw the entire level from a high vantage point.

Zelda is an incredible improvement post patch in terms of framerate. Very pleased with that.

Can't say I'm not disappointed with these estimates that the Switch CPU is only 30/40% of PS4 CPU especially when we were told by certain posters it would be around 30% faster than it just a few months ago...

The CPU performance is probably the major reason a lot of CPU intensive multiplatform games just aren't being considered for down ports :-(

 

[JonnyDBrit]

I agree with the folks here suggesting not to use Z:BoTW as the baseline for the Switches performance...

That being said, what upcoming title for the switch would you guys recommend keeping an out for a true representatation of the consoles strengths/ weaknesses?

Arms? Splatoon 2? MK8D?

In terms of multi-platform performance comparisons? Sonic Forces could be one to watch. Unlike Snake Pass it won't be ported to the system at the last minute, nor would it be an older game receiving a late port like Skyrim. Add in Generations-derived gameplay which can be heavy on motion blur, vast levels, and various particle effects, and I can imagine it being a good stress test.

 

[TLZ]

I'm plying Snake Pass on the Switch just now in handheld mode. It's a little blurry at first but it's amazing how fast your eyes get used to it.

It's a very nice looking game, the grass especially is technicallly impressive.

There are quite a lot of framerate drops through especially when moving the camera around and especially when the engine has to draw the entire level from a high vantage point.

Zelda is an incredible improvement post patch in terms of framerate. Very pleased with that.

Can't say I'm not disappointed with these estimates that the Switch CPU is only 30/40% of PS4 CPU especially when we were told by certain posters it would be around 30% faster than it just a few months ago...

The CPU performance is probably the major reason a lot of CPU intensive multiplatform games just aren't being considered for down ports :-(

Yea I hooked up my og Xbox last night in my bedroom and played shenmue 2 on 42" full hd tv and even though it was blurry at first was amazing how my eyes got used to it and now I think it's gorgeous.

 

[AlStrong]

It may be worth it. You get a 60% clock speed power efficiency with Pascal. Bandwidth can be held a bit more back in handheld mode

Perhaps. I suppose it's a moot point if they just end up licensing Parker and just buy chips again as opposed to commissioning R&D + verification of a another design.

Tx2 power consumption is not higher than tx1

You're referring to the figures that nvidia released for the chip's TDP? (Not power consumption nor includes a screen)

I'm wondering about the baseline - a wider memory bus adds to the overhead static power so overall consumption may not go down as much even at lower memory clocks (undocked) despite seeing a huge efficiency gain at higher clocks on 16nmFF.

A lower memory clock might do the trick, but I don't know the crossover point where it'd make sense or they just use the extra bandwidth mainly in docked mode.

¯_(&#12484_/¯

 

[dr_rus]

I see. Should we expect models and background geometry to be within the similar range as games for the XB1/PS4?

No idea really. I think that geometry detail is mostly CPU limited these days, even on newer thin APIs like Vulkan, and since Switch is quite a bit lacking in CPU department even when compared to the two Jaguar based home consoles I think that it will probably be generally simpler in geometry. Geometry also consume memory which Switch has less than XBO/PS4. But then there's a better Maxwell tessellator and NV's own API which may make things easier on CPU so it's really hard to guess just from basic specs.

 

[Log4Girlz]

I wonder what this thing is capable of. BOTW is first Gen and built for Wii U.

 

[ShadowFox08]

Not anymore i don't think...I mean at launch with 6 cores sure, but they did unlock the 7th core for gaming right?

It may be better than the Wii U's CPU, especially with an easy to make architecture, but its still bad for a home console. But at least in relative to the PS4's CPU, it isn't worse per se. The CPU, GPU, and RAM all each seem to scale down by 2.5 to 3x from the PS4 nearly perfectly.

So the discussion about Switch's CPU and bandwidth being potential major bottle necks is true then. But how much better is PS4's bandwidth over the Switch again? Is it more than a factor of 3?

I'm plying Snake Pass on the Switch just now in handheld mode. It's a little blurry at first but it's amazing how fast your eyes get used to it.

It's a very nice looking game, the grass especially is technicallly impressive.

There are quite a lot of framerate drops through especially when moving the camera around and especially when the engine has to draw the entire level from a high vantage point.

Zelda is an incredible improvement post patch in terms of framerate. Very pleased with that.

Can't say I'm not disappointed with these estimates that the Switch CPU is only 30/40% of PS4 CPU especially when we were told by certain posters it would be around 30% faster than it just a few months ago...

The CPU performance is probably the major reason a lot of CPU intensive multiplatform games just aren't being considered for down ports :-(

Any PS4/xbone game should be scaled without too much trouble the switch. Devs that say they aren't porting to the switch because of "power reasons/isn't possible" are straight up lying and just don't think it would sell.

Again, if the PS4 has a factor of 2.5 to 3 over the Switch in CPU, RAM, and GPU, then the CPU isn't relatively underpowered when compared to PS4/Xbone, and I don't think ports should be a problem. On the flipside, no single aspect of the Switch hardware stands out from its competition for any single area in hardware other than power efficiency. Anyway, from the ports we've seen so far, the Switch has held its own. Games like am Setsuna, Snake Run have held up on their own on the switch. It will be interesting to see more demanding games that take more bandwidth and other areas. AAA 3rd party platforms especially.


X2/pascal switch version can't come out any sooner. It should scale much better with third party ports.

 

[icecold1983]

No idea really. I think that geometry detail is mostly CPU limited these days, even on newer thin APIs like Vulkan, and since Switch is quite a bit lacking in CPU department even when compared to the two Jaguar based home consoles I think that it will probably be generally simpler in geometry. Geometry also consume memory which Switch has less than XBO/PS4. But then there's a better Maxwell tessellator and NV's own API which may make things easier on CPU so it's really hard to guess just from basic specs.

not in gpu driven rendering. which is still uncommon due to devs having to worry about pc

 

[lwilliams3]

No idea really. I think that geometry detail is mostly CPU limited these days, even on newer thin APIs like Vulkan, and since Switch is quite a bit lacking in CPU department even when compared to the two Jaguar based home consoles I think that it will probably be generally simpler in geometry. Geometry also consume memory which Switch has less than XBO/PS4. But then there's a better Maxwell tessellator and NV's own API which may make things easier on CPU so it's really hard to guess just from basic specs.

Thanks for the reply. Consider the comparisons with the CPUs, I suspect that the Switch will at least be a few notches above the Wii U in potential geometry.

It may be better than the Wii U's CPU, especially with an easy to make architecture, but its still bad for a home console. But at least in relative to the PS4's CPU, it isn't worse per se. The CPU, GPU, and RAM all each seem to scale down by 2.5 to 3x from the PS4 nearly perfectly.

So the discussion about Switch's CPU and bandwidth being potential major bottle necks is true then. But how much better is PS4's bandwidth over the Switch again? Is it more than a factor of 3?



Any PS4/xbone game should be scaled without too much trouble the switch. Devs that say they aren't porting to the switch because of "power reasons/isn't possible" are straight up lying and just don't think it would sell.

Again, if the PS4 has a factor of 2.5 to 3 over the Switch in CPU, RAM, and GPU, then the CPU isn't relatively underpowered when compared to PS4/Xbone, and I don't think ports should be a problem. On the flipside, no single aspect of the Switch hardware stands out from its competition for any single area in hardware other than power efficiency. Anyway, from the ports we've seen so far, the Switch has held its own. Games like am Setsuna, Snake Run have held up on their own on the switch. It will be interesting to see more demanding games that take more bandwidth and other areas. AAA 3rd party platforms especially.


X2/pascal switch version can't come out any sooner. It should scale much better with third party ports.

Well, compromises had to be made to make the Switch a real hybrid that doesn't completely break the bank. As for things that the Switch hardware stands out, it also has a more modern GPU featureset than the original PS4/XB1. Things like 2x fp16 (which PS4Pro and Scropio also have) helps it punch beyond its weight compared to the paper specs or the original systems. The RAM bandwidth would be difficult to compare, but it does have about 60% of the system RAM of the PS4.

I suspect the next revision of Switch will be influenced by Xavier. Pascal/X2 wil not be that much a leap from the current system.

not in gpu driven rendering. which is still uncommon due to devs having to worry about pc

Probably was very common for the Wii U.

 

[icecold1983]

Thanks for the reply. Consider the comparisons with the CPUs, I suspect that the Switch will at least be a few notches above the Wii U in potential geometry.




Well, compromises had to be made to make the Switch a real hybrid that doesn't completely break the bank. As for things that the Switch hardware stands out, it also has a more modern GPU featureset than the original PS4/XB1. Things like 2x fp16 (which PS4Pro and Scropio also have) helps it punch beyond its weight compared to the paper specs or the original systems. The RAM bandwidth would be difficult to compare, but it does have about 60% of the system RAM of the PS4.

I suspect the next revision of Switch will be influenced by Xavier. Pascal/X2 wil not be that much a leap from the current system.



Probably was very common for the Wii U.

Was the gpu in wii u capable of gpu driven rendering? I dont think it was. Gcn introduced the necessary hardware for it to be feasible.

 

[Marmelade]

So the discussion about Switch's CPU and bandwidth being potential major bottle necks is true then. But how much better is PS4's bandwidth over the Switch again? Is it more than a factor of 3?

The RAM bandwidth would be difficult to compare, but it does have about 60% of the system RAM of the PS4.

60%?
Are you only talking usable memory in games?
Regarding PS4's RAM bandwidth it's 176GB/s

 

[AmyS]

Pascal / Tegra Parker / X2 wouldn't make a major performance upgrade.

Nintendo will or should wait until Volta / Tegra X3 variant of Xavier.

 

[dr_rus]

Pascal / Tegra Parker / X2 wouldn't make a major performance upgrade.

Nintendo will or should wait until Volta / Tegra X3 variant of Xavier.

It's rather unlikely that Xavier can even be considered a SoC for consumer (mobile) devices.

 

[lwilliams3]

60%?
Are you only talking usable memory in games?
Regarding PS4's RAM bandwidth it's 176GB/s

Yes. The 60% number is for the RAM quantity used for games (3.25GB vs 5.5GB). The Switch has a much lower RAM bandwidth number compared to the PS4, but the use of TBR makes a significant difference.

Was the gpu in wii u capable of gpu driven rendering? I dont think it was. Gcn introduced the necessary hardware for it to be feasible.

You're probably right. I just guessed eariler that
devs would have pushed the Wii U's GPU more.

 

[Donnie]

Can't say I'm not disappointed with these estimates that the Switch CPU is only 30/40% of PS4 CPU especially when we were told by certain posters it would be around 30% faster than it just a few months ago...

The CPU performance is probably the major reason a lot of CPU intensive multiplatform games just aren't being considered for down ports :-(

I wouldn't dwell too much on the 40% number, its purely theoretical based on performance per core and multiplying cores. It assumes that 6 cores will give a 100% performance increase over 3 cores. That doesn't happen in practice, so its really just a baseline number

So the discussion about Switch's CPU and bandwidth being potential major bottle necks is true then. But how much better is PS4's bandwidth over the Switch again? Is it more than a factor of 3?.

Its impossible to compare, as the systems don't treat main memory in the same way. Switch main memory is only used for certain tasks (a bit like XBox One's main memory), while PS4's main memory is used for everything.

 

[Liabe Brave]

But how much better is PS4's bandwidth over the Switch again? Is it more than a factor of 3?

Yes, on paper it's about a factor of 7. But Switch uses bandwidth differently due to tile-based rendering, so the real-world comparison isn't clear.

 

[Pasedo]

I was talking to someone at work today who had his Switch with him and I asked whether the graphics were any good and he said 'amazing, it's got an Nvidia chip'. Clearly the target audience who know very little about specs but perception wise think it's amazing.

 

[Minsc]

I was talking to someone at work today who had his Switch with him and I asked whether the graphics were any good and he said 'amazing, it's got an Nvidia chip'. Clearly the target audience who know very little about specs but perception wise think it's amazing.

For a handheld, the graphics are basically top of the line (it's like stepping 10 years in to the future from Nintendo's previous handheld, a leap equal to about two generations). Amazing is a perfect summary. Sure, smartphones may have newer and better tech, but even four or five years from now, they probably still won't have games as good looking and complex as BotW running on them (unless CEMU gets ported over I guess).

 

[NOLA_Gaffer]

I was talking to someone at work today who had his Switch with him and I asked whether the graphics were any good and he said 'amazing, it's got an Nvidia chip'. Clearly the target audience who know very little about specs but perception wise think it's amazing.

When you see games like Fast RMX running on what's essentially a tablet, yeah, it's pretty amazing.

 

[lwilliams3]

I wouldn't dwell too much on the 40% number, its purely theoretical based on performance per core and multiplying cores. It assumes that 6 cores will give a 100% performance increase over 3 cores. That doesn't happen in practice, so its really just a baseline number

How good are most games on completely utilizing all six cores for the XB1/PS4?

For a handheld, the graphics are basically top of the line (it's like stepping 10 years in to the future from Nintendo's previous handheld, a leap equal to about two generations). Amazing is a perfect summary. Sure, smartphones may have newer and better tech, but even four or five years from now, they probably still won't have games as good looking and complex as BotW running on them (unless CEMU gets ported over I guess).

I agree. If Nintendo developed their handheld at a similar pace as the DS and 3DS, I suspected that it would have been close the PSVita in raw power (which was less than 30GFLOPS at the highest setting) with 480p/540p screen(s). Even the successor to that system may not have surpassed the Wii U in raw power, and that is not a criticism to Nintendo's pacing as much as acknowledging how wide the power gap it is between the PS2 and PS3 or GCN and Wii U. I didn't imagine anything with the graphical fidelity of BOTW, and that is not even efficiently using the system's power in portable mode.

 

[GankzyMcfly]

I see alot of people talking about switch being absolutely amazing with regards to it being a portable... my question is how does it perform as a purely home console? (with a current disability i wouldn't get much use from its portability functions)... I could swear i remember reggie saying switch was a home console first and foremost but here on gaf youd think it was the 3ds's successor.

Ultimately i can't justify the bloated extra costs for the screen i won't use and the joycon's id likely replace with a pro controller. If they released a box with no screen or joycons put a disclaimer on it, a usb for external storage and shaved 100 of the price tag, id buy that day 1.

 

[Minsc]

I see alot of people talking about switch being absolutely amazing with regards to it being a portable... my question is how does it perform as a purely home console? (with a current disability i wouldn't get much use from its portability functions)... I could swear i remember reggie saying switch was a home console first and foremost but here on gaf youd think it was the 3ds's successor.

Ultimately i can't justify the bloated extra costs for the screen i won't use and the joycon's id likely replace with a pro controller. If they released a box with no screen or joycons put a disclaimer on it, a usb for external storage and shaved 100 of the price tag, id buy that day 1.

It performs as an improved Wii U essentially, there's lots of technical discussion if you read the thread (or at least the last 10 pages). Disregarding its most important aspect because you don't need it is really skewing the importance of the device in regards to its peers.

Switch is the sole portable gaming solution going forward, without it, there'd basically be no portable gaming consoles outside of smartphones/tables. That it is a huge leap forward, especially for a Nintendo system, where their handhelds generally are behind the times, is really a surprise.

 

[joesiv]

Things like 2x fp16 (which PS4Pro and Scropio also have) helps it punch beyond its weight compared to the paper specs or the original systems.

Actually, now that the specs of the Scorpio have been officially released, it may not actually support double rate FP16. it seems more heavily based on Polaris rather than Vega.

 

[lwilliams3]

Actually, now that the specs of the Scorpio have been officially released, it may not actually support double rate FP16. it seems more heavily based on Polaris rather than Vega.

Oh.. I need to catch up on Scorpios' info. That would be a missed opportunity IMO, because games that takes full advantage of PS4Pro specs will not get the additional 2xFP16 bonus if it's ported to Scorpio. That also open the doors of posters pointing out that PS4Pro (and Switch) devs can narrow the gap in power by using FP16.

 

[joesiv]

Oh.. I need to catch up on Scorpios' info. That would be a missed opportunity IMO, because games that takes full advantage of PS4Pro specs will not get the additional 2xFP16 bonus if it's ported to Scorpio. That also open the doors of posters pointing out that PS4Pro (and Switch) devs can narrow the gap in power by using FP16.

Indeed...

 

[Donnie]

How good are most games on completely utilizing all six cores for the XB1/PS4?

Its not something I could quantify specifically for XBox1/PS4 games. But generally, as an example, in a favourable situation (where say 90% of code can be parallelised) three cores will generally give you something near to three times the performance of a single core. While six cores will give you around four and a half times performance increase over a single core (absolute best case scenario you might get 5 times). As the cores multiply in number the increase per additional core diminishes further.

 

[dr_rus]

Actually, now that the specs of the Scorpio have been officially released, it may not actually support double rate FP16. it seems more heavily based on Polaris rather than Vega.

I find that very hard to believe. That would mean that MS had Scorpio GPU silicon before Sony had Neo's, and that just doesn't compute considering their launch frames. FP16x2 will most certainly be in Scorpio, I think, and it will most certainly have more Vega features than PS4Pro, not less. HDMI 2.1 support for example can be one such feature.

 

[lwilliams3]

Its not something I could quantify specifically for XBox1/PS4 games. But generally, as an example, in a favourable situation (where say 90% of code can be parallelised) three cores will generally give you something near to three times the performance of a single core. While six cores will give you around four and a half times performance increase over a single core (absolute best case scenario you might get 5 times). As the cores multiply in number the increase per additional core diminishes further.

Hmm. In that scenario, the Switch would performance at about 50% of the PS4's CPU. I thought it may be something like that.

 

[Kilrogg]

I was talking to someone at work today who had his Switch with him and I asked whether the graphics were any good and he said 'amazing, it's got an Nvidia chip'. Clearly the target audience who know very little about specs but perception wise think it's amazing.

I showed BotW running on my Switch at a birthday party the other day: everyone thought it looked really impressive. Well, except one guy who thought it looked too cartoony and Horizon looks better.

For a lot of people, the kind of graphics BotW pushes is good enough, even for a home system. The sheer scope of the game, some great art direction and a few flancy weather effects, lighting effects and particles are enough to impress most people

Spoiler

including me

.

 

[Pasedo]

Well actually I was wrong. Spoke with him again today and he told me he also has a specced up PC and startling rattling on abouts it's hardware, visuals and frame rate prowess and said when he wants eye candy he plays on that. And he likes how his mate who also has a Switch can plug it into his dock when he is over and they can play games from his library. I guess the PC/Nintendo combo gamer is still alive and well

 

[Donnie]

BOTW looks amazing to be fair. Despite how involved I was in the game itself I found myself stopping quite often and marveling at how beautiful the scene in front of me was, using the capture button to take pics as if it was something I might never see again. No screen shots I'd previously seen of the game on my PC came close to doing justice to it.