Xbox Series S [XBSS] (Lockhart) General Rumors and Speculation *spawn*

Status
Not open for further replies.
Just to put into perspective.
Lockhart will sit around 14-18CUs.
Scarlett is 56 CUs - 4 for redundancy.

That's a lot of CUs to toss away, not to mention an oversized bus and cache to follow.

I agree, that’s why I’m not dismissing it in the long term. They had a lot more than 30 bad ones at the beginning. They were spinning chips a year ago if they follow SOP. There’s a lot of bad stock to work with that’D have 50% capacity.
 
I agree, that’s why I’m not dismissing it in the long term. They had a lot more than 30 bad ones at the beginning. They were spinning chips a year ago if they follow SOP. There’s a lot of bad stock to work with that’D have 50% capacity.
nah, it won't work is what i"m saying. It's too big of a chop for this to happen. There are great deal of many other products a lower yield scarlett would be appropriate for.
 
So.... let's assume 14Gbps because they'll be buying in bulk for both SKUs, ergo:

1.
192-bit @ 336GB/s
4x2GB + 2x1GB = 10GB, 6 chips x 32-bit
6GB @ 192-bit, 4GB @128-bit
Requires similarly split pool considerations after the OS reservation. (bad imo)

2.
160-bit @280GB/s
5x2GB = 10GB, 5 chips x 32-bit

3.
128-bit @ 224GB/s
6x1GB + 2x2GB, 8 chips x "16-bit" clamshell
8GB @ 128-bit, 2GB @ 32-bit
Devs only have to worry about <8GB down from SX's 13.5GB
In the future it can reduce to 4 chips without clamshell (1x4GB + 3x2GB)

------

Knowing the 4TF target though, I don't imagine they would have gone particularly wide in the first place sooooooooo.........

If it's really 10GB, I'd bet on option 3.
A bit over 1/3rd the bandwidth of the Series X that has a 3x more powerful GPU, saving as much as possible on the PCB traces and amount of chips.
 
If it's really 10GB, I'd bet on option 3.
A bit over 1/3rd the bandwidth of the Series X that has a 3x more powerful GPU, saving as much as possible on the PCB traces and amount of chips.
I'm going with option 2. When your slow pool is 300+ GB/s you can get away with it, but why mess with a 56 GB/s pool when you have a clean, and possibly cheaper 5-chip option?
 
I'm going with option 2. When your slow pool is 300+ GB/s you can get away with it, but why mess with a 56 GB/s pool when you have a clean, and possibly cheaper 5-chip option?
Because 56GB/s for the CPU is more than enough (more than a PC with dual-channel DDR4 3200MT/s) and with clamshell they can do less traces, with 4GB chips down the road they can do it even cheaper.
 
Because faulty processors will be so few and far between, you'll be throwing away good chips to make a cheap box.

Let's say you can make 1,000 XBSX SOCs on a wafer, and there's 30 defects. That means for each wafer, you can make 970 XBSX's and 30 LHs. That's not enough LHs, so you have to take some of those good SOCs to make LHs. Now you're putting a $100 chip into a $300 console instead of a $50 SOC.

Now if instead MS make a LH SOC and can fit 2,000 of them onto a wafer, they can produce more economical LH consoles and not cannibalise their XBSX production.

Or, you could make 1,000 XBSX SOC on a wafer, say you get 30 defects, when running @ speed X, then keep increasing X until your get the desired ratio, of pass / fail units.
In this case XBSX vs LH soc's. It helps you push the high end further up, while also satisfying whatever internal need you have for LH chips.
so XBSX ends up @ X + 200Mhz, and you get the 500/500 ratio you want?
I mean it's possible...
But I don't think it is likely, and it is still a woefully in-efficient way to use the die space.
 
Or, you could make 1,000 XBSX SOC on a wafer, say you get 30 defects, when running @ speed X, then keep increasing X until your get the desired ratio, of pass / fail units.
In this case XBSX vs LH soc's. It helps you push the high end further up, while also satisfying whatever internal need you have for LH chips.
so XBSX ends up @ X + 200Mhz, and you get the 500/500 ratio you want?
I mean it's possible...
But I don't think it is likely, and it is still a woefully in-efficient way to use the die space.
A wafer is a fixed size. A product line will only produce so many completed wafers per day. If you can have a wafer with a 10 chips on it with 9 good ones and 1 bad one that can go into your smaller console or a wafer with 30 chips in which maybe 2 are bad. You now have 28 chips for that lower priced console. Now lets say you can do 100 wafers per day and you want to launch two consoles as price efficiently as you can. What do you do ?
 
Because 56GB/s for the CPU is more than enough (more than a PC with dual-channel DDR4 3200MT/s) and with clamshell they can do less traces, with 4GB chips down the road they can do it even cheaper.
This layout, OS is 2GB that would be across the slower pool, with 0.5GB of fast memory also allotted for OS for graphics related tasks.

So the game wouldn't be using slower pool anyway.

I'm not sure if 224GB/s would be enough though, but more than happy to be told its plenty.
Especially as will be doing high bandwidth tasks like RTRT, and bandwidth can seriously hobble the gpu and stop it from reaching its potential.
Probably one of my biggest worries is Lockharts memory bandwidth to be honest. Everything else I think is easily dealt with by devs.
 
This layout, OS is 2GB that would be across the slower pool, with 0.5GB of fast memory also allotted for OS for graphics related tasks.

So the game wouldn't be using slower pool anyway.

I'm not sure if 224GB/s would be enough though, but more than happy to be told its plenty.
Especially as will be doing high bandwidth tasks like RTRT, and bandwidth can seriously hobble the gpu and stop it from reaching its potential.
Probably one of my biggest worries is Lockharts memory bandwidth to be honest. Everything else I think is easily dealt with by devs.
I think mem bandwidht and mem size, GPU adjusted power will be very much similar to One-X.... this console is IMHO an One-X replacement, it will run same specs games (at least at the beg)... don't think MS force to support 4 much different consoles... at least 2 of 4 should be very similar...
 
8 mb of L3 seems really good. Hope also Ps5 has the same.... this is gonna help a lot with bandwidth.
 
Or, you could make 1,000 XBSX SOC on a wafer, say you get 30 defects, when running @ speed X, then keep increasing X until your get the desired ratio, of pass / fail units.
In this case XBSX vs LH soc's. It helps you push the high end further up, while also satisfying whatever internal need you have for LH chips.
so XBSX ends up @ X + 200Mhz, and you get the 500/500 ratio you want?
I mean it's possible...
But I don't think it is likely, and it is still a woefully in-efficient way to use the die space.
The components are built around the SoC hitting a certain number. It could throw costs way out in the method you described. Just easier to make 2 separate ones.
 
it definitely exists.
We do not know for sure if it will be announced/released.
I'm going to assume this decision hinges on price on PS5.
You think Microsoft ordered and paid for the development of a full next-gen console but is holding out the decision to release it based on the PS5's price?

I don't think that makes sense considering the cheer amount of money this already cost Microsoft. The decision to release Lockhart was already made, probably well over a year ago.
 
You think Microsoft ordered and paid for the development of a full next-gen console but is holding out the decision to release it based on the PS5's price?

I don't think that makes sense considering the cheer amount of money this already cost Microsoft. The decision to release Lockhart was already made, probably well over a year ago.

At the wrong price and placement, the project could cost them even more money. With XSX as their main developed console, and lockhart deriving from that, the additional costs werent that huge for MS i think.
 
I think mem bandwidht and mem size, GPU adjusted power will be very much similar to One-X.... this console is IMHO an One-X replacement, it will run same specs games (at least at the beg)... don't think MS force to support 4 much different consoles... at least 2 of 4 should be very similar...
From what I gather RTRT is bandwidth heavy, a lot of decent graphics cards are hampered by there bandwidth, and don't won't that too be the case here.
You think Microsoft ordered and paid for the development of a full next-gen console but is holding out the decision to release it based on the PS5's price?

I don't think that makes sense considering the cheer amount of money this already cost Microsoft. The decision to release Lockhart was already made, probably well over a year ago.
Wouldn't be the first time they've made such a decision.

This time though, I fully expect them to release it, but they don't want to talk or acknowledge it until their sure of price points.
Only reason not to release, is if the PS5DD was so cheap that Lockhart was unable to put enough daylight between it.
I don't see why PS5 in general would be significantly cheaper than xsx.
 
PS5DD is still a full PS5, is on another league compared to LH.
If the main selling point must be the price, they really need to be priced at about half the price of it.
 
PS5DD is still a full PS5, is on another league compared to LH.
If the main selling point must be the price, they really need to be priced at about half the price of it.

I think the form factor is still a mystery too. Might not be a standard TV set-top.

BTW, it was free(or low monthly fee) with a 2 year Game Pass subscription I think that could be really disruptive.

Tommy McClain
 
Status
Not open for further replies.
Back
Top