Vita 2 / PS4 Go?

  • Thread starter Deleted member 13524
  • Start date
See Tottentranz's post for his breakdown, but he thinks that the PS4 SoC can get down to 6w at 7nm.

This gives you a PS4 capable tablet running at 11.5w in portable mode when you factor in the other components and HBM. That's doable in a slightly bigger form factor that the Switch.

Assuming TT's assumptions on 7nm are correct, my pondering was that if you can actually get PS4 SoC power consumption that low, then you'd be able have a Pro equivalent chip in there. You run 18CU's when portable and 36 when docked. It'd be well within laptop fan territory for cooling when docked.

I'm still struggling with TT's assumptions on power improvement on 7nm, but no one from the B3d hive mind has pushed back on that yet. :)
 
Last edited:
ToTTenTranz has a well reasoned breakdown but I think it's too optimistic.. my thoughts:

- Rocket League @ 63W doesn't seem anywhere close to a high load situation for PS4Slim. You can throw out the AC Unity title screen number because the disk is spinning (but I doubt title screens tend to be very full tilt either), but here are three other games measuring 80-90W: http://www.trustedreviews.com/playstation-4-review And at least one of these games is digital only, but I suspect in practice none of them spin the disc very much. These are peak values, but I think that's applicable for this exercise, so I would start with something like 85W.
- 80% AC/DC PSU efficiency is likely low. I've seen claims of 90% for PS4 (and presumably Slim wouldn't be worse) attributed to Sony, although I can't find the original source: http://www.psdevwiki.com/ps4/ADP-240AR This is at least consistent with the NRDC's report "They use very efficient power supplies to minimize energy losses in power conversion" for PS4/XB1: https://www.nrdc.org/experts/noah-h...one-game-consoles-mixed-bag-energy-efficiency The NRDC has recommended > 85% PSU efficiency under typical high load and recommended improvement for PS3 Slim's PSU (itself 80%), so I doubt they'd praise a PSU that only achieved that level.
- PS4Slim's HDD (or at least the one in the unit DF tested) is MQ01ABD050: http://wfcache.advantech.com/www/certified-peripherals/documents/96nd750g-st-to5k_Datasheet.pdf Idle power consumption is 0.55W and standby is 0.18W. This probably applies to the Rocket League numbers, although for the Trust Reviews numbers I'd include the full HDD power consumption of ~1.85W since they're peaks and probably got some HDD activity on them.
- The memory power consumption numbers seem reasonable and consistent with estimations I've seen from Anandtech - with the caveat that it's under full load (may not apply to Rocket League). I'd feel comfortable using it with the 85W numbers.
- So we're looking at (85 * 0.85) - 16.5 - 1.85 = ~54W after 85% power supply loss, 16.5W memory, and 1.85W HDD. Blu-ray takes a bit even when idle, I think maybe around 1W, and there are some other components in there, I'd guess a good estimate for the SoC chipset numbers would be around 45-50W including DC power regulation losses. Or something a lot higher than ToTTenTranz's 30W SoC number.
- Replacing 8x Jaguar @ 1.6GHz with 4x Zen @ 1.6GHz could be plausible, although there might be some compatibility challenges. I'm not sure the scaling would work out well. According to The Stilt's measurements on AT (https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/), Zen is hitting Vmin around 2GHz, which is reflected in the efficiency knee it hits around 35W. According to a later post in the thread (https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/page-6#post-38774364) the clock speed is estimated to be around 1900MHz at 35W for 8 cores. So I'd guess one CCX at 1.6GHz would likely consume at least 12W, possibly significantly more. Even with improvements from Zen 2 and a double shrink to 7nm I have a hard time seeing this number get much below 6W, when it really needs to be below 2.5W, maybe 3W at the highest. A more aggressively optimized 8x Jaguar may actually use less power, but I doubt AMD really wants to work on that.

I just don't see a 6W PS4-level SoC as that viable. And even if it is I don't think a Sony tablet that's significantly larger than Switch would be too popular.

EDIT: If Sony ever really does want to go with something like this I'd think they'd be best off eating the compatibility hit of moving to ARM CPUs.
 
Last edited:
Exophase said:
So we're looking at (85 * 0.85) - 16.5 - 1.85 = ~54W after 85% power supply loss, 16.5W memory, and 1.85W HDD. Blu-ray takes a bit even when idle, I think maybe around 1W, and there are some other components in there, I'd guess a good estimate for the SoC chipset numbers would be around 45-50W including DC power regulation losses. Or something a lot higher than ToTTenTranz's 30W SoC number.
Click to expand...
DC Power regulation on ULP devices is on a whole other level and I think I was pretty conservative in that area.
We're also not considering the very probable fact that there was close to zero effort on Sony's engineering teams to apply any kind of power saving measures during the shrink from the CH1000 SoC to the CH1200. There seems to be absolutely nothing new in that chip apart from the move to 16FF. No new video codec, no higher clocks that could've come for free (and usable in boost mode), no display update to HDMI 2.0, etc.

Regardless, even if you think the SoC is spending 45W, that's 50% more, so in the end instead of having a 6W SoC we'd have a 9W one. I don't really see that as a dealbreaker. Instead of a 8" tablet it's a 9" one. Or it's 3mm thicker and weighs 450g instead of 400g.

Exophase said:
Replacing 8x Jaguar @ 1.6GHz with 4x Zen @ 1.6GHz could be plausible, although there might be some compatibility challenges. I'm not sure the scaling would work out well.
Click to expand...
I'm not sure it would work out well either. Microsoft would be in much better position to achieve this because of the Xbone's virtualized environment.
But even if some (non-VR) games showed a tangential reduction in framerate, or games with dynamic resolution showed somewhat lower detail in some areas, the sheer value of being able to play PS4 games on the (actual) go would trump all that, IMO.

Exophase said:
According to The Stilt's measurements on AT (https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/), Zen is hitting Vmin around 2GHz, which is reflected in the efficiency knee it hits around 35W. According to a later post in the thread (https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/page-6#post-38774364) the clock speed is estimated to be around 1900MHz at 35W for 8 cores. So I'd guess one CCX at 1.6GHz would likely consume at least 12W, possibly significantly more. Even with improvements from Zen 2 and a double shrink to 7nm I have a hard time seeing this number get much below 6W, when it really needs to be below 2.5W, maybe 3W at the highest. A more aggressively optimized 8x Jaguar may actually use less power, but I doubt AMD really wants to work on that.
Click to expand...
You're forgetting there's a Raven Ridge APU with 1 CCX + 12 CUs beginning at 4W, coming this year on 16FF.
Even if the 4W version only has half a CCX enabled and 4 CUs, we're still looking at much lower power consumption values than the 12W-per-1.6GHz-CCX you're suggesting.
Obviously, desktop Summit Ridge's power-performance curves are not the same that will be applied in mobile Raven Ridge. You can't take Ryzen 7's curves and linearly scale them down to predict the performance on mobile SoCs.
 
Is that 4W 2017 Raven Ridge APU figure coming from anything more definitive than that slide leaked in August 2016? I'll believe it when I see it (and when power is measured)
 
Lisa Su has already confirmed there are design wins for 2in1 devices coming out in 2017.
 
ToTTenTranz said:
Lisa Su has already confirmed there are design wins for 2in1 devices coming out in 2017.
Click to expand...

If you're okay with a 9" tablet using a 9W SoC why does a 2-in-1 device need a 4W one? Is there any indication that that is what these design wins are using? Last year's Envy x360 is called a 2-in-1. It has a 15.6" screen and uses a 15W Bristol Ridge APU.

Not going to say you can't make a 4W Zen APU but I doubt it'll be 1.6GHz base even with just two cores.

Intel has 2C4T 4.5W SoCs, but:

- The base clock is right now at most 1.2GHz for Broadwell, 900MHz for Skylake and 1GHz for Kabylake
- Base clock only holds when GPU isn't very heavily used, you can't get dual CPU + GPU base clock at the same time
- Intel's 14nm CPUs attain lower Vmin than Zen so there's more potential to scale them lower
- The Y-series/m3 are made with a different physical layout and process with even lower Vmin and lower power consumption at the expense of peak clock speed
- When it comes to multiple SKUs the best parts, like the 1.2GHz Core M, are likely binned which is a luxury that a custom console chip won't have

Now the 4W Zen APUs could be made with a different physical layout and process than the higher power variants but this would be quite a bold shift for AMD who has used the same dies in as many places as they can. It's a really big investment for a market segment that is really hard for them to break into.
 
Last edited:
This article from a recent TCMS update briefing states that 16ff+ to 1st gen 7nm will be a 60% power reduction, so nice to remove the need to stack assumptions on the process power reductions at least.

I do feel that Sony has to pursue PS4 compatibility for their next portable. It carves out an immediate market for itself Vs the Switch and regular tablets.

A dedicated portable that's somewhere between last gen and PS4 will be in a worse position than Vita. Third parties will 'wait and see' before supporting and Sony can't divided their first party support in a significant way.

It'd also argue that Sony's first parties enhance the value proposition for Playstation but aren't system sellers in their own right. The masses buy PS/Xbox for third party AAA.
 
cheapchips said:
I do feel that Sony has to pursue PS4 compatibility for their next portable. It carves out an immediate market for itself Vs the Switch and regular tablets.

A dedicated portable that's somewhere between last gen and PS4 will be in a worse position than Vita. Third parties will 'wait and see' before supporting and Sony can't divided their first party support in a significant way.
Click to expand...

They wouldn't need to wait and see if such a portable's games could be played on the PS4: there would immediately be an install base in the 10's of millions.

A portable PS4 is a pretty sexy idea, but it seems fairly unfeasible any time soon. By the time AMD are at 7nm, the PS5 should be Sony's main focus.

I'm of the opinion that late next year is the latest point at which they should launch a new portable: the PS4Pro and PSVR will be well established and relatively cheap, yet the PS5 would still be a while away.
 
Please actually include a line explaining a video. It's some random person suggesting there's a next gen PS4, possibly portable. So vague it's not worth mentioning. It's also OT for this thread if about a portable. This is a technical discussion thread so please keep out totally baseless rumour and anything non-technical.
 
Is there an official news and rumors thread for this sort of things? I think we lost any remaining catch-all sony thread. :-?

(btw, don't bother with the video, it's 8 minutes of random guesses with no substance)
 
Even if PS5 is ready to launch by the time Sony is able to make a PS4Portable, provided the portable console is fully compatible with all existing PS4 content, then it could still be worth doing, since you'd be launching a portable with literally hundreds of games available at launch.

Even to existing PS4 owners it would be a rather attractive prospect. Then if you factor in the fact that most AAA third party games at the PS5's launch will be cross-gen (for business reasons), it means that PS4Portables will have access (at least for a while) to the latest big third party games on their shiny new portable console.

In terms of software, it would launch in a better position that the Switch launched today.
 
Using the same PS4PRO APU (but the GPU downclocked to just 300 MHZ and CPUs to 1,6 Ghz... maybe with 4 GDDR5 chips of 2 Gbytes, 256 Giga SSD) should be quite easy to build something like a PS4 Go (720p) with a quite durable battery... Developers have to patch the PS4 games... but this should be a quite easy task. SIlicon is there already... or maybe at 10nm even better.... better chips you use for the PS4pro worst chips for the PS4 GO. An even better job would be using HMB2.. but costs will lift up a lot.
 
HBRU said:
Using the same PS4PRO APU (but the GPU downclocked to just 300 MHZ and CPUs to 1,6 Ghz... maybe with 4 GDDR5 chips of 2 Gbytes, 256 Giga SSD) should be quite easy to build something like a PS4 Go (720p) with a quite durable battery... Developers have to patch the PS4 games... but this should be a quite easy task. SIlicon is there already... or maybe at 10nm even better.... better chips you use for the PS4pro worst chips for the PS4 GO. An even better job would be using HMB2.. but costs will lift up a lot.
Click to expand...

If devs have to patch the games the platform won't get any significant momentum.
It's either a portable PS4 in raw specs or they might as well just make a portable console using GPU/CPU/RAM architectures more appropriate for ULP usage.
 
PS4 Slim uses 86 watts of power during games [~10W less when disc drive is not in use]. Even if we discount the need for complicated motherboard with so many GDDR5 chips, I really don't think that even shrink to 7nm would help it achieve power window for portability.

It ain't happening.

The best that can happen is 7nm super slim PS4 that would drive the price of the console down, with or without BD drive [digital only game access] and/or HDD drive [flash storage fused to the mobo like on the latest PS3]... The question remains if Sony will be willing to invest in 7nm project, or they will achieve lower BOM by yanking BD/HDD.
 
You must log in or register to reply here.
Back
Top