Digital Foundry Article Technical Discussion [2021]

Status
Not open for further replies.
But PS5 doesn't have RDNA1 CU's.....so it's a pointless comparison.
will not go into what ps5 cu looks like or not ;d this benchmark for me at least show that difference between rdna1 vs rdna2 rops is hard to notice in game benchmarks, but not saying its not impossible if somebody want to found such an example, could be interesting
 
will not go into what ps5 cu looks like or not ;d this benchmark for me at least show that difference between rdna1 vs rdna2 rops is hard to notice in game benchmarks, but not saying its not impossible if somebody want to found such an example, could be interesting

So a benchmark that's not been designed to test and show ROP performance shows the difference between RDNA vs RDNA2 ROP performance?

And then somehow said benchmark is being used to show the difference between two consoles that feature hybrid combinations of both RDNA1 and RDNA2?

Makes sense.
 
Lets see if I can recall correctly. They use RDNA2 rops just like the desktop GPUs, but are cut in half and then double-pumped. This was for some silicon savings.

Some folks like to color all their posts with hyperbole adjectives to describe technology from a company that is not the one they worship. So take all their posts on the matter with the appropriate levels of salt.

Its still something we all wanted to see improvements upon. Its not going to magically dissapear without quality moderation.
 
Does it require VRS, which isn't seeing widespread adoption yet?

The AMD RDNA2 RB+ are required for VRS hardware support. Without those you have to use software approaches for VRS.

AFAIK, the double-pumping happens regardless of VRS usage, however there are some scenarios it does not apply. As Iroboto already posted, depth calculations only happens at 1x rate.
 
And PS5 could have a 44%+ higher frame rate, it was very strange PS5 running at the resolution it did in Hitman as it clearly has the head room to go higher.
Math wise. 44% more pixels is significantly less work and bandwidth than 44% more frame rate.

So I would have hard doubts about this.
 
Math wise. 44% more pixels is significantly less work and bandwidth than 44% more frame rate.

So I would have hard doubts about this.

During the tall grass section PS5 holds 60fps and XSX drops hard.....in this section PS5 can be 44% faster......maybe higher if the cap was removed.

Frame rate is there to see in the DF video.
 
During the tall grass section PS5 holds 60fps and XSX drops hard.....in this section PS5 can be 44% faster......maybe higher if the cap was removed.

Frame rate is there to see in the DF video.
Sometimes. Bottlenecks do this. It’s not indicative of anything typically wrong. But I’m just saying 44% more frame rate is 44% more cpu, 44% more fixed function, 44% more compute and 44% more bandwidth. Resolution typically is only hit on some bandwidth and compute.

When you see XSX dip like this, it’s just 1 part of the XSX GPU that has run into a hard bottleneck, that’s not indicative of what PS5 is capable of or useful to determine PS5 performance.
 
Sometimes. Bottlenecks do this. It’s not indicative of anything typically wrong. But I’m just saying 44% more frame rate is 44% more cpu, 44% more fixed function, 44% more compute and 44% more bandwidth. Resolution typically is only hit on some bandwidth and compute.

When you see XSX dip like this, it’s just 1 part of the XSX GPU that has run into a hard bottleneck, that’s not indicative of what PS5 is capable of or useful to determine PS5 performance.
I wouldn't read to much into this. It is still a beta, anything can happen. But on the other hand, their games were never really technical masterpieces.
I find it really astonishing that they still don't integrate a 30fps cap from the beginning into the game. VRR makes things a bit better, but for all those that don't have a VRR TV a 30fps would make the game much smoother in quality mode.
I'm also missing a 40fps mode for 120Hz displays. More games should integrate something like this.
 
A difference of 44% is higher then any paper difference between the two machines (Largest is memory bandwidth at 25%)
It isn't a 44% difference in throughput, though. 1800p@60fps is still 8% less pixels per second than 2160p@45fps. I don't disagree that PS5 can likely go higher in FPS or resolution, but to act as though the framerate in Hitman indicates a performance advantage for either console without accounting for resolution is missing a big part of the equation.
 
AFAIK Sekiro run better on PlayStation than Xbox this game is based on same engine. There could be simply matter of better optimization on Sony machine.
I know I'm late saying this but it also runs better in part because it is pushing less pixels. Checkerboard rendering versus none checkerboard rendering. DF pointed this out when they did BC test on both platforms and how the pursuit of native 4k on Xbox One X kind of hurts BC performance on the Series consoles.
 
It isn't a 44% difference in throughput, though. 1800p@60fps is still 8% less pixels per second than 2160p@45fps. I don't disagree that PS5 can likely go higher in FPS or resolution, but to act as though the framerate in Hitman indicates a performance advantage for either console without accounting for resolution is missing a big part of the equation.
Down to ~41fps on XSX in another transparency heavy scene, so PS5 actually pushes even a bit more pixels (about 2%) here as a minimum and possibly much more (PS5 being locked 60fps). DF actually tried to compare both versions in a cutscene showing something like 15% performance edge on XSX using one frame, not 44%. They oddly used only one frame to compare both versions, but at least it was a valid comparison.

I maybe should not have used this game as an example but my point was more to compare how unstable the framerate was on XSX specifically in those ROPs limited scenes (compared to the others scenes on the same console). Some kind of performance 'unbalance' that was also very obvious in AC Valhalla (which they eventually 'patched' by significantly reducing the minimum resolution, down to 1080p if needed).

HrHeF6a.png
 
Im confused so layman talk - the XSX should be outperforming the PS5 by a noticeable margin due to paper specs (iirc 18%?). Surely there is something either quite wrong with XSX or something very beneficial for PS5 when we see any area outperforming in PS5? People arguing that ‘PS5 is only pushing 8% less work’ are missing the point no?

The argument around alpha effects and ROPs keeps popping up…maybe where there’s smoke there’s fire? The fact it’s been patched out of the XSX doesn’t mean it doesn’t exist, what gains could PS5 have seen with similar extra work? Some said that at the start of the gen it would be close and PS5 would win some face offs and as the gen went on the XSX would take the lead…and that seems to be the case where older engines favours the PS5.

Probably completely wrong…just a thought.
 
Tbh I don’t think much will change. From software make amazing games but they run kinda bad.
Maybe so, but people need to keep things in perspective.

Months away from launch.
What is the lead platform. This can have impact, even at, and after launch.

This is just a snapshot of where the game is right now for each platform.

It's ok to discuss possible hardware reasons, but in the end this is far from final code and don't even know if one version is further along than the other.
 
Status
Not open for further replies.
Back
Top