Current Generation Hardware Speculation with a Technical Spin [2022] [XBSX, PS5]

New Year, New Thread, Same sensible technical speculations.

Old Thread @ https://forum.beyond3d.com/threads/...chnical-spin-post-launch-2021-xbsx-ps5.62271/

 

Might be worth keeping an eye on, if he provides a little more information.

 

Cool. In the comments he mentions using a wifi power brick for the power draw info.

Edit: Also, he says 170W for the XSX (I guess versus the ~80W max I saw in the video).

 

Surprised at how good Cyberpunk 2077 looks on the XBS-S (LOD transitions are a bit rough, but still looks pretty good), especially with that power consumption.

Regards,
SB

 

First full comparison is up.

 

Can we distinguish between real and rendered versions of the actors by the power draw?

The PS5 is generally +25W over the XSX, though the average margin seems to tighten to roughly 10-15W when more stuff is happening. XSX seems to max out at 200W to the XSS's 80W, and generally stays in the +100-120W range under load. That seems like a fair difference for slightly higher clocks, more RAM, and 3x the GPU.

It'd be useful to know ambient temp and any extra console storage.

 

Interestingly some parts are consuming a bit less power on PS5. A few examples:
0:16 XSX 107w PS5 105W.
2:57 XSX 192W PS5 186W.
4:43 XSX 189W PS5 175W
4:50 XSX 189W PS5 185W

Max power consumption is actually very close between both at 6:58 XSX 219W vs PS5 227W.

 

I wish they would have included a FPS benchmark along side the wattage consumption.

 

Yup, but who knows if that’s the game or random OS or networking stuff. (Let’s add network traffic to see if it correlates to the anomalies. > ) The low frequency of the power data plus console variance and slightly out of sync scene timing (is the PS5 a frame or two ahead?) means I’m just looking for general clues, though a graph would be cool.

Just play DF’s video side by side.

 

Is there a DF video illustrating a full and complete loop of the demo with frame counter? The first video only covered a few sections/details on performance (second one by Alex, mostly SSD performance and imagery quality) nothing suitable for this purpose. Maybe I will have to dig into VG Tech's videos for such a comparison.

 

I think it could be caused by downclock of APU (CPU or GPU, don't remember) whenever possible when it's not impacting performance. Cerny talked about this.

 

Seems so. Power draws kicks up at 1:10 or one second before the rendered Reeves appears, steps down at 1:28 after the real Reeves appears in the mirror and revs right back up at 1:36 when the rendered Reeves reappears during the bullet time scene.