Predict: The Next Generation Console Tech

[Docwiz]

What happened to your post that listed these rumors?

I actually do believe them..

Heres why ...

Those rumors talk about a radical new processor design that is backed by atleast 9 patents.

The guy on all those patents, Eric Mejdrich, started out at IBM then moved to MS 2 yrs ago. He's the architect for the chip design of the next xbox.

Other IBM distinguished engineers, some also on those patents, are now also at MS in the xbox devision.

MS also has some patents that leverage the same chip design, basically carrying on research with those chip designs.

The idea behind the processor is that its like stem cells. The basic unit is called a "Basic throughput engine (BTE)". It has a very simple scalable design that can be programmed to be specialized.

So when its manufactured you only need to FAB 1 design and test that 1 design. BUT when you assemble it on a SOC it can specialize to be several different components on that SoC.

This BTE can become a specialized in Physics processing, AI, Sound, Search, RayTracing, Animation etc. etc.

This BTE can also become something that specializes in processing Vectors, which the patents call a VTE, vector throughput engine.

I reiterate the fact that this one building block, BTE, can become specialzed to be so many things. And it only needs to be tested as a BTE.


This BTE needs huge amounts of memory, L1/L2 and RAM , to do its work. Traditionally what used to be done in software can now all stay on the hardware. So a games engine that needed to store a list of objects in a scene in RAM and managed by software, can do all that in the hardware and never leave the HW. It can just push these objects around the SoC via the huge pipes and processed thru the BTE's .

Another brilliant design around these BTE's is that they understand richer types of data structures, we can store rich objects, even trees of objects. All we need to do that is large amounts of memory and plumbing to allow the movement of these big chunks of data.

Again this all stays on the hardware, never leaving.

As a developer this is pretty amazing whats described in those several patents...

And 3 of those patents explicitly mention how it will achieve RayTracing, and the science/maths is there ... The HOW is explicitly explained, I do believe it is going to arrive in Xbox next..

Nothing out in the market comes close to what these new processors can do...

The CELL from IBM was a great processor by design, but in reality it was crazy hard to programme for. It did not have a great design around caching and object/object trees.

This design from IBM/Microsoft/AMD fixes all those failings of the CELL..

The magic around the new xbox, will be around this new processor architecture PLUS the software to make use of it. Basically MS could remove something that has been the pain point of game development for the last 2 decades, memory pointers. This chip will be shared memory across the entire SoC, all the components can share memory and pass the same rich object/object tree between them.

It can work on all objects within a scene per frame all on hardware never leaving.

If MS truly pulls this off, they will have a brilliant chip design for ALL there devices NOT JUST XBOX.

This chip will make its way into its phones, surface tablets, servers, desktops..

.... Its high stakes for MS .. they need there own chip for all the devices they plan on making in the future, its the only way they can truly own the entire experience from CHIP to OS, they never have to rely on dodgy chips from OEM's that take forever to release updated drivers that ultimately make there OS run crap! ..

MS wants to do chips ... trust me on that ...

Well, I do agree with something that nobody here has given us.....

Nobody has explained why we have so many engineers from AMD, IBM, and Microsoft and they are only producing a console with a GPU of 1.2 TF, that has made no sense to me at all.

Think about this for a minute. Why would you have all of those people working on a console that is top secret to only produce a GPU of 1.2 TF and gimped speed of only using DDR3 memory.

I am not buying those rumors at all, because logically there has to be more too it. Maybe people are getting confused from developer kits and not the final hardware.

Think rationally and logically for a minute, why do we have all of these people and we know these people have plenty of patents and we know they are huge teams and we had AMD saying a couple of years ago that we would have "avatar" type graphics with Durango and hype or not they only said it about Durango and nobody else. Why do they have huge teams working on all of this if it's just going to be an average console?

I don't think that the final hardware is anything like the devkits.

 

[Shifty Geezer]

It all started in 2005/6 , they've been working on this already for many many years! Only in the last year have they needed to go into mass production..

Googling 'IBM "Basic Throughput Engine"' throws up only two results - this thread, and this patent. A cursory glance at the patent suggest "basic throughput engine" (not capitalised, so not the proper name of a component unlike SPU and PPU) is patent-lawyer speak for 'processor' (patent speak tries to be as non-committal and all-encompassing as possible). The claim is only for a raytracing architecture with nothing about adaptive processors as you seemed to describe. And a search of the IBM website for BTE throws up 3 responses, none relating to anything substantial (typo, random message reference number, and company name), let alone an amazing, secret adaptive processor technology.

I think you are very much mistaken about the tech and what you believe it is doesn't exist.

 

[Carl B]

...Orbis FPGA-driven programmable I/O bus...

You heard it here first folks!

 

[Shifty Geezer]

"LOL @ rest of internet" is not a valid contribution to this thread. Bring in new rumours, sure, but "Look what this crazy guy has posted" is not wanted here. This is a proper thread in the Technical Forum.

 

[Bagel seed]

A telling comment from aegies?

no, the problem is that people are comparing closed system performance with pc system performance. PC GPUs operate around 60%.

1.2TF / 0.6 = 2TF implied?

How accurate is that 60% quote anyways.

 

[Carl B]

...This is a proper thread...

Let's not get ahead of ourselves now.

 

[itsmydamnation]

A telling comment from aegies?



1.2TF / 0.6 = 2TF implied?

How accurate is that 60% quote anyways.

not at all, it is a vast over simplification, but the point he is trying to make is correct.

 

[Shifty Geezer]

Let's not get ahead of ourselves now.

The rate of noise attraction is getting to the point where the management effort doesn't seem justified. Maybe I should just abandon it and go read some tech website for real info about the new consoles and let everyone post their "LOL, 1.2 TFlops, Internet's is crazy, hey, listen to my impossible dream machine based on a technology I don't understand but think I do, here's what we've been talking about on GAF and we'll repeat ourselves here" 'conversation'. It's not like there's anywhere else on the internet for that sort of talk, so maybe I should be more welcoming?

 

[babybumb]

A telling comment from aegies?



1.2TF / 0.6 = 2TF implied?

How accurate is that 60% quote anyways.

Honestly if MS is only going to get 1.2TF fitted by AMD what special sauce does Sony use? Does MS not have access to Sony APU tech?

 

[Carl B]

A telling comment from aegies?

All Flops discussion should only be in theoretical maximums when it comes to reverse engineering these things, as that's the only thing that's quantifiable. In that context, the 1.2T number would stand. If you would want to equate a 1.2T chip in a closed environment with a 2T chip in the broader PC ecosystem, that's probably a different thread topic altogether.

 

[Proelite]

A telling comment from aegies?



1.2TF / 0.6 = 2TF implied?

How accurate is that 60% quote anyways.

I think he lifted that part up from a post made here on b3d?

 

[itsmydamnation]

if it counts for anything, i vote use the ban hammer Shifty

 

[Docwiz]

Googling 'IBM "Basic Throughput Engine"' throws up only two results - this thread, and this patent. A cursory glance at the patent suggest "basic throughput engine" (not capitalised, so not the proper name of a component unlike SPU and PPU) is patent-lawyer speak for 'processor' (patent speak tries to be as non-committal and all-encompassing as possible). The claim is only for a raytracing architecture with nothing about adaptive processors as you seemed to describe. And a search of the IBM website for BTE throws up 3 responses, none relating to anything substantial (typo, random message reference number, and company name), let alone an amazing, secret adaptive processor technology.

I think you are very much mistaken about the tech and what you believe it is doesn't exist.

There are some patents for Eric's own Ray tracing and also there is a Vector Ray Tracing Engine made by IBM.

No matter the rumors though, it does not make sense to me that the Next Xbox is some average 1.2 TF GPU with gimped memory speed.

Why on earth would so much money be spent and so many man hours be spent on something that is so AVERAGE?

I think that people are getting confused over the Development kits and the final hardware and that explains all of the confusion of specs and it also explains why nobody has the correct answers on this.

It's not that I don't want to believe it. Because we can all make the case that maybe Microsoft is building something cheap to put some new exotic controller in the box or something.

However, all the money being spent and the amount of people working on it from three different companies AMD, IBM, and Microsoft are all working together on this and with huge teams of people and people that have specialized patents just don't add up with what is out there currently. There has to be more to it than this.

 

[Bagel seed]

not at all, it is a vast over simplification, but the point he is trying to make is correct.

Well then it's a point that goes without saying, since PS2 days.

The rate of noise attraction is getting to the point where the management effort doesn't seem justified.

I think we're doing a good job here despite the occasional noise... we just need to stop wording opinions or rumors like facts.

I think he lifted that part up from a post made here on b3d?

Right. His signal to noise ratio is way too low. All we heard from him is monster chip and nothing really quantifiable.

 

[Carl B]

The rate of noise attraction is getting to the point where the management effort doesn't seem justified. Maybe I should just abandon it and go read some tech website for real info about the new consoles and let everyone post their "LOL, 1.2 TFlops, Internet's is crazy, hey, listen to my impossible dream machine based on a technology I don't understand but think I do, here's what we've been talking about on GAF and we'll repeat ourselves here" 'conversation'. It's not like there's anywhere else on the internet for that sort of talk, so maybe I should be more welcoming?

The noise attraction is definitely enormous right now, and the incestuous nature of Internet board discussions is not helping things; items brought up here in other threads as pure mental exercise get taken, regurgitated, expanded upon, and show up back here two days later warped and presented as fact by different posters. Said tech websites won't be any better right now though, as they're drawing their info from the same (non)sources.

None of that is to say give up and stop swinging the hammer.

 

[bgassassin]

Now, I remember there was that Serenox poster that claimed a 450mm APU. That's the only number I remember, but he also claimed stuff like an NVIDIA GPU and that both Sony and MS have basically the same hardware. I don't believe there's any validity to his rumor.

Thanks for the background.

That's the impression I get. Recently lherre said something along the lines of "both (ps4/720) are powerful". At a basic level it implies some equivalency, and I havent heard anything different from lherre.

I dont see why the Durango specs as rumored call for a large die anyway. It's basically a 7770+8 Jag cores+any alleged custom hardware. Total die size should be quite svelte. Which is why although I've doubted the APU stuff mostly, I'm starting to become amenable to it. Dont know if I'd bet on it, but I dont think it's crazy, either.

In fact to me if the Jag cores are as tiny as people say, it probably almost doesnt make sense to not glue them onto the GPU and make manufacturing easier theoretically. But I wonder where the ESRAM fits into it all considering they still havent integrated EDRAM in 360. Probably on a separate die on the same package, just like the current 360 setup...

Yeah, I wanted to bring that piece of info back up because I think too much focus was being placed on the GPU. If MS is getting debatable performance compared to PS4, then there should be more focus on the total path MS is pursuing to get there and not a part of it.

Well, I do agree with something that nobody here has given us.....

Nobody has explained why we have so many engineers from AMD, IBM, and Microsoft and they are only producing a console with a GPU of 1.2 TF, that has made no sense to me at all.

Think about this for a minute. Why would you have all of those people working on a console that is top secret to only produce a GPU of 1.2 TF and gimped speed of only using DDR3 memory.

I am not buying those rumors at all, because logically there has to be more too it. Maybe people are getting confused from developer kits and not the final hardware.

Think rationally and logically for a minute, why do we have all of these people and we know these people have plenty of patents and we know they are huge teams and we had AMD saying a couple of years ago that we would have "avatar" type graphics with Durango and hype or not they only said it about Durango and nobody else. Why do they have huge teams working on all of this if it's just going to be an average console?

I don't think that the final hardware is anything like the devkits.

Replace Microsoft with Nintendo and the same argument could be made about Wii U since it's development started in '09. They probably need all those people to build a "large", stable SOC. There's more to it than how the GPU's power is influenced. At the same time what kind of difference would be expecting between the dev kit(s) and final hardware other than certain components not being ready for the PC kit? It's likely even with said components missing they were simulated in some fashion.

A telling comment from aegies?



1.2TF / 0.6 = 2TF implied?

How accurate is that 60% quote anyways.

Here's a thread I made awhile back dealing with that subject seeking the answer.

http://forum.beyond3d.com/showthread.php?t=62135

 

[MasaC]

Why on earth would so much money be spent and so many man hours be spent on something that is so AVERAGE?

To hold the production costs down but still get a lot of bang for the buck and release at reasonable $300-$400?

 

[liquidboy]

Googling 'IBM "Basic Throughput Engine"' throws up only two results - this thread, and this patent. A cursory glance at the patent suggest "basic throughput engine" (not capitalised, so not the proper name of a component unlike SPU and PPU) is patent-lawyer speak for 'processor' (patent speak tries to be as non-committal and all-encompassing as possible). The claim is only for a raytracing architecture with nothing about adaptive processors as you seemed to describe. And a search of the IBM website for BTE throws up 3 responses, none relating to anything substantial (typo, random message reference number, and company name), let alone an amazing, secret adaptive processor technology.

I think you are very much mistaken about the tech and what you believe it is doesn't exist.

Parallelized Ray Tracing
==================
http://www.google.com/patents/US20110285710

Ray tracing image processing system
===========================
http://www.google.com/patents/US8139060

Full Vector Width Cross Product Using Recirculation for Area Optimization
======================================================
http://www.google.com/patents/US200...#v=onepage&q=vector throughput engine&f=false


Vector Permute and Vector Register File Write Mask Instruction Variant State ...
=====================================================
http://www.google.com/patents/US200...#v=onepage&q=vector throughput engine&f=false


Two-Tiered Dynamic Load Balancing Using Sets of Distributed Thread Pools
=======================================================
http://www.google.com/patents/US200...PgD&ved=0CEIQ6AEwBTgK#v=onepage&q=104&f=false


Using ray tracing for real time audio synthesis
===================================
http://www.google.com/patents?id=26oJAgAAEBAJ&printsec=abstract&zoom=4#v=onepage&q&f=false


Utilizing Ray Tracing for Enhanced Artificial Intelligence Path-Finding
==================================================
http://www.google.com/patents?id=58utAAAAEBAJ&printsec=abstract&zoom=4#v=onepage&q&f=false




I'll link the other patents when I get the chance...

I'm not lying when I say that these BTE's, as they are referenced in the patents, are multi-perpose cores and can become specialized to do a plethora of things...

 

[Carl B]

Why on earth would so much money be spent and so many man hours be spent on something that is so AVERAGE?

I think you need to evaluate the target result of all those man-hours and millions spent though: a profitable console. Performance is being maximized insofar as it allows reductions in budget elsewhere (whether transistor or otherwise), while maintaining a graphically competitive offering.

 

[Docwiz]

Thanks for the background.



Yeah, I wanted to bring that piece of info back up because I think too much focus was being placed on the GPU. If MS is getting debatable performance compared to PS4, then there should be more focus on the total path MS is pursuing to get there and not a part of it.



Replace Microsoft with Nintendo and the same argument could be made about Wii U since it's development started in '09 least two years. The probably need all those people to build a "large", stable SOC. There's more to it than how the GPU's power is influenced. At the same time what kind of difference would be expecting between the dev kit(s) and final hardware other than certain components not being ready for the PC kit? It's likely even with said components missing they were simulated in some fashion.

yeah, I don't think the same argument can be made comparing Nintendo with Microsoft because the same claims were not met. Nice try though.

I don't think that Durango is just going to be so cheap and under powered like people think. It has to be more than that because there is a lot of money and talent and so many people building something that can be reused. You don't use that kind of money, talent, and personnel to make something cheap and under powered and just average.

This doesn't come from the land of wishes, it comes from common sense.

If they have a 1.2 TF GPU, it won't be long until tablets and other mobile devices catch up, not buying it. That maybe the cheaper GPU in the system, but there has to be something more here.