Define "not full 480". It's 2304 P10 cores, albeit at 911 Mhz.
Leak from March which was accurate on the GPU.
It's definitely Polaris (but not full 480); though not sure about the CPU frequency. Probably only guesstimates at this point for any numbers beyond 1.6
Also, where are people reading the CPU is 2.2 ghz? I still can't find anything official outside of speculation.
Define "not full 480". It's 2304 P10 cores, albeit at 911 Mhz.
This means that Sony can equally pick and choose which parts of the GPU they want to update, which is important given the forward compatibility mandate of the PS4 Pro. Adopting features from Polaris does not mean that this is a full Polaris GPU...
Beyond that, however, there’s nothing that definitively points to the PS4 Pro’s SoC integrating other, newer components from Polaris. An updated memory controller would make a lot of sense if the memory clockspeeds are higher, but whether the shader cores or geometry engine have been touched are both very good questions. The PS4 has previously been described as using a GCN 1.1-era shader core, so it’s not impossible that it simply uses a larger version of such core (as opposed to Polaris’s GCN 4) to ensure compatibility. Essentially Sony & AMD’s options run the gamut from a truly updated PS4 SoC to a Polaris 10 stuffed into an SoC, and depending on Sony’s technical needs, both options would be valid.
Essentially Sony & AMD’s options run the gamut from a truly updated PS4 SoC to a Polaris 10 stuffed into an SoC, and depending on Sony’s technical needs, both options would be valid.
I had that feeling too (that the Pro is at 28 nm) but it's probably just GloFo being GloFo and they have the voltage set very high to maximize yield. Actual gameplay is probably in the 200-230 range FTW. The Slim is 60-80.
The AT article is basically what made me start to think it isn't based on Polaris, let alone it isn't 16nm. I get the feeling PS4 Pro might still be 28nm (unless I missed something official from Sony).:
1. The GPU is a Polaris family architecture because Mark Cerny stated so during the conference.
2. Since it's a Polaris-based GPU, the SoC has to be either 14nm or 16nm, not 28nm. One of the major reasons why Sony and MS both waited 3 years to release slim consoles is because they waited for a shrink from 28nm to 14/16nm.
3. The original PS4's PSU is rated up to 250W, but the console only used ~ 150W of power.
"And that's where the new CUH-1200 model really comes into its own. According to Pocket News's metrics, standby power is anything from 30 to 50 per cent lower (depending on mode) compared to the launch unit, while the main menu is around 11 per cent more efficient. Perhaps not surprisingly, the biggest gain comes during gameplay, where the launch unit draws 148.6W, while CUH-1200 brings that down to 122W - that's an 18 per cent drop,"
There are various reasons why Sony would include a more powerful PSU than the maximum power draw of a console such as the PSU will run cooler and less stressed over the lifetime of the console as opposed to running at 80-90% of its power-rated capacity. The reasons the PS4 Pro is larger in size isn't because it's built on 28nm but because they integrated the 310W unit inside the console; and the cooling system is likely more robust than on the original PS4 to accommodate for an SoC that will use > 150W of power. However, you are drawing an incorrect conclusion that because the PSU is rated at 310W, that the system's maximum load is around 300-310W. Using the same logic, you would have claimed that the original PS4 would have used 200-250W of power?
"It was not possible to get an idea of acoustics in the PlayStation Meeting environment, but max power consumption is rated at 310W - we assume that this is the maximum load capable by the PSU, not a typical system draw,"
http://www.eurogamer.net/articles/digitalfoundry-2016-three-hours-with-playstation-4-pro
Sorry, but none of the points you made regarding the system's power usage, the tech process involved and the GPU architecture are logically thought out.
If it's 16nm it isn't GloFlo. Unless , you now think it's 14nm and done at GloFlo?
Yeah, the Pro seems to be GloFo 14FF; Slim TSMC 16FF.
It's already gimped by the lack of improvement on the CPU.
I'm of opinion this thing is based on GCN 1.0/1.1 and Hawaii - not Polaris/GCN 4.0. Would explain the power draw, and now seeing physical pictures of it's bigger than the PS4.
The AT article is basically what made me start to think it isn't based on Polaris, let alone it isn't 16nm. I get the feeling PS4 Pro might still be 28nm (unless I missed something official from Sony). The power consumption is almost double of PS4 Slim which is 16nm.
Might want to rewatch that presentation.
Sorry, but none of the points you made regarding Mark Cerny saying it is a Polaris based design is logically thought out.
Admit you are 100% wrong on all of the points you made. Sony's own slides list boosted clocks related to the CPU (because the first point was directed ONLY at the GPU). Of course you would have known this had you actually watched the presentation.
<snip>
There is a difference between having a reasonable opinion based on some research and data and simply stating a baseless opinion because you feel like it. This is a technical forum which means you should expect to be challenged on the points you've made, especially since none of them has any reasonable logic or data behind them. It's amazing you are now publicly telling me to rewatch the presentation, where Mark did state that PS4 Pro is based on Polaris, which means you never watched the presentation carefully or you simply didn't watch it at all.
"Inside, the key to the PS4 Pro is an upgraded graphics processor. Sony’s Mark Cerny says the PS4 Pro relies on AMD’s new Polaris GPU architecture" ~ Source
https://twitter.com/PlayStation/status/773600156356866048/photo/1
I guess at this point I'll just wait until someone rips it open. Because no one is reporting it is GloFlo 14nm.
why is that entire page in the preview dedicated to Polaris and talking about the custom Sony implementation of Polaris? It is neither a 480, 470, or 460, because it is an as-yet unseen AMD APU based on Polaris. That isn't really disputed that this point, from what I can tell. It seems to hover between 470-480 performance, but it is Polaris.
The power consumption is almost double of PS4 Slim which is 16nm. I find it hard to believe that a Polaris like chip @ 16nm under clocked with a probable 2.2 Ghz Jaguar (still not confirmed) is eating that much juice.
Especially when a higher clocked Intel + higher clocked 480X (with more cores) + higher clocked RAM don't even eat that much power.
It has a 100% bigger GPU clocked 12.5% higher, this would mean roughly 185% higher power usage (power scales roughly with the cube of frequency, when the necessary voltage increase is taken into account).
The CPU is clocked 37.5% higher, which again would mean a rough increase in power usage of 160%.
It is of course possible that the above numbers are too high if the PS4 Slim is overvolted in general, as then the power will not scale with the cube of frequency.
The max power usage has only gone up by 88% (from 165W to 310W), so there really isn't anything particularly unusual about the power usage of the PS4 Pro relative to the PS4 Slim.
You can't really compare like this since the number we have for the PS4 Pro isn't an average number, but rather a max number (i.e. akin to TDP). As RS mentioned above, the average usage of the PS4 was much lower than the rated 250W. If the PS4 Pro is lower by a similar amount, then it would only hit about 190W, or about 2/3 of the PC setup you linked.
Which scenario would use more power?
Game running @ 720p/30 or game running @ 1080p/60?
This thing is going to use much more power beside the increased GPU clocks/cores. If some of the claims they make are true, some of the older games would get a 60FPS patch, even if they don't get an IQ modification.
The PS4 power numbers differ between the two revisions. The latest was down even more to about 130W during gameplay. The Power example I gave for the PC system is not using a frame limiter where the games used to measure the power of the PS4 were limited to 30 FPS which would affect power consumption. Going back to my example of the PC system. Cap the FPS, lower the clock speed, disable some cores and reduce voltage (since you won't be needing it as high) and that thing comes down - alot. I've read enough people downvolting their Hawaii's to reduce power consumption by almost 100W in some scenarios.
Which do you think is more likely?
They ported Jaguar to 14nm or they ported Polaris to 16nm?
I guess the thing you're trying to get at here is that the load will be different, but relatively speaking that won't necessarily be the case. In other words the PS4 Slim running 720P/30 may very well be at the same percentage load as the PS4 Pro running 1080P/60.
But of course if the PS4 Pro runs a higher relative load (and thus use more power), that would just make the power numbers seem even more mundane, and thus all the more likely that it's on 14/16nm.
The fact that the power numbers differs between the two revisions is not really relevant, since I'm sure the 250W number for the PS4 was set for the original revision.
Bringing down the PC system by a lot (using the methods you described) is also what you would need to match the PS4 Pro (if it does indeed run at 190W), since you need to cut off roughly 110W (relative to the 190W number). So given that the RX 480 is already on 14nm, this just further supports that the PS4 Pro is on 14/16nm and not 28nm.
No idea, but it certainly isn't 28nm.
Polaris @911 MHz would mean something like 0.9v which results in ~60W power consumption. Just for the GPU alone.
I wouldn't be surprised if the SOC had more ROPs given the aim to push higher resolutions, but I wouldn't be surprised if it didn't either.
And something to keep in mind is that the PS4 Pro is meant to be backwards compatible: it needs to run existing PS4 binaries. The ISA for GCN1.2/GCN4 shader cores is different from the ISA for GCN1.1. For PCs this isn't an issue since everything is abstracted away by APIs, but for something like a console that is commonly programmed at a low level, it may be that Sony requires a GCN 1.1 GPU for perfect compatibility. This being the whole point of the semi-custom business: customers pick and choose the parts they want and need.EDIT; I'd just like to point out, it wasn't until I read AT's article that I even pondered perhaps this isn't a full out Polaris SOC based chip. Maybe, Sony to get this to market faster (and most likely cheaper) got AMD to Frankenstein them something.
And something to keep in mind is that the PS4 Pro is meant to be backwards compatible: it needs to run existing PS4 binaries. The ISA for GCN1.2/GCN4 shader cores is different from the ISA for GCN1.1. For PCs this isn't an issue since everything is abstracted away by APIs, but for something like a console that is commonly programmed at a low level, it may be that Sony requires a GCN 1.1 GPU for perfect compatibility. This being the whole point of the semi-custom business: customers pick and choose the parts they want and need.
Assuming it's on anything besides TSMC 28nm, AMD still would have needed to rework large parts to port them to the smaller nodes. So by no means would this mean they couldn't do their DVFS changes and other optimizations. They just wouldn't be able to change the underlying microarchitecture of the shader core.
(Personally, I've never understood the sometimes fervent position that the PS4 Pro must be pure Polaris. All the good bits of Polaris were outside the shader core anyhow, and the PS4 is clearly getting many of those. Otherwise you have GCN 1.1 on a smaller process, which is what Polaris's shader core essentially is anyhow)