Next Generation Hardware Speculation with a Technical Spin [post E3 2019, pre GDC 2020] [XBSX, PS5]

Status
Not open for further replies.
Okay, perhaps I've misinterpreted that upper limit. That's only an informal limit, but there are hard limits in stand-by and media playback modes.

I'd still be hard pressed to believe Sony and MS wouldn't wanna avoid to exceed that magical barrier.
 
There are going to be realistic upper limits simply due to heat and noise concerns. A 400 W console is going to need some significant cooling, so 200 W, whereabouts other consoles sit, is what we'd expect as an upper limit. I don't know that there are any real limits on, say, a 250 W box though if one of them went to town. Or even 300 W - just depends on what it costs additionally to deal with that power on top of the cost of the hardware that uses that power, and a 200 W machine should be (a lot?) cheaper than a 300 W machine.
 
I just hope the ps5 will be less noisy than my ps4 pro when she's pushed a lot.
Curious about back compat' too.

Flops&stuff, yeah, not so much in the end. Being not tied to based ps4/xbox will already do wonder. Add a good cpu, more ram, ssd, more modern gpu, I really believe it will be a great jump, 9tf or 13tf...
 
Think what they could do with 1.84 of GCN TF, see latest GoT trailer. A CPU ment for more then tablets, SSD, more and faster ram, and about 10TF of navi2 power, advanced toolsets and we will see great graphics improvements across ps5, xbox sx and pc. If not nintendo also gets more serious.
 
It's the only solution that works. But the yield still needs to work in favour of Anaconda; well with respect to time it will eventually work it's way towards Anaconda.
If the die is small in nature it's possible to have a good yield on perfect chips; but large dies is going to make this strategy very expensive; there will need to be redundant CUs in that scenario.

I suspect yield % will determine if they release a lockhart console (poor yields for perfect chips means green light); but I'm fairly positive there will be lockhart xcloud whether developers/people/concern trolls cry about it or not. The reality is the mass market is not ready in 2020 for 4K streaming over mobile - data consumption is just too high and MS needs to wait for the industry to catch up (in which makes sense over time as yields improve for Anaconda).

1080/1440p streaming is very suitable and will be for some time. Why waste perfect Anaconda chips in server blades doing that.

The topic of a lockhart console is still controversial for a binning strategy though (though obviously doesn't apply to xcloud). They must secure the devices against the consumers who own the product. You don't want people buying Lockhart and finding a way to 'unlock' the chip to be an unstable Anaconda. We haven't had that happen today; but if you're going to drop from 12TF to 4TF, that jump may be big enough for a market of folks who can take the difference in savings to 'upgrade' a console.
 
It's the only solution that works. But the yield still needs to work in favour of Anaconda; well with respect to time it will eventually work it's way towards Anaconda.
If the die is small in nature it's possible to have a good yield on perfect chips; but large dies is going to make this strategy very expensive; there will need to be redundant CUs in that scenario.

I suspect yield % will determine if they release a lockhart console (poor yields for perfect chips means green light); but I'm fairly positive there will be lockhart xcloud whether developers/people/concern trolls cry about it or not. The reality is the mass market is not ready in 2020 for 4K streaming over mobile - data consumption is just too high and MS needs to wait for the industry to catch up (in which makes sense over time as yields improve for Anaconda).

1080/1440p streaming is very suitable and will be for some time. Why waste perfect Anaconda chips in server blades doing that.

The topic of a lockhart console is still controversial for a binning strategy though (though obviously doesn't apply to xcloud). They must secure the devices against the consumers who own the product. You don't want people buying Lockhart and finding a way to 'unlock' the chip to be an unstable Anaconda. We haven't had that happen today; but if you're going to drop from 12TF to 4TF, that jump may be big enough for a market of folks who can take the difference in savings to 'upgrade' a console.

Aren't disabled parts of chip done so physically, and if so, isn't that irreversible?
 
Aren't disabled parts of chip done so physically, and if so, isn't that irreversible?
I'm not sure. The modding community has this crazy shit in that video where they were drilling in to the chip physically to bypass stuff. Like after and x-ray of the chip, they made templates to find the exact location to drill a whole into the silicon to bypass things.

In the past I could unlock cores on my CPU just by bridging some pins.
I personally don't know ;) but as the security experts have said (on xbox), the hacking community has some serious commitment to making things work the way they want to. They really needed to engineer a very custom method of ensuring that their console could not be hacked and that's actually where a majority of customizations are as I understand it.
 
I just hope the ps5 will be less noisy than my ps4 pro when she's pushed a lot.
Richard Leadbetter had some anecdotal info about it, him and the other journalist have a quiet one, including Richard's early model, and another staffer at work got a noisy one. He said all the ones produced in the last 2 years have been reported to be quiet. It seems to be a QC issue in the first year or so.

Some journalists said it was related to thermal compound (it was for a batch from the launch ps4), but I'm convinced this time it's not, as some users on reddit replaced the paste and it worked for a few days only.

My theory is the heatsink supplier that had a QC problem, where the surface was uneven or pitted. That would falsely show up as thermal paste issue, and come back as the new paste flows. If I had a noisy one I would have certainly inspected the surface. Power consumption differences would be interesting to test also.

Sure it's nice they solved the issue, but it sucks for those who got a noisy one... And Xb1x tech support doesn't seem any better for noisy xb1x, altough not nearly as common judging from forum posts, and for completely different reasons.
 
I suspect yield % will determine if they release a lockhart console (poor yields for perfect chips means green light); but I'm fairly positive there will be lockhart xcloud whether developers/people/concern trolls cry about it or not. The reality is the mass market is not ready in 2020 for 4K streaming over mobile - data consumption is just too high and MS needs to wait for the industry to catch up (in which makes sense over time as yields improve for Anaconda).

1080/1440p streaming is very suitable and will be for some time. Why waste perfect Anaconda chips in server blades doing that.

Yeah I speculated the same sort of thing as well. Anaconda is probably overkill for what Microsoft wants to use in their Xcloud data centers. Which is probably the main reason for Lockhart.

Microsoft vision for streaming right now seems to be more supplemental and not a direct alternative to consoles. No need for a massive power hungry chip if your primary aim is to just be able to stream next gen games at 720p/1080p quality to mobile devices...
 
XSX in cloud for streaming is overkill for right now.
But if they are going to be running other workloads 60% of the time?
Early on xcloud is for mobile, doesn't mean they wouldn't want to support XSX level games on TV couple years in, which in that case may make better sense to deploy it instead of having to go around upgrading to it.
I can definitely see the reasoning behind using Lockhart in the cloud though.

As for buying Lockhart and flashing bios to get XSX, considering how much security they put into every aspect of their design (Inc bios) , I don't see it happening. Especially as they would probably disable at the hardware level also.
Probably costs probitive to try converting Lockhart which was their goal with security for the XO in regards to running unsigned games etc.
 
For example:
  • XSeX = 64CU's with 8 disabled = 56 total CU's. Clocked high @ 1800MHz.
  • Lockhart = 64CU's with 16 disabled = 48 CU's. Clocked mid @ 1500MHz.
  • X1X = The shit chips which are only just over 60% functional = 40 CU's. Clocked low @ 1200MHz.

I am curious about lockhart APU. A perfect 7nm+ 64CUs APU has 350mm2, then a defected APU also occupies 350mm2.

In other words if perfect APU costs $160 then the defected APU also costs $160 which is costly for a $299 console.
 
In other words if perfect APU costs $160 then the defected APU also costs $160 which is costly for a $299 console.

You have to look at it relative to a non-binnable design, which would otherwise have conservative targets to mitigate throwing away entire chips ala traditional retail console hardware specs (i.e. redundancy in case of defects, lower clocks/voltage to encompass worst case thermal output etc).

There are trade-offs to maximize usable chips/yield.
 
Status
Not open for further replies.
Back
Top