Xbox Series X [XBSX] [Release November 10 2020]

[snc]

Before launch many people said XSX GPU was more like 2070, I said it would be more like 2080 Ti. Well not completely apples to apples, but we have a closer compare now with 6700XT. In AMD's 6700XT "game clock" (AMD's conservative estimation of the clock in real world gaming use, different than max boost clock) 6700Xt is 12.4 TF or so. So, fairly close to XSX. In Techspots review in 14 game avg at 1440P, it was very close to a 2080Ti 2080Ti avg 117 FPS, 6700XT avg 114 FPS. 2080 Super 101 FPS, 2080 96 FPS, 2070 Super 89, 2070 75. Probably only I care about this.

but performance not scaling so well on wider navis
https://www.computerbase.de/2021-03/amd-radeon-rdna2-rdna-gcn-ipc-cu-vergleich/2/

https://imgur.com/XmU1JLE

 

[Deleted member 86764]

Before launch many people said XSX GPU was more like 2070, I said it would be more like 2080 Ti. Well not completely apples to apples, but we have a closer compare now with 6700XT. In AMD's 6700XT "game clock" (AMD's conservative estimation of the clock in real world gaming use, different than max boost clock) 6700Xt is 12.4 TF or so. So, fairly close to XSX. In Techspots review in 14 game avg at 1440P, it was very close to a 2080Ti 2080Ti avg 117 FPS, 6700XT avg 114 FPS. 2080 Super 101 FPS, 2080 96 FPS, 2070 Super 89, 2070 75. Probably only I care about this.

I'd expect the XSX to perform a decent amount better than the 6700xt over the course of the generation.

 

[Allandor]

but performance not scaling so well on wider navis
https://www.computerbase.de/2021-03/amd-radeon-rdna2-rdna-gcn-ipc-cu-vergleich/2/

https://imgur.com/XmU1JLE

There is always a limit for "general purpose" tasks scaling. But as you can see in the graph, CUs scale perfectly up to 60. After this point higher resolutions are needed to scale a bit better. Also those big navi cards have all the same bandwidth, which should also get upscaled to provide additional needed bandwidth.

But those are just PC configurations. Not optimized console code. You can look at GCN and see how good it scales on pc vs consoles.
Also PS5 with it's dynamic clocks make a comparison really hard. Because you "never know" how fast something can be done. It can also go lower than 10tf if the code is demanding. Those comparisons only make sense if you can fix the clockrate or you have many different benchmarks.

 

[snc]

There is always a limit for "general purpose" tasks scaling. But as you can see in the graph, CUs scale perfectly up to 60. After this point higher resolutions are needed to scale a bit better. Also those big navi cards have all the same bandwidth, which should also get upscaled to provide additional needed bandwidth.

But those are just PC configurations. Not optimized console code. You can look at GCN and see how good it scales on pc vs consoles.
Also PS5 with it's dynamic clocks make a comparison really hard. Because you "never know" how fast something can be done. It can also go lower than 10tf if the code is demanding. Those comparisons only make sense if you can fix the clockrate or you have many different benchmarks.

you have results for 4k, I wouldn't call 36% (down from 50%) perfectly scaling for 60

 

[smithg]

Before launch many people said XSX GPU was more like 2070, I said it would be more like 2080 Ti. Well not completely apples to apples, but we have a closer compare now with 6700XT. In AMD's 6700XT "game clock" (AMD's conservative estimation of the clock in real world gaming use, different than max boost clock) 6700Xt is 12.4 TF or so. So, fairly close to XSX. In Techspots review in 14 game avg at 1440P, it was very close to a 2080Ti 2080Ti avg 117 FPS, 6700XT avg 114 FPS. 2080 Super 101 FPS, 2080 96 FPS, 2070 Super 89, 2070 75. Probably only I care about this.

Another interesting detail to come out of the 6700XT launch is that apparently RDNA1 and RDNA2 IPC are virtually identical according to this testing:

 

[iroboto]

guarantee you that many are confused when faster is base

It was the last card to be announced though.
It may just be a matter of preference, but I would start at the flagship announced, with the understanding that AMD has no plans to release a card that will top it's performance. Therefore, it's performance is stable, as a relative anchor. That's just me.
If I wanted to deep dive into 6700XT performance, I would make everything relative to it. On the question of CU scaling.. that's a tough one.
I would be tempted to request a card that had 1 CU and try to make a unit test, but that's not how video cards work.
So the reverse option is to take 80CU as being the fullest glass, or a unit vector, and make everything else relative to that for CU scaling.

 

[Deleted member 13524]

but performance not scaling so well on wider navis
https://www.computerbase.de/2021-03/amd-radeon-rdna2-rdna-gcn-ipc-cu-vergleich/2/

https://imgur.com/XmU1JLE

I'm actually pretty surprised with how well the 6700XT behaves at native 4K. Or rather how much better the 6800 series should be at that resolution but isn't.

 

[pjbliverpool]

Before launch many people said XSX GPU was more like 2070, I said it would be more like 2080 Ti. Well not completely apples to apples, but we have a closer compare now with 6700XT. In AMD's 6700XT "game clock" (AMD's conservative estimation of the clock in real world gaming use, different than max boost clock) 6700Xt is 12.4 TF or so. So, fairly close to XSX. In Techspots review in 14 game avg at 1440P, it was very close to a 2080Ti 2080Ti avg 117 FPS, 6700XT avg 114 FPS. 2080 Super 101 FPS, 2080 96 FPS, 2070 Super 89, 2070 75. Probably only I care about this.

But 1440p can be partially CPU limited on a 2080Ti class GPU which would potentially mask some of the performance gap.

At 4k, a much more GPU limited scenario, the 2080Ti is 11% faster than the 6700XT in TPU's testing, or the 6700XT is only 12% faster than the 2080.

Now if we consider that at its boost clock which it seems to be able to hit or thereabouts in most reviews, the 6700XT has about 9% more compute than the XSX and over 40% more fill rate, plus 1.5TB/s of Infinity Cache then it starts to look like the XSX GPU may perform closer to a 2080 than a 2080Ti at 4k if it were in a PC. And 4k is the target resolution of the XSX in many games so I think that's a valid comparison resolution.

 

[Riddlewire]

 

[zed]

I've put it to the vote
https://forum.beyond3d.com/threads/faster-than-means.62327/

 

[invictis]

Got two Series Xs, one for me, one for the kid, and have to say they are beautiful consoles. I also have to say the unboxing experience with them is the best I have had or seen with any console period. Well done Microsoft.

They are both whisper quiet, well built, and you can see the progression from the One X to the Series X. Its like Microsoft has made them very similar so to separate them both from the OG Xbox One.

 

[Deleted member 11852]

They are both whisper quiet, well built, and you can see the progression from the One X to the Series X. Its like Microsoft has made them very similar so to separate them both from the OG Xbox One.

Ignoring Series S, it's interesting both companies picked a design that was intended to be positioned vertically. Sure, you can lay them on their side... if you're some kind of monster.

 

[iroboto]

Ignoring Series S, it's interesting both companies picked a design that was intended to be positioned vertically. Sure, you can lay them on their side... if you're some kind of monster.

I'm feeling attacked

 

[Allandor]

Ignoring Series S, it's interesting both companies picked a design that was intended to be positioned vertically. Sure, you can lay them on their side... if you're some kind of monster.

Well, it makes sense for Series X as it follows the natural air-flow (at last on earth ^^). But I really don't get why Sony chose to create a top-down solution for the air flow. Yes the impact is really minimal once the cooling is spinning. But working against the natural airflow (hot goes up) does not make it automatically better. So the cooler must always spin a bit faster than the other way around. This is especially surprising as they seem to want to keep the console as quite as possible.

 

[cheapchips]

Well, it makes sense for Series X as it follows the natural air-flow (at last on earth ^^). But I really don't get why Sony chose to create a top-down solution for the air flow. Yes the impact is really minimal once the cooling is spinning. But working against the natural airflow (hot goes up) does not make it automatically better. So the cooler must always spin a bit faster than the other way around. This is especially surprising as they seem to want to keep the console as quite as possible.

How much faster? Or, since we're in the X thread, how many fan RPMs does the the SX save due to convection? Has anyone measured SX fan RPMs on it's side vs vertical? Has anyone measured SX fan RPMs when held upside-down?

This is an important issue and I demand answers!

 

[Deleted member 11852]

Well, it makes sense for Series X as it follows the natural air-flow (at last on earth ^^). But I really don't get why Sony chose to create a top-down solution for the air flow. Yes the impact is really minimal once the cooling is spinning. But working against the natural airflow (hot goes up) does not make it automatically better.

In any upright object you generally want as much of the the weight on the bottom as is possible. The PSU and copper fins around the APU and memory are heavy so if the fan is right at the bottom then you've created a structural and stability problem to solve. Microsoft's approach is a simple and rigid shape and that's fine given this has been Microsoft's design language for a while. Sony's design language is different. I.e.. mental. Sacrifices were obviously made for this.

However hot air convection, hot being relative to the ambient temperature, is pretty slow and even is ludicrously slow fan blowing downwards can overcome the natural heat transfer coefficient.

 

[Allandor]

In any upright object you generally want as much of the the weight on the bottom as is possible. The PSU and copper fins around the APU and memory are heavy so if the fan is right at the bottom then you've created a structural and stability problem to solve. Microsoft's approach is a simple and rigid shape and that's fine given this has been Microsoft's design language for a while. Sony's design language is different. I.e.. mental. Sacrifices were obviously made for this.

However hot air convection, hot being relative to the ambient temperature, is pretty slow and even is ludicrously slow fan blowing downwards can overcome the natural heat transfer coefficient.

That's right. But at the same time, the air you heated up goes up and might get pulled back into the system again.

Sony could also use the cooling to pull the air from the bottom and blow it out at the top with the same design. It might just require additions air channels inside the console so the power-supply does not heat up the air before the gpu get's it.

 

[Deleted member 11852]

That's right. But at the same time, the air you heated up goes up and might get pulled back into the system again.

Sure, but the heat doesn't go up unless the fan is entirely off. The airflow produced by the fan in the PS5 is far too powerful. The heat wouldnt't rise for even a microsecond. Air convection is a process easily easily disrupted by turbulence and that's 100% of the job of the fan.

Sony could also use the cooling to pull the air from the bottom and blow it out at the top with the same design. It might just require additions air channels inside the console so the power-supply does not heat up the air before the gpu get's it.

They would have had to redesign the console. In vertical orientation - even when sitting on the stand - there is virtually no space around the bottom to draw air in. It's usually sitting on a flat surface, the sides panels have no vents so there is only the narrow front and rear apertures (the black middle between the panels). It's much easier to blow air out of a small gap than draw air in so it feels like the decision to have the upper third of the PS5 to be two long curves arranges of vents to feed the downward-blowing fan was intentional.

It's not how I would have designed it but Sony have a history of form over function and they do like their distinctive, often asymmetrical, designs.

 

[see colon]

However hot air convection, hot being relative to the ambient temperature, is pretty slow and even is ludicrously slow fan blowing downwards can overcome the natural heat transfer coefficient.

Not only that, but hot air only rises because heat only travels toward cold, and air expands when it's heated causing it's pressure to increase, and high pressure pushes toward low pressure. I'd be curious to see an experiment that measure air movement from an Series X with the fan disabled with it vertical and... invertical? Is that a word that describes vertical but upside down. Theoretically, the hot air inside the Xbox would be vented from every opening that has lower pressure and temperature than the internal, and the main opening is the vented top (or bottom in the invertical position).

 

[Deleted member 11852]

Not only that, but hot air only rises because heat only travels toward cold, and air expands when it's heated causing it's pressure to increase, and high pressure pushes toward low pressure.

I can see what you're trying to say but not quite. Warm air rises above cold air because it's less dense. It's not so much warm air rising but the colder air above (more dense, heavier) dropping. This only works in the vertical plane because of gravity and it's the same principle in air or fluids.