BadBreathOfTheWild
Member
You can disagree with as much heart as you want but this just isn't how it works, and is what I suspected was the misunderstanding.
It is is no way bad hardware design, it's how it works. First of all neither PS5 nor XSX has anything even approaching good cooling from the PC gaming side of things.
How many Watts you can pull through your chip is not dependent on your cooling but the firmware controlling that chip.
A good stable GPU overclock is one where the limiting factor is power usage at the clock speed you want it to be at. Not a thermal limit.
Secondly, the final fixed clock speed you decide on in something like a game console is entirely dictated on what you expect your worse case load to be, and you don't do that by assuming it will be running some synthetic benchmark tool.
On my PC I can have very different stable fixed clocks depending on whether I want to just basically boot at sit at the desktop (useless), play games, or post some benchmark result.
What can work all day for gaming can lock up in an instant with Linpack. A clock that can sit for hours on Linpack is lower than what I can safely use to run my games.
Lastly, the above is even more critical when talking about an APU sharing a die and heatsink for both CPU and GPU.
When asking "which one gets downclocked" it seems to be asked as if the same fundamental issue of where you spend your heat/power budget doesn't also apply to fixed clocks on a shared APU when running some nonsensical code loop that is designed to do nothing but run the chip hard and in a way they're not designed to do.
In the Eurogamer interview, Leadbetter couldn't wrap his head around it either and had how variable clocks in the PC work in his head when trying to ascertain the base-clock.
When Cerny realised what it is he was actually asking, he also explained what would happen if you ran the same kind of code it would take for the PS5 to appreciably down-clock on a PS4 with fixed clocks, and the answer that it would potentially crash.
Watts consumed is a function of clock-speed and the instructions being run. Not just clock-speed. Watts consumed is capped in the APUs firmware.
Watts consumed is also an indicator of work being done, regardless of clock-speed.
Variable clock and fixed power budget is about making the cooling requirements fully known.
If throughout the course of running expected game code the developer isn't having to choose which device gets priority because there isn't enough power being drawn then the question becomes a bit pointless.
The worse case scenario for power draw isn't some AAA scene with a load of enemies on screen where even a fixed-clock system might chug (as happens this generation), but a low triangle map screen where the GPU can race.
One of the most intense GPU benchmarks you can run is just a furry looking cube or torus. You can hear a GPU literally give off a very high pitch whine when it's rendering something extremely simple at an insane frame-rate.
Slightly off-topic but with the weather we're having here at the moment my brother was wondering what he could do to cool his room down a bit. He noticed that his GPUs weren't down-clocking and saving power, even when just in Windows and it was adding a lot of heat to the room (no AC here). It turns out it was just because they were set to 240Hz refresh rate for his monitors. Dropping it to 120Hz allowed them to go into a resting state and consume less power.
Does it really matter which part of the chip PS5 prioritises during a "worst case scenario" low triangle map screen, so long as it chose the right one based on which part of the APU is being stressed? Does it really matter in that kind of scene so long as it's repeatable and the same each time so that it can be tested?
If this needed to be done in typical game code or graphically complex scenes then it would be worth worrying about, but we've been told twice that isn't where down-clocking can occur in the current design.
Cerny could of course be lying, but it's hard to base an argument on that at the moment.
Almost entirely off-topic:
After getting back into PC gaming again this year my PC would randomly crash and reboot during games. Running a GPU benchmark was totally fine with the clock I was at, same for the CPU. They could run for hours without issue, but certain games could cause the whole thing to shut down.
Then I found running the CPU and GPU benchmark at the same time would do it.
The culprit was the PSU. The fix was unplugging the PCIe power connectors from the PSU side and plugging them back in again. Enough to knock off any oxidisation that had obviously increased resistance and caused voltage to drop to push the power out.
I had to shop around a few places in Osaka to find a preorder for a day one PS4 Pro. That mobile gaming is more popular there than console gaming is fair comment, but it's likely the same where you live, too.
The Switch is also extremely popular outside of Japan and if it continues to sell as it does will end up very successful.
To say they couldn't care less about console gaming is wrong.
Last edited:
The marketing team of Xbox Game Pass is pretty great, they always have fun stuff
