Xbox : What should MS do next? *spawn

Because it will cost more than the current xbox one soc.

The Xbox one SOC will be tiny on 14nm and it already plays xbox one games just fine. They should be able to stay cheaper than sony through the generation while a change of SOC and ram will just increase price.

What you describe above would not sell for $249 or less unless ms eats some cost , everything would be more expensive than the xbox one.

Also why would they need a more powerful low end console when they have scorpio coming out. The goal is to have a good price difference between the two. I would suspect scorpio will come in at $400-$500 depending on its power and you'd want the xbox one at $200 .

If they can't do it at 14nm then I would suspect they'd want the xbox one at 10nm to be their media stick / box

It wont be more expense, that is the point. You are looking at maybe an extra 5 dollars max to the cost of the SOC.

I see no reason why they couldn't have a $199 driveless SKU that breaks even. Correct me if I am wrong, but doesn't MS buy wafers not individual chips?
 
Brit comes up with lots of slang. He's the one that coined the name "Scropio".
And XBone during the reveal event! I still don't get how they didn't see that one coming given XBox 360 was always XB360. Major forehead smack moment there.
 
And XBone during the reveal event! I still don't get how they didn't see that one coming given XBox 360 was always XB360. Major forehead smack moment there.

Speak for yourself. I always called it X360. I'm too lazy to type the extra B. :p

But yeah, they should have known that someone somewhere (or multiple people multiple wheres) would conjugate that into XBone. Although again I'm lazy so I just use XBO.

Regards,
SB
 
It wont be more expense, that is the point. You are looking at maybe an extra 5 dollars max to the cost of the SOC.

I see no reason why they couldn't have a $199 driveless SKU that breaks even. Correct me if I am wrong, but doesn't MS buy wafers not individual chips?

Your going to stop producing an SOC of which they've made 20m + and put in a new gpu portion which will be the same or bigger than the gpu + sdram in the xbox one (when shrunk) . Your also going to move from DDR3 to GDR 3 ram. Those are all more expensive.

Sdram on the Xbox one apu will shrink really well compared to the rest of the chip. GDR 5 ram also is more expensive and uses more power than DDR 3.


You will also have 20m + Xbox ones , X amount of Xbox one enhnanced and then Scorpio to program for. That will be 3 systems with 2 of them releasing a year apart if even. It makes little sense to me. IMO the goal would be to minimize the amount of different render targets and provide a complete top to bottom option. That's why I've suggested at $100 or even $150 a intel nuc sized xbox one with just a 2.5 inch drive (or even some amount of nand ) $200-$250 xbox one slim with 4k capability and then the $400-$500 scorpio.

We run into a problem where is xbox tv that thurrot has mentioned uses a different SOC they will already have 3 different SOCs , I don't see how adding a 4th will help.

Also if the industry will shift to 3 year cycles they will want the xbox one gone as soon as possible i'd imagine. So I think we'd see 2017 Scorpio , 2020 Xbox three. With 2013 xbox one retiring or if they weren't able to do it earlier move it to the media stick since 10nm or less will be possible then. When would you retire this hybrid sku ?
 
Brit comes up with lots of slang. He's the one that coined the name "Scropio".

Baffin - is the term used for creating a console that has secret sauce so powerful it eventually needs Scropio resin to be mixed to start the reaction that unlocks a dGPU by etching the upper dies of the dark silicon. Or it could just be the sudden onset of heatstroke I am suffering.

Scropio does have a nice ring to it, almost sounds like Xbox Scrapio...

In all seriousness, if the Surface team is truly in on all this I can't wait to see what the rumored Xbox Slim/TV whatever and Scorpio will look like. Curious on the controller, as I don't mind the current design (unless voice recognition is added).
 
Question for more knowledgeable folks, as part of the ole 1-2 punch is there any reason why MS would NOT go with Boffin in the slim? That would add very minimal space to the die, and that plus FinFet upclocks and architecture changes to improve bandwidth through compression should be enough power (16 CU x 1200/800 MHz = 2.4 TF with higher IPC or ~ 50% increase over PS4) to get consistent 1080P60. At 1200 MHz, esRAM would have theoretical peak > 300 GB/s. Faster DDR4, main memory up to 80 GB/s or more. All of that selling for $249 or less.
How do you know how many CUs are in Baffin? ;)

:eek:
 
Is possible just to shrink Xbox One SOC for Xbox One Slim and make it 10% faster? Targeting the same profile as Xbox One, but with smoother FPS.
 
Is possible just to shrink Xbox One SOC for Xbox One Slim and make it 10% faster? Targeting the same profile as Xbox One, but with smoother FPS.
anything is possible, it's whether it's the correct and practical choice for its user base; it's not. MS will receive massive backlash for the split in models performance. Either the entire base model moves up in performance with the slim, or the slim comes down to the performance of the existing hardware. I'm fairly certain it will be the latter. The former would require an overclock, which I like, but it's not practical.
 
Your going to stop producing an SOC of which they've made 20m + and put in a new gpu portion which will be the same or bigger than the gpu + sdram in the xbox one (when shrunk) . Your also going to move from DDR3 to GDR 3 ram. Those are all more expensive.

Sdram on the Xbox one apu will shrink really well compared to the rest of the chip. GDR 5 ram also is more expensive and uses more power than DDR 3.


You will also have 20m + Xbox ones , X amount of Xbox one enhnanced and then Scorpio to program for. That will be 3 systems with 2 of them releasing a year apart if even. It makes little sense to me. IMO the goal would be to minimize the amount of different render targets and provide a complete top to bottom option. That's why I've suggested at $100 or even $150 a intel nuc sized xbox one with just a 2.5 inch drive (or even some amount of nand ) $200-$250 xbox one slim with 4k capability and then the $400-$500 scorpio.

We run into a problem where is xbox tv that thurrot has mentioned uses a different SOC they will already have 3 different SOCs , I don't see how adding a 4th will help.

Also if the industry will shift to 3 year cycles they will want the xbox one gone as soon as possible i'd imagine. So I think we'd see 2017 Scorpio , 2020 Xbox three. With 2013 xbox one retiring or if they weren't able to do it earlier move it to the media stick since 10nm or less will be possible then. When would you retire this hybrid sku ?


You are putting more stuff in the slim than I meant. Go faster DDR4, cost neutral, keep esRAM, a 16 CU Baffin would only add maybe 10-15 MM^2 over the existing GPU so you would still have a sub 200mm die. That means that the SOC will cost ~ $25 if yields are good + whatever royalties AMD gets. Thus the SOC should be sub-$50. We will found out next week, but to me it is a no-brainer to go that route.

all of the xboxs are the same, just different power levels.
 
well leaks that have since proven out show 2 Baffin SKUs, 16 and 20 CU versions.
Fair enough. :p
You are putting more stuff in the slim than I meant. Go faster DDR4, cost neutral, keep esRAM, a 16 CU Baffin would only add maybe 10-15 MM^2 over the existing GPU so you would still have a sub 200mm die. That means that the SOC will cost ~ $25 if yields are good + whatever royalties AMD gets. Thus the SOC should be sub-$50. We will found out next week, but to me it is a no-brainer to go that route.

hm... Durango has 14CUs, which is not dissimilar from Bonaire. Orbis has 20CUs, similar to Pitcairn (w/ GCN1.1 upgrades, which I think are outside the CU blocks themselves). Anyways, it seems like AMD just did a copy/paste of those configs, so maybe they'd do the same for a hypothetical Baffin integration on there. Activating 16/20 with higher clocks would be neat. I suppose they don't have to go too far with the main memory given the bandwidth compression.

I'm just not sure how they'd go about the need for higher resolution buffers (VR or 1080p upgrades) since the ESRAM wouldn't be changing size here.

----

Perhaps worth noting that Polaris 10 is supposed to be 232mm^2 with apparently 40CUs (8RBEs?)
 
Failing a wonderful all HBM2 utopia ...

256-bit LPDDR4 with 64 ROPs and 64 MB of 500 GB/s esram (buses and internal whatnots permitting) would be my dreamboat. Lots of BW, very little power spent on memory and more to power more CUs, and more shrinkable (probably) than GDDR5(X).

Stab in the dark, but I can't see LPDDR4 being more expensive than GDDR5X, it might use a smaller interface on the die, and it'd certainly use less power. And the esram would keep sebbbi busy, and the results of that are always interesting.
 
Then again, we're talking say, 8 chips per console vs 1 per phone.

hum...

Aren't iPhones onto two chips on a 64-bit interface now? I think Samsung are about to release their first 6GB phone, so maybe 4 chips?

Graphics cards use lots of chips, but phones are as common as bad opinions. And Intel were about to adopt LPDDR4 for their next gen Atom .... though who knows what the hell is happening with that now.:confused:
 
I thought it was one for iPhone, 2 for iPad? I think it's just the high end phones at the moment (Samsung too, of course), but yeah, it'll be around for a while. GDDR5X is seemingly only going to be in the upper echelon of GPUs as well, so there's that too. :)
 
I thought it was one for iPhone, 2 for iPad? I think it's just the high end phones at the moment (Samsung too, of course), but yeah, it'll be around for a while. GDDR5X is seemingly only going to be in the upper echelon of GPUs as well, so there's that too. :)

You could be right. I'm not totally Au Fait with smart phones. I'm still rocking a Nokia. :oops:
 
Back
Top