By using this site, you agree to our Privacy Policy and our Terms of Use. Close
potato_hamster on 14 May 2018
Edit Thread
Pemalite said:
potato_hamster said:

First, I love how my post was about how the technology isn't there and then in response, you go about talking about hypothetical chips that do not exist that supposedly proves it does.
Consoles typically use semi-custom designs, Thus the "Hypothetical" chips generally don't exist until they hit the market anyway.
There is only one exception that comes to mind though... The Switch.


Sure, consoles use semi-custom designs. The key there being semi-custom, you know, meaning a customized version of something that actually exists. Except in this case it would be a semi-custom design of something that doesn't yet exist that people are assuming exists secretly, or will exist within the next 6-18 months or so, and be at a price point comparative to current APUs. Kev here is acting like there's a decent chance Sony already has this all nailed down and could be announcing it at E3.

Pemalite said:

potato_hamster said:

Increasing performance increases power draw.
Not always. Back in the Dothan/Bonais days... Intel was rapidly increasing chip performance to save on power draw... As the mantra of "Hurry up and finish and go back to idle" was a key point in chip design philosophy... In short the faster the chip, the faster it can get the processing task done and the faster the chip can go to idle, switch to a lower power state and save on power.


Well wonderful that you found an exception to a broadly established rule of thumb per se. Unless you're advocating that it would make any sense to design a PS4 mobile APU that is design to spend as much time idling as possible, please make your case, and if not, why bother bringing it up?

Pemalite said:

potato_hamster said:

You're acting like Sony can just take something like the 2300U, increase a little of this, decrease a little that, and blammo! PS4 APU with a 12W TDP, and about the same price as the PS4's current APU. Sorry, it's not that simple. Decreasing power draw decreases performance. You might be able to take a more... bleeding edge approach to the design, the that will almost undoubtedly decrease yields and blow up the cost of the chip. Working chips pay for dead ones. That's how it works.
The 2300U is a baseline of what to expect with AMD's current technology on the current fabrication process with it's current chip designs.
Obviously a semi-custom design will be tweaked to meet various goals.

There would not for instance be a need to have the CPU's operate at such a high clock and thus voltage (Remember, voltage has a direct relationship with power consumption!)  as Zen is significantly more efficient than Jaguar. I mean significant.


You keep talking like it's just arbitary to effectively emulate something closely resembling the PS4's APU on something like a 2300U, with the appropriate TDP and cost, do you have an examples of something similar being done using similar technologies, or is this just something that you think hypothetically should be easily manageable?

Pemalite said:

potato_hamster said:

This isn't rocket science. Yet you're advocating for a higher performance, lower TDP chip, presumably for the same price seeing as you think Sony can make this PS4 portable for less than $400 even accounting for the fact that any memory and storage solution would be undoubtedly higher than what's in the PS4 and the fact that it would have to have a 1080p screen and a battery. It doesn't add up.
* It wouldn't have to have a 1080P screen. Super Sampling is a thing you know.

The Playstation 4 SoC is based on Graphics Core Next 1.0, which is slow, old, power hungry and inefficient.
Vega is leagues ahead of it in every scenario. (And probably more so once AMD sorts out it's Primitive Shaders and Draw Stream Binning Rasterizer bullshit.)


I mean, it kinda would have to have a 1080p screen, maybe not in the the "they could make it work on a 720p screen" but definitely in the "how in the fuck are we going to market device for a target audience that definitely won't accept a 720p screen on a handheld  in 2019 for $400.?"

Pemalite said:

potato_hamster said:

Furthermore, where do you think I to supposed to logically land going from 14 nm to 7 nm? What are you basing that off of? If you're basing it on previous advancements, such as say going from 20 nm to 14 nm, is it still reasonable to apply here? Do we know enough about this new 7nm fabrication process to be able to predict what kind of TDP improvement we should get? Will that 7nm process be stable enough to produce cost-effective yields in the next year or so?
Early reports are placing the 7nm process at a 2.6x improvement over 14nm.

TSMC was pegging their 7nm process to be a 1.63x improvement over their 10nm process, which in turn had a 14nm BEOL.
TSMC's 16nm process had a 20nm BEOL.

You need to keep in mind that these aren't actually 7nm and 14nm fabrication processes though, they have been bastardized into advertising numbers to trick the less educated.

But the improvement from TSMC's 16nm to 7nm is bloody massive. Global Foundries, Samsung and so on are seeing similar improvements.
The jump to "7nm" is probably going to be one of the most significant fabrication improvements we have seen in a long time.


https://www.anandtech.com/show/12677/tsmc-kicks-off-volume-production-of-7nm-chips

"TSMC’s CLN7FF process technology will rely on deep ultraviolet (DUV) lithography with argon fluoride (ArF) excimer lasers operating on a 193 nm wavelength. As a result, the world’s largest contract maker of semiconductors will be able to use existing manufacturing tools to make 7 nm chips. Meanwhile, to keep using DUV lithography the company and its customers have to use multipatterning (triple and quadruple patterning), which increases design and production costs as well as product cycles."

If I'm reading this correctly (and forgive me as I am not anywhere close to an expert at these manufacturing processes) it appears to me that TSMC's 7nm process might actually be more expensive than previous fabrication processes? Doesn't that fly in the face of previous claims about how much you can expect such an APU based on this technology to cost? Let's also keep in mind that production of these chips is only getting underway right now, and they plan on ramping up production of different 7nm processes oeer the next year or so. If Sony is going to produce a PS4 portable, the window for releasing it more than likely shuts firmly as soon as the PS5 releases, so unless these new hypothetical 7nm AMD 2300 APUs, and/or PS4 portable semi-custom APU is one of the 18 customer products being manufactured in 2018, or having it manufactured by one of the other producers ASAP, this hypothetical handheld would most likely be released in 2019-2020, which falls squarely in PS5 territory. The Vita experienced very similar issues launching a year before the PS4. People didn't really see the appeal of paying $100 less than a PS4 to play dated-looking ports of PS3 games. Wouldn't a PS4 portable have to endure similar challenges?

Pemalite said:

potato_hamster said:

What kind of modifications (if any) can  AMD make to the design of the chip to make it power efficient than their current offeringgs?
Is this hypothetical? Or do you want a list? I can provide a list.


Considering AMD doesn't currently have any 7nm APUs they're offering to the public, it's going to be pretty hard to improve on something that doesn't exist. But if you think you're up for improving an APU that might not currently exist and whose design definitely isn't public, please go ahead. I'm very curious how you would improve something you can't possibly know enough about. Why don't you tell me how you'd simultaneously improve the fuel economy and  BHP of the 2019 Toyota Supra while you're at it.

Last edited by potato_hamster - on 14 May 2018