Factor 5: Nintendo Switch Performance Is Somewhere Between Wii U And Xbox One

Julian Eggebrecht, Co-Founder of Factor 5, has recently participated in an interview with German website Spieleveteranen to talk all about the Nintendo Switch. Eggebrecht says that the console’s performance is somewhere between the Wii U and Xbox One and that the Switch is as easy to develop for as the PlayStation 4. He also revealed the news that Factor 5 have got all their rights back, so hopefully this means we will get a new Rogue Squadron title for the platform. I guess we shall have to wait and see.

Source / Source

Advertisements

77 thoughts on “Factor 5: Nintendo Switch Performance Is Somewhere Between Wii U And Xbox One

  1. That’s helpful.
    I would say they’re all definitely somewhere between NES and Tianhe-2.

    But isn’t Rogue Squadron SW game?
    I’d guess Disney has the final say on that matter.

    Liked by 9 people

    1. Roughly weaker than Xbox One, but it’s pretty close.
      Switch is really only 250GigFlops away from the Xbox One in power but thanks to the clockspeed of the GPU and CPU, it’ll be pretty close but not too close.

      Honestly, for a Tablet that is both a Home Console and Handheld in one, this is amazing for the performance. Razer did something similar but it wasn’t much of success.

      Liked by 3 people

      1. Where are you getting that number from? It’s max clock speed in handheld mode is only 384 Mhz making it 196 full-precision GFLOPS and 393 half-precision GFLOPS.

        The number you cited is the theoretical max half-precision performance of the Tegra X1 (1000Mhz x 256 ALUs x 4 half-precision FLOPS per ALU = 1,024 GFLOPS) which the Switch doesn’t even do in docked mode.

        Liked by 1 person

      2. Shut up you dork! You don’t understand how this stuff works! The Wii U has 352 GFLOPS (half-precision, I guess) at 550 MHz (techpowerup.com). The Switch’s die size is approx. the same, but the process size is at least 20nm and the clock speed is 768 MHz. Now let’s do simple math: 352 GFLOPS * 4 / 550 * 768 = 1966 GFLOPS. The Switch GPU would have almost 2 TFLOPS half-precision and almost 1 TFLOPS single-precision! Let’s assume that the die size is only 100mm2, that would still make it perform at 1347 and 673 GFLOPS respectively.
        In handheld mode that would be 673 half-precision GFLOPS AT THE DAMN LEAST!

        Liked by 1 person

      3. Yes, the clock speed is 768 MHz… IN DOCKED MODE.
        In handheld mode, the GPU is running at around 307.2 MHz only. Malice Chris was claiming that it would reach 1 TFlops in handheld mode, which simply isn’t true.
        Not even commenting on the calculations themselves here.

        Liked by 2 people

      4. The Wii U didn’t support double-speed half-precision FLOPS because only AMD’s most recent GPUs even support them at single-speed.

        I’m trying to make some sense of your calculation. Where is the four coming from and why are you dividing by the Wii U’s clock speed? I’m assuming the 4 is supposed to represent 4 times the density or something because the chip in the Switch is the same size as the Wii U’s GPU but 20nm instead of 40nm? Even that doesn’t make sense. Like I don’t know what you think those numbers show. You’re equating the die size to GFLOPS which themselves are unreliable ratings of performance. Even if something like that were true, you’d have to use the combined size of the Wii U’s GPU + CPU (146 + 27 = 173 square millimeters) since the Switch uses a system-on-a-chip.

        That 352 GFLOPS rating is also heavily contended. Those numbers came from a clock speed uncovered by Marcan and ALU count derived from discussion of the Latte’s die shot on Neogaf and Beyond3D. The 8 blocks on the GPU that were most likely the shader cores happened to closely represented an AMD chip where each block contained 20 cores so people assumed there was 160 cores which they’re pretty sure are VLIW5 based on the GPU’s that were in Wii U dev kits. That would make it a 176 GFLOPS GPU. This confused some people who were closely equating GFLOPS with actual performance and thought there was no way that 176 GFLOPS GPU could perform as well as the 200+ GFLOPS GPU in the Xbox 360 even though the Wii U’s GPU was based off a more modern architecture than the X360. The 352 GFLOP number came about when people found that each block of shaders was larger than the blocks in the reference die they were comparing it too which inferred that each block had more than 20 ALUs. It wasn’t double the size but those that brought that up were committed to the idea that there was 40 cores per block and thus 320 cores altogether which would make it 352 GFLOPS (320 cores x 550mhz x 2 FLOPS per ALU = 352 GFLOPS).

        There’s a few problems with that assumption though. First, the chips they used for reference was made on a 55nm process and they just shrunk it to the size it would be at 40nm so they can use it for comparison. Things don’t always shrink perfectly when going to a smaller process so part of the differential could come from that. The other problem was that the 55nm part was manufactured by either TSCM or Global Foundries while the Wii U’s GPU was manufactured by Renasas, so they were comparing processes by two different semiconductor foundries as well.

        Even if each block did have more than 20 cores, they only noticed about an 80% increase in the size of the block so it wouldn’t have been doubled. Since it was VLIW5, the amount of cores would increment in fives meaning there could have been 25 per block (220 GFLOPS), 30 per block (264 GFLOPS), or 35 at max (308 GFLOPS).

        Again though, GFLOPS are a theoretical max and are not representative of real performance.

        The calculation that I gave you is the calculation that every GPU company except ARM uses to come up with their GFLOPS ratings.

        So at 307.2MHz, it’s 307.2 x 256 x 2 = 157.286 GFLOPS. At 384, it’s 384 x 256 x 2 = 196.608 GFLOPS. At 768, it’s 768 x 256 x 2 = 393.216 GFLOPS. To get half-precision GFLOPS, you just multiply each by two.

        I’ll give you more examples. The PlayStation for has 1152 cores clocked at 800 Mhz so it’s 1152 x 800 x 2 = 1843 GFLOPS. The PS4 Pro has 2304 cores at 911 Mhz. Plug it in and you get 4197 GFLOPS.

        https://en.wikipedia.org/wiki/PlayStation_4_technical_specifications <- For reference.

        Liked by 2 people

      5. Source? I actually work as a Switch dev and this doesn’t quite match up with the data I have. I know that CPU clock speed is somewhat configurable on the Switch, but from what I know, GPU clock speed isn’t. At least not in the way you’re describing; you can’t make handheld mode magically use the same GPU speed as docked mode (not only would that drain it’s battery way too quickly, but it would also make the system overheat and/or require that the fan gets active even in portable mode, requiring even more battery usage). I’m also quite sure that the 1 TFlops in portable mode aren’t correct. You’re probably thinking of the Tegra X1 here, but the Switch doesn’t use a Tegra X1, it’s using a custom Tegra chip (which is probably based on a Tegra X1 in some way, but it’s not the exact same chip). Of course not saying that you’re incorrect here (since your profile claims you’re working as a game developer, so you might have access to a Switch devkit yourself). Just saying that this doesn’t quite match up with the information I have, so I wonder if you have a actual source for those 1 TFlops specs?

        Liked by 3 people

      6. The CPU speed is configurable? Digital Foundry claims the CPU speed is completely consistent at all times though it wouldn’t surprise me if Nintendo offered the option of lower clock speeds to improve battery life.

        And you’re absolutely correct about the fan thing. The fan uses up about 1.65 watts which would be significant in handheld mode.

        Like

      7. I probably worded it badly. The CPU speed is consistent, yeah, but if I remember correctly, you can actually configure it to run at an even lower clock speed (in BOTH modes). The CPU speed will still be consistent between handheld and docked modes, just running at an overall lower speed. Basically, this is for less-demanding games which don’t need that much CPU power. You can configure the CPU to run on an even lower speed and make it use less battery that way. I don’t think it’s possible to have the CPU run at different speeds between the two modes. That would make achieving a reliable performance quite tricky.

        Also another small addendum to my last comment: I said you couldn’t configure GPU speed. This is true only for handheld mode. You can actually configure GPU speed in docked mode. Again only to make it run slower, though. So basically, if you want, you can configure your game to run at the same GPU speed in docked mode as in handheld mode (not that you’d gain much from this besides probably consuming less power).

        Liked by 3 people

      8. That’s what I figured. Basically, a bunch of options for power management. I’d ask if Nintendo actually tests to see if your game can run at a lower clock speed before they approve it for the eShop but I understand that you can’t be an open book about this stuff do to NDAs. Thanks for clarifying though. There’s a lot of people who immediately assume that an clock speed customizations mean overclocking.

        Like

      9. Right now don’t actually know that much, anyways, to be honest. We’re still far from submitting our first Switch game (in fact, we’ve barely started porting our code library to the Switch and it’s running with a low priority, since we still have other projects to work on). So it’ll probably take a good while before we actually get to see what their submission process is like for the Switch.

        Liked by 1 person

      10. Not really. The Switch is nowhere close to as powerful as the XBO.

        The Xbox One has a 1,310 GFLOPS GPU with the Xbox One S having 1,404 GFLOPS GPU. Those are full-precision FLOPS that can be used for vertex shading (polygons) and fragment shading (pixels).

        The Switch has a 393 GFLOPS GPU when docked. Each of it’s GPU ALUs can do twice as many FLOPS (786 GFLOPS) at half-precision, which is good for fragment shaders, but only if the instructions arriving at the ALU are the same.

        So you can say the Switch’s GPU lies somewhere between 393 and 786 GFLOPS or anywhere from 27% to 60% the performance of the XBO with it being closer to the bottom.

        However, this is the wrong way to compare performance between the XBO and Switch because GFLOPS are the theoretical max performance of a GPU, not the actual performance. GFLOPS are derived from an algorithm (number of ALUS x clock speed x 2 FLOPS per ALU). That’s assuming the GPU is infinitely efficient and it has an infinite amount of memory bandwidth.

        In this case, the Switch has 25.6 GB/s of memory bandwidth for textures, geometry, AI, and the frame buffer. The Xbox One has 68.3 GB/s of bandwidth for textures, geometry, and AI and 218 GB/s for the frame buffer. That’s a little over 11 times more bandwidth.

        In reality, the Switch’s docked performance is around the same as some phones except the software made for Switch can actually take advantage of that power.

        Liked by 5 people

      11. I don’t think it’s as close to the xbox1 as you think it is. I think it’s fan hype. I’m going to need to see some facts to back that up. It can’t be close. The laws of nature just wouldn’t support it yet.
        And the xbox one is how old? There is really very little to be gained to compare Switch with anything and just let it be what it is, existing in Nintendo’s ecosystem. Hopefully it will get 3rd party support and games that would have been current-gen 4 years ago. But it will need to keep selling to pull that off.

        Liked by 4 people

      12. Still 2x more powerful than Wii U that was more powerful than PS3. They ported Zelda, they are porting Mario Kart, and all are better than the originals in many ways. Fast RMX too.
        Anyway the point is ‘exclusive games’, if people is just waiting for third party ports they are losing their time. History will always repeat itself. Nintendo want exclusive games on their systems, not ports (still some ports will come).

        If you can’t accept a tablet form factor it’s not the console for you.

        It’s 1/4 the Xbox One in gflops, though it has just half its RAM.

        Like

      13. For me, it’s about the games. And Nintendo systems just don’t have the games I want to play anymore. Zelda (got it), Star fox, that will be a long time, Metroid (God only knows how long that will be) – I like Xenoblade, but I don’t need it. I’m not paying $3-$370 in hardware, just to play one game. Hopefully it will get expanded 3rd party support and new games will come.

        Like

      14. King Kalas X3 {Greatness Awaits at Sony PlayStation 4! Hopefully it will also await us at Nintendo Switch if Nintendo doesn't FUCK things up again!} says:

        Or they have a touch of masochism in them. Or they got a PS4 (me), Xbox One, or PC to pick up any slack the Switch might have.

        Like

  2. We’ve already seen that it’s much more capable than the Wii U, so I’d say it’s close to Xbox One in power as I believe the Switch uses about 1.025TFlops which is about 250GFlops away from the Xbox One which is amazing for a Tablet.

    Honestly, this gives me hope for Third Party games to come to the Switch :D

    Like

    1. I’m not trying to pick on you. To your last comment, I think Switch Sales volume will contribute to 3rd party support far, far more than how strong it is. It will be greatly obsolete by the end of the year when the other systems jump another Gen ahead of the Switch – but that’s not a HUGE deal if Switch builds a solid install base. System power is important, but the size of the install base is much more important. SO, if the switch continues to sell really well, I think you’ll get 3rd party support, no matter how far behind it falls technologically. – I’m intending on that to be a positive statement, that you won’t need to be reliant on the system’s strength to get 3rd party support. :)

      Liked by 3 people

      1. Ok. Well, you got me there! But the Wii was VERY fucking underpowered. It was so so far behind. Switch … I don’t think, is nearly as behind as the Wii was. I think it’s much closer to where it needs to be to accommodate the dev’s minimum requirements. The Wii was Standard Definition for crying out loud. Laughable!

        The Switch is HD, running decent mobile hardware, and designed to be easy to develop for – so hopefully if the install base is there, it’s not the hurdle the wii and wiiu was to develop for.

        Liked by 3 people

      2. The Wii did get third party support just not multi-platform games because the Wii had a GPU that worked very differently from modern GPUs. Modern GPUs use shaders which are programs that run on a GPU. The Wii had a TEV unit which was kind of like turning certain switches on on the GPU to create different effects.

        Liked by 2 people

      3. ||Most children believe that exclusive third class weaponry means nothing because it’s not a multiplatform generic mainstream one the Xbots and Sonyans get…||

        Liked by 1 person

    2. Using Dragon Quest Heroes as an example, the Switch uses the PS3’s assets and runs at 30 fps max with consistent dips to 20 fps in battles in docked mode while running around fields generally hits a smooth 30 fps. In docked mode, it has dips while walking around the field. The PS4 version runs at the same resolution (1080p) but at 60fps at all times, with much higher polygon counts, better draw distance on ground clutter, higher quality textures, more varied textures, and better lighting.

      Liked by 1 person

  3. Don’t forget Nvidia TFlops always = more performance than AMD equivalent, you only need to look at GPU’s like the 980Ti vs the Fury x to see that.. The Fury has more TFlops and usually doesn’t perform as well.

    So i wouldn’t be surprised if they could get Xbox One level performance out of it.

    Like

    1. That’s because GFLOPS aren’t the way GPUs should be compared. Nvidia’s GPUs have better task scheduling and maybe better cache layout result in Nvidia’s chips getting closer to their max theoretical performance.

      Liked by 2 people

    2. I just don’t think that’s physically possible. Not without pumping more voltage into the switch to power the GPU and CPU. Even if those COULD perform anywhere near the xbox one (which is older tech) I think the Switch would need a lot more juice.

      I could be wrong, but until I see something more factual than hopeful Switch fan theories, I can’t get on board with these statements.

      Liked by 3 people

      1. For what it’s worth, benchmarks for Tegra X1 devices and for the GPUs that most closely resemble the XBO and PS4 GPUs are available and they’re nowhere close. Even when comparing the TX1’s GPU configuration with the Maxwell GPU that best matches the performance of the afore mention AMD GPUs, the TX1 would be about 1/3rd the performance and we have every reason to believe the Switch is clocked lower than the TX1 even in docked mode.

        Liked by 2 people

      2. It’s a custom Tegra, that’s a whole SoC with CPU, GPU, memory controller, Audio DSP, ISP, MIPI DSI and CSI interfaces, h.264, HEVC, and VP9 encoders and decoders, etc. In other words, it has way more stuff than the Switch has so it’s likely just customized to get rid of the unnecessary interfaces. When they made the X1 and P1, they were the best mobile GPUs that Nvidia can make. There are no improvements that they can make for Nintendo’s purposes unless you’re inferring that Nvidia was holding back for their own product.

        Liked by 2 people

      3. Yeah, that makes sense. But like I said, at least the architecture is friendly this go around. If Swtich can pull off a large install base… for whatever reason, then I think it’s at least possible for devs to port to it. Whereas for Wii and in some cases WiiU – it was just so difficult. – But I think the whole, “How strong is it?” thing should just die. It’s a tablet. what do you want from it? Can devs develop on it? Yes? then good, now hope for the sales to grow so it looks like aplace developers can make money making games for it!

        Liked by 2 people

      4. I don’t believe anything about the Wii U’s architecture would have been complicated, it’s just that the CPU didn’t have proper SIMD units because it was based on an old PowerPC design. The TX1 is superior but both should be pretty straight forward to develop for.

        Liked by 1 person

      5. The wiiu got a couple great ports I suppose. But not some of the really big ones. And I’ll just have to take your word for it, you seem to know more about the console’s guts than I’ve taken the time to learn! XD

        Liked by 2 people

      6. Whether or not the Wii U was that challenging to develop for, I think devs wanted something nearly effortless to port to make it worth their while. And the Wii may as well have sold poorly since nobody bought games for it. The majority of owners just thought it was the Wii Sports machine. I actually think the Switch will get ports because it offers the closest thing you can get to a console experience in handheld form. Obviously, this is contingent upon Skyrim selling well, which it will.

        Liked by 2 people

  4. I remember Factor 5 doing this gen-surpassing work on Rogue Leader for GameCube.
    Bump/normal maps and all kinds of effects you didn’t see used much until well into the 360’s lifespan.
    (It’s a beaut in high res on emulators too)

    Would be… would be interesting to see what they could come up with if they did specific work on the Switch, or in porting to the Switch.

    Liked by 2 people

  5. Factor 5 was incredible in it’s day. It was like a second Rareware for Nintendo. Any news that is good for Factor 5 is good for the game industry. However, like Rareware, Factor 5 is a shell of its former self. It’s talent was long ago scattered throughout the gaming industry. Also, the rights Factor 5 has recaptured do not include rights to the Star Wars franchise. So any hope of a rebirth of the immensely talented and capable Factor 5 of old are very misplaced.

    Liked by 1 person

  6. Well this is what I thought about the Switch every sense it was announced as a handheld/home console hybrid system. Making it more powerful than the PS4 would’ve made the system cost insane amount but I’m pretty down the line there will be a more powerful version of the Switch that can display games in 4K but still for what the Switch is, for it being more powerful than the Wii U and closer in power to the Xbox One is still impressive

    Liked by 1 person

    1. It’s not the cost that’s the problem, the chip just would have used too much power and generated two much heat to be used as a handheld or in anything device that’s that small. They would have had to go to a 10nm process and clocked it at like 2Ghz and it would be comparable to the PS4’s GPU on paper but it would still fall short of it. In fact, it would fall waaay short of it if they didn’t also increase memory bandwidth by about 7 times (six times as many LPDDR4 chips clocked 16% faster).

      Liked by 1 person

    2. damage control as its finest..
      you said the switch would be more powerful than the ps4, even calling it “a monster”.
      sasori, forgive him for he is a virgin.
      May the glory of Obinna Mii Bless your red ring of death.
      Amen

      – Church of Sasori

      Like

    1. It varies up the experience and potential puts you in situations where you need to get around using bombs, magnesis or stealth to get weapons to attack enemies. You have to let go of your attachment to any particular weapons.

      Like

  7. regardless of power, if the Switch is successful and wildly popular, developers will make games or versions of games that fit the platform. They did it for Wii, for better or for worse, and with this one at least being HD and having that portable element (not to mention it being easier to develop for) we should see companies put a solid effort into not just ports, but different experiences for franchises that were not recently available on Nintendo consoles.

    I hope anyway

    Liked by 1 person

      1. Well, I am not sure. But for Nintendo’s sake, I am glad it pushed Switch sales so they can move ahead. I love my Wii U and I bought BotW for it, but if it is true that mostly Nintendo diehards bought the system (WiiU that is) then those people would know and choose to buy it for WiiU or also move ahead…

        In theory, the WiiU version SHOULD have sold more because the install base is ten times as high or more… Would love to see those numbers when they come out

        Liked by 1 person

  8. It’s a bit like saying a piece of string is as long as a finger or as long as an arm. Eggbrecht answers nothing with that statement.
    Still saying that I hope Factor 5 can somehow resurrect themselves. Would like to see some more Star Wars games that aren’t multiplayer only COD clones.

    Like

  9. The Switch is fairly weak even in docked mode. Only a little stronger than the Wii U and much weaker that the xbox1. That makes it a strong handheld but a weak home console. Oh well.

    Like

  10. King Kalas X3 {Greatness Awaits at Sony PlayStation 4! Hopefully it will also await us at Nintendo Switch if Nintendo doesn't FUCK things up again!} says:

    If Factor 5 have gotten ALL of their rights back, and if they include the Rogue Squadron series, hopefully Disney will be smart & let them have at it with that franchise & let them continue it. And don’t let EA step on them, either. Now they just need to get all their great minds back into the company & we could see Factor 5 return to it’s glory days. *keeps fingers crossed & hopes down to a minimum just in case it doesn’t happen*

    Like

    1. King Kalas X3 {Greatness Awaits at Sony PlayStation 4! Hopefully it will also await us at Nintendo Switch if Nintendo doesn't FUCK things up again!} says:

      Anyway, I’d buy another Rogue Squadron game in a fucking heart beat. I’ll even take HD remasters or even HD remakes of the previous entries.

      Like

Leave a Reply

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s