Predict: The Next Generation Console Tech

[pipo]

At the same time though I don't see anyone remaining with a platform and platform holder on the sole basis of a gamer score.

<snip>

I personally see many being happy to switch consoles if given enough reason to. I certainly am, and that's considering I've never owned an xbox this gen and have never really seen any reason to.

That sums it up nicely. Gamerscore and the achievements involved are what drives a lot of these guys. Going back again and again to squeeze every point out of a game and comparing these with others.

High score table evolved. I don't really care, as my personal 4350 points make painfully clear, but I must say MS did a good job on that.

 

[Brad Grenz]

I suppose if you're one of the handful in the race to have the highest gamerscore, you probably are to damaged in the brain to ever give that up.

 

[upnorthsox]

Yep, I get that distinct feeling too. Like he's just repeating the rumors. He hadnt said a single thing that really proves anything or is new info. Either he's repeating BG's info from reading it online (possibly even this thread ), or BG's info is pretty accurate, one of the two.

EXCEPT I guess, that apparently he maintains it's Intel CPU.



Too be fair, DD4 probably is not in the kit at all, since it barely exists if it exists. Even if they plan DDR4 for the final, it's likely not in the kit.

I think if DaE does have a kit, he might be a lot more into the programming nuts and bolts and a lot less into whats under the hood. It just doesn't concern him that much. That's the feeling I get. Also, if the kit is ill-gotten somehow, he may have incomplete or missing documentation, and according to BG MS is pulling a Wii U and not even including GPU specs in the target specs, so thats another aspect that could be affecting his knowledge.

He said he was going to post kit related pics on his twitter if I understood correctly, and didn't. Another opportunity to verify, that he hasn't come through on.

Could be that the final SOC will have edram and the only/best way to emulate that is with an intel chip.

 

[itsmydamnation]

Could be that the final SOC will have edram and thUe only/best way to emulate that is with an intel chip.

given that the intel l3 is inclusive and amds is exclusive while having latency closer to edram then Intels l3 bulldozer would make more sense then sb/ib intel chip.

Posted from iPhone

 

[upnorthsox]

given that the intel l3 is inclusive and amds is exclusive while having latency closer to edram then Intels l3 bulldozer would make more sense then sb/ib intel chip.

Posted from iPhone

Does bulldozer have L3 edram? How much?

 

[SKYSONY]

According to BG MS is pulling a Wii U and not even including GPU specs in the target specs, so thats another aspect that could be affecting his knowledge.

Can you elaborate on this, please? I guess I miss it.

 

[tunafish]

Does bulldozer have L3 edram? How much?

No, but it does have 8MB of SRAM L3. Of course, any realistic console would probably have less.

 

[ultragpu]

Tell him to run Dxdiag and have him post the processor information, the total system memory, the display device name, and the total amount of video memory.

I am under the assumption that there is a windows environment on the dev kit.

If he doesn't know what Dxdiag is then he is not a developer. He would be lying if he said so.

Hey aren't you the guy who claimed PS4's spec was a beast on gaf? Care to share something new? We have been patient you know.

 

[SKYSONY]

Hey aren't you the guy who claimed PS4's spec was a beast on gaf? Care to share something new? We have been patient you know.

Proelite opened this thred in GAf a while ago, it doesn´t seem he has insider info.

http://neogaf.net/forum/showthread.php?t=478941

 

[itsmydamnation]

Does bulldozer have L3 edram? How much?

Now that im not on a an iphone......

Edram isn't attractive because of its performance, its because of its performance to cell size. That is why companies like IBM are using it as an L3/4 cache, not because its fast compared to other caches but because you can have a big chuck of it for far less transistors then sram while it is still way way faster then main memory.

SB/IB have a very fast L3 by design. it has a big ring bus between its segments and its the first cache of any real size, the L2 of SB/IB is fast but being 256K a core for 2 threads is pretty small. Then there is the fact that the L1/L2 is inclusive in the L3, this is for maintaining low memory latency while maintain data coherency across a multicore chip. That in itself is very different to what you would use a block of edram for which would be some kind of eviction cache or scratch pad.

Bulldozer on the other hand has a very different cache design, in effect AMD's L2's acts like intels L3's. the L1 is inclusive in the L2 and the L3 is exclusive, it is an eviction cache. On top of this AMD's L3 are much slower then Intels because it doesn't need to be its job is for throughput across many threads not minimizing latency, its why trinity doesn't have L3 its largely useless for general consumer workloads because each core already has 2Mb of fast L2.

So when looking at which SOC has a design that fits what you would use edram for AMD's cache design is way closer then Intels while still being miles away anyway .

 

[aaronspink]

Edram isn't attractive because of its performance, its because of its performance to cell size. That is why companies like IBM are using it as an L3/4 cache, not because its fast compared to other caches but because you can have a big chuck of it for far less transistors then sram while it is still way way faster then main memory.

Actually there is a performance cross over point where edram better performance than sram primarily due to decreased wordline and dataline lengths.

SB/IB have a very fast L3 by design. it has a big ring bus between its segments and its the first cache of any real size, the L2 of SB/IB is fast but being 256K a core for 2 threads is pretty small. Then there is the fact that the L1/L2 is inclusive in the L3, this is for maintaining low memory latency while maintain data coherency across a multicore chip. That in itself is very different to what you would use a block of edram for which would be some kind of eviction cache or scratch pad.

Once of the primary benefits of an inclusive L3 is that it can effectively act as a snoop filter for all the L2 with an additional couple of bits per tag.

 

[XpiderMX]

Proelite opened this thred in GAf a while ago, it doesn´t seem he has insider info.

http://neogaf.net/forum/showthread.php?t=478941

Durango 1-1.2TFlops total processing and Orbis 1.8 TFlops only the GPU?

1.0TF/4GB vs +1.8TFlops/2GB?

It looks weird...

 

[Rangers]

Can you elaborate on this, please? I guess I miss it.

He just dropped the comment a couple times on GAF i Think.

It's something I had speculated before he even said it. Seems to me one reason we cant get good info on the Wii U GPU/CPU is the official target specs just dont include it. I speculated it would be best for Sony and MS to follow suit, it seems to have worked spectacularly well for Nintendo. This way only a few people working directly to the hardware likely know the capabilities, and it's easier to prevent leaks.

The PS4 target specs (the version that is on VGleaks) actually contain GPU clocks, SP's, etc. Since we only hear shadowy things like "around 1 teraflop" for Durango, I wondered if they are pulling a Nintendo.

 

[ERP]

This way only a few people working directly to the hardware likely know the capabilities

Yeah because developers never talk to each other....
I doubt there are any more dev's with info about PS4 or Durango than their are Nintendo devs that know about WiiU, in fact I'd bet that there are more of the latter currently.
Most of the leaks I've seen over the years don't come from technical people anyway, from the inside I could often see what started a particular rumor, but it was often misunderstood by whoever started the rumor.

 

[itsmydamnation]

Actually there is a performance cross over point where edram better performance than sram primarily due to decreased wordline and dataline lengths.

Interesting what ballpark size are we talking 16/24/32mb or even larger?

Once of the primary benefits of an inclusive L3 is that it can effectively act as a snoop filter for all the L2 with an additional couple of bits per tag.

yep, its another one of those interesting points about bulldozer. that L2 probe and data transfer has a real latency cost when your pulling data from another modules L2. Its funny the more of the intricacies we find out about bulldozer the more im convinced that the technology in it is very, very good, they just got the target completely wrong.

 

[bgassassin]

Can you elaborate on this, please? I guess I miss it.

My last post a couple of posts before one of yours.

 

[Rangers]

Yeah because developers never talk to each other....
I doubt there are any more dev's with info about PS4 or Durango than their are Nintendo devs that know about WiiU, in fact I'd bet that there are more of the latter currently.
Most of the leaks I've seen over the years don't come from technical people anyway, from the inside I could often see what started a particular rumor, but it was often misunderstood by whoever started the rumor.

maybe it's literally as simple as, if they're not written down, they dont spread as easily. especially when it becomes "i heard from this guy" with no real way to verify and nothing to keep the info from being distorted or discern real from fake other than nebulous word of mouth.

And your comment about leaks from non-technical people just reenforces the point why not printing specs internally might be a good idea.

 

[ERP]

Not to be flippant, but I suspect the reason there are so few leaks from technical people is that for the most part they are better compensated and have more to lose.
Obviously the less people that know the better in terms of keeping a secret, but I'd bet we've already heard a lot about WiiU's CPU, but it's drowned out in the noise. I know that the CPU speed has been posted accurately at various locations. Most developers aren't going to care what it's derived from, just how it performs.

 

[Rangers]

Not to be flippant, but I suspect the reason there are so few leaks from technical people is that for the most part they are better compensated and have more to lose.
Obviously the less people that know the better in terms of keeping a secret, but I'd bet we've already heard a lot about WiiU's CPU, but it's drowned out in the noise. I know that the CPU speed has been posted accurately at various locations. Most developers aren't going to care what it's derived from, just how it performs.

1.6 ghz is my guess Because, Nintendo.

But yeah, thats what I mean about not printing the specs. Without a paper reference somewhere in the chain, even a secret one, it becomes exponentially harder to discern the wheat from the chaff in the wild, is my nascent theory.

 

[Rangers]

2013 is a go, repeat, 2013 is a go

Microsoft unintentionally hints at new Xbox console again after stating that recent 720 comments were misunderstood


A job vacancy for a senior role at Microsoft's entertainment division has suggested that the next generation Xbox will be released before 2014.
The advert, seeking a Central Marketing Group director, claims that "over the next eighteen months Microsoft will release new versions of all of our most significant products including Windows (Client, Server, Phone and Azure), Office and Xbox, along with completely new offerings like Microsoft Surface".

The statement suggests that Microsoft will launch the next Xbox before 2014. CVG sources have previously suggested the system's release date will likely fall in the latter half of 2013. Development kits were issued to several third party studios in the UK at the start of 2012, the sources have said.