Fake
Member
I agree, but is Nintendo we are talking about. They either gonna regress allocating that amount of resource for the streaming stuff or be proud about how succeful was.
DF - Nintendo Switch 2 Confirmed Specs: CPU, GPU, Memory, System Reservation + More
- Thread starter HeisenbergFX4
- Start date
I agree, but is Nintendo we are talking about. They either gonna regress allocating that amount of resource for the streaming stuff or be proud about how succeful was.
Gambit2483
Member
Indeed yes, there are many RAM intensive games that turn off your ability to record video footage.
Kataploom
Gold Member
Came back here to talk about this elephant the room... Not sure they're actually right about it and it is way too optimistic like literally out of the realm of any realistic possibility.Wait a fucking minute
DF says 10 gigarays/s portable and 20 gigarays/s docked
That's almost impossible
A 3080 & 2080 Ti are 10 GR/s, 2080 is 8 GR/s.
This thing has 12 RT cores at low clocks vs 3080's 68 RT cores @ 1.4 GHz.
How in the fuck
Edit: Here Kepler made an estimation, not sure if this is the right way to calculate this tho:
Last edited:
LordOcidax
Member
The magic of a custom chip.Wait a fucking minute
DF says 10 gigarays/s portable and 20 gigarays/s docked
That's almost impossible
A 3080 & 2080 Ti are 10 GR/s, 2080 is 8 GR/s.
This thing has 12 RT cores at low clocks vs 3080's 68 RT cores @ 1.4 GHz.
How in the fuck
Drell
Member
Yeah I did't mention it but thought about it too. I hope that if really nobody use this feature (and I very much doubt it), just like MS with Kinect 2 Nintendo and Nvidia could be wise and free these ressources for the games.
Buggy Loop
Member
Oh, just that
We've had this conversation. Its groundhog day!
Pick 3060 - 12.74 TFlops, divided by 2 according to you
= 6.37 TFlopsPick 1080 - 8.873 TFlops, right? Motherfucking raw raster TFlops
You're telling me that a 3060 Cuda core hack managed to outperform a Pascal raw cuda core raster of 8.873 TFlops by 25% on 14 games average
Or at the heels of the 5700XT's 9.754 TFlops
And averages will of course vary game to game, what about more computational heavy games?
Like this one, the 3060 at the heel of a 6600XT's 10.6 TFlops in Cyberpunk
Or lean even more into computation shaders with mesh shaders, why not?
a 3050 at 9.098/2=4.55 TFlops outperforming the 9.754 TFlops 5700 XT
Amazing!
Maybe we should derate RDNA TFlops
FalconPunch
Member
What a load of rubbish. On your chart, the Rtx 3080 has 30tfs and the 6800xt has 20tfs. Is the 3080 50% faster in raster?Oh, just that
We've had this conversation. Its groundhog day!
Pick 3060 - 12.74 TFlops, divided by 2 according to you
Pick 1080 - 8.873 TFlops, right? Motherfucking raw raster TFlops
You're telling me that a 3060 Cuda core hack managed to outperform a Pascal raw cuda core raster of 8.873 TFlops by 25% on 14 games average
Or at the heels of the 5700XT's 9.754 TFlops
And averages will of course vary game to game, what about more computational heavy games?
Like this one, the 3060 at the heel of a 6600XT's 10.6 TFlops in Cyberpunk
Or lean even more into computation shaders with mesh shaders, why not?
a 3050 at 9.098/2=4.55 TFlops outperforming the 9.754 TFlops 5700 XT
Amazing!
Maybe we should derate RDNA TFlops
At the end of the day, the only thing you did in your comparison is compare unlike things. Referencing Pascal? Who is talking about pascal? The only relevant architectures here are rdna 2 and ampere. Using AW2 that is specifically developed to use Nvidia's features… Yea definitely a representation of the average performance of the GPU.
Anyone with a semblance of intelligence can see through your nonsensical comparison.
Last edited:
LordOcidax
Member
Buggy Loop
Member
This guy doesn't read whitepapers clearly, K KeplerL2
You can't even have a recipe of architecture to architecture Giga rays/s without running the same demos, and certainly nothing to compare to AMD as it depends on the geometry of their demo and only the Turing whitepaper has an indication of that demo with each geometries. In no way is AMD bound to give numbers with those numbers and I've never seen it. Anyone can go make a box in a demo and get GR/s and it means what? Compared to what? Only comparison we ever had is Nvidia's whitepapers for their own demo.
Unless he goes and runs these exact demos, Gigarays from his bench versus the numbers that nvidia would put on a whitepaper for an architecture cannot be compared.
Unless you think a 2060 laptop has 3.6 times the giga rays/s than Nvidia's own whitepaper on 2080 Ti.
Buggy Loop
Member
So, in handheld mode is around 500-700 Gflops almost a Switch 1 docked
Wow, Switch 1 was so close to run Cyberpunk 2077 phantom liberty afterall!
I'm aware about RT, IQ, resolution, physics and frame rate but I'm not much of a techy especially pertaining to numbers like flops, memory size etc, May I ask if this a good specification?
Anyway what matters with Nintendo are there great exclusives. Just asking. Thanks.
Anyway what matters with Nintendo are there great exclusives. Just asking. Thanks.
Last edited:
LordOcidax
Member
Is a Switch 1.5 after allWow, Switch 1 was so close to run Cyberpunk 2077 phantom liberty afterall!
FalconPunch
Member
I mean, it should be obvious by now that teraflops are a useless metric to capture performance but it's the nintendo clowns that keep parroting that narrative. Might as well call the cohort, the Buggy Loop mafia. It's an apt username since the whole line of reasoning is completely buggy and stuck in a loop of rubbish.So, in handheld mode is around 500-700 Gflops almost a Switch 1 docked
Buggy Loop
Member
Those benchmarks I put up are gonna make you cry?
Again
a 3050 at 9.098/2=4.55 TFlops outperforming the 9.754 TFlops 5700 XT RDNA 1 by 13.5%
a 3060 at 12.74/2 =6.37 TFlops outperforming the 10.6 TFlops 6600 XT RDNA 2 by 10.5%
Amazing!
Fucking Nvidia magic
Last edited:
LordOcidax
Member
Yes, when the specs leaked some people here said that those specs were too good to be true…
Last edited:
LordOcidax
Member
I think that you are the only people that have come to the 1.5 TF conclusion number when the Switch 2 is docked… IN THE WHOLE INTERNET… Thats may be some kind achievement.
Boy bawang
Member
We have known those specs for a very long time, but the RAM and CPU allocation is a tad disappointing.
Games look great so far though, and I hope that Nintendo will allow more flexibility when it comes to resources allocation in the future.
This console will give you overall performance of about 2/3 of a Series S when docked and a bit more than 1/3 in portable mode. Best one can hope at that price point in 2025, I guess...
Games look great so far though, and I hope that Nintendo will allow more flexibility when it comes to resources allocation in the future.
This console will give you overall performance of about 2/3 of a Series S when docked and a bit more than 1/3 in portable mode. Best one can hope at that price point in 2025, I guess...
FalconPunch
Member
The only magically thing happening here is your ability to delude yourself.. RDNA 1 suffering due to a lack required hardware features against ampere is amazing? Sure..... Magical indeed. Meanwhile in the real world where like for like comparisons exist, we can see in the charts below that you're fishing for edge cases. Like I said, anyone with even mustard seed levels of intelligence can see through the nonsense you're spewing...Those benchmarks I put up are gonna make you cry?
Again
a 3050 at 9.098/2=4.55 TFlops outperforming the 9.754 TFlops 5700 XT
Fucking Nvidia magic
FalconPunch
Member
I mean, we already know the switch 2 is weaker than the rtx 2050 mobile. DF already confirmed it and the specs were leaked out so we know it's a derivative. We know how that level of hardware performs. The only ones who are surprised are nintendo clowns.
S0ULZB0URNE
Member
No we don'tI mean, we already know the switch 2 is weaker than the rtx 2050 mobile. DF already confirmed it and the specs were leaked out so we know it's a derivative. We know how that level of hardware performs. The only ones who are surprised are nintendo clowns.
Buggy Loop
Member
FalconPunch
Member
Sure you don't but Digital foundry does. They have access to developers who confirm this off the record. I mean, you can inhale as much hopium as you want but the specs are the specs. We have die shots, clock speeds, it's pretty much a done deal.
I see so it is a decent upgrade. I might get one but probably next year or 2. I still have so many backlogs on my ps5 and will wait for a new Zelda, new Earth bound, and Metroid 4.
S0ULZB0URNE
Member
LordOcidax
Member
In the video Richard said that those kinds of comparisons are irrelevant when it comes to the Switch 2 (Basically all their past videos are now trash)… Also looks like he have some insight information from devs about the Switch 2 performance, his tone changed a lot about the topic… I think that you are the one that is going to be surprised next month.I mean, we already know the switch 2 is weaker than the rtx 2050 mobile. DF already confirmed it and the specs were leaked out so we know it's a derivative. We know how that level of hardware performs. The only ones who are surprised are nintendo clowns.
FalconPunch
Member
With the way things are going, if you want it, get it right away. I will not be surprised if it gets hit with a price increase shortly after launch due to "tariffs and rising economic conditions". It's the reason I'm buying one at launch.. That and the desire to get a first edition unit in the event that any hardware flaws are uncovered.
Hawk The Slayer
Member
As long as the console is strong enough to make a whole big game that looks like the HD Zelda tech demo and functions well than it's a pass.
Buggy Loop
Member
And what does the 6600XT lack here then?
Not as magical as your 1/2 TFlops claim without any fucking numbers to support it
So surely easy to prove on average that Ampere is half TFlops compared to AMD
Still waiting
And you're fishing your claims out of your fucking ass
I backed with numbers.
You have a MOUNTAIN of evidence to make that claim that Switch 2 dock would be 1.5 TFlops because Ampere is half TFlops on desktop.
hint to save everyone time : you'll never be able to prove your stupid theory.
FalconPunch
Member
I won't be surprised next month. I've seen the gameplay footage and the performance is just bad. I'm buying it despite the poor performance....
As much as it hurts, it's more than likely correct when comparing flops between ampere and rdna 2. You gotta multiply ampere flops by 0.5-0.67 to get a comparable flop number between architectures when it comes to rasterization. This is not new news to anyone in the pc gpu space..
LordOcidax
Member
Example? Almost all the games showed in the creator voice videos looks and run really good.
FalconPunch
Member
The ramblings of an insane individual. I swear you don't even look at your graph. The 3060 has 12.6 teraflops, the 4060 has 15.11 teraflops, the 6600xt has 10.6 teraflops. Look at the performance difference between gpus? Practically non-existent. Like I said, you clearly don't know what you're talking about and only reference the two same graphs. We have an unbelievable amount of data points for all these gpus and they don't tell the story you're trying to tell. Keep pushing the incorrect narrative though like you're stuck in a buggy loop....And what does the 6600XT lack here then?
Not as magical as your 1/2 TFlops claim without any fucking numbers to support it
So surely easy to prove on average that Ampere is half TFlops compared to AMD
Still waiting
And you're fishing your claims out of your fucking ass
Last edited:
S0ULZB0URNE
Member
Buggy Loop
Member
By your own fucking claim earlier in the thread, 3060 is 6.3 TFlops. YOUR CLAIM. Half TFlops.
"the 6600xt has 10.6 teraflops. Look at the performance difference between gpus? Practically non-existent."
You just fucking self owned yourself
Really time to stop posting. This is pathetic trolling.
edit, forgot to say earlier, alan wake 2 chart you were complaining that the 5700XT lacks features for mesh shaders, there's the a 3060 at 12.74/2 =6.37 TFlops outperforming the 10.6 TFlops 6600 XT RDNA 2 by 10.5%. What's the excuse here again? Your half TFlops argument is a complete dead end.
Last edited:
GigaBowser
The bear of bad news
FalconPunch
Member
The 3060 is 8% slower on average than the 6600xt despite having 20% more "teraflops" and 100gb/s more memory bandwidth than the 6600xt. On top of this, the 6600xt is equipped with an insufficient amount of infinity cache at only 32mbs compared to the bigger rdna2 gpus. It causes performance issues even at 1080p talk less of 1440p. I'd say do the math but you've proven that you're absolutely clueless.
By your own fucking claim earlier in the thread, 3060 is 6.3 TFlops. YOUR CLAIM. Half TFlops.
"the 6600xt has 10.6 teraflops. Look at the performance difference between gpus? Practically non-existent."
You just fucking self owned yourself
Really time to stop posting. This is pathetic trolling.
I'd continue this debate but, I cannot continue a discussion with someone who doesn't even have elementary knowledge on the subject matter at hand.
Last edited:
RodriGames40
Member
If these specs weren't as good as some want to pretend, we wouldn't be seeing them having meltdowns over them.
Last edited:
FalconPunch
Member
Cyberpunk and Hogwarts Legacy. I watch them on a 77inch oled and the flaws are so easy to see even without my glasses. When I compare them to other pc handhelds, consoles, it's just bad. And just so
Buggy Loop
can continue his meltdowns, I'll be posting a picture of my Switch 2 when I receive it just like I did with my 2 ps5 pros, 4090, etc....That certainly an interesting interpretation. One clearly rooted in delusion. If you think I'm spending $700 CAD of my own hard earned money and complaining for no reason, then I think there's nothing left to say.
Last edited:
LordOcidax
Member
Both games looks really good in the creator voice videos…Cyberpunk and Hogwarts Legacy. I watch them on a 77inch oled and the flaws are so easy to see even without my glasses. When I compare them to other pc handhelds, consoles, it's just bad. And just so
That certainly an interesting interpretation. One clearly rooted in delusion. If you think I'm spending $700 CAD of my own hard earned money and complaining for no reason, then I think there's nothing left to say.
Yes there's a difference between measured throughput vs theoretical peak, but the Switch 2 numbers are clearly theoretical peak like MS did for the Xbox Series X since it's obviously not faster than the RTX 2080 Ti.This guy doesn't read whitepapers clearly, K KeplerL2
You can't even have a recipe of architecture to architecture Giga rays/s without running the same demos, and certainly nothing to compare to AMD as it depends on the geometry of their demo and only the Turing whitepaper has an indication of that demo with each geometries. In no way is AMD bound to give numbers with those numbers and I've never seen it. Anyone can go make a box in a demo and get GR/s and it means what? Compared to what? Only comparison we ever had is Nvidia's whitepapers for their own demo.
Unless he goes and runs these exact demos, Gigarays from his bench versus the numbers that nvidia would put on a whitepaper for an architecture cannot be compared.
Unless you think a 2060 laptop has 3.6 times the giga rays/s than Nvidia's own whitepaper on 2080 Ti.
Last edited:
Buggy Loop
Member
3060 56 fps vs 6600XT 55 fps for 15 games at AMD Unboxed
Not a good start of your counter argument
17.7% but I know.. maths
But yet at half TFlops, as per your claim, it manages to beat the 6600XT 10.6 TFlops with just 6.3 TFlops. Amazing.
So your entire argument is that it performs like a 56/55*10.6 = 10.79 TFlops GPU on average and not 12.6 TFlops.
Is this close to half TFlops?
Yet here you are saying to everyone that Switch 2 at half rate 1.5TFlops docked (ha ha ha) is outperformed by Steam deck's 1.6TFlops RDNA 2 with that kind of logic you show above. How absolutely fucking braindead.
Oh so AMD gimps their card capabilities now
How much the 3060 has for L3 cache again? oh shit, 0 MB.
Oooops, maybe that explains the requirement of having more bandwidth. Different architectures and all. I've only said it like every pages in this thread so far?
What's the infinity cache MB do the RDNA 1.5 & 2 the consoles and Steam deck have? Seems like a very sensitive architecture and goes quickly gimped without a fat pool of cache.
You have the work to do for that half TFlops claim troll. Don't even bother replying unless you've put it on paper.
FalconPunch
Member
Well if that's how you feel, more power to you. To give you a point of reference as to where I'm coming from. I think cyberpunk looks bad on ps5, series x, series s, steam deck, etc. This is due to low resolutions, low crowd density, the use of fsr2, inconsistent frame rates, bad TAA ghosting and a really bad ssr implementation by cdpr. It only starts looking good to me on pc with RT enabled and DLSS to address some of the many rendering issues. If I think those look bad, there's no way I can think the switch 2 looks good. Especially when many of those flaws are magnified by the lower input resolution.
That being said, I still preordered cyberpunk because I want to support CDPR as they did the right thing by not using the stupid "game key carts"…
Kataploom
Gold Member
Actually it's mostly Switch 2 optimistic ppl the ones saying "don't look too much into TFLOPS" because on paper the console looks around Steam Deck in raw performance yet it cleans the floor with it comparing third party games...
FalconPunch
Member
Well it's to be expected is it not? If games were coded directly for the steam deck as it were for the switch 2, I hazard that we'd see vastly different results. As for cleaning out the steam deck(Oled) in 3rd party games, there's not nearly enough data to make that declaration. That being said, I have no vested interest in the steam deck as I sold mine in the lead up to the switch 2 release.
We'll see more after the games are released. I think we'll need to see where the cuts were made to the switch 2 versions and if the steam deck even has access to those cuts for performance purposes. The steam deck is just a portable pc and the switch 2 is a console. In terms of performance, a console always gets more out of the hardware over the long run.
Kataploom
Gold Member
The thing is: What matters is how it looks and play, doesn't matter why... Switch 2 versions of games look just like current gen consoles versions but at lower frame rate and/or resolution... I'm curious to see your insight of the console when it comes out, I'll get it at lunch too
bender
What time is it?
Wow, Switch 1 was so close to run Cyberpunk 2077 phantom liberty afterall!
FalconPunch
Member
Agreed, I wish people would focus on that instead of trying to play a power narrative that they're not qualified to discuss.
I strongly disagree with this statement. However I've only looked at 4 3rd party games and half of them do not measure up. There's also no access to unedited extended gameplay footage of 3rd party games. The little snippets I've seen don't lead me to that conclusion. That said, I don't plan to purchase many 3rd party games on switch due to the excessive use of game key cards. As for first party games, it's nowhere near current gen at all. It's very far away.
I'm not going anywhere so I'll be here to comment for sure. I'm looking forward to Digital Foundry's analysis along with Brazil Pixel.
Last edited:
Buggy Loop
Member
Yes there's a difference between measured throughput vs theoretical peak, but the Switch 2 numbers are clearly theoretical peak like MS did for the Xbox Series X since it's obviously not faster than the RTX 2080 Ti.
That's only the accelerate ray/box and ray/tri testers in hardware without doing BVH processing like Nvidia does and those are not giga rays/s, they are intersections/sec.
Its 1 intersection per TMU per clock.
To get to Giga rays/sec you need a number of intersections per ray and that's ray-bounce, / geometry dependent / how the BVH is efficient with how many tests you need depending on geometry and how it stalls waiting for instructions or data.
Giga rays/sec are benchmark results. There's no theoritical peak to calculate.
So your numbers you calculated kind of land neatly close to what DF mentions but if DF mentions truely Giga rays/sec, which only Nvidia uses as far as I know, then it doesn't make an ounce of sense with what you calculate.
This seems like an impossible comparison. Giga rays/s are effectively Nvidia's invention for old scientific paper ray tracing demos (stanford) and I'm not sure how Switch 2 can have those numbers be so high compared to Nvidia's own 2080Ti benchs I agree its sus, I'm sure there's a problem of communication between DF and whoever is leaking, but also they cannot be compared to intersections/sec like you did. We comparing potatoes to apples.
Last edited: