XBox One 20nm APU revision possibilities *spawn*

mosen

Regular
Lol you are famous now mosen. Smaller, cheaper, cooler Xbox One processor in development • Eurogamer.net

http://www.eurogamer.net/articles/d...aper-cooler-xbox-one-processor-in-development

Unless Microsoft are well on the way of deploying a slim Xbox One, this is a bit of a bugger for Microsoft. People desperately wanting a smaller Xbox One may now be minded to wait for the smaller unit, even if one isn't imminent. Microsoft may deny it but then they denied they'd remove Kinect the weeks later it was gone. A smaller Xbox One hugely appeals to me.

I blame Mosen for destroying Microsoft.

Bad Mosen, bad! :yep2:

That was not intentional. :| I only wanted to share this information with you to see your opinions and read your discussions, as most of you are far more informed than me. Sorry, Microsoft, AMD and Mr. Daniel McConnell, I'm really sorry. :|

But I should say thanks to Mr. Leadbetter because of mentioning my username in his article. :D
 
Last edited by a moderator:
Okay, these are good insights as to technical obstacles to a Slim upgrade, thanks for sharing them.

I agree with this perspective as it is the more likely and realistic expectation to follow course, I'm just leaving open consideration for the possibility that precedent is broken and something unexpected and unconventional unfolds for a Slim upgrade.

I would guess the combination of a smaller, cheaper hardware and exclusives is the much more expected approach MS will take for Xbox One marketing for the rest of its years on the market.

If the older consoles don't have the shaders physicly turned off and its just in software it may be better for them to unlock them and just exchange out consoles in which the chip can't run fully unlocked.


But for ms i'm sure if they aren't able to turn the tide with halo 5 and a slim xbox one next year they will simply turn towards a replacement.
2017/18 would be a good time for one ( i wouldn't mind 2016 but alot of people here would cry bloodly murder)

You'd have stacked ram , ssd will be dirt cheap (you can already get a 480/512 gig drive for under $200 ) sub 10nm chips will be avalible and vr should be established a sucess or failure . So they can dicate the console on that .
 
Dont forget that the 360 had a chip shrink or 2 without changing the actual form factor of the system.
I belive the Halo3 edition of the 360 was the first version released with a smaller cpu and gpu.
 
Dont forget that the 360 had a chip shrink or 2 without changing the actual form factor of the system.
I belive the Halo3 edition of the 360 was the first version released with a smaller cpu and gpu.

didn't the xbox 360 get smaller chips in the old chasis and then would shrink the casing after ? So i would think alot of the smaller chips would release in the original casing and at some point ms would switch over to a smaller form factor. They would want to reap the savings as soon as possible
 
Sony, Kutaragi I believe, said they intended to launch PS3 at 65nm. 360 didn't get a die shrink for two years. PS3 got its a year after release, but in line with 360 and the 65 nm process.

A die shrink one year after launch is unusual. It only happened for PS3 because the intended process was a year late and Sony had to fall back to 90nm.

I don't think the reporting is any different. Websites ran with reporting that the 40 GB PS3 had 65nm, then didn't, then did again when Hirai finally cleared it up. Perhaps the number of websites and clueless assumptions are more numerous, but that's just because the internet has grown in volume (in every sense of the word) in seven years.

Playstation 2 had a die shrink soon and the development on the dies was quite aggressive, different times of course with regards to the rapid advancements in the area, but still.

EE+GS.jpg


The work on the chip is always done much prior to it hitting the market, just like the work on the original 28nm chip was done prior to the launch of the console. The new chip isn't out yet, only reported. Considering that there have already been products (Apple A8 chips) shipping from TSMC for quite some time using that process, I wouldn't use the word unusual here, not because it isn't a rarely occuring event (console launch + new node), but because it's just logical for a die shrink to appear soon after a new node becomes commercially viable and this time it is going to happen relatively soon after the launch of the consoles. (a reason I was hoping for more aggressive/less elegant launch systems with regards to power/noise)

Having seen an Xbox One in person only recently I have to say the unit gave me a pretty good impression, looks and feel premiumish, I don't think the size is the problem here and I think they already could have designed it to be much slimmer with the current 28nm chip as evidenced by PS4s higher power draw, internal power supply and much lower internal volume.

Maybe at next E3 we will see a new smaller unit introduced and perhaps the new chips could even be introduced to the current form factor a bit sooner? If the chip can be produced on the same process as the A8 chips, I'm guessing it can, but not certain about that...
 
Last edited by a moderator:
If the older consoles don't have the shaders physicly turned off and its just in software it may be better for them to unlock them and just exchange out consoles in which the chip can't run fully unlocked.

Just give up this pipe dream, Xbox one is what it is.

What they should have done was enable the redundant CU's BEFORE the generation, as both PM rumors in my inbox and Eurogamer Xbox architects interview confirmed they heavily considered, but decided against. If they could have done that and kept the clock at at least 853 it could have closed the gap considerably. The clock was only +7%, the CU's were +16% flops by themselves.

It's too late now, you cant fracture the userbase. and I've considered MS "taking back" the already out the door consoles, it quickly runs into the billions of dollars. Not possible. 6 million (bare minimum) consoles out there times what, AT LEAST $300 per, 1.8 billion dollars. It's really probably more like 8 million consoles shipped at least by now (October is complete and my estimate was 7.5 shipped by end of September) and ~$400 per, which is $3.2 billion.
 
didn't the xbox 360 get smaller chips in the old chasis and then would shrink the casing after ? So i would think alot of the smaller chips would release in the original casing and at some point ms would switch over to a smaller form factor. They would want to reap the savings as soon as possible

They didnt switch to a smaller form factor until the second version of the combined cpu/gpu soc.
 
i don't disagree and i said as much in my post that ms wont do that and just intsead launch a new system faster.

As for your caculations why would you figure them that way. If the APU is $50 bucks (just a number) then they would be able to replace just the apu with a new one .

I'm also not sure why your going with 7.5m if they are just disabled in software ( which is sugested by your pm comments) then they can turn them on . Yields should have been getting better over the last year so I'm sure a certian percentage would run without a problem .

But like i said , they are better suited to just get a new console gen running in 5 years.

Like I said by 2017/18 we should know if VR has taken off , we should have stacked ram , sub 10nm processors. They can regroup on that and the xbox one shouldn't be a total loss. DDR 3 and the edram shouold allow it to keep a price advantage through the generation and it should still sell large amounts of software. They can even keep it (if they can get the box to sub $200 prices ) when they launch a new system.
 
I'm with those who think this will most likely be a drop in replacement, possibly even before alternative ddr3/4 configurations are used (as they're still looking for extra folks to work on that).

As already pointed out a slim is already kinda-feasible, if MS are prepared to bump the noise up more towards ps4 levels.

Dropping the new chip into the existing platform would probably let them switch to a cheaper cooler, much like the 360 Falcon did for the CPU. Heatpipes are surprisingly expensive, and the Bone uses 2 of them. A hunk of aluminium with a copper core might be heavier but it'll save 10+ dollar off the bom.
 
If they could significantly reduce the size of the console with the 20nm APU, the whole XB1 proposition would be more tempting for many people, including me.

If only for the asian markets sake.
 
I think any significant size reduction would need to be tied to not just heat and power but also dropping down from 16 memory chips on a single side to a more manageable figure like 8. And going by mosen's findings, it seems that the first apu revision design is done and dusted, but work on ddr3 memory revision is ongoing. I guess we'll know soon if those 32-bit ddr3 memory chips made it in for the first revision.

A smaller box would be good news, but in the short term just saving some money on the chip and the cooler would be a good start.
 
I think any significant size reduction would need to be tied to not just heat and power but also dropping down from 16 memory chips on a single side to a more manageable figure like 8. And going by mosen's findings, it seems that the first apu revision design is done and dusted, but work on ddr3 memory revision is ongoing. I guess we'll know soon if those 32-bit ddr3 memory chips made it in for the first revision.

A smaller box would be good news, but in the short term just saving some money on the chip and the cooler would be a good start.
To me the most essential feature of a new slim mode would be keeping the console as silent as it is now.
 
I'd like a pretty quiet machine too, but looking at the ps4 - which can get pretty noisy and had what is likely a cheaper cooler - I'm not sure that matters so much to most core games.

The first Xenon 360s were truly awful - still sold well. My Falcon was much quieter but I still didn't like the dual fan drone. But I still bought one ...

My 360S is practically silent and my pc is fairly quiet and never intrusive despite having one million fans inside, so it's obviously important to me, but for the right price even a quietness-chump like me might compromise and go for something none-silent again...
 
I'd like a pretty quiet machine too, but looking at the ps4 - which can get pretty noisy and had what is likely a cheaper cooler - I'm not sure that matters so much to most core games.

The first Xenon 360s were truly awful - still sold well. My Falcon was much quieter but I still didn't like the dual fan drone. But I still bought one ...

My 360S is practically silent and my pc is fairly quiet and never intrusive despite having one million fans inside, so it's obviously important to me, but for the right price even a quietness-chump like me might compromise and go for something none-silent again...
I have a Slim and it still sounds like a reactor to me -compared to the Xbox One- with no disc spinning in the drive. I couldn't sleep when the Slim was downloading something in the background --while technically switched off.

I agree with you on the core part, most people don't care over time, get used to it, or just don't mind at all, but I have a pretty good hearing -bad eyesight though, I was born myopic- and the noise of the Slim is very distracting. And that's considering how the X360 improved over time on that regard.
 
Just give up this pipe dream, Xbox one is what it is.

What they should have done was enable the redundant CU's BEFORE the generation, as both PM rumors in my inbox and Eurogamer Xbox architects interview confirmed they heavily considered, but decided against. If they could have done that and kept the clock at at least 853 it could have closed the gap considerably. The clock was only +7%, the CU's were +16% flops by themselves.

It's too late now, you cant fracture the userbase. and I've considered MS "taking back" the already out the door consoles, it quickly runs into the billions of dollars. Not possible. 6 million (bare minimum) consoles out there times what, AT LEAST $300 per, 1.8 billion dollars. It's really probably more like 8 million consoles shipped at least by now (October is complete and my estimate was 7.5 shipped by end of September) and ~$400 per, which is $3.2 billion.

I agree with this, but if these market trends continue, the potential lost software and XBL Gold revenue, by people switching to PS4 and/or buying 3rd party games only on PS4, could be in the billions anyway. It’s almost a damned if you do, damned if you don’t situation.

The sad/ironic part is that adding a small bit of VRAM (1 or 2 GB GDDR5) and replacing the on-die ESRAM with more CU’s like PS4 probably would have added less cost than the $50 they are forced to eat this holiday, just to stay competitive. (someone with more cost/tech knowledge correct me if I’m wrong!)
 
News & Rumors: Xbox One (codename Durango)

What is the probability that the holiday discount bundles priced at 350 might be to clear out all the old Xbox One 28nm SKU to sell the 20nm SKU? Maybe the price will be a permanent 350 soon? And I don't mean slim, I'm just thinking SOC and cooling changes.
 
Just give up this pipe dream, Xbox one is what it is.

What they should have done was enable the redundant CU's BEFORE the generation, as both PM rumors in my inbox and Eurogamer Xbox architects interview confirmed they heavily considered, but decided against. If they could have done that and kept the clock at at least 853 it could have closed the gap considerably. The clock was only +7%, the CU's were +16% flops by themselves.

It's too late now, you cant fracture the userbase. and I've considered MS "taking back" the already out the door consoles, it quickly runs into the billions of dollars. Not possible. 6 million (bare minimum) consoles out there times what, AT LEAST $300 per, 1.8 billion dollars. It's really probably more like 8 million consoles shipped at least by now (October is complete and my estimate was 7.5 shipped by end of September) and ~$400 per, which is $3.2 billion.

MS already stated that the 7% clock outperformed the extra 2 CU due to either diminishing returns on CUs or some other bottleneck (possibly ROPS). I could see both a clock increase and a CU unlock in the spring. Two ways to do this - A - each system is either stock, upclocked, UP CUed, or both and have games respond dynamically like COD to the system. B - assuming that the successful conversion rate is sufficiently high allow folks to elect the upgrade and replace those that fail.

equivalence of 7% upclock to 2 CU implies a scaling factor of 0.96. If that scaling factor applies, then an upclock to 1 ghz would enable the XB1 12 CU to outperform the 18 CU of the PS4.
 
MS already stated that the 7% clock outperformed the extra 2 CU due to either diminishing returns on CUs or some other bottleneck (possibly ROPS). I could see both a clock increase and a CU unlock in the spring. Two ways to do this - A - each system is either stock, upclocked, UP CUed, or both and have games respond dynamically like COD to the system. B - assuming that the successful conversion rate is sufficiently high allow folks to elect the upgrade and replace those that fail.

equivalence of 7% upclock to 2 CU implies a scaling factor of 0.96. If that scaling factor applies, then an upclock to 1 ghz would enable the XB1 12 CU to outperform the 18 CU of the PS4.

Interesting.

I d like to hear from somewhere with better insghts on this so we can clear up if this is indeed the case
 
Back
Top