So we are at 8tf and 13tf for PlayStation. I'm going to put it somewhere in the middle of the two extremes haha. 10.5tf.
You think it's going to end after they announce the numbers? Not a chance. They could announce it is 12 with the same exact chipset as the XSX and people would still be arguing about it.Same here, just want the stupid 9.2 v 12 endless repeating debate to end
I agree, but the entire reason we're talking about March 3rd is because of a Giant Bomb podcast. Thank you for proving my point.Keep in mind that the last two major announcements Sony made regarding the PS5 (the two Wired articles) were not preceeded by any announcements. They just decided to pull the trigger on two (seemingly) random, regular Tuesdays. March 3rd also happens to be a... you guessed it, Tuesday.
lolYou think it's going to end after they announce the numbers? Not a chance. They could announce it is 12 with the same exact chipset as the XSX and people would still be arguing about it.
Xeontech got hit hard damn.It's Friday! I'm drinking! Let's have some sacrifices! Who else is as excited as I am?
Well, don't everyone speak up at once.....
Hnetuduft, Miruna, Nazgûl, Turbowaffles, Xeontech. Please step forward. For the crime of failing to guess the correct reveal date for the PlayStation 5, I hereby sentence you to the following avatarsto wear shamefully until the PS5 releases:
Hnetuduft, when you gotta go, you gotta go. This should help. Sorry, it was the only color they had left in stock.
Miruna, scare children and loved ones with this creepy old man mask!
Nazgûl, call the fire department! The cat got stuck in a jar!
Turbowaffles, send that smile away for a while with this totally un-nutritious breakfast!
Xeontech, remember Kidz Bop? Did you know there's a lot of them?
Well, that's all for today folks. Come back tomorrow for even more silliness because it's a weekend and whatnot. Until then, let's get some dinner.
That's a Volta successor based on the core clock and RAM size. But yes, it's a behemoth.Amper, probably 3080 ti or Titan, freaking 118SM (CU). Yes, 118, big Navi, doesn't sound so big now right?
For prespective, TF count for each frequency:
1500Mhz - 22.6TF
1600Mhz - 24.2TF
1700Mhz - 25.7TF
1800Mhz - 27.2TF
1900Mhz - 28.7TF
2000Mhz - 30.2TF
Lmao. Damn you, Sony
Oh my sweet summer child no.Lmao. Damn you, Sony
So I'll use this avatar until Sony reveals the P55 eh
Well, at least i love cats
Amper, probably 3080 ti or Titan, freaking 118SM (CU). Yes, 118, big Navi, doesn't sound so big now right?
For prespective, TF count for each frequency:
1500Mhz - 22.6TF
1600Mhz - 24.2TF
1700Mhz - 25.7TF
1800Mhz - 27.2TF
1900Mhz - 28.7TF
2000Mhz - 30.2TF
Oh shit lol :(Oh my sweet summer child no.
You wear this mark of shame until the PS5 RELEASES
You gotta read the fine print next time
whats the Nvidia to AMD compute unit conversion?Amper, probably 3080 ti or Titan, freaking 118SM (CU). Yes, 118, big Navi, doesn't sound so big now right?
For prespective, TF count for each frequency:
1500Mhz - 22.6TF
1600Mhz - 24.2TF
1700Mhz - 25.7TF
1800Mhz - 27.2TF
1900Mhz - 28.7TF
2000Mhz - 30.2TF
Console War, console war never changes.You think it's going to end after they announce the numbers? Not a chance. They could announce it is 12 with the same exact chipset as the XSX and people would still be arguing about it.
118 x 64 = 7,552 CUDA cores. 16.6TF @1.1GHz.whats the Nvidia to AMD compute unit conversion?
eg.
AMD =20 workgroups (x 2 = 40 compute units [x 64 = 2560 shader cores] )
What does 118 compute units translate to with Nvidia?
That's a Volta successor based on the core clock and RAM size. But yes, it's a behemoth.
Yes, you are probably right. But there are two, 118 SM and 108 SM GPUs. 108 SMs is probably the ballpark of the 3000 Titan, 3080 ti is probably ~100 SM.
It was 2:1 with pascal, but with Turing they've changed it to 1:1. So 118 SMs are just like 118 CUs. If they can hit 2Ghz, we are talking about a 30TF GPU, Jesus.whats the Nvidia to AMD compute unit conversion?
eg.
AMD =20 workgroups (x 2 = 40 compute units [x 64 = 2560 shader cores] )
What does 118 compute units translate to with Nvidia?
Thanks.118 x 64 = 7,552 CUDA cores. 16.6TF @1.1GHz.
Note that Arcturus is 128 Vega CUs.
64 Cuda cores per SM x 2 instructions per Cuda core per cycle X clock speed (number of cycles per second).Thanks.
Is there a reason they are all being multiplied by 64? Like it an instruction set type thing?
64 ALUs per CU. 1 FMADD and 1FMUL per ALU gives you the 2x.Thanks.
Is there a reason they are all being multiplied by 64? Like it an instruction set type thing?
We're so close to Tim Sweeney's magical 40TF number. By 3nm, we should have a consumer GPU that can hit that.Yes, you are probably right. But there are two, 118 SM and 108 SM GPUs. 108 SMs is probably the ballpark of the 3000 Titan, 3080 ti is probably ~100 SM.
It was 2:1 with pascal, but with Turing they've changed it to 1:1. So 118 SMs are just like 118 CUs. If they can hit 2Ghz, we are talking about a 30TF GPU, Jesus.
Thanks.
Is there a reason they are all being multiplied by 64? Like it an instruction set type thing?
GDDR6 | DRAM | Samsung Semiconductor Global
Meet upgraded GDDR6's industry's fastest speed, greater power efficiency, and incredible versatility, with low voltage for high-bandwidth, high-computing tasks.www.samsung.com
Samsung has updated their GDDR6 catalog to include 16gbps chips. 8Gb chips are in mass production, and 16Gb chips are in sample.
For reference
256bit bus with 16gbps chips = 512GB/s. Oberon is rumored to be ~530 GB/s using underclocked 18gbps GDDR6 chips.
320bit bus with 16gbps chips = 640GB/s. Arden is rumored to be 560GB/s based on rumors from AMD.
The Pro's LPDDR memory is managed off-SoC.Sounds like 8, 16Gb chips will suffice. For additional tasks, LPDDR4 (which is also at 16Gb density atm) might come into play.
This is what the One X's mobo looks like w/ 12, 8Gb ram chips:
So, hypothetically speaking, if 12 chips (setting aside the 384-bit bus width and it being of GDDR5 variety) is considered the upper end, 3 configs come to mind:
1. 24 GB GDDR6 (feels expensive, really expensive)
2. 20 GB GDDR6 + 4GB LPDDR4 (again, feels very expensive)
3. 16GB GDDR6 + 8GB LPDDR4 (to handle a suite of multimedia and social media features)
Then again, I am not expecting it to be more than 10 Chips total, with 8, 16Gb GDDR6 or 16GB GDDR6 + 2, 16Gb DDR4 modules or 4GB; a total of 20GB. I have no idea how the buses would be design for a split memory architecture (Oh hey, PS3!).
Photorealism gaming could actually be possible in this lifetime, insane!We're so close to Tim Sweeney's magical 40TF number. By 3nm, we should have a consumer GPU that can hit that.
Sweeney: 40 TFLOPS Can Render Photo-Realistic Dynamic Scenes, But Humans Require More Than Computing Power
According to Epic's Founder Tim Sweeney 40TFLOPS would be enough to render photo-realistic dynamic scenes, but humans are more complicated.wccftech.com
Sorry, the RAM is DDR3, not LPDDR.Apologies for potential inconvenience. I am still confused by the difference between a SoC and APU. Thus by extension, would you summarize how it pertains to LPDDR4 (that I didn't even know existed on Pro) and Pro?
Danke, merci et thanks.
Sony should do the reveal in a pastebin
Sorry, the RAM is DDR3, not LPDDR.
APU is a subset of SoCs. SoCs is a catch-all term for any packaged part that contains more than a CPU. APU has conventionally meant CPU with iGPU.
Pheonix
Yes, totally, I'm sure NV will have a 40TF card when they reach 5nm. And who knows what RT improvements they've done with Amper. We might see the 3080 ti more than double the 2080 ti RT performance.We're so close to Tim Sweeney's magical 40TF number.
Sweeney: 40 TFLOPS Can Render Photo-Realistic Dynamic Scenes, But Humans Require More Than Computing Power
According to Epic's Founder Tim Sweeney 40TFLOPS would be enough to render photo-realistic dynamic scenes, but humans are more complicated.wccftech.com
Yup. I figure it will be time to update the 1070 I have in my laptop by then (Ampere).Yes, totally, I'm sure NV will have a 40TF card when they reach 5nm. And who knows what RT improvements they've done with Amper. We might see the 3080 ti more than double the 2080 ti RT performance.
Seems like most people don't expect any major third-party company will release any next-gen only games within the first year (Gearbox aside with Godfall).
With that said, I still stand by my opinion that at least Capcom will.Which major third-party companies do you expect to release next-gen only games within the first year?
The title really speaks for itself on this one. With next-gen on the horizon, there are still a ton of owners of the previous systems who may not make the jump right away. With this & Square-Enix's recent announcement of sticking with cross-gen for a while in mind, what other major third-party...www.resetera.com
Thanks.
Is there a reason they are all being multiplied by 64? Like it an instruction set type thing?
With remote play you could play two different PS4 games at the same time, I like it. Maybe they could bring back the two HDMI ports from the prototype PS3 too. :)4SEs. 72 CUs 1.55 ghz. 14 tflops. wide and slow. just like cerny likes it.
If we're talking Street Fighter, Capcom will probably avoid that for SFVI. At most, we may see Capcom reach a marketing deal with Sony. But otherwise, I'd imagine that SFVI & [Insert Versus Game Here] will be PS5/XSX/PC.I fully expect third party exclusives from Japanese companies. It happened this gen on both sides, Capcom had exclusives on both PS4 and XB1 and I don't see why it wouldn't happen this gen.
anexanhume DukeBlueBall
What I forgot to say was that the 108 SM variant running at 1.01Ghz (13.9TF) beats the Titan RTX (16.3TF) in the benchmark, so it looks like Amper isn't just a node size bump but another big architectural improvement. Seems like it's the 1000 series all over again.
Keep in mind this thing probably has 1TB/s or greater memory bandwidth.anexanhume DukeBlueBall
What I forgot to say was that the 108 SM variant running at 1.01Ghz (13.9TF) beats the Titan RTX (16.3TF) in the benchmark, so it looks like Amper isn't just a node size bump but another big architectural improvement. Seems like it's the 1000 series all over again.
You know March 3rd was FFVIIR original release day before the delay. Funny too there's a not so secret FFVIIR demo no one wants to comment on. You cannot stealth drop it now, since we know about it. But we weren't supposed to know about it. Would've been a cool "available now" demo to drop during a PS SoP or whatever. And if this thing is really March 3rd kinda interesting to drop it on the day people were originally planning to play FFVIIR. Also good marketing opportunity while the entire world is tuned in for PS5 reveal. Just saying. It's like poetry, it rhymes.
seems normal for Teslas. these ain't consumer cards with those ram amounts
I was with you up until you assumed that an early March PS SoP would be the PS5 reveal.