Digital Foundry Article Technical Discussion [2020]

[BRiT]

I wonder what the cap on download speeds is about? Server resources? Could they easily have it at 1Gbps?

I wondered that as well. I am thinking the puny Jaguar cores may be limiting it on the OneX. I assume there are crc checks done with the downloads that may be done differently or not at all with file copies.

Curious when or if digital downloads will be part of the preview program for SeriesX.

 

[iroboto]

It only takes however long it takes to download. The past week I've been seeing 250 Mbps to 400 Mbps on OneX game updates even for large 20 gig patches.

Or do you mean for physical games? Well good news, you can now pre-install games on the Xbox Beta App on Android,, so they download with patches before you even get the physical media. After that point, the disc only acts as a license entitlement.

What likely caused installs to go slow in the past was the game was actually downloading patches at install time. One hack was mentioned to install physical games offline to improve that portion. Then when you go online the patches would be downloaded and installed.

Hmm was that the issue? It had nothing to do with just really slow installing ?

 

[BRiT]

Hmm was that the issue? It had nothing to do with just really slow installing ?

I know they changed the display on how games install a couple times over the past 4 years or so, in the insider program.

I hope they now display it as 2 portions, installing from disc and then downloading patches. AFAIK, the trouble is they don't really have a download then install for games, when its downloading its directly into the resulting game image. So you can't do the 2 steps in parallel. Only the firmware/dashboard updates have separate step process.

I'll try to find a physical disc to give it a try to see what a OneX handles it. Nearly 99.9% of all my physical discs are OG Xbox or X360 titles, but I may have one or two One titles still in shrink-wrap.

 

[BRiT]

I had to open up my mint in shrink-wrap Assassin Creed Unity title for this experiment. I hope you're happy.

I placed it in the OneX drive, after a few moments it popped up a display

"Assassin Creed Unity needs an update. The size of the update is 6.4 GB (connect time charges may apply). You'll need it to play on this console.​

[Update Now] [Update Later]."​

I left the prompt up, and after a while the Toaster popped up showing "Ready to Start".

Backing out of the Update Prompt shows the following information from Manage Queue screen

"Ready to start. Completes in about 19 minutes 11 seconds"
"Disc: 13.85 GB of 38.58 GB. (36%)"
​

This entire time the disc is spinning and I hear drive seek noises, but can't tell if it's Optical or Hard Drive.

I cancelled the install, ejected, then reinsert the disc, answered Update Now. It displays same Disc line stats but also had more info shown with the following:

"220 Mbps"
"Ready to start. Completes in about 11 minutes 30 seconds"
"Disc: 5.07GB of 33.93 GB (25%)"
"Network: 4.06 GB of 6.38 GB"
​

After the network part finished, the time kept increasing as it installed from disc (and dont know why download size changed) :

"Ready to start. Completes in about 13 minutes 45 seconds"
"Disc: 10.68 GB of 33.96 GB. (42%)"
"Network: 6.34 GB of 6.34 GB"
​

Doing a pure Digital install at only 7% installed so far: 250 Mbps, Completes in 22 minutes.
At 50%: 303 Mbps, 20.01 GB of 40.31 GB, completes in 10 minutes 39 seconds.


So it seems network makes the install faster.

 

[iroboto]

I had to open up my mint in shrink-wrap Assassin Creed Unity title for this experiment. I hope you're happy.

I placed it in the OneX drive, after a few moments it popped up a display

"Assassin Creed Unity needs an update. The size of the update is 6.4 GB (connect time charges may apply). You'll need it to play on this console.​

[Update Now] [Update Later]."​

I left the prompt up, and after a while the Toaster popped up showing "Ready to Start".

Backing out of the Update Prompt shows the following information from Manage Queue screen

"Ready to start. Completes in about 19 minutes 11 seconds"
"Disc: 13.85 GB of 38.58 GB. (36%)"
​

This entire time the disc is spinning and I hear drive seek noises, but can't tell if it's Optical or Hard Drive.

I cancelled the install, ejected, then reinsert the disc, answered Update Now. It displays same Disc line stats but also had more info shown with the following:

"220 Mbps"
"Ready to start. Completes in about 11 minutes 30 seconds"
"Disc: 5.07GB of 33.93 GB (25%)"
"Network: 4.06 GB of 6.38 GB"
​

After the network part finished, the time kept increasing as it installed from disc (and dont know why download size changed) :

"Ready to start. Completes in about 13 minutes 45 seconds"
"Disc: 10.68 GB of 33.96 GB. (42%)"
"Network: 6.34 GB of 6.34 GB"
​

Doing a pure Digital install at only 7% installed so far: 250 Mbps, Completes in 22 minutes.
At 50%: 303 Mbps, 20.01 GB of 40.31 GB, completes in 10 minutes 39 seconds.


So it seems network makes the install faster.

lol. I think it's way too slow, but it might just be transferring material to HDD say 90GB of it, is still going to be a slow process any way around it. How fast is a blu ray drive?

 

[BRiT]

lol. I think it's way too slow, but it might just be transferring material to HDD say 90GB of it, is still going to be a slow process any way around it. How fast is a blu ray drive?

The disc when I ejected it was very hot, maybe it was just warm, don't remember how hot the DVDs were from X360 games.

BluRay movie playback requires 2x speed, where 1x is 54 Mbps. So 2x would be 108 Mbps, 4x at 216 Mbps, 6x at 324 Mbps, 7x at 378 Mbps, 8x at 432 Mbps.

My download speeds on the OneX hovers between 4x and 7x.

They say the fastest BluRay drives are 16x which would be 864 Mbps, or still slower than Gigabit internet.

UltraHd BluRay on Wiki lists these specs, but none of the games are on UHD BluRay media:

50 GB (dual-layer, 92 Mb/s)
66 GB (dual-layer, 123, 144 Mb/s)
100 GB (triple-layer, 123, 144 Mb/s)​

 

[Jay]

Spiderman RT primer

 

[Deleted member 7537]

Spiderman RT primer

Apart from other value their videos provide, the best thing DF does is explaining complex concepts in lay terms.

 

[London Geezer]

Apart from other value their value their videos provide, the best thing DF does is explaining complex concepts in lay terms.

Yep. It may be obvious to us that RT reflections need to be scaled down in resolution, objects reflected etc on next gen consoles, but clearly this is not so obvious to everyone.

 

[snc]

As I wrote in different subject, strange comparison to rtx 2060 (ps5 gpu should be closer to 2070super) but very informative video.

 

[iroboto]

As I wrote in different subject, strange comparison to rtx 2060 (ps5 gpu should be closer to 2070super) but very informative video.

I believe in reference to RT capabilities. If I understood the data points correctly; 2060 surpasses both consoles in this particular function.

 

[function]

As I wrote in different subject, strange comparison to rtx 2060 (ps5 gpu should be closer to 2070super) but very informative video.

Comparing Minecraft RTX and the Minecraft RT demo on XSX, it seems unlikely that either next gen system will be able to match a 2060.

This video is about RT, and the performance cost of RT, and some of the ways to balance cost and quality.

Incidentally, reducing what you test rays against (detail and distance), the number of rays you use, and the distance and number of bounces you track, are all ways which can be used to manage RT cost (including memory) on the Series S.

 

[snc]

I believe in reference to RT capabilities. If I understood the data points correctly; 2060 surpasses both consoles in this particular function.

should clarify it more also still we don't know rdna2 rt possibilities (but probably worse than geforces knowing amd rt patent)

 

[Jay]

should clarify it more also still we don't know rdna2 rt possibilities (but probably worse than geforces knowing amd rt patent)

Nothing was about raster performance?

Have to admit I thought it was made pretty clear, but I'm sure they'll take what you say on board to improve their content.

Incidentally, reducing what you test rays against (detail and distance), the number of rays you use, and the distance and number of bounces you track, are all ways which can be used to manage RT cost (including memory) on the Series S.

Considering the expected RT performance of the consoles, we'll be getting a big mix of many solutions in the hybrid engines.
So as you say, not only will that help the xss, in some situations probably will used a different solution all together when tuning.

 

[PSman1700]

Didnt expect more either.

 

[Shortbread]

Spiderman RT primer

Very excellent and informative video. Can't wait to see the final build of the game. From what Insomniac developers have stated over at Era or by tweets, the release shots are captured from an older build. That things have improved. What those improvements are, we'll have to waint and see. But I don't expect much beyond what Alex has already suggested.

Also, console gamers who are expecting perfect RT and FPS from the next-generation of consoles are setting themselves up for disappointment. For these consoles to have RT at all, especially at launch, is quite suprising and welcoming.

 

[Rootax]

should clarify it more also still we don't know rdna2 rt possibilities (but probably worse than geforces knowing amd rt patent)

Yeah, same here, I was surprise to see that in a df video. I get raw numbers, but AMD and nVidia seems to do rt in a very different fashion, and console are closed box so they can optimise further I guess.

And we need to see rdna2 on pc...

 

[snc]

Nothing was about raster performance?

well he argumented it by bandwidth which is quite bizzare considering rtx 2080 also has 448gb/s and other argument was that rtx 2060super has performance similar to rx 5700 which is 8tf rdna 1 which is even more bizzare argument

 

[function]

well he argumented it by bandwidth which is quite bizzare considering rtx 2080 also has 448gb/s and other argument was that rtx 2060super has performance similar to rx 5700 which is 8tf rdna 1 which is even more bizzare argument

RX5700 XT is ~ 9.8 TF at advertised top boost (some cards can clock higher). At an average game clock of ~ 1.8 gHz that's still clearly above 9 TF.

2060S is likely to be rather ahead of next gen consoles in terms of RT performance. Using a 2080 would not make this investigation of performance vs compromise any more meaningful. If anything, it would only make it less so.

 

[snc]

RX5700 XT is ~ 9.8 TF at advertised top boost (some cards can clock higher). At an average game clock of ~ 1.8 gHz that's still clearly above 9 TF.

he compared to rx5700 not xt