AMD/ATI for Xbox Next?

[lefizz]

But if you go with a cpu with the same architecture and i guess its about 90% odds on to be AMD gpu then the BC should be pretty trivial to do.

As far as i can see Intel and Nvidia are out of the picture due to issue with the first xbox that is unless Intel is basiclly giving away Larrabee to try and start the ball moving for the PC market..

 

[Dr. Nick]

But if you go with a cpu with the same architecture and i guess its about 90% odds on to be AMD gpu then the BC should be pretty trivial to do.

As far as i can see Intel and Nvidia are out of the picture due to issue with the first xbox that is unless Intel is basiclly giving away Larrabee to try and start the ball moving for the PC market..

Intel's main issue this gen is that they didn't like the idea of Microsoft owning the CPU ip. If they were willing to play ball they might have been chosen...they might be willing to play ball now same thing goes with NVIDIA. Not saying it's going to happen just that Microsoft did shop around this gen.

 

[Acert93]

"Owning the IP" means MS can shrink the design, incorporate other chips (e.g. a GPU) on a single process, etc. MS wouldn't own the actual patents; the problem from what I remember is that putting a GPU and CPU on a single die could expose libraries and whatnot that are sensative. Some companies like to control their designs and exposure.

 

[Dural]

Yes, I will. Even more-so now. I now have a big investment in DD titles, that I didn't have before. So when I sell my old system to get the new system, I have now lost all that investment if it's not BC. Explain that to mom & pop, they won't be too happy that's for sure.

Tommy McClain

Yeah, I've also invested a lot in digitial distribution this gen and if I can't play those same games on the next console I'll be pretty pissed.

 

[Silent_Buddha]

Wont they need to go with a similar CPU as last time if BC is such a supposed big concern?

Dont see why the GPU has to be from ATI for BC supposedly, but they can just switch CPU vendors?

Neither the CPU nor the GPU emulation was a problem for BC on the X360. Both were easily worked around.

The problem was Microsoft having to pay Nvidia licensing fees for every title with BC compatability. After doing BC for the majority of big selling Xbox titles, the return on investment wasn't enough to continue adding BC for lesser selling titles.

Assuming there aren't any licensing fees with CPU or GPU, BC going forward is almost guaranteed for Xbox Next. The question is, will they need to do emulation or not? With ATI designing the GPU, there's a good chance they wouldn't need any for the majority of titles. For the CPU, they probably will if they switch to an x86 CPU.

Regards,
SB

 

[Shifty Geezer]

Are you sure that's true? I thought the reason there's a limited selection of BC titles is because each one had to have its own idiosyncracies managed. I can't imagine the licensing costs to nVidia would run into multiple dollars rending the otherwise virtually free BC conversion no longer cost effective. Whereas the effort to iron out BC bugs for low-selling titles would make one hesitant to bother.

 

[AzBat]

Neither the CPU nor the GPU emulation was a problem for BC on the X360. Both were easily worked around.

What? Says who? There's only 462 games on the backward compatibility list out of the 796 titles that are currently listed on Xbox.com. It took them years just to get that many & there are some really popular titles that will never get support*(Chronicles of Riddick, all Brothers in Arms, Mech Assault 1, Midnight Club 2/3, Rainbow Six 3, Deus Ex Invisible War, Time Splitters 1/2). It was not easy by any stretch of imagination.

* I didn't mention Madden since it was probably requested by EA not to provide support. In fact, most yearly sports titles didn't get support.

Tommy McClain

 

[fehu]

Well a 4 core would be around 284mm2 which is a bit too large imho but awesomely powerful with the 16 threads.

284mm2 only if you split in two a full power7
to be correct must first subtract the memory controller, about 28MB of edram, a lot of external bus, something that i can't figure out, reduce it at least at 32nm, and then calculate the real area

and maybe with ibm and ati learning to make cooler chip (in particular in idle), microsoft can be tempted to accept a design bigger than 160mm2

but by the time the x720 come out we will be talking about power8

 

[upnorthsox]

284mm2 only if you split in two a full power7
to be correct must first subtract the memory controller, about 28MB of edram, a lot of external bus, something that i can't figure out, reduce it at least at 32nm, and then calculate the real area

and maybe with ibm and ati learning to make cooler chip (in particular in idle), microsoft can be tempted to accept a design bigger than 160mm2

but by the time the x720 come out we will be talking about power8

It's pretty much 2 4 cores interconnected so it should be a pretty clean split:

Splitting a 4 core though would be run into the issues you bring up.

 

[Acert93]

Do you want 4 cores like that, or many more smaller cores (less efficient cores but distinct each with dedicated execution units; i.e. PPEs). Robust cores are nice, but you also tend to have a lot of silicon realestate idle. MS will want something accessible and performant on "typical" code with legs to grow.

While AMD64 cores may kill BC, a core like that with some advanced vector units would probably fit the bill. Maybe IBM can remove the nasty "gotchas" on the PPEs and deliver an OOOe chip that is BC and still small.

 

[fehu]

next gen i think that a lot of task will migrate to the fat gpu, so the cpu will keep only the big non parallelizable threads (bigger core no more spe), or not?

in any case ibm has the know how to extract smt4 from ooo cores, so they'll be hardly in idle due to hardware design

 

[RudeCurve]

In terms of BC I think it would be a great idea to off load BC onto a daughterboard that you could optionally buy and plug into your new console. Imagine the daughterboard taking the form factor of something like a game cartridge that you simply plug into a AUX slot on the back of the console.

If you look back at PS2 BC on PS3 there were some hardware integrated onto the motherboard, but in the future I propose the BC hardware be modularized. I know this is a great and easily doable, but it's really up to the console manufacturers to decide whether or not it's worth it. Personally as a consumer I think it's worth it if the BC cart is priced right.

What do you guys think of this idea?

 

[AlNom]

What do you mean by a daughterboard (what's in it)? The cost to the motherboard design and chassis might be prohibitive there... besides the uncertainty of selling the hardware and producing it versus a software solution. Even the presence of PS2 hardware in PS3 didn't guarantee 100% compatibility...

 

[RudeCurve]

I would like the daughterboard to contain something like what SONY had EE+GS+audioIC+RAM. The power supply and AV connections etc would be shared with the main console.

The other benefit is of course the main console could have a totally new architecture without having to worry about compromises made in order to have good BC.

 

[grandmaster]

Problem being that the cost would be fairly close to the price of the console being emulated. Which you would already own, presumably. While it would be more elegant to have everything in one box, I can't see people laying out the money to do so. Plus of course it will instantly limit the market for downloading BC games from PSN/Live or whatever.

 

[Acert93]

If MS has chosen AMD for the GPU you would think a thin wrapper mapping onto the new GPU should be simple enough in most cases. Presumably the new GPU will have more features, more units, and higher frequencies. eDRAM may be harder to emulate, although that depends on the design (and if they can compensate with some compression and brute force).

The CPUs could be much harder going from back and forth again between Big/Small Endian and how would AMD deal with the 128bit VMX?

Going with an evolutionary PPE design may not be ideal, but if they get OOOe and fix some of the craptastic stuff and design with a focus on the GPU they could get their BC all around pretty easily it would seem. Evolutionary GPU and evolutionary PPC.

I don't see why they cannot do a next gen design and get BC if they stick with the same vendors. The question is can IBM deliver the CPU with a good bang/buck ratio?

 

[AlNom]

The CPUs could be much harder going from back and forth again between Big/Small Endian and how would AMD deal with the 128bit VMX?

The endian stuff should be relatively trivial, but yeah... VMX128... :s

Going with an evolutionary PPE design may not be ideal, but if they get OOOe and fix some of the craptastic stuff and design with a focus on the GPU they could get their BC all around pretty easily it would seem. Evolutionary GPU and evolutionary PPC.

I don't see why they cannot do a next gen design and get BC if they stick with the same vendors. The question is can IBM deliver the CPU with a good bang/buck ratio?

They'll at least be in a better position than the Xbox<->360 transition; they won't be rushing it out of the gate.

With their Games-On-Demand service, it'd be a pretty nice scenario with full backwards compatibility, offering cheaper (old) games right off the bat and maybe promoting enhanced graphics.

Of course, on the developer side, the cross-generation transition will be that much easier, and perhaps the best MS can do on the software side besides their multiplatform (PC,360) endeavours. Here's an easy way to convince developers to continue the using MS hardware as the target platform in light of reducing costs.

Publishing could be interesting too, offering a 360 SKU and the next gen SKU (higher price) for early adopters. With the PC platform in mind, the next gen SKU would be closer to that in terms of say textures and models or DX11 (who knows).

 

[liolio]

Actually if MS were to use a x86 cpu getting the game to work and sell them for dowmload would be "a piece of cake" as a lot of games are available on PC use SSE instead of VMX etc.
Actually the more I think about it the more I would favor a full AMD system (putting aside business consideration). I've wondered if ATI would accept to sell the chip at a reasonable price as it would also add a lot of meat to their 'game compliance standard' ( I don't remember the exact marketing name) by having their hardware really close to the next XBox. I could see Ms and ATI happy to further close the gap between Pc and console gaming and leveraging a bigger as silicon gets cheaper try to get mid/low mid Pc config as a compliant gaming platform.

 

[Shifty Geezer]

Problem being that the cost would be fairly close to the price of the console being emulated.

Perhaps not. No optical drive, PSU, HDD, perhaps by the time PS2 was £100, a PS3-plugin would cost £50? I quite like the idea of a compromise of this format if a true software BC isn't possible, or would be at a significant performance disadvantage comapred to a clean-break hardware solution.

Although personally I have no interest in BC if it isn't improving on the old experience. My ideal is hardware that can run the old games at superior quality. That means extending the current designs rather than starting from scratch. I certainly see that possible with PS3>PS4. Wouldn't be too hard with XB360>neXBox although the CPU may be a comparative bottleneck?

 

[AlNom]

I'd wonder how adverse AMD would be to modifying one of their CPUs to suit MS's needs (considering their own apparent issues with their PC designs). I mean similarly to what happened with Xenon. i.e. extended vector processing, hardware security measures, cache locking by GPU, implications with the GPU handling the northbridge/memory control... Does AMD have the resources for such a side-project to do a good job...

I've not been particularly impressed by their power consumption in the desktop space, nor their cache density, but comparing that to what IBM can do would be very difficult to say...

Just some concerns...