• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

RedGamingTech: "Why Series X is having performance problems."

Can anyone explain why Sony's devkit is so easy to work with but the XBox GDK is rocket science?

MS created an entirely new GDK called GameCore. Essentially a PC/Xbox combo. Thus all games work for both platforms and it’s incredibly easy to switch between them. Overtime this will align PC an Console development and make it very easy to have both with almost no porting.

Sony took PS4 SDK and upgraded it for PS5. Thus everything is familiar and just improved. Anyone on PS5 sdk is essentially using something they’ve seen for 7 years. Xbox is a new platform with stuff to be worked out.
 

M1chl

Currently Gif and Meme Champion
Debunked to who? You can't use the same bandwidth with 2 different speed in the same time, it's physically impossible.
They are conected separately, right? Or are those memory chips on same "ring"? Because on PC you can get slower speed through ram and higher speed through GPU-VRAM... I think it's fairly possible. How is it managed by MS, tho not sure... When working on Kingdom Come: Deliverance, I saw some of the memory management and MS does not give you the access to it, so X1 and X1X has the same APIs, one is DDR3+ESRAM and other is GDDR5 only and sometimes that was a pain, because you really cannot do anything else, then trying to "flag" object as to where they should be store, but all in all MS decides what things it's best. So when I am talking about "tools" it means their Perfomace Stack (back-end) rather than front end what devs see.

I know you did not reacted to me, but I probably miss my opportunity at the begining of this thread to elaborate...
 

assurdum

Banned
They are conected separately, right? Or are those memory chips on same "ring"? Because on PC you can get slower speed through ram and higher speed through GPU-VRAM... I think it's fairly possible. How is it managed by MS, tho not sure... When working on Kingdom Come: Deliverance, I saw some of the memory management and MS does not give you the access to it, so X1 and X1X has the same APIs, one is DDR3+ESRAM and other is GDDR5 only and sometimes that was a pain, because you really cannot do anything else, then trying to "flag" object as to where they should be store, but all in all MS decides what things it's best. So when I am talking about "tools" it means their Perfomace Stack (back-end) rather than front end what devs see.

I know you did not reacted to me, but I probably miss my opportunity at the begining of this thread to elaborate...
I'm not saying is not possible. I said you can't have full 560 GBs speed in the gpu if you use the 332 GBs cpu in the same time. Isn't it the same bus? Just to saying it's not me who noticed this problematic, I read a bunch of opinions about it but I'm open to hear the counterpoint.
 
Last edited:

Azurro

Banned
That FUD has been debunked.
The XSX can use the full 10gb at 556gbs, while using the slower pool at the same time. Accessing one doesn't limit the other.
The fact remains that the GPU is fed by 556gbs of RAM. That is faster than the PS5 does.
I'm glad you brought up the 3070 in your reply, saying that if the speed of 448gbs is enough for the 3070 then it's fine for the PS5. So when your next argument comes that limiting the fast pool to 10gb isn't enough because a game might need more and then the XSX is going to have to use the slow RAM which will cause a bottleneck, when I bring up how the 3070 only has 8gb of VRAM you will also have to accept that 10gb is plenty for the XSX.

You are a bit late to the speculation party, let's discuss performance on screen, and the Xbox doesn't show the advantages it was touting for months. Whether it's a bottleneck somewhere caused by using an apu designed for servers for the new system, whether it's less efficient caching, or that the CUs can't be fed properly, or it's the tools themselves but what is for sure is that there's an issue somewhere.
 
Not comparable.
3070 is 8GB of VRAM only.
Series X: 10GB to be used as VRAM and System RAM
Lol, the XSX has 16gb not 10.
The system RAM (System OS) of 3.5gb is taken from the slower 6gb pool, while the remaining 2.5gb is used for CPU and Audio.
The 10gb of fast RAM is for the GPU.
 
I wouldn't be too happy about only having a dominant player in the console space, In the end the only loser will be the consumer. We've already seen arrogant Sony once, do we really want to go there again when its time for PS6?
Whe have seen the arrogants from MS for years even now...a knife cuts on 2 sites you know....
 

M1chl

Currently Gif and Meme Champion
I'm not saying is not possible. I said you can't have full 560 GBs speed in the gpu if you use the 332 GBs cpu in the same time. Isn't it the same bus?
Alright I don't know how are those chips wired, however it seems they are all connected to the chip 1 by 1, so how is that not possible to utilize them at their full speed, if they are one ring or let's say if it's "one memory bank" than sure, you can't, but if those are 2 separated banks, then I don't see why not. But I probably need to take a look at chip diagram : )
 
You are a bit late to the speculation party, let's discuss performance on screen, and the Xbox doesn't show the advantages it was touting for months. Whether it's a bottleneck somewhere caused by using an apu designed for servers for the new system, whether it's less efficient caching, or that the CUs can't be fed properly, or it's the tools themselves but what is for sure is that there's an issue somewhere.
Actually, the guy I was answering was the one who was late to the party.
You are correct, there is an issue.
Now MS has to step up and sort it out. If it's a tools issue, then devs will work that out.
If it's a bug in the system, MS will need to sort that out.
Or, it could be that as the PS5 dev kits were out well before the XSX ones, the XSX was just ported over from the PS5.
We really don't know the reasons yet, hence the speculation.
Time will tell all.
It's going to be really interesting to watch all this pan out, which is good because the games so far haven't been very interesting.
 
Last edited:

assurdum

Banned
Alright I don't know how are those chips wired, however it seems they are all connected to the chip 1 by 1, so how is that not possible to utilize them at their full speed, if they are one ring or let's say if it's "one memory bank" than sure, you can't, but if those are 2 separated banks, then I don't see why not. But I probably need to take a look at chip diagram : )
They are separated banks but how can be the whole bandwidth 560+332 in the same time if the whole it's 560 GBs ? I'm genuinely asking because I want to know.
 
Last edited:

M1chl

Currently Gif and Meme Champion
They are separated bank but how can be the whole bandwidth 560+332 in the same time if it's 560 GBs ? I'm genuinely asking
Oh well that's what I not disputed, I though that it is, that if whole RAM is utilised, then you get speed of the slower bank on the whole mem pool. I probably did not have proper reading comprehension in place. So yeah I am not disputing that...
 

assurdum

Banned
Oh well that's what I not disputed, I though that it is, that if whole RAM is utilised, then you get speed of the slower bank on the whole mem pool. I probably did not have proper reading comprehension in place. So yeah I am not disputing that...
Don't get me wrong, I have nothing against your opinion but neither MS has been exactly transparent about this aspect and looks the coincidence, AC Valhalla when is CPU stressed drops the performance around 30% less in the fps of the ps5, the exact difference with the series X CPU bandwidth. And if I'm not wrong Dirt 5 has the same drops sometimes. Maybe it's really just a coincidence but make me think a bit about it.
 
Last edited:
Debunked to who? You can't use the same bandwidth with 2 different speed in the same time, it's physically impossible. It would means series X has 560 GBs+332GBs in the same time (I don't remind the exact counts of the cpu side) so 893 GBs available. That's quite false.
You are behind on the topic.
The FUD you are pushing on this has been done and dusted ages ago.
Try and find another angle.
I will help you out here.
Fact, over the last couple of generations Sony's dev tools are the best in the business.
For all of MSs software abilities, Sony actually has the better development environment out of both companies.

You will be on a better winner with this line of attack.
 

M1chl

Currently Gif and Meme Champion
Don't get me wrong, I'm nothing against your opinion but neither MS has been exactly transparent about this aspect and look the coincidence, AC Valhalla when is CPU stressed drops the performance around 30% less in the fps of the ps5, the exact difference between the series X CPU bandwidth. Maybe it's a coincidence but make me thing about it.
I not having opinion so far, we probably got to need some leak for this. I don't really know what is going on and since the CPU is basically same, could be myriads of issue, however knowing the development on Xbox One vs PS4, I am not exactly sure if the issue is HW. I mean MS DirectX is always pain in the ass with draw calls, this is well known thing and if there is something on performance stack not exactly how it should be, it could do crazy things, because I am suspecting that memory management is all on MS, so there I see some potential problems. Just my guess. Could be multipple issues, but I don't necessary thing that HW is that. But who knows...
 

Azurro

Banned
Actually, the guy I was answering was the one who was late to the party.
You are correct, there is an issue.
Now MS has to step up and sort it out. If it's a tools issue, then devs will work that out.
If it's a bug in the system, MS will need to sort that out.
Or, it could be that as the PS5 dev kits were out well before the XSX ones, the XSX was just ported over from the PS5.
We really don't know the reasons yet, hence the speculation.
Time will tell all.
It's going to be really interesting to watch all this pan out, which is good because the games so far haven't been very interesting.

No, not really, you are arguing about how a set of specs are clearly better, when it's quite clear by the results that it is actually performing worse. I am not the type to speculate and die on my sword on it, but here we go: I do think the fault lies with the design of the chip. There's probably a very good reason why the arrays of CUs are of 10 and yet MS decided to use 14 because it's an APU meant for XCloud, I'm guessing the CPU isn't capable of keeping them fed. The decision to go with an XCloud chip probably affects other areas, such as not having unified cache for the CPU core clusters. The split memory decision is also probably not doing any favours to the developers, as they have to be very careful not to utilise more than 10 GBs because otherwise performance will tank, as it falls into the slower bandwidth.

The tools themselves will probably mature as well, so that will mitigate problems with performance, but PS5 games are not using the Geometry Engine either which presents a lot of opportunities for optimisation but apparently does require a learning curve.

I think they are going to stay mostly the same during the generation.
 

mhirano

Member
Lol, the XSX has 16gb not 10.
The system RAM (System OS) of 3.5gb is taken from the slower 6gb pool, while the remaining 2.5gb is used for CPU and Audio.
The 10gb of fast RAM is for the GPU.
I was using the 'fast 10GB' as example because surely the VRAM comes from that.
In PC space we have GPU VRAM and System RAM (the latter usually 8-16GB), so we have in an average Gamer PC 8GB of VRAM plus 16GB of SystemRAM.
There is no way Series X uses all of the fast 10GB as VRAM, cause games need system RAM for things like gamelogic, levels, asssets and so on.
And, of course, using that slow 6GB for that would bottleneck everything.
 
Last edited:

assurdum

Banned
You are behind on the topic.
The FUD you are pushing on this has been done and dusted ages ago.
Try and find another angle.
I will help you out here.
Fact, over the last couple of generations Sony's dev tools are the best in the business.
For all of MSs software abilities, Sony actually has the better development environment out of both companies.

You will be on a better winner with this line of attack.
Why indeed doesn't explain how the series X can reach 560+332 GBs bandwidth in the same time if the whole is 560 GBs? Fact, you don't know how works and you won't to hear it could be a possible problematic. So it's better to lead the discussion in the console war camp, it seems
 
Last edited:

assurdum

Banned
I not having opinion so far, we probably got to need some leak for this. I don't really know what is going on and since the CPU is basically same, could be myriads of issue, however knowing the development on Xbox One vs PS4, I am not exactly sure if the issue is HW. I mean MS DirectX is always pain in the ass with draw calls, this is well known thing and if there is something on performance stack not exactly how it should be, it could do crazy things, because I am suspecting that memory management is all on MS, so there I see some potential problems. Just my guess. Could be multipple issues, but I don't necessary thing that HW is that. But who knows...
To be clear I'm not saying a splitted speed is an impossible hardware issue unfixable. But I don't think it's that comfortable at the first try.
 
Last edited:

M1chl

Currently Gif and Meme Champion
To be clear I'm not saying a splitted speed is an impossible hardware issue unfixable. But I don't think it's that comfortable at the first try.
I agree, however I think that this is simply done by MS APIs automatically, because that whole Series S|X think. Hard to say, it's frustrating not getting more info for us nerd. Same thing with PS5 die shot.
 
I was using the 'fast 10GB' as example because surely the VRAM comes from that.
In PC space we have GPU VRAM and System RAM (the latter usually 8-16GB), so we have in an average Gamer PC 8GB of VRAM plus 16GB of SystemRAM.
There is no way Series X uses all of the fast 10GB as VRAM, cause games need system RAM for things like gamelogic, levels, asssets and so on.
And, of course, using that slow 6GB for that would bottleneck everything.
The GPU will be fed from the 10gb pool of fast RAM. It won't use the slower pool. The slower pool will be for things that don't require the same speed as the GPU, such as audio and CPU.
Yes, if the GPU did use the slower pool it would be a bottleneck for sure.
But it doesn't use it.
So in short, the RAM on the XSX is in no way a bottle neck.
I don't think either console will have an issue with RAM speed. They both went around it a different way, and both ways will be fine.
The XSX has a more powerful GPU and so actually needs more bandwidth than the PS5 does.
 

kyliethicc

Member
That FUD has been debunked.
The XSX can use the full 10gb at 556gbs, while using the slower pool at the same time. Accessing one doesn't limit the other.
The fact remains that the GPU is fed by 556gbs of RAM. That is faster than the PS5 does.
I'm glad you brought up the 3070 in your reply, saying that if the speed of 448gbs is enough for the 3070 then it's fine for the PS5. So when your next argument comes that limiting the fast pool to 10gb isn't enough because a game might need more and then the XSX is going to have to use the slow RAM which will cause a bottleneck, when I bring up how the 3070 only has 8gb of VRAM you will also have to accept that 10gb is plenty for the XSX.
Its not FUD and you're dead wrong, but have fun. This is a waste of time.

And no 8 GB of VRAM is not enough for 4K gaming in the long run. Not when baseline console games will be using 13 GB of RAM. But the good news is PCs have 16 or 32 GB of DRAM as well, not just 1 pool of RAM like consoles.
 
Xbox Series X is performing worse then the competition, in spite being more powerful.
No AAA launch exclusive titles for the console (must be a first in console history).

Simply put, Microsoft wasn't ready to launch the Xbox Series.

Theoretically once all devs are on GameCore the performance gap will not only be closed but opened up on Sony. It might take a year.
 
Why indeed doesn't explain how the series X can reach 560+332 GBs bandwidth in the same time if the whole is 560 GBs? Fact, you don't know how works and you won't to hear it could be a possible problematic. So it's better to lead the discussion in the console war camp, it seems
The RAM speed isn't 560+332. Not sure where you got that from.
The speed of the optimum 10gb of fast RAM is 560gbs. That is where the GPU will be fed from.
The speed of the slower pool of 6gb (minus the 2.5gb used for OS) is 336gbs.
The speed of either pool doesn't alter if the other pool is being used at the same time.
It's not a bottleneck at all.
 
Its not FUD and you're dead wrong, but have fun. This is a waste of time.

And no 8 GB of VRAM is not enough for 4K gaming in the long run. Not when baseline console games will be using 13 GB of RAM. But the good news is PCs have 16 or 32 GB of DRAM as well, not just 1 pool of RAM like consoles.
I think you need a rest.
The system RAM on a PC isn't VRAM. It can do audio and the CPU, and obviously the OS.
But the only pool of RAM the GPU access is the VRAM.
The RTX 3080, which is a 4k card, has 10gb of VRAM, the same amount as the XSX will be using for its GPU.

The RAM on both consoles are just fine.
Looking to try and find a bottleneck on the XSXs RAM is a losing position.
 

assurdum

Banned
The RAM speed isn't 560+332. Not sure where you got that from.
The speed of the optimum 10gb of fast RAM is 560gbs. That is where the GPU will be fed from.
The speed of the slower pool of 6gb (minus the 2.5gb used for OS) is 336gbs.
The speed of either pool doesn't alter if the other pool is being used at the same time.
It's not a bottleneck at all.
You continue to not understand what I'm saying. You can't use the same bandwidth in two different speed in the same time. Which means when cpu works with the bandwidth, GPU have to wait that CPU to free bandwidth left if want to reach the whole 560 GBs. Now it's a very approximate explanation, that means gpu slowdown to the bandwidth left until CPU free the bandwidth used. I don't know if I'm clear .
 
Last edited:
You continue to not understand what I'm saying. You can't use the same bandwidth in two different speed in the same time. Which means when cpu works with the bandwidth GPU have to wait that CPU to free bandwidth if want to reach the 560 GBs. Now it's a very approximate explanation, that means gpu slow down until CPU leave the whole bandwidth. I don't know if I was clear .
That's not how it works. That was explained along time ago.
 

assurdum

Banned
That's not how it works. That was explained along time ago.
It doesn't explain that. You continue to not listen. I'm talking when CPU and GPU needs of the bandwidth in the same time. What you show to me it's the banks partition but MS has never explained clearly what can happen when CPU and GPU work contemporary with the bandwidth in the same instant. That's why I continue to repeat it can't be 560+332 in the same time. It's something the infamous interview with the Crytek developer has remarked. He said when you use two separate speed in the same bandwidth as MS did, you need even of more bandwidth compared to a unified one.
 
Last edited:

kyliethicc

Member
I think you need a rest.
The system RAM on a PC isn't VRAM. It can do audio and the CPU, and obviously the OS.
But the only pool of RAM the GPU access is the VRAM.
The RTX 3080, which is a 4k card, has 10gb of VRAM, the same amount as the XSX will be using for its GPU.

The RAM on both consoles are just fine.
Looking to try and find a bottleneck on the XSXs RAM is a losing position.
The entire PC community agrees the 10 GB of the 3080 is not enough long term for 4K gaming, Nvidia are just cheap bastards.

Thats why they're about to release a 20 GB 3080 Ti for $1000 in a few months. Thats why the $1500 3090 has 24 GB. They can fuck over the customers and make them pay out the ass for a proper amount of VRAM.

Feeding a GPU with data via bandwidth is an entirely different debate then the size of the VRAM buffer a on PC vs APU RAM.

You have zero understanding of how memory is accessed by the console APU if you think it can always magically access all 10 GB at 560 GB/s every cycle. Thats not how memory interfaces work. It needs to use the entire 320 bit bus for 560 GB/s, all 20 channels in 16 bit mode. The issue is for that to happen, none of the 6 GB of RAM can be accessed (which is on the same 6 chips as 60% of the 10 GB is.) As soon as say 1 GB of the 6 GB is being accessed, that takes up some of those bus lanes and slows down the bandwidth. This is why consoles like the PS4 PS5 and One X use unified bandwidth with evenly interleaved chips of RAM. The uneven split setup of the Xboxes is a major issue.

This Xbox 1st party lead engineer programmer at fucking Id Soft agrees with me and he knows whats talking about.


E5CvjRu.jpg


Educate yourself please man for fuck's sake.
 
Last edited:

Shmunter

Member
That's not how it works. That was explained along time ago.
I’m not quite sure that explains much as it’s completely speculative.

Devs coming into this gen have been saying TF as a measure of performance is a fools errand. But people look for simple indicators and cannot wrap their heads around a console being a gaming ‘system’ and a sum of its parts, not just 1 or 2 components.

Even assuming the split memory speed can be accessed from the slow portion without penalty, there is a degree of management involved that is not required on other platforms. How easy is the management? is it possible to mange It well? what is the penalty for managing the memory layout?

On Beyond3d the eggheads are speculating that 3rd party will struggle with the architecture as game design will require upfront memory layout considerations. Where this leaves 3rd party remains to be seen, will ports be reworked for XsX or left as are and suffer? Will 3rd party target XsX as the main platform to engineer the games accordingly and ported to other platforms from there?

See how it plays out. Early indicators are not flattering.
 

assurdum

Banned
The entire PC community agrees the 10 GB of the 3080 is not enough long term for 4K gaming, Nvidia are just cheap bastards.

Thats why they're about to release a 20 GB 3080 Ti for $1000 in a few months. Thats why the $1500 3090 has 24 GB. They can fuck over the customers and make them pay out the ass for a proper amount of VRAM.

Feeding a GPU with data via bandwidth is an entirely different debate then the size of the VRAM buffer a on PC vs APU RAM.

You have zero understanding of how memory is accessed by the console APU if you think it can always magically access all 10 GB at 560 GB/s every cycle. Thats not how memory interfaces work. It needs to use the entire 320 bit bus for 560 GB/s, all 20 channels in 16 bit mode. The issue is for that to happen, none of the 6 GB of RAM can be accessed (which is on the same 6 chips as 60% of the 10 GB is.) As soon as say 1 GB of the 6 GB is being accessed, that takes up some of those bus lanes and slows down the bandwidth. This is why consoles like the PS4 PS5 and One X use unified bandwidth with evenly interleaved chips of RAM. The uneven split setup of the Xboxes is a major issue.

This Xbox 1st party lead engineer programmer at fucking Id Soft agrees with me and he knows whats talking about.


E5CvjRu.jpg


Educate yourself please man for fuck's sake.
Well to be honest he is referring to the series S where the situation is really dramatic.
 

assurdum

Banned
Sure but the split bandwidth is same on both. Yes only the S has insufficient amount of RAM, but both share the same split bandwidth bottleneck.
Yeah but the series S has a lower bandwidth compared the series X. I'm not saying I disagree with you but I think he is referring just the series S situation.
 

kyliethicc

Member
Yeah but the series S has a lower bandwidth compared the series X. I'm not saying I disagree with you but I think he is referring just the series S situation.
Bandwidth is relative to GPU. The S at 224 is good enough for 1080, the PS4 Pro only had 217 and was running games over 1080p. And the Sbox runs Dirt 5 at sub-720p lol so its not a high resolution gaming console.

The huge flaw for the S is it only gives devs 7.5 GB of RAM when last gen was 5 GB. That's basically the same for the devs. Barely any better. And the bandwidth is not unified. 2 GB of the 10 is at 56 GB/s cause of the 128 bit bus and uneven RAM setup. 56 GB/s is fucking DDR speeds, not GDDR speeds. Thats a bottlneck cause it takes away total bandwidth from the devs, because its a poor setup.

Series X is just as poor, it just doesn't lack for total amount of memory. But the fact the XSX GPU has 52 CUs, and will in real world performance when running games it will have less bandwidth than the PS5 (which has far fewer shaders that need to be fed work to do, and each CU in the PS5 is running faster) is bad. Its a clear issue.

Thats why it will take lots of time for devs to find workarounds for their code to get the system running at an optimal level. Thats what's hes saying. It will be an issue for both Xboxes for 3rd party devs because the PS5 is not like this, so they'll need to adjust their code on just xboxes, which is not the lead platform for most 3rd parties.
 
Last edited:

quest

Not Banned from OT
Sure but the split bandwidth is same on both. Yes only the S has insufficient amount of RAM, but both share the same split bandwidth bottleneck.
You don't think Microsoft did not run the split ram through a simulator 10s of thousands of times or more before adding significant BOM of the console? They could of easily just did what Sony did and shave off some money on the BOM. It was done because both ran into issues of getting higher clocks on ram. This was the solution to squeeze some extra performance out of the clocks they could squeeze out of the ram. You think they did this for marketing and intentionally tanked performance?
 

Astral Dog

Member
I still believe the Series X will improve over time(maybe not by much, but it should improve if the dev kits weren't ready or whatever) , but this is a huge win for PS5 regardless, it won't be seen as weaker next to SeX
 

Shmunter

Member
You don't think Microsoft did not run the split ram through a simulator 10s of thousands of times or more before adding significant BOM of the console? They could of easily just did what Sony did and shave off some money on the BOM. It was done because both ran into issues of getting higher clocks on ram. This was the solution to squeeze some extra performance out of the clocks they could squeeze out of the ram. You think they did this for marketing and intentionally tanked performance?
They skipped it. COVID
 

kyliethicc

Member
You don't think Microsoft did not run the split ram through a simulator 10s of thousands of times or more before adding significant BOM of the console? They could of easily just did what Sony did and shave off some money on the BOM. It was done because both ran into issues of getting higher clocks on ram. This was the solution to squeeze some extra performance out of the clocks they could squeeze out of the ram. You think they did this for marketing and intentionally tanked performance?
lol no. That would not have been a solution for what the XSX SOC was made to do. It had to have a 320 bit interface.

And the XSS needed to be a very small die, so they could only afford to dedicate the die space to 4 memory phys.

There were cost and design limitations that forced them to make compromises.
 

Lysandros

Member
Yep, on paper Series X only has a 17% advantage in terms of GPU power, but I always figured real world situation would be half that.
Incorrect. XSX has 18% more (max theoretical) vector ALU throughput while being in deficit in rasterization, culling and pixel fillrate by 22%. It doesn't have an advantage over PS5 in '(whole) GPU power' even on paper spec.
 

M1chl

Currently Gif and Meme Champion
Regarding memory situation, here is a thread about that, it's late I can't sleep, however it does not seems to be the case, that whole subsystem would suddenly slowed down with whole RAM utilization:

However I still feel dumb, that I could not fully grasp the situation...
 

mhirano

Member
Would Series X 12 Teraflops be something like AMD Vega GPUs compared to NVIDIA GTXs?
Like: Vega had amazing compute performance but was lacking on rasterization and then got it ass kicked by GTX in games.
In the end, the Vega' large compute advantage was only good for Ethereum mining...
 
Have the performance issues that are shown on some 3rd party titles also seen in MS titles like Gears 5 and Forza Horizon 4? Surely if the hardware was defective we'd see the same tearing and dropped frames? Has anyone here even played those updated titles or are 3rd party titles the only true test of hardware performance?
 

mhirano

Member
Have the performance issues that are shown on some 3rd party titles also seen in MS titles like Gears 5 and Forza Horizon 4? Surely if the hardware was defective we'd see the same tearing and dropped frames? Has anyone here even played those updated titles or are 3rd party titles the only true test of hardware performance?
No one has said that torn frames and bad performance were defective hardware but software bugs (bad API/tools).
And the best way to compare two systems is 'apples to apples', aka multiplatform games. This is the gold standard since the x360/PS3 gen (Digital Foundry FaceOffs).
 
Last edited:
No one has said that torned frames and bad performance were defective hardware but software bugs (bad API/tools).
And the best way to compare two systems is 'apples to apples', aka multiplatform games. This is the gold standard since the x360/PS3 gen (Digital Foundry FaceOffs).
Torn and dropped frames wouldn't need a comparison to detect. If there is a performance problem with the Xbox it will show on all software not just 3rd party software right?
 

AgentP

Thinks mods influence posters politics. Promoted to QAnon Editor.
That FUD has been debunked.
The XSX can use the full 10gb at 556gbs, while using the slower pool at the same time. Accessing one doesn't limit the other.
The fact remains that the GPU is fed by 556gbs of RAM. That is faster than the PS5 does.
I'm glad you brought up the 3070 in your reply, saying that if the speed of 448gbs is enough for the 3070 then it's fine for the PS5. So when your next argument comes that limiting the fast pool to 10gb isn't enough because a game might need more and then the XSX is going to have to use the slow RAM which will cause a bottleneck, when I bring up how the 3070 only has 8gb of VRAM you will also have to accept that 10gb is plenty for the XSX.

A PC GPU is only holding graphics data in the VRAM. The consoles GDDR is shared for all data.
 

mhirano

Member
Torn and dropped frames wouldn't need a comparison to detect. If there is a performance problem with the Xbox it will show on all software not just 3rd party software right?
Not necessarily. An exclusive can be cut down until performance targets are met.
We would never know Dirt 5 Xbox version was trash if not for the PS5 version showing us how it was supposed to look.
 
Last edited:
Top Bottom