Digital Foundry Article Technical Discussion [2021]

Status
Not open for further replies.
It’s always entertaining watching forum dwellers debate over whether the people making these pieces of hardware - where hundreds of millions have been invested in over many years - know how to do their job.
The wounds are still fresh from my vacuum cleaner level ps4 pro fan noise. I also had a launch ps3 crap out. And let’s not even talk about the 5 times Microsoft had to replace my Xbox 360. Microsoft even gave me a bunch of games and a new controller for the issues, but the trust was already broken. All these devices are well maintained and in open ventilation.
I’ve owned every PlayStation model, even the revisions (with the exception of the ps2 slim). I have all of the original controllers. All of them have lasted with the exception of the ps4 controllers. I’ve replaced my ps4 controllers twice due to drift and a broken trigger. If anything, Sony’s track record is getting worse with my experience.
 
I’ve owned every PlayStation model, even the revisions (with the exception of the ps2 slim). I have all of the original controllers. All of them have lasted with the exception of the ps4 controllers. I’ve replaced my ps4 controllers twice due to drift and a broken trigger. If anything, Sony’s track record is getting worse with my experience.
My original PS1 had problem reading disks which got fixed and worked like a charm
My PS2 had problems reading disks properly so I readjusted the lens which worked for DVD games but stopped reading CD games.
I had a drift/analogue issue with one of my 4 PS3 controllers :p
None with my PS4 controllers which I ve played the crap out of
I had an original PS3 Fat die from the YLOD which got fixed and later sounded like this while idle on the main menu. Never used the fat again after that.
.

Seems like a mixed bag
 
Last edited:
Shocking that Sony engineers know how hat they are doing? ;)
No not really, I made the guess using some math that a 5 degree exhaust likely equated to a 5 degree increase at the SoC. I got that number, which means the new cooler cools as well as the old one while being cheaper. That's a fairly reasonable target for a company to aim for and in hind sight when looking to reduce costs, that should have been the general message. I think the idea that it would be a further 10 degree cooler at the SoC while exhibiting only 5 degree more exhaust was too optimistic.
 
The wounds are still fresh from my vacuum cleaner level ps4 pro fan noise. I also had a launch ps3 crap out. And let’s not even talk about the 5 times Microsoft had to replace my Xbox 360. Microsoft even gave me a bunch of games and a new controller for the issues, but the trust was already broken. All these devices are well maintained and in open ventilation.
I’ve owned every PlayStation model, even the revisions (with the exception of the ps2 slim). I have all of the original controllers. All of them have lasted with the exception of the ps4 controllers. I’ve replaced my ps4 controllers twice due to drift and a broken trigger. If anything, Sony’s track record is getting worse with my experience.
I think the difference here is that Sony specifically stated they had worked to improve the noise levels of the PS4 & pro. I can't say I have had many issues with consoles, generally been quite lucky - I think PS360 largely suffered due to less lead in the solder so I suppose you can cut a little slack...but other than that noise has never really bothered me too much and generally no issues with controllers (one or two with controller drift, certainly on XBO). The more tech you put in the more can do wrong...I've got a filtered cabinet now so that will help with nose levels.
 
I think the difference here is that Sony specifically stated they had worked to improve the noise levels of the PS4 & pro. I can't say I have had many issues with consoles, generally been quite lucky - I think PS360 largely suffered due to less lead in the solder so I suppose you can cut a little slack...but other than that noise has never really bothered me too much and generally no issues with controllers (one or two with controller drift, certainly on XBO). The more tech you put in the more can do wrong...I've got a filtered cabinet now so that will help with nose levels.
They already said something similar for the PS4 ;)
The problem with PS4 & PS4 Pro consoles were not so much the noise when the console is new, it is the noise the console makes after a few month of usage. E.g. I bought a used PS4 Pro. It was really, really loud. I completely disassembled the console to get all the dust out of it. Replaced the non working cooling-pads on the memory modules, ....
After all this, the console was quitter, until I realized that the improved cooling (better thermal pads and paste) wasn't really good for the console at all. Now the PSU in this thing get's really hot because the chip is not so hot that the fan would spin up. This means there is much less air-flow threw the PSU which than increases the heat in the PSU. That again seems to lead to the coil-whine problem that disappears once I play a game that stresses the GPU so the fan-speed increases and the much higher airflow cools down the PSU and the coil-whine stops.

And yes I have my problems whenever a more cost-effective solution is used in a new revision. Yes, the cooler of the initial PS5 might be a little bit bigger than needed and the temperatures of the new revision weren't that much higher that I would really worry about. But if I had the choice, I would always try to get the old model but as the console is still not available anywhere ... I have no choice ;). But for my taste the PS5 is just to big, so I wait for a Slim version (and for games).
 

An amazing amount of information crammed into that video. Spectacular work as ever. The performance summary seems to be that without RT the PS5 is between a 2060S and 2070S, and with RT it's slower than a 2060S. So largely in line with expectations. Looks like the beta patch is a must for a good experience on PC though. At least outside of 60fps.
 
I'm surprised that texture quality on PS5 isn't highest setting.
Well, typically its because SSD is not responsible for rendering. Bandwidth is still a requirement to put something onto the screen, so it may largely come down to how well they handle texture streaming which can affect how large the texture pools are as well as how much bandwidth is needed. Then we get into resolution, AF and other things that can greatly eat up bandwidth and having highest quality textures may be out of the question.
 
I'm surprised that texture quality on PS5 isn't highest setting.

The texture quality setting of very high evidently is higher than the amount of memory available for caching textures on PS5. The high SSD I/O speed probably just negates or greatly reduces the performance penalty for streaming in textures as it appears there's texture pop-in similar to PC when the texture quality setting is higher than available VRAM.

And while the I/O allows you to load textures in really fast, you likely don't want to be reading, writing, reading, writing, reading writing, to a texture on SSD hundreds of times a second for rendering purposes so VRAM quantity can still be limiting factor if you have too many assets of a given size.

I wonder if a lower texture quality setting would have been better in order to avoid texture pop in?

As the generation goes on, I'm imagine that developers might get a better grasp of what is the right balance for texture streaming and VRAM used for rendering work on those textures.

Regards,
SB
 
The texture quality setting of very high evidently is higher than the amount of memory available for caching textures on PS5. The high SSD I/O speed probably just negates or greatly reduces the performance penalty for streaming in textures as it appears there's texture pop-in similar to PC when the texture quality setting is higher than available VRAM.

And while the I/O allows you to load textures in really fast, you likely don't want to be reading, writing, reading, writing, reading writing, to a texture on SSD hundreds of times a second for rendering purposes so VRAM quantity can still be limiting factor if you have too many assets of a given size.

I wonder if a lower texture quality setting would have been better in order to avoid texture pop in?

As the generation goes on, I'm imagine that developers might get a better grasp of what is the right balance for texture streaming and VRAM used for rendering work on those textures.

Regards,
SB
I can’t imagine it being a cache size issue. @Dictator by any chance do you know the VRAM amount for each level of texture quality ?
IMO, larger textures should slam bandwidth more than being a cache size issue. Are we sure a 3070/3080 (8GB/10GB) can’t run highest quality textures for instance ? Because if they can’t it’s a VRAM deficiency. If the 3080 can it’s a bandwidth deficiency, if both 3070 and 3080 can, it’s a processing deficiency somehow on PS5 since 3070 has the same bandwidth as PS5 and less VRAM.
 
The ps5 uses the highest res texture up close as shown in the video, it is just that due to streaming pressure the highest res texture are not always visible (as shown in the Video).
It makes sense - for texture to be shown, they have to be in Main RAM - streaming speed or whatever be damned.
That was noted as somewhat of an issue on PS4 specifically. I wonder what can be done to lessen that problem?
It happens with all APUs. Both CPU and GPU want to access the Memory -That means the GPU, a bandwidth sensitive component, has less than it might otherwise have had if it had its own dedicated Memory.
@iroboto
The amount of vram the game "needs" cannot really be measured, as it Caches as much as it can greedily and we have no insight on cached amount vs active amount in any given scene or view change. But the evidence points that the highest texture at 1440p with no RT require more than 8GB on PC and more than however much the PS5 has available.
 
Last edited:
Thanks for clarifications and information Alex. PS5 is very impressive for a 500 dollar machine but it's still clearly limited in certain ways especially compared to PC.

Of course as time goes on surely devs will be able to optimize more refined results out of it as well
 
An amazing amount of information crammed into that video. Spectacular work as ever. The performance summary seems to be that without RT the PS5 is between a 2060S and 2070S, and with RT it's slower than a 2060S. So largely in line with expectations. Looks like the beta patch is a must for a good experience on PC though. At least outside of 60fps.
5700xt level perf + not so great rdna2 rt implementation, nothing new, almost 100% inline with paper specification ;)
 
Status
Not open for further replies.
Back
Top