Estimate a BOM delta for PS5 and XBSX

[Shifty Geezer]

Yes, but a standard that absolutely needs to be achieved doesn't prove that's why that choice is made, because it's not the only option. And if you want to sell a profit-driven peripheral, it helps to dress it up in some explanation like "dependable performance." I'm not saying that's the case, but it's a genuine possibility that needs to be factored in.

 

[Silent_Buddha]

There's an economic reason to restrict games to a proprietary drive - one can't rule that out.

or a baseline standard/restrictions that absolutely needs to be achieved and they had no other method to do so.

Or both. One doesn't preclude the other.

Personally, I think it's a combination of profitability (for MS and 3rd party making the drive), affordability (for consumers) and how easily you can guarantee minimum sustained performance for an external add on storage drive.

The last is key if you don't want a game to suddenly run erratically when a user installed an external drive to run their games on.

Regards,
SB

 

[PSman1700]

so it must run even hotter if pushed to it's 5.5 GB/s throughput.

We will have to see what the min-rate is for the SSD then (like with the var clocks). Maybe MS thought of this and used another hw block to mitigate for the lower ssd transfer rates.If the SSD doesn't have to work as hard but the end result is the same that might be a fine solution in the end.

Thinking on this, Sony will need to have more cooling power for their console whether the external drive is present there or not, on top of cooling that high clocked CPU and GPU. The reason being that consumer SSDs _will throttle_ even if Sony's doesn't. How does Sony guarantee a high level of sustained bandwidth to the graphics system, if they don't make that 3P SSD which could throttle?

Well, there were reports/rumors that sony had invested more then ever in a cooling solution. Seems many resources have gone into developing the cooling solution.

Normally I would just hand wave it off as well; but if the storage device is highly integrated with the graphics system;

It seems that both systems are designed around the SSD, better then a common/current PC does.

 

[Deleted member 13524]

Also the SOC cost difference can easily be made up for buy tempest or their custom ssd controller.

Custom flash controller and faster NAND might make up for the difference, but the Tempest SPU is inside the SoC and it should be pretty small since it's just a single RDNA2 CU without cache.

 

[MrFox]

Cooling nand chips is trivial here and doesn't do much to the BOM.

There's no such thing as a guaranteed bandwidth without mentioning the access block size and queue length. Any drive can be completely crippled with some agressive access patterns, or give the peak controller badwidth with a sequential pattern.

Sony is in an better position to maximize the guaranteed performance, since they have the priority queues necessary to do so. It should be telling that to have the same performance as their 5.5GB drive, even without any of the OS concurrent accesses, they will need a 6-7GB standard drive because the additional priority queues are that important for getting the same performance from an off-the-shelf controller.

 

[iroboto]

Cooling nand chips is trivial here and doesn't do much to the BOM.

There's no such thing as a guaranteed bandwidth without mentioning the access block size and queue length. Any drive can be completely crippled with some agressive access patterns, or give the peak controller badwidth with a sequential pattern.

Sony is in an better position to maximize the guaranteed performance, since they have the priority queues necessary to do so. It should be telling that to have the same performance as their 5.5GB drive, even without any of the OS concurrent accesses, they will need a 6-7GB standard drive because the additional priority queues are that important for getting the same performance from an off-the-shelf controller.

Block sizes and queue lengths are going to be set by both companies and going to be set for the best performance for their intended usage.
The bandwidth numbers provided at the best absolute impossible to obtain scenarios. The guarantee is against heat throttling.
I personally don't believe it's trivial. We won't know until we learn more though.

 

[disco_]

MS guaranteed this number. Meaning it won't throttle below this number. The number isn't impressive except to think out loud that they felt guranteeing this number for developers was important. That means they can depend on it working for 3 hrs straight, hard.

As I understand it, throttling on nvme drives occurs at north of 60C to protect the silicon. We do want it to last 5-7 years not 20 months.
And most consumer and even enthusiast nvme drives are not meant to be running capacity for 3 hrs straight, articles have them hit at 15 minutes of max performance before they need to throttle down.

I love the ideas and creativity there is out there for the fast SSD usage, but the usage of it in the way people are discussing about how it could be used over the course of a 1 - 3 hr play session makes the cooling system of PS5 and the claims of it's performance worthy of dissection. Mainly because they are accomplishing innovation on several fronts, not just bandwidth.

Sony isn't guaranteeing 5.5?
That number should be over 80-90*, not 60*. My evo+ throttled around 83* reported in AIDA. Reorienting one of my fans so that the edge of the air stream went across the nvme slot dropped temps to low 70*s.

 

[iroboto]

Sony isn't guaranteeing 5.5?
That number should be over 80-90*, not 60*. My evo+ throttled around 83* reported in AIDA. Reorienting one of my fans so that the edge of the air stream went across the nvme slot dropped temps to low 70*s.

Depends on your SSD controller when it wants to throttle. Some do it earlier. Some do it later. 83 is pretty high. Curious to see how long your drive will last. Keep it cool!

And your drive isn’t used as part of the graphics system. Not in the way we are discussing it being an extension of virtual memory which is my main concern. The 180 snap around load everything into memory -all- the time. The constant last minute loading all sorts of huge textures in and out of memory. These are wicked concepts. I just want to see the strain on the system now (for both)

Sony did not guarantee anything in their announcements.

 

[Globalisateur]

Depends on your SSD controller when it wants to throttle. Some do it earlier. Some do it later. 83 is pretty high. Curious to see how long your drive will last. Keep it cool!

And your drive isn’t used as part of the graphics system. Not in the way we are discussing it being an extension of virtual memory which is my main concern. The 180 snap around load everything into memory -all- the time. The constant last minute loading all sorts of huge textures in and out of memory. These are wicked concepts. I just want to see the strain on the system now (for both)

Sony did not guarantee anything in their announcements.

At least Sony disclosed typical values: 8-9 GB/s and not only theoretical max. Which should be enough. Nobody is going to guarantee anything in computing in any domain.

 

[PSman1700]

They didn't claim any 'at least' rate, they didn't state 5.5gb/s sustained/never throttle lower rate. Typical, theoretical, burst etc, no idea why MS didn't state anything like that officially. WindowsCentral did but that's not really official.

 

[MrFox]

I personally don't believe it's trivial. We won't know until we learn more though.

We can deduce how little heat is generated by looking at the teardown: MS is only using the EMI tab as a thermal connection for the expansion, this tab have a low thermal transfer capability, and it's still enough to guarantee it won't throttle... therefore cooling nand is trivial.

There's a number of watts over a certain surface area. Cooling low thermal density chips is easy and inexpensive. Cooling a dense SoC is difficult.

 

[iroboto]

We can deduce how little heat is generated by looking at the teardown: MS is only using the EMI tab as a thermal connection for the expansion, this tab have a low thermal transfer capability, and it's still enough to guarantee it won't throttle... therefore cooling nand is trivial.

There's a number of watts over a certain surface area. Cooling low thermal density chips is easy and inexpensive. Cooling a dense SoC is difficult.

No it goes right into the HSF unit

 

[MrFox]

No it goes right into the HSF unit

It only contacts with the EMI tab from the pictures. Is there more contact points?

Regardless it's a trivial problem to cool the internal chips, it's a few watts over a large area. This here is a way to avoid needing any heatsink on the expansion cart itself.

 

[iroboto]

It only contacts with the EMI tab from the pictures. Is there more contact points?

The EMI tab is slabbed against the entire aluminum plate which sucks air through through that main channel.

"""
PC NVMe SSDs often lose performance simply because they get too hot - and this required some innovative engineering for the new Xbox. "We have these competing set of springs that we call thermal bias springs," says Jim Wahl. "What that does is that it actually biases the card up against this top thick heatsink, so that the card is transferring heat through its connector into the chassis off to cooling air as it goes through the system... there's tons of engineering that just sort of sharpens the pencil and gets it right."
"""

This is the piece of metal it's pressed against

The heatsink is connected to this unit.

The holes in the metal are for air pass through, the external SDD is placed onto that metal slab and that metal slab is fed air through the rear holes when the pulls air through the holes pictured above. Diagram shows the air entering from 2pm through the back holes

 

[MrFox]

The EMI tab is slabbed against the entire aluminum plate which sucks air through through that main channel.

"""
PC NVMe SSDs often lose performance simply because they get too hot - and this required some innovative engineering for the new Xbox. "We have these competing set of springs that we call thermal bias springs," says Jim Wahl. "What that does is that it actually biases the card up against this top thick heatsink, so that the card is transferring heat through its connector into the chassis off to cooling air as it goes through the system... there's tons of engineering that just sort of sharpens the pencil and gets it right."
"""

They needed a heatsink on the card. They found another way to not need a heatsink and used the existing heatsink with tabs on the connector. It's not a complicated process. The internal flash would be straight against the heatsink regardless. Or a small one in the inlet airflow path. Or use the EMI aluminum plate to cool the chips like the ps4. To get back on topic about the BOM.

Ah nice! This is so much clearer!

So it's a cast aluminum frame like the PS3. I thought it was folded metal frame.

 

[iroboto]

They needed a heatsink on the card. They found another way to not need a heatsink and used the existing heatsink with tabs on the connector. It's not a complicated process. The internal flash would be straight against the heatsink regardless. Or a small one in the inlet airflow path. Or use the EMI aluminum plate to cool the chips like the ps4. To get back on topic about the BOM.

It’s also running at 1/2 speed the PS5 one is.
and they know exactly how to cool it. Sony can cover their own stuff too.

What will the external storage space for PS5 look like inside the console to ensure a 7GB/S nvme drive isn’t throttling?
Not saying they aren’t going to do it; but I don’t think any of this is trivial.

 

[iroboto]

At least Sony disclosed typical values: 8-9 GB/s and not only theoretical max. Which should be enough. Nobody is going to guarantee anything in computing in any domain.

If I recall correctly, All previous console generations have provided fixed values.

This will be the first time there will be more than 1 device other than the main CPU and GPU components that can throttle. So we are in new territory a bit here; with a possible increase in console failure rates for both. For XSX I don’t trust Seagate (LOL). And for PS5 I just want to see the box and get back some typical values before I assume any further.

And this is the first time a console have gone without fixed clocks. So I’m not sure what you mean by saying that nobody will guarantee anything. Lots of hardware has operating temperatures in they guarantee performance within a specific band of operating temps.[/QUOTE]

 

[MrFox]

It’s also running at 1/2 speed the PS5 one is.
and they know exactly how to cool it. Sony can cover their own stuff too.

What will the external storage space for PS5 look like inside the console to ensure a 7GB/S nvme drive isn’t throttling?
Not saying they aren’t going to do it; but I don’t think any of this is trivial.

It's also a 2280 instead of a 2242 style, considering they need at least 8 channels, so twice the surface area. I see two simple ways to deal with it.

Option 1 is to provide a tray in which to clip a 2280 (the way they always did it for the hdd on ps3, ps4), the tray can have a full length Sil-Pad to interface heatsink. Cost is minimal.

Option 2 is to provide a predetermined laminar airflow in a cavity where you put any 2280, and have an inch of clearance for whatever vendors use. Cost is also minimal.

Pass/Fail tests for available nvme drives.

 

[iroboto]

It's also a 2280 instead of a 2242 style, considering they need at least 8 channels, so twice the surface area. I see two simple ways to deal with it.

Option 1 is to provide a tray in which to clip a 2280 (the way they always did it for the hdd on ps3, ps4), the tray can have a full length Sil-Pad to interface heatsink. Cost is minimal.

Option 2 is to provide a predetermined laminar airflow in a cavity where you put any 2280, and have an inch of clearance for whatever vendors use. Cost is also minimal.

Pass/Fail tests for available nvme drives.

BOM is minimal (maybe), but the design will be intriguing. You must be curious how it’s all going to come together.

 

[MrFox]

BOM is minimal, but the design will be intriguing. You must be curious how it’s all going to come together.

Well yes of course, but not that part! The two options I mention I personally consider trivial, it's the main SoC cooling I'm very curious about, because it's really difficult to save costs there. The rest of the parts to cool are just going through the motions of using the flow available here and there, it just falls into place when they chose a starting point...