Support NeoGAF

[Twin Tails]

Nice, one of the main reason I game on consoles is to never think about specs. Glad to see these consoles reaching final specs

 

[StevieP]

Good luck with that, you know every MS thread turns into a versus thread.

You mean pretty much any thread in here. lol
No console or company is spared on this forum (for example, the last couple pages have been Wii U trolling as well as calling MS a bunch of liars).

As I said, on the bright side - now that everything is out soon people can discuss games

 

[digoutyoursoul]

Keaton was a better Batman.

PS4 > PC confirmed.

Nice outfit

 

[Slurmer]

As Batman Actors:

PC: Christian Bale
PS4: Michael Keaton
Xbox: Val Kilmer
WiiU: George Clooney

Ouya: Ben Affleck

 

[IN&OUT]

There is so many computer engineers online you would think MS would've known better.

Apparently they didn't read the Wikipedia teraflop article.

Your statement is true, but that doesn't mean we are baboons who knows nothing at all, we can compare and draw conclusions.

 

[ShapeGSX]

The slide is lying. In order to achieve that they'd need a dual bus ESRAM. X1 does not have one. It's the exact same bullshit Leadbetter was trying to spoonfeed us.

SRAM is generally dual ported (read and write). I'm not sure there are single ported SRAM cells (not my area). It would be pretty stupid to waste all those transistors on 2 ports and then only have one bus.

On what are you basing your single bus assumption?

 

[Winterfang]

PS4: 10
Xbox One: 8
Wii U: 3.5

 

[Flatline]

My point is that they've never really told us with any suitable detail how that ESRAM was designed. This is literally the first time we're learning that its somehow 4 separate 8MB ESRAM chips for the combined 32MB that we know about. Before we were all thinking it was one huge 32MB chunk of ESRAM. We've never been told outside of some leaks, and even those didn't go into the level of detail required, to tell us just what kind of memory we were dealing with. Who made it for them? How was it designed? How about some details on how it was incorporated into the GPU? There's a lot we don't know.

This wasn't done a couple of months before release. This was clearly done from the early architectural design stages of the system, or we wouldn't be looking at it now. They simply never went into the required detail on any of this stuff. Remember, it was Sony being more forthcoming on the various details of the PS4 architecture largely because they're so proud of the turnaround from the PS3. Microsoft was being more cryptic and shy about giving out certain details, and now we have some more details.

Dude, they would have mentioned the dual bus ESRAM, you're reaching Reiko levels of rumor mongering. The number they've mentioning is theoretical bullshit unless they added a second bus, it's that simple. Simulaneous reads and writes does not equal double the speed unless the Xbone can somehow ignore the laws of physics.

 

[cchum]

Should've used the Xperia Z.

You....you're right. Fixed.

 

[Portugeezer]

PS4:

Xbox One:

Wii U:

PC:

 

[Bravocado]

I love this. I love this so much.

 

[Moral Panic]

The slide is lying. In order to achieve that they'd need a dual bus ESRAM. X1 does not have one. It's the exact same bullshit Leadbetter was trying to spoonfeed us.

What makes it bullshit? You have no evidence and are going against direct figures output by the people who do have the facts.

Well, according to sources who have been briefed by Microsoft, the original bandwidth claim derives from a pretty basic calculation - 128 bytes per block multiplied by the GPU speed of 800MHz offers up the previous max throughput of 102.4GB/s. It's believed that this calculation remains true for separate read/write operations from and to the ESRAM. However, with near-final production silicon, Microsoft techs have found that the hardware is capable of reading and writing simultaneously. Apparently, there are spare processing cycle "holes" that can be utilised for additional operations. Theoretical peak performance is one thing, but in real-life scenarios it's believed that 133GB/s throughput has been achieved with alpha transparency blending operations (FP16 x4).

Are you telling me this is a completely made up statistic and they did not get any extra performance out of the chips?

 

[BlackClouds]

PC:
image redacted

oh yeah

 

[squidyj]

PS4 = Stone sculpture
Xbox One = Soundtrack from 'Predator'
WiiU = That one tiny gun from Men in Black
PC = That one pair of headphones that only works on one side now but they're too expensive to replace and the warranty ran out

the right ear on my headphones just went out and I can't bring myself to buy cheap ones but school is also starting and I'm getting a ps4 this year .

 

[IcyEyes]

I thought so then, too. However, newer information since then, through both DF's development sources, and especially now at hot chips seems to directly contradict that initial assumption that they were simply just cobbling bandwidths together.

Unless they're outright lying, which I really don't think is the case, the ESRAM peak theoretical bandwidth is indeed 204GB/s. I'm looking at the diagram from hotchips. I see 68GB/s for the DDR3 off by itself somewhere, I see a 30GB/s coherent bandwidth path way off by itself somewhere, and then further down the diagram I see 4 x 8MB blocks worth of ESRAM totaling 32MB that is labeled as having a minimum bandwidth of 109GB/s and a peak bandwidth of 204GB/s. This isn't the first time that a bandwidth figure from an MS console has been questionable. The 360's 256GB/s for its EDRAM was highly questioned in the early days of the 360, as I remember reading quite a bit of the doubt going around until it was later explained, I believe, by Dave Baumann on beyond3D. People thought Microsoft was playing funny business then, too, but it turned out to be true.

Now obviously I don't know what's going on, but neither do a lot of other people. As an example, Mark Cerny himself stated that it was possible to have an embedded memory chip with upwards of 1TB/s worth of bandwidth for the PS4, but they opted for a simpler design that was easier for developers to extract power from. He said he didn't want developers having to crack some puzzle to get at the system's power. After an unbelievable bandwidth figure like that? 204GB/s doesn't seem all that pie in the sky. I know there's a tendency to be very skeptical of what Microsoft is saying, but they've more or less acknowledged, although not without some kicking and screaming, in so many ways that they aren't packing the kind of raw horsepower that the PS4 is. I don't see what further meaning there is in lying about their memory bandwidth now, as if it would honestly make a big difference to public perception of the console now, especially at hot chips of all places. They gave an honest presentation on Xbox 360 silicon at hot chips when the 360 was launching, and I don't see any reason why they wouldn't do the same this time with the Xbox One. They gain nothing at all with an attempt to wave a bigger penis on their ESRAM bandwidth.

I couldn't agree more. Both are great systems. Developers have what they need to blow our socks off in the coming months and years.

I love to read post like this.

Ps I totally agree with the bold part.

 

[Portugeezer]

oh yeah

 

[BleachAndPepsi]

Whaaaaaat the fuuuuuck??

 

[inpHilltr8r]

Semi-knowledge and implications from that:
Afaik you have to use a whole CU because you can't give multiple different instructions to different cores in one CU. So you are not able to get just 10gflops processing power from one CU while the rest of it does other stuff at the same time. You can switch context but that probably wastes cycles. You probably can use one CU for applying DSP effects to more than one channel by switching contexts though.

You can process more than one sample at once. So if CU is 64 floats wide, then you could process 64 samples of same channel, rather than 1 sample of 64 channels.

 

[Flatline]

SRAM is generally dual ported (read and write). I'm not sure there are single ported SRAM cells (not my area). It would be pretty stupid to waste all those transistors on 2 ports and then only have one bus.

On what are you basing your single bus assumption?

There's a lot of evidence from leaks. Plus if their ESRAM was dual bus they wouldn't be bragging about simultaneous reads/writes as if it's something they invented.

What makes it bullshit? You have no evidence and are going against direct figures output by the people who do have the facts.



Are you telling me this is a completely made up statistic and they did not get any extra performance out of the chips?

Are you seriously quoting the Leadbetter fluff piece as proof?

 

[TheCloser]

Whaaaaaat the fuuuuuck??

Miley Cyrus fail lol.

 

[Winterfang]

the right ear on my headphones just went out and I can't bring myself to buy cheap ones but school is also starting and I'm getting a ps4 this year .

Just buy one of those cheap 10 dollar ones that Walgreens sell. Gummy something is the name, they have a short life span but are surprisingly loud and comfy.

 

[SenjutsuSage]

Proof or just stop. Its getting really sad. Seriously, just stop. It is beyond embarrassing. You like the xbox one, its not a crime. Stop spreading bs all around and trolling like reiko.

First: Take a chill pill.

Second: How is it bs on my part when I'm not the one making wild assumptions about what kind of ESRAM Microsoft put inside their console, and what the capabilities of that ESRAM are? The guy said with certainty what kind of ESRAM Microsoft was using inside their console. Now, in my bewilderment that he could possibly know this information when it has never been shared publicly, I asked him how he arrived at such a conclusion.

Reasonable? I think so. I like the Xbox One. You're damn right it isn't a crime. I also like the PS4, too. I don't troll anything. I just come on here and say my piece just like anybody else. If you don't like what I have to say, feel free to put me on ignore, because I don't see how anything I'm doing should illicit such, usually immature, responses. Now I'm actually trying to have a decent discussion, and talk about this stuff with people in a decent manner, and I'm quite sure you don't see me trying to turn anyone inside out because they disagree with me lol. There's no need to get so edgy when it comes to these consoles. We have no explanation at all for why that bandwidth on the ESRAM is what it is. The closest thing to anything of the such is a Eurogamer article where development sources informed the site of what was taking place with the ESRAM once Microsoft got their hands on near final silicon. People called that article into question for a variety of reasons, and now we have Microsoft more or less showing a similar such increase at their Xbox One silicon presentation at hot chips, a presentation they also made for the Xbox 360, and one that was made honestly.

If this Xbox One presentation is anything like the Xbox 360 hot chips presentation, there will be a pdf file available online, possibly in 4 days that we can hopefully all read and get more information about what's going on with the system. Who knows, there may legitimately be things that Microsoft may not want to share about the Xbox One silicon. Is that unprecedented? Not really. There's things that both Nvidia and AMD still don't share publicly regarding their GPU architectures and how they construct or design certain things.

 

[Codeblew]

That really just proves that they'll talk around/down the differences and performances gaps as much as they can, but not necessarily translate that into development. If the performance gap is really as significant as is suggested by the data, developers are going to have an increasingly difficult time maintaining parity because they'll be doing more work to restrain the performance on one platform. I don't really think anyone's interested in that for the sake of "keeping customers happy" which amounts to only the most obsessed on either platform who will care about DF rundowns. The vast majority will either have a PS4 or an XBO and not care a wit for the performance of a multiplat game on the other platform they didn't buy. Those are the customers that devs/pubs most want to keep happy.

It is actually pretty easy to restrain performance. Just sleep in the main loop between frames.

 

[Moral Panic]

There's a lot of evidence from leaks. Plus if their ESRAM was dual bus they wouldn't be bragging about simultaneous reads/writes as if it's something they invented.




Are you seriously quoting the Leadbetter article as proof?

The Leadbetter article supports the information being presented now 100%. So MS haven't changed their tune. With no other evidence the only option is for me to believe it/them. It's up to you to bring some evidence as to how it is impossible rather than just calling bullshit.

 

[Rekkapryde]

PC:

Now that's what I'm talkin about.....

 

[Flatline]

The Leadbetter article supports the information being presented now 100%. So MS haven't changed their tune. With no other evidence the only option is for me to believe it/them. It's up to you to bring some evidence as to how it is impossible rather than just calling bullshit.

Of course it does, because Leadbetter got his info directly from Microsoft and it was proven in that thread. So you're practically "confirming" Microsoft PR using different Microsoft PR.

 

[3rdamention]

Doesn't this basically mean we know more about the Xbone than we do about the PS4? Interesting for everyone who said they'll never go into a specifics about specifications because they wont admit it's weaker.

We know exactly what Microsoft is telling us, so far..

 

[AgentP]

SRAM is generally dual ported (read and write). I'm not sure there are single ported SRAM cells (not my area). It would be pretty stupid to waste all those transistors on 2 ports and then only have one bus.

On what are you basing your single bus assumption?

They never say it is dual ported, in fact they say they "discovered" this new found ability that increases the BW. If this was something as simple as able to read and write over the bus at the same time, then is would 2x the compute bandwidth (2x109=218 not 204). I think they are playing games with numbers, some specific situation where they can have a higher effective bandwidth, something that will be rare but helps with PR (hence the DF leak).

But over all the details are nice, nothing we didn't know from the leaks, but at least final verification of what is inside. Knowing the final die size is nice, not quite the 400mm^2 we thought, but still huge.

I wonder what the 8GB flash is for, most cheap flash is slower than a HD (just lower seek times).

 

[Moral Panic]

Of course it does, because Leadbetter got his info directly from Microsoft and it was proven in that thread. So you're practically "confirming" Microsoft PR with different Microsoft PR.

So? MS says they used a certain technique to get extra throughput. You haven't addressed this point at all.

 

[la flama blanca]

Given your posts in any technical threads I'd amend that to:

"There's a lot SenjutsuSage doesn't know"

Starting with basic math.

SenjutsuSage is pretty well known for his BS.

 

[Flatline]

So? MS says they used a certain technique to get extra throughput. You haven't addressed this point at all.

Let me think, I'm either gonna believe that they used a certain magical technique that practically emulates the results of a dual bus ESRAM without it actually existing, or you know, I could believe that the laws of physics still apply. I'll go with the second.

 

[Moral Panic]

Let me think, I'm either gonna believe that they used a certain technique that practically emulates a dual bus ESRAM, or you know, I could believe that the laws of physics still apply. I'll go with the second.

The technique they describe doesn't break the laws of physics. Try again.

 

[Phire Phox]

Whaaaaaat the fuuuuuck??

That was a pic of Miley Cyrus performing at the Video Music Awards Last Night.

 

[AgentP]

The technique they describe doesn't break the laws of physics. Try again.

What exactly did they describe, some hand waving about magic holes?

 

[Chobel]

The technique they describe doesn't break the laws of physics. Try again.

What technique?

 

[Boogybro]

 

[Flatline]

The technique they describe doesn't break the laws of physics. Try again.

The only thing this "technique" achieves is provide Microsoft with highly theoretical numbers for PR. If these numbers were actually true they would break the laws of physics.

 

[SenjutsuSage]

Let me think, I'm either gonna believe that they used a certain magical technique that practically emulates the results of a dual bus ESRAM without it actually existing, or you know, I could believe that the laws of physics still apply. I'll go with the second.

So, ESRAM that's capable of reading and writing simultaneously simply doesn't exist?

 

[kasane]

9/10 everyone should be seen.

 

[ShapeGSX]

SRAM is generally dual ported (read and write). I'm not sure there are single ported SRAM cells (not my area). It would be pretty stupid to waste all those transistors on 2 ports and then only have one bus.

On what are you basing your single bus assumption?

Sorry, actually, the standard 6 transistor SRAM cell that would most likely be used in such a large cache is indeed single ported. I'm more used to the multi ported variety that are used in small designs.

 

[cchum]

.

Hahahhahahaha...I like you

 

[Hyunashi]

The technique they describe doesn't break the laws of physics. Try again.

I would love to know the technique they used.

 

[StevieP]

Guys c'mon - has there been a single console outside the Gamecube in recent memory that didn't use theoretical bullcrap numbers in PR releases?

 

[phosphor112]

I don't like those numbers. They don't make sense still.

128 x 853 = 109.1, not 204...

 

[KidBeta]

So, ESRAM that's capable of reading and writing simultaneously simply doesn't exist?

It does but its not something you discover after getting samples back, its something you build into the chip.

 

[BleachAndPepsi]

 

[SenjutsuSage]

I would love to know the technique they used.

That's kind of a straw man argument, don't you think? So because there is an inability to counter the evidence with what he rightfully points out is stated in that article, and now apparently supported by an official hot chips presentation from Microsoft on the Xbox One Silicon, you've resorted to piling on him to name or describe a technique that very few, or none, discussing this matter currently are capable of understanding just to shift the discussion away from the points he has made.

It does but its not something you discover after getting samples back, its something you build into the chip.

And I'm simply asking in response to that how can any of us possibly know for certain that Microsoft didn't build these capabilities into the chip in the first place? We have never known or understood with any sort of low level depth how Microsoft built or designed the various parts of the Xbox One, least of which the ESRAM. If their hot chips presentation is to be believed, and the information DF presented in that article about the X1's ESRAM really has been communicated to game developers (lying to developers you need to make games on your system is never good of course), then doesn't the prospects that Microsoft might actually be telling us the truth about the ESRAM occur in the slightest? It can't always be that they are lying to everybody. The main point is what do they gain at this point? The ESRAM bandwidth does nothing to change the well known advantages that the PS4 is already known to possess.

 

[George Claw M.D.]

A better pic of the GPU diagram from TechHive (from here):

Unfortunately it seems to be the only big slide they have online.

All dem swizzles. Did Snoop Dogg work on the Xbone?

 

[KidBeta]

That's kind of a straw man argument, don't you think? So because there is an inability to counter the evidence with what he rightfully points out is stated in that article, and now apparently supported by an official hot chips presentation from Microsoft on the Xbox One Silicon, you've resorted to piling on him to name or describe a technique that very few, or none, discussing this matter currently are capable of understanding just to shift the discussion away from the points he has made.

Its because aside from handy wavey new age bullshit, nothing in the article suggests any real evidence. Until you show us _HOW_ they achieved this, the evidence is near nill.

Also.

'extraordinary claims require extraordinary evidence'

they barely have any evidence let alone extraordinary evidence.

 

[Portugeezer]

Whaaaaaat the fuuuuuck??

I think it's Justin Bieber