• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

DigitalFoundry: X1 memory performance improved for production console/ESRAM 192 GB/s)

Status
Not open for further replies.

benny_a

extra source of jiggaflops
With their own number thrown in to make it work.
Can you be very explicit where the own number are being thrown in there?

I see a few variables, none of which are from outside the leaks or this article's source.

Humor me, think of it not as convincing me but of making a case to the big amount of people silently reading this that aren't sure about the situation.
 

charsace

Member
I see forwarding getting called a few times on beyond3D.
It sounds like coalesced read and writes.

What do you think of my reasoning? I suck with these types of calculations so I'm not claiming to be the end all be all right way to do it guy, but the way people are going about calculating the numbers based on the wording in the article just seems off. Maybe I'm wrong.

Can someone that understands the down clock math break it down for me?
 

Ding-Ding

Member
No I didn't.

I said it'll take at least 2.5TF GPU in a pc to do what the One's GPU is capable. A sentiment that I still stand by.

Aha, I remember you saying something about 2.5TF. Its difficult to make out sometimes as you flipflopped all over the place when proved wrong.
 
I wouldn't expect those awards to be for graphics.

TitanFall doesn't even really look good. In fact, I'd be willing to be that it will look noticeably better on PS4. Especially since it's coming later.

Forza looks great as always, but there wasn't anything that stood out about it graphically in terms of being next-gen.

Not to mention I don't know if there was even an 'approximate' XB1 machine on the floor. Weren't they using Windows 7 PCs with higher spec Nvidia cards?


Read lots on the web, that Forza 5 is at least one of the games which ran on native Xbox one. If Forza is not next gen then all of the PS4 games on the e3 show floor are not next gen. Was there a time someone mentioned Xbox is 6 months behind? Underpowered or not I will enjoy the games.
 
Its hilarious how the day after Cerny talks about XB1's eSRAM this article is put out and the math doesnt quite add up.
Cerny was talking about the benefits and drawbacks of using eDRAM in the PS4, not the eSRAM in the Xbone.

It will be interesting to see what actually happens a year from now, though. To my mind this doesn't change much, especially given the real world performance number of 133 gb/s, but who knows.
 
750MHZ isn't even so bad. People were talking about it losing 100+ MHZ a couple of weeks ago. I think the math adds up too well for it to be 800MHZ. I am no engineer though.
 

charsace

Member
What's your point? They're pretty much admitting that they're adding read/write to give us theoretical bullshit numbers which was the basic premise of my argument. Even if their PR is somehow true the downclocking is still real.
This is isn't PR though. This isn't a Press Release from MS. The author of the article says, "Well-placed development sources." How you get from that that the article is MS PR? You present no proof that the article is PR yet you call it that.
 

allan-bh

Member
This is a discussion forum. Its pretty much THEE place to talk about this stuff.

Yes, the problem is when people try to turn what they think in fact.

Somehow the thread was taken by a certainty that there was a 50MHz downclock through an extremely simplistic calculation.
 
Why is a specific number necessary to be rumored?

Do you understand the math that is being done here or do you think the values are chosen at random to make the Xbox One look bad?

I feel the posts about the downclock in this thread are very clear and specific with how they arrived at that conclusion. They are based on the numbers that seem to come from the source of the article and aren't the part that is editorialized.

Indeed. It's 100% more clear than how they just magically "found" an "88%" increase in performance... you can't.. just... multiply shit to hardware like that. Especially these random ass numbers.

That's why the 50mhz decrease makes sense.

It's not going to give a huge difference in performance.
~1.23
vs
~1.15

It's only a ~.77 drop.

It'll likely save them millions (if not more) and most devs won't be missing that very small amount of power.
 
This is isn't PR though. This isn't a Press Release from MS. The author of the article says, "Well-placed development sources." How you get from that that the article is MS PR? You present no proof that the article is PR yet you call it that.

Well, but it is available to the public and thus may be said to fall into the realm of Public Relations, which is what PR actually stands for.
 
Keep in mind that SenjutsuSage has been banned in the past for talking about his source. (Edit: Incorrect. See below.)


Yes and people seem to make a strong case. Do you have a counter interpretation that makes sense?

I wasn't banned for anything relating to my source. I was banned for getting a bit overzealous in defense of my opinion over the previous Xbox One's DRM policies. So banned for "spinning," not for anything else.
 

benny_a

extra source of jiggaflops
I feel that I've missed something, what is the calculation based on if not the figures aligning at the end?
This post explains it best:

Let's go through this logically, I'll lay it out so everyone can understand.

Previously MS engineers though read/write was only unidirectional, and the APU to ESRAM bandwitdth was pegged at 102GB/s, the reality was that it was bidirectional, regardless of what they thought which meant it was actually 204GB/s which lines up perfectly to 800MHz for the GPU.

Now we have information saying that MS engineers have discovered that information is bidirectional (not that it is now, just that they found out it is) and the consolidated read/write bandwidth is 192GB/s which is 96GB/s in each direction. That figure is lower than the old 102GB/s figure and it implies a GPU clock of 750MHz.

So yes, 192 is higher than 102, but it is not comparable as the latter is unidirectional bandwidth and the former is bidirectional. The fact that MS engineers didn't know or realise that you could run read/write operations simultaneously is irrelevant because it was still possible, this is not a new addition, more a new discovery. Think of it like a scientific discovery, just because an apple fell on Newton it doesn't mean he invented gravity, it existed before that, he just discovered it.

So we've actually gone from 204GB/s to 192GB/s or on the old measure, 102GB/s to 96GB/s, it's not that hard to understand. Leadbetter has this one wrong and he should try to correct it.

Just to explain where these variables are from:
102GB/s bandwidth was the old eSRAM figure cited everywhere before for Xbox One.
192GB/s is the new figure from the source in the article this thread is about.
Because 192GB/s is now bidirectional and 102GB/s was unidirectional you divide it by 2 and end up with 96GB/s.

This is not the only interpretation but the one that seems most reasonable to me.

The unused cycles and unfilled holes and such don't have any place in a theoretical peak discussion so they seem like misdirection.
 
Lemme get this straight. Yesterday , Cerny talks very precisely about the high system bandwidth of the PS4, then out of nowhere Microsoft responds the next day by saying their bandwidth is somehow like 80 percent bigger that they expected and thus similar to ps4. They also claim that us gamers should not worry about specs anyway... And that this is no longer about bits, even though they talked about transistors at their reveal...

And some people believe them... scary stuff.

It's all part of the game. MS won't go down without a fight.
 

Jomjom

Banned
Man, that blows.

So even if the RAM in the Xbone had the same amount of bandwidth that the RAM in the PS4 does, it would STILL have a 3 gig deficit?

Yeah but that 3 gigs is what will allow the instant switching and the SNAP features and whatnot.
 

allan-bh

Member
I find hard to believe Microsoft engineers didn't fully understand their own design and suddenly increased performance just 'happened'

In fact always have been there, not suddenly happened.


I think is plausible, before it was just theory, now they have seen in practice.
 
This is isn't PR though. This isn't a Press Release from MS. The author of the article says, "Well-placed development sources." How you get from that that the article is MS PR? You present no proof that the article is PR yet you call it that.

The press release doesn't say it is still at 800 Mhz.
 

Freki

Member
Do either of you mind explaining how anyone got the down clock number from reading the article? I personally don't get that from what I've read.

102GB/s was the embedded memory peak bandwith up until today.

800mhz*128byte * 1 (read or write operation) =102.4GB/s

800mhz*128byte * 2 (1 read and 1 write operation) =204.8GB/s

DF stating the new peak memory bandwith beeing 192GB/s while maintainig 800mhz clockspeed just doesn't add up...

I personally stop here and leave it at that - the numbers don't add up.



Some people are taking this a step further though:

750mhz*128byte * 2 (1 read and 1 write operation) =192GB/s

this means that a 50mhz downclock would alling with the peak memory performances given in the article.
 
Is it true that 3 gigs of memory are reserved for the OS? Or was that just a rumor?

We know this so far which I posted earlier:

https://www.gameinformer.com/b/feat...-ones-chief-product-officer-marc-whitten.aspx

When asked a direct question if the OS took 3GB of RAM, well, you can see the answer. It RAM issue starts at the 9:44 mark. When he is asked Whitten just nods and says he isn't going to talk about Sony - which is mentioned in the question - and then he says we think the way MS balances the system is key. He acknowledges the 5GB without doubt. Then of course he says they will have the best looking games in the living room.

Then there is the other article I listed:

http://kotaku.com/the-five-possible-states-of-xbox-one-games-are-strangel-509597078
 

benny_a

extra source of jiggaflops
I wasn't banned for anything relating to my source. I was banned for getting a bit overzealous in defense of my opinion over the previous Xbox One's DRM policies. So banned for "spinning," not for anything else.
Yeah sorry for that. I've read the part where you were banned again and I felt you stated your position reasonably while being taunted by a specific poster.

Man, that blows.

So even if the RAM in the Xbone had the same amount of bandwidth that the RAM in the PS4 does, it would STILL have a 3 gig deficit?
It's not yet known how much the PS4 will reserve.

There is some conjecture out there that it won't be that much, because the system was designed a long time with 4GB in mind, so reserving 3GB would be absurd but it's speculation at this point.
 

Espada

Member
Stealth_Cobra said:
Lemme get this straight. Yesterday , Cerny talks very precisely about the high system bandwidth of the PS4, then out of nowhere Microsoft responds the next day by saying their bandwidth is somehow like 80 percent bigger that they expected and thus similar to ps4. They also claim that us gamers should not worry about specs anyway... And that this is no longer about bits, even though they talked about transistors at their reveal...

And some people believe them... scary stuff.

People actually believed that Power of the Cell secret sauce nonsense back in 2007. There's no limit to how deluded people will get in order to maintain their image of whatever side they're on.
 
People act like the two GPUs are light years apart, which cracks me up. In reality they are very similar. That "50% power" difference is funny too. At 1080p you simply aren't going to see much of a difference.

In reality you get a tiny FPS (frames per second) boost that may not even matter and similar very high settings vs possible high settings (using PC "wording" there). We'll probably have 3rd party games where the X1 runs a game at 68fps and the PS4 runs it at 85fps...both would be locked to 60fps on both. They're so similar I doubt you'll see any worthy games sub-30fps.

Granted, sub-30fps seemed to be a norm for multiple PS4 games at E3.

Actually, it's pretty easy to see how in a closed box the differences will be very noticeable in the long run when those same differences already are pretty noticeable in a platform like the PC.

See, it's like... trying to convince other people that there isn't much of a difference between a car hitting 180 mph, and another who is only able to hit 125 mph. It's even worse when you realize that not only can the other car hit 180 MPH, the track he can drive in is bigger (Available ram).

In a closed box system a gap bigger than the PS3 - Xbox 360, but much more easily accessible, will show itself sooner or later whether you like it or not.
 

Acheteedo

Member
This post explains it best:

Just to explain where these variables are from:
102GB/s bandwidth was the old eSRAM figure cited everywhere before for Xbox One.
192GB/s is the new figure from the source in the article this thread is about.
Because 192GB/s is now bidirectional and 102GB/s was unidirectional you divide it by 2 and end up with 96GB/s.

This is not the only interpretation but the one that seems most reasonable to me.

The unused cycles and unfilled holes and such don't have any place in a theoretical peak discussion so they seem like misdirection.

Thanks for the explanation, makes more sense now. I think Eurogamer has earned a lot of credit with rumours in the run up to this new generation, this would mark a colossal departure from their record to date, no?
 

SRTtoZ

Member
Cerny was talking about the benefits and drawbacks of using eDRAM in the PS4, not the eSRAM in the Xbone.

It will be interesting to see what actually happens a year from now, though. To my mind this doesn't change much, especially given the real world performance number of 133 gb/s, but who knows.

What I meant was he was comparing the complexity of both solutions and the fact that 3rd party dev teams asked for unified memory. I should have worded it better.
 
Unless the GPU has been upgraded? Maby Reiko's 64MB of ESRAM/12GB RAM/2TFGPU is true?

I know, i'm in wonderland right now. I'll tell alice you guys said hi.
 

Dlacy13g

Member
Man, that blows.

So even if the RAM in the Xbone had the same amount of bandwidth that the RAM in the PS4 does, it would STILL have a 3 gig deficit?


Actually it would be a 2 gig deficit if we are all on the same page of the PS4 reserving 1GB for OS. People seem to magically forget when doing a comparison that the PS4 will need to pull off some of the 8GB for their OS.
 
102GB/s was the embedded memory peak bandwith up until today.

800mhz*128byte * 1 (read or write operation) =102.4GB/s

800mhz*128byte * 2 (1 read and 1 write operation) =204.8GB/s

DF stating the new peak memory bandwith beeing 192GB/s while maintainig 800mhz clockspeed just doesn't add up...

I personally stop here and leave it at that - the numbers don't add up.



Some people are taking this a step further though:

750mhz*128byte * 2 (1 read and 1 write operation) =192GB/s

this means that a 50mhz downclock would alling with the peak memory performances given in the article.

This should be in the op
 

kaching

"GAF's biggest wanker"
But I guess you know how the ESram works in Xbox One.
Given that the article has to air-quote such a technical term as "holes", does anyone, even amongst MS engineers, actually know how XBO's Esram works?

What's wrong with people trying to fill in the blanks? If MS doesn't want to be forthright and transparent about it, speculation and best guesses are what naturally ensue. It's no worse than taking what this article says at face value as a significant improvement like several people are.
 
What do you think of my reasoning? I suck with these types of calculations so I'm not claiming to be the end all be all right way to do it guy, but the way people are going about calculating the numbers based on the wording in the article just seems off. Maybe I'm wrong.

Can someone that understands the down clock math break it down for me?

oke when microsoft said the had more then 200 GB/s internal bandwidth everybody took up their pitch fork and said creative math and stuff.

Now people are using creative math to conclude that there is an 50mhz downclock.
It depend how the hardware works if it does like they think it does the math is right and microsoft engineer never lied at the hardware panel.

In the DF article it is mentioned there is no downclock but something about improved drivers and being sure that final silicon works. Microsoft told the devs that they esram now has 192GB/s of theoretical bandwidth but only practical usage right now was something about doing blending actions and they got 133GB/s out of the esram.

but the math was
204GB/s for a read and write so 102GB
now its 192 GB for read and write so someone concluded 96GB/s
96GB/s divided by the 128 gave an 750mhz clock count.

They creative math is there but the trues are conflicting. It is assuming stuff and saying its true which and what they assume goes against the rumors in the article no downclock just some extra bandwidth they didn't expect or weren't sure if it would make it in so they said to developers you only have 102GB/s. Time will tell specs aren't fixed until console are in the tech journos hand. It on both the specs sheets of the x1 and p4
Hell it could be even clocked like thuway said 100~300 mhz

"Subject to change"
 

charsace

Member
Well, but it is available to the public and thus may be said to fall into the realm of Public Relations, which is what PR actually stands for.
It is PR. People are acting like the author got the news straight from MS. I don't understand how the people in this thread are certain that MS gave the author this info? There are many sources outside of MS that can have this info.
This post explains it best:



Just to explain where these variables are from:
102GB/s bandwidth was the old eSRAM figure cited everywhere before for Xbox One.
192GB/s is the new figure from the source in the article this thread is about.
Because 192GB/s is now bidirectional and 102GB/s was unidirectional you divide it by 2 and end up with 96GB/s.

This is not the only interpretation but the one that seems most reasonable to me.

The unused cycles and unfilled holes and such don't have any place in a theoretical peak discussion so they seem like misdirection.
This doesn't deal with the problem that is laid out in the article at all. "Spare," is a key word in the explanation of how they are getting more out of the esram. "Spare," doesn't equate to something always happening.
 

benny_a

extra source of jiggaflops
Thanks for the explanation, makes more sense now. I think Eurogamer has earned a lot of credit with rumours in the run up to this new generation, this would mark a colossal departure from their record to date, no?
Based on the comments in the article and on here, the issue is not with Eurogamer or DigitalFoundry but with the specific author of this article.

This doesn't deal with the problem that is laid out in the article at all. "Spare," is a key word in the explanation of how they are getting more out of the esram. "Spare," doesn't equate to something always happening.
Wouldn't theoretical peak performance include what is possible in the best possible scenario? Which with 800MHz clock would mean 204GB/s.

Technically 2, since PS4 OS takes up 1.
Where is that number of the PS4 OS from?
 
Status
Not open for further replies.
Top Bottom