• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Sony Reveals the PlayStation 5 Pro, Launches November 7th

Joined
Oct 6, 2023
Messages
31 (0.07/day)
Location
Bahrain
Processor AMD Ryzen 9 7950X
Motherboard ASUS ROG Crosshair X670E Hero
Cooling NZXT Z73 RGB
Memory 64GB G.Skill Trident Z RGB DDR5 6200MT/s CL30
Video Card(s) Zotac NVIDIA RTX 4090 Trinity
Storage 2TB Samsung 990 Pro M.2 SSD + 15.36TB Micron 9300 Pro U.2 SSD
Display(s) LG C9 UHD 120Hz OLED TV
Case Lian Li O11D XL ROG
Audio Device(s) Denon AVR + 5.1 Setup
Power Supply Corsair HX1000i
Mouse Logitech G903
Keyboard ASUS ROG Strix Flare
Well I do work on those kind of memory buses so yes, if a memory bus is shared between several devices, depending on the access cycles there will be some memory controllers (usually it's divided into 32b controllers) that will be fully parked by the CPU while the GPU is working. I guess some game engine might be able to share the usage smoothly but it's definitely not full bandwidth for the GPU all the time.

"Close to/slightly behind the 7800XT in rasterization" Then how do you explain the dev of the PS5 himself said "up to 45% compared to previous gen" and not 75% ??

"You're the one who keeps saying the PS5 GPU is ~RX6700," This is from the numbers given here in the TPU database. Check the links in my original message. Given the actual in game perfs it's seems quite likely even.
Developers can access the full bandwidth at any time. Even on a PC the CPU doesn't need that much bandwidth, thats why CPUs come only with 2x 32bit memory controllers.

Up to 45% is probably because not everything in the graphics pipeline was increased by 67% like they did with the core count, who knows, its anybody's guess until they clarify their testing methodology.

We had two leaks on the PS5 Pro specs which turned out to be correct, they are:-
First leak said the GPU is 33.5TFLOPs which makes the GPU 60CU @ 2.18 GHz, 50 MHz lower than the PS5 <== my guess is that they're unlikely to drop the clockspeed to maintain compatibility with non-patched PS5 games.
Second leak said the GPU actually can boost to 36TFLOPs so that puts the GPU 60CU @ 2.344 GHz <== more likely accurate since the PS5 Pro has a boost mode, if it works like how the boost mode on the PS4 Pro did, it bumps up the GPU clock slightly to boost frames.

7800XT = 60CU @ 2.43 GHz with 19 Gbps VRAM.
PS5Pro = 60CU @ 2.344 GHz with 18 Gbps shared VRAM.

Not much behind the 7800XT, at least on paper.
 

f0ssile

New Member
Joined
Jan 5, 2023
Messages
24 (0.03/day)
Location
Italy
We don't know what architecture it's based on so it's hard to say still between the 45% uplift in raw perfomance it no longer needing to rely on FSR and the uplift in RT with some sort of dedicated hardware it's not crazy priced for a whole system. I doubt they are making much if any money on it and consoles haven't been sold at a loss since the PS360 days.

It's an APU after all I can't imagine them trying to make anything bigger than 60CU in 2024

Still nowhere near a 4070ti lol even the vanilla one.

The biggest letdown with it honestly is that it still uses Zen2 it's odd that upgrading to Zen 3 even would break compatibility but probably has to do with how their api works I guess.
It was clear to me what was not known, as it was clear to me that, not knowing (therefore from presumptions), they were pushing only in the upper part.

They will be (are) things that will become useful tomorrow, when even the console user will be able to say he has them, so it will be the usual story...
Do you know the frequent 60fps, become visible thanks to the sensory OC as ephemeral as nature... and the loadings that became unbearable the day after iSSDs entered consoles? At the same time, I predict a decreasing importance of the processor department. It will be pathetic, it will mean that it will not count...

Well, if that's the case we understand each other. :)
 
Joined
Oct 12, 2021
Messages
41 (0.04/day)
System Name Deneb
Processor AMD Ryzen 7 7800X3D
Motherboard Asus ProArt X670E-Creator WiFi
Cooling CPU: Noctua D15. Case additional : ML140 Pro blue PWM (400-2000) x1, ML140 Pro RGB PWM (400-1200) x2
Memory G.Skill Flare X5 DDR5 6000Mbps EXPO 2x16Go
Video Card(s) Geforce RTX 4070 Asus Dual 12Go
Storage SSD Solidigm (SK Hynix) P44 Pro 2To (Gen4). HDD WD80EAZZ (8To, CMR) x2 in AMD RAID1
Display(s) Asus VG27AQ 27" 1440p 165Hz ELMB Sync, Freesync/Gsync compatible
Case Fractal Design R6 Tempered Glass Black
Audio Device(s) Motherboard soundcard. Logitech Z623 2.1 THX speakers.
Power Supply Corsair RMx 2018 850W
Mouse Logitech G502
Keyboard Filco Majestouch Convertible 2 (USB/BT) TKL Cherry blue
Software Windows 11 Pro
Benchmark Scores Cinebench R23 18440, Geekbench 6 CPU 2671/14823, GPU 181588, 3DMark Speedway 4723, Steel Nomad 4019
Developers can access the full bandwidth at any time. Even on a PC the CPU doesn't need that much bandwidth, thats why CPUs come only with 2x 32bit memory controllers.

Up to 45% is probably because not everything in the graphics pipeline was increased by 67% like they did with the core count, who knows, its anybody's guess until they clarify their testing methodology.

We had two leaks on the PS5 Pro specs which turned out to be correct, they are:-
First leak said the GPU is 33.5TFLOPs which makes the GPU 60CU @ 2.18 GHz, 50 MHz lower than the PS5 <== my guess is that they're unlikely to drop the clockspeed to maintain compatibility with non-patched PS5 games.
Second leak said the GPU actually can boost to 36TFLOPs so that puts the GPU 60CU @ 2.344 GHz <== more likely accurate since the PS5 Pro has a boost mode, if it works like how the boost mode on the PS4 Pro did, it bumps up the GPU clock slightly to boost frames.

7800XT = 60CU @ 2.43 GHz with 19 Gbps VRAM.
PS5Pro = 60CU @ 2.344 GHz with 18 Gbps shared VRAM.

Not much behind the 7800XT, at least on paper.

Developers can access the full bandwidth at any time.
I just explained you why but no, sometimes a good amount of bandwidth is used by the CPU.
thats why CPUs come only with 2x 32bit memory controllers
With dual channel, that's already 128b. While not all the bandwidth is used in gaming, there is a decent amount of bandwidth used by the CPU. The bandwidth usable by the GPU might be a bit higher but at least it should be close to the 160b bus of the 6700 for the original PS5.

Up to 45% is probably because not everything in the graphics pipeline was increased by 67%
+45% is exactly the performance bump between the RX 6700 and the RX 6800 (see the perf comparison diagram at this link), the latter having 67% more units. So it makes perfect sense, this is just how RDNA architecture scale on modern games.

First leak said the GPU is 33.5TFLOPs
I have never found anything about this but either it is 16b TFLOPs, or the architecture has poor efficiency in gaming compared to theoretical TFLOPS, because it is not far from the 7800 XT perfs, although the 7800 XT has +72% perfs in gaming compared to the 6700. The RX 6800 has 32.3 TFLOPs in 16b so it would make sense that it is 16b.

50 MHz lower than the PS5 <== my guess is that they're unlikely to drop the clockspeed to maintain compatibility with non-patched PS5 games.
Not sure what you mean here. Clockspeed has nothing to do with compatibility.

Not much behind the 7800XT, at least on paper.
Unlikely, 7800XT has far more than +45% more perfs than the original PS5.
 
Joined
Oct 6, 2023
Messages
31 (0.07/day)
Location
Bahrain
Processor AMD Ryzen 9 7950X
Motherboard ASUS ROG Crosshair X670E Hero
Cooling NZXT Z73 RGB
Memory 64GB G.Skill Trident Z RGB DDR5 6200MT/s CL30
Video Card(s) Zotac NVIDIA RTX 4090 Trinity
Storage 2TB Samsung 990 Pro M.2 SSD + 15.36TB Micron 9300 Pro U.2 SSD
Display(s) LG C9 UHD 120Hz OLED TV
Case Lian Li O11D XL ROG
Audio Device(s) Denon AVR + 5.1 Setup
Power Supply Corsair HX1000i
Mouse Logitech G903
Keyboard ASUS ROG Strix Flare
I just explained you why but no, sometimes a good amount of bandwidth is used by the CPU.

With dual channel, that's already 128b. While not all the bandwidth is used in gaming, there is a decent amount of bandwidth used by the CPU. The bandwidth usable by the GPU might be a bit higher but at least it should be close to the 160b bus of the 6700 for the original PS5.
Yes the CPU on the PS5 magically needs over 120GB/s of bandwidth all the time so that it fits your narrative that PS5 GPU = 6700XT.
+45% is exactly the performance bump between the RX 6700 and the RX 6800 (see the perf comparison diagram at this link), the latter having 67% more units. So it makes perfect sense, this is just how RDNA architecture scale on modern games.
Again with the flawed comparison that a PS5 = 6700XT just because the TFLOPs rating is close, it doesn't work that way.
I have never found anything about this but either it is 16b TFLOPs, or the architecture has poor efficiency in gaming compared to theoretical TFLOPS, because it is not far from the 7800 XT perfs, although the 7800 XT has +72% perfs in gaming compared to the 6700. The RX 6800 has 32.3 TFLOPs in 16b so it would make sense that it is 16b.
Lol RDNA3 has dual issue FP32 SM/ALUs.
Not sure what you mean here. Clockspeed has nothing to do with compatibility.
The way compatibility mode works on the PS4 Pro is that it disables the extra CU (vs PS4) and reduces the GPU clock to PS4 clocks so that the game runs without issues, boost mode worked by bumping only the GPU clock to PS4 Pro clocks to give small additional performance (it caused some issues on some games, that why its optional). Most likely PS5 compatibility and boost mode will work the same.

Thats why you can't have the GPU clock at lower clocks than the original PS5.
Unlikely, 7800XT has far more than +45% more perfs than the original PS5.
Did you compare both on equal settings/testing methodology to come to this conclusion? At least did you do a scientific estimation? I already know the answer, its no to both, you estimation is based on the flawed estimation that a PS5 = 6700XT plus Sony claims of +45% boost.
 
Joined
Feb 1, 2019
Messages
3,655 (1.70/day)
Location
UK, Midlands
System Name Main PC
Processor 13700k
Motherboard Asrock Z690 Steel Legend D4 - Bios 13.02
Cooling Noctua NH-D15S
Memory 32 Gig 3200CL14
Video Card(s) 4080 RTX SUPER FE 16G
Storage 1TB 980 PRO, 2TB SN850X, 2TB DC P4600, 1TB 860 EVO, 2x 3TB WD Red, 2x 4TB WD Red
Display(s) LG 27GL850
Case Fractal Define R4
Audio Device(s) Soundblaster AE-9
Power Supply Antec HCG 750 Gold
Software Windows 10 21H2 LTSC
Nope, all features are optional and not required to play and enjoy games on the platform.
PC isn't cheaper on the long run, overall its more expensive.

PS+ features are all optional and not required to play games, you won't enjoy the gaming less if you don't have cloud saves, free packs ....etc

Wrong again, online multiplayer works even without a PS+ subscription for "free for all games" like Rocket League, Call of Duty: Warzone, Fortnite and Apex Legends ....etc
Not to mention even if you want to play a non-free game online, if you know a PS+ plus subscriber then they can share that feature with one PS5 console and one PS4 console, so you get it for free.

1) Yes +45% performance is for rasterization, RT performance boost is up to 3x.
2) RX6700 has a 160-bit memory bus versus 256-bit for the PS5, its like a RX6700 but with a %40 memory overclock, which depending on the game it can give a big boost to FPS versus a RX6700.
3) RX6800 has poor RT performance, something that the PS5 Pro won't have, the comparison to the RX6800 is flawed at best.
4) PS5 Pro GPU is an 7800XT with a downclock of 11% on the core and 8% on the memory, with a better RT performance. Performance will very close to the 7800XT rather than a RX6800.

I have no problem with people preferring PCs over consoles, in the end each person requirement/budget is different, I do have problem with people who spread wrong information just to make the PC look better/consoles look worse.

The fact is even the "badly" priced PS5 Pro offers tremendous value as a gaming system versus a PC.
I think on cost it depends.

If you get lucky with the controller and it lasts for the entire console gen, dont buy expanded storage, and dont buy mid gen refresh, dont pay for the subscription services, then yeah console is cheaper

However

Lets say on average you get 1 year between stick drift, you buy mid gen console upgrade, you pay for PSN plus or whatever it is called these days, buy expanded storage, buy the stand, buy the optical drive. Assume 8 years for the gen.

psn plus premium x 8 £960
ps5 original release £450
ps5 pro £700
ps5 pro stand £25
ps5 pro optical drive £100
premium to buy games on sony platform £10 x number of games buy.
6 dualsense controllers (as one will come with base ps5, and another ps5 pro) £360
We assume you buy 20 games during the lifespan. £200

Total £2795 over 8 years of use.

Some form of psn payment is needed if you care about your saves since they disabled usb backups. I have ignored the upcoming price bump for controllers and assumed ps5 pro stand cost same as for the base model. Would get some money back selling base ps5 on mid gen upgrade, but of course that applies to PC also selling old parts.
 
Joined
Oct 12, 2021
Messages
41 (0.04/day)
System Name Deneb
Processor AMD Ryzen 7 7800X3D
Motherboard Asus ProArt X670E-Creator WiFi
Cooling CPU: Noctua D15. Case additional : ML140 Pro blue PWM (400-2000) x1, ML140 Pro RGB PWM (400-1200) x2
Memory G.Skill Flare X5 DDR5 6000Mbps EXPO 2x16Go
Video Card(s) Geforce RTX 4070 Asus Dual 12Go
Storage SSD Solidigm (SK Hynix) P44 Pro 2To (Gen4). HDD WD80EAZZ (8To, CMR) x2 in AMD RAID1
Display(s) Asus VG27AQ 27" 1440p 165Hz ELMB Sync, Freesync/Gsync compatible
Case Fractal Design R6 Tempered Glass Black
Audio Device(s) Motherboard soundcard. Logitech Z623 2.1 THX speakers.
Power Supply Corsair RMx 2018 850W
Mouse Logitech G502
Keyboard Filco Majestouch Convertible 2 (USB/BT) TKL Cherry blue
Software Windows 11 Pro
Benchmark Scores Cinebench R23 18440, Geekbench 6 CPU 2671/14823, GPU 181588, 3DMark Speedway 4723, Steel Nomad 4019
Yes the CPU on the PS5 magically needs over 120GB/s of bandwidth all the time so that it fits your narrative that PS5 GPU = 6700XT.
No it's based on my personal experience with DRAM interfaces. Accessing DRAM with several devices at the same time requires arbitration and above all, changing address domains which decrease bandwidth by a huge amount. Not to mention that GDDR isn't optimized for CPU usage but for GPU usage, highly parallel / vector processing. You won't get that much bandwidth out of it with a CPU, so that the effective bandwidth of 64b of that GDDR bus won't be faster than one 64b channel of DDR4 SDRAM. In the end, if optimistic you might still keep 192b usable for the GPU out of the 256b, which is arguably not that far from a 160b bus, especially since stuff is quite downclocked on the RX 6700 (not XT btw). And I do think it's more likely to be less than that.

Again with the flawed comparison that a PS5 = 6700XT just because the TFLOPs rating is close, it doesn't work that way.
RX 6700, not 6700 XT. Literally everything is close from the data we have, not only the TFLOPs. But sure it is not exact science, that I agree, but it's better than nothing.

Lol RDNA3 has dual issue FP32 SM/ALUs.
Yes and the 7800 XT is RDNA 3, and has those TFLOPs and +72% perfs more than the RX 6700, so that doesn't match up, that's what I'm saying.

The way compatibility mode works on the PS4 Pro is that it disables the extra CU (vs PS4) and reduces the GPU clock to PS4 clocks so that the game runs without issues, boost mode worked by bumping only the GPU clock to PS4 Pro clocks to give small additional performance (it caused some issues on some games, that why its optional). Most likely PS5 compatibility and boost mode will work the same.

Thats why you can't have the GPU clock at lower clocks than the original PS5.
That doesn't make much sense. In anyways, I don't see that having any sort of priority here, they will make the device the most efficient as they can, setting whatever clock they need to set. Keeping clocks the same for compatibility is nothing more than a quick trick to make it work just in case it wouldn't in some minor and very badly developed games, there are obviously better ways to ensure compatibility.

Did you compare both on equal settings/testing methodology to come to this conclusion? At least did you do a scientific estimation? I already know the answer, its no to both, you estimation is based on the flawed estimation that a PS5 = 6700XT plus Sony claims of +45% boost.
Again it's RX 6700, not XT. Well it's a rough estimation but it does look pretty good and arguably better than just saying PS5 Pro = 7800 XT which is much farther from the truth. Because I question what sort of scientific estimation YOU made to come up to say this.

And actually since you insist so much I did end up looking this up a bit and I found this video (at 57:30). Digital Foundry apparently did a previous test showing the PS5 was indeed close to the RX 6700 : "when we tested it, very very close". Obviously the PS5 GPU isn't the same, it's older than the RX 6700 and it's a custom device, with no Direct X support. And in the video link I give you, they say exactly the same as I did, 45% perf upgrade puts it close to a RX 6800 (or a very badly downclocked 7800 XT), but neither the RX 6800 nor the 7800 XT have the RDNA4 RT features supposed to be included with the PS5 Pro, plus there is no machine learning specialized units on RDNA2 or 3 GPUs (as it is supposed to be included on the PS5 Pro GPU) so since there is no RDNA 4 GPU we can't compare to any existing PC GPU yet. But in terms of non RT perfs they do say it's close to an RX 6800 and that's exactly what I was saying.

PS: Linus Tech Tips is even putting the PS5 and PS5 Pro lower than Digital Foundry : https://youtu.be/FglPnj1eu2o?si=RvPVZZR6N84WpP_l
For him the PS5 was close to a 5700 XT and the PS5 Pro would be close to a 7700 XT. I prefer Digital Foundry comparison though.
 
Last edited:
Joined
Jun 18, 2021
Messages
2,567 (2.01/day)
Some form of psn payment is needed if you care about your saves since they disabled usb backups

No they didn't. People keep mentioning this, it's not true. The options are not very clearly marked but you can definitely backup your console without live services (this is not new, this is there from day one!). I think this confusion started because the PS5 didn't have the same direct option the PS4 had to allow for migration, but you can still backup your console to a USB drive including your saved games.



They should design this better but the options are clearly there!

psn plus premium x 8 £960

Your math is way off, you don't need PS Plus premium for saves or to play online, PS Plus essential is enough and goes for 59£ a year (48£ if you buy during black friday or days of play in june). PS Plus Premium is the equivalent-ish to Xbox Game Pass with access to games on a subscription base.

The final long term price can vary wildly, if you're counting on buying a controller per year you should also count on buying a mouse per year and a keyboard every two years when comparing to PC :D I can tell you I still have my Dualsense from launch day, I don't feel any drift but if I did I'd take the thing apart and just fix it (I know the average won't but anyway...)
 
Joined
Oct 6, 2023
Messages
31 (0.07/day)
Location
Bahrain
Processor AMD Ryzen 9 7950X
Motherboard ASUS ROG Crosshair X670E Hero
Cooling NZXT Z73 RGB
Memory 64GB G.Skill Trident Z RGB DDR5 6200MT/s CL30
Video Card(s) Zotac NVIDIA RTX 4090 Trinity
Storage 2TB Samsung 990 Pro M.2 SSD + 15.36TB Micron 9300 Pro U.2 SSD
Display(s) LG C9 UHD 120Hz OLED TV
Case Lian Li O11D XL ROG
Audio Device(s) Denon AVR + 5.1 Setup
Power Supply Corsair HX1000i
Mouse Logitech G903
Keyboard ASUS ROG Strix Flare
No it's based on my personal experience with DRAM interfaces. Accessing DRAM with several devices at the same time requires arbitration and above all, changing address domains which decrease bandwidth by a huge amount. Not to mention that GDDR isn't optimized for CPU usage but for GPU usage, highly parallel / vector processing. You won't get that much bandwidth out of it with a CPU, so that the effective bandwidth of 64b of that GDDR bus won't be faster than one 64b channel of DDR4 SDRAM. In the end, if optimistic you might still keep 192b usable for the GPU out of the 256b, which is arguably not that far from a 160b bus, especially since stuff is quite downclocked on the RX 6700 (not XT btw). And I do think it's more likely to be less than that.
That defeats the whole idea about the benefits of a shared memory.
RX 6700, not 6700 XT. Literally everything is close from the data we have, not only the TFLOPs. But sure it is not exact science, that I agree, but it's better than nothing.
Yeah the 6700XT was a typo from me, I meant RX6700.
Again, TFLOPs rating is one part of the graphics pipeline, not all of it, there are other things that affects performance depending on the game/moment, if at the exact moment it was compute limited then yes a PS5 will perform like a RX6700, but thats not always the case.
Yes and the 7800 XT is RDNA 3, and has those TFLOPs and +72% perfs more than the RX 6700, so that doesn't match up, that's what I'm saying.
You're the one who said the PS5 Pro TFLOPs rating must be FP16 since its seems high, its not, RDNA3 has dual issue FP32 SM which gives theoretical TFLOPs a 2x boost.
That doesn't make much sense. In anyways, I don't see that having any sort of priority here, they will make the device the most efficient as they can, setting whatever clock they need to set. Keeping clocks the same for compatibility is nothing more than a quick trick to make it work just in case it wouldn't in some minor and very badly developed games, there are obviously better ways to ensure compatibility.
It doesn't make sense to you because you are used to PC Gaming where developers has to account for different GPUs, CPUs, OS ........etc, so that when you pop-in an upgraded GPU it just works, thats not the way its done on the consoles otherwise developers, for example, will be able to easily let PS4 games work at 4K60 on the PS5 without any development effort.
Again it's RX 6700, not XT. Well it's a rough estimation but it does look pretty good and arguably better than just saying PS5 Pro = 7800 XT which is much farther from the truth. Because I question what sort of scientific estimation YOU made to come up to say this.
The leaked PS5 Pro GPU specs match AMD Navi32 die inwhich the 7800XT is based on, it matches the 7800XT on everything except slightly lower GPU and Memory clockspeed plus the RT and upscaling hardware.
And actually since you insist so much I did end up looking this up a bit and I found this video (at 57:30). Digital Foundry apparently did a previous test showing the PS5 was indeed close to the RX 6700 : "when we tested it, very very close". Obviously the PS5 GPU isn't the same, it's older than the RX 6700 and it's a custom device, with no Direct X support. And in the video link I give you, they say exactly the same as I did, 45% perf upgrade puts it close to a RX 6800 (or a very hardly downclocked 7800 XT), but neither the RX 6800 nor the 7800 XT have the RDNA4 RT features supposed to be included with the PS5 Pro, plus there is no machine learning specialized units on RDNA GPUs (as it is supposed to be included on the PS5 Pro GPU) so since there is no RDNA 4 GPU yet we can't compare to any existing GPU yet. But in terms of non RT perfs they do say it's close to an RX 6800 and that's exactly what I was saying.
Digital Foundry has an AMD 4800S Desktop kit which is essentially a Microsoft Xbox Series X chip with a disabled GPU slapped on motherboard for PC users in China, they installed a RX6700 on it.
While that system is the closest you can get to try to run PS5-esque hardware on a PC it doesn't come with some flaws.

1) Since the iGPU is disabled the CPU can get all the 560GB/s bandwidth of the GDDR6 memory.
2) The CPU boosts to 4.0 GHz much higher than the PS5.
3) When they try to test games on the two system, they try to visually match PS5 settings on their simulated PC, that doesn't always work since sometimes the PS5 setting is not available to choose on a PC, so they try the closest setting.
4) When they choose a scene on a PS5 where the frames isn't capped, for example somewhere where the PS5 can't hit 60fps and test on that, the problem is that on a PS5 you don't know if its a GPU limit or a CPU limit or whatever, which can result in improper results thanks to point #1 & #2 above.
5) The whole system is different, graphics API, optimizations, different drivers ...etc can play a role in the performance as well, too many variables.

For example in their latest video the PS5 was 5% to 10% faster than the simulated RX6700 based PC.

While I do commend Digital Foundry for their efforts (Which they can't do any better) its still far from a proper comparison, inwhich you keep all variables the same expect the GPU.
 
Joined
Oct 12, 2021
Messages
41 (0.04/day)
System Name Deneb
Processor AMD Ryzen 7 7800X3D
Motherboard Asus ProArt X670E-Creator WiFi
Cooling CPU: Noctua D15. Case additional : ML140 Pro blue PWM (400-2000) x1, ML140 Pro RGB PWM (400-1200) x2
Memory G.Skill Flare X5 DDR5 6000Mbps EXPO 2x16Go
Video Card(s) Geforce RTX 4070 Asus Dual 12Go
Storage SSD Solidigm (SK Hynix) P44 Pro 2To (Gen4). HDD WD80EAZZ (8To, CMR) x2 in AMD RAID1
Display(s) Asus VG27AQ 27" 1440p 165Hz ELMB Sync, Freesync/Gsync compatible
Case Fractal Design R6 Tempered Glass Black
Audio Device(s) Motherboard soundcard. Logitech Z623 2.1 THX speakers.
Power Supply Corsair RMx 2018 850W
Mouse Logitech G502
Keyboard Filco Majestouch Convertible 2 (USB/BT) TKL Cherry blue
Software Windows 11 Pro
Benchmark Scores Cinebench R23 18440, Geekbench 6 CPU 2671/14823, GPU 181588, 3DMark Speedway 4723, Steel Nomad 4019
That defeats the whole idea about the benefits of a shared memory.
Kinda, but this is still good in small form factor embedded systems or when the 2 devices need to actively work on the same set of data. But yes, usually the CPU and the GPU have their own DRAM, it's the case for most HPC systems, you may look at Nvidia Grace-Hopper systems for instance. But the benefit of shared memory is power consumption and cost, those huge memory buses cost a lot in wiring (requires much bigger IC packages and much more board layers), silicon space for the controllers, at the expense of performance.

Yeah the 6700XT was a typo from me, I meant RX6700.
Again, TFLOPs rating is one part of the graphics pipeline, not all of it, there are other things that affects performance depending on the game/moment, if at the exact moment it was compute limited then yes a PS5 will perform like a RX6700, but thats not always the case.
Yes but Digital Foundry did base themselves on gaming tests and found out the RX 6700 was indeed very close to the original PS5. Besides, if the architecture wasn't the same then you'd be right to dissociate TFLOPs and gaming perfs, but since it was supposed to be based on a similar architecture, it was a strong hint that they would be close.

You're the one who said the PS5 Pro TFLOPs rating must be FP16 since its seems high, its not, RDNA3 has dual issue FP32 SM which gives theoretical TFLOPs a 2x boost.


It doesn't make sense to you because you are used to PC Gaming where developers has to account for different GPUs, CPUs, OS ........etc, so that when you pop-in an upgraded GPU it just works, thats not the way its done on the consoles otherwise developers, for example, will be able to easily let PS4 games work at 4K60 on the PS5 without any development effort.
That's a good point and I can understand there are these sorts of compatibility problems that you don't encounter in the PC world. But still, requirements on frequencies seem way beyond the minimum separation between software and hardware. From a hardware engineer point of view, that's really not doable to account for this sort of things. Or then you make a DSP or this sort of specialized ASIC, but those are not programmed the same way a game is programmed, not even close.

The leaked PS5 Pro GPU specs match AMD Navi32 die inwhich the 7800XT is based on, it matches the 7800XT on everything except slightly lower GPU and Memory clockspeed plus the RT and upscaling hardware.

Digital Foundry has an AMD 4800S Desktop kit which is essentially a Microsoft Xbox Series X chip with a disabled GPU slapped on motherboard for PC users in China, they installed a RX6700 on it.
While that system is the closest you can get to try to run PS5-esque hardware on a PC it doesn't come with some flaws.

1) Since the iGPU is disabled the CPU can get all the 560GB/s bandwidth of the GDDR6 memory.
2) The CPU boosts to 4.0 GHz much higher than the PS5.
3) When they try to test games on the two system, they try to visually match PS5 settings on their simulated PC, that doesn't always work since sometimes the PS5 setting is not available to choose on a PC, so they try the closest setting.
4) When they choose a scene on a PS5 where the frames isn't capped, for example somewhere where the PS5 can't hit 60fps and test on that, the problem is that on a PS5 you don't know if its a GPU limit or a CPU limit or whatever, which can result in improper results thanks to point #1 & #2 above.
5) The whole system is different, graphics API, optimizations, different drivers ...etc can play a role in the performance as well, too many variables.

For example in their latest video the PS5 was 5% to 10% faster than the simulated RX6700 based PC.
Hmm, do you have the links ? I may be lazy but since those are quite old I couldn't find them easily. But arguably one of their latest video is the one I sent you and they do say the RX 6700 is "very very close", and the PS5 Pro should be close to a RX 6800 if not taking into account RT and super sampling, so I'm sorry to trust them more than you since they indeed put quite some effort into testing this.

While I do commend Digital Foundry for their efforts (Which they can't do any better) its still far from a proper comparison, inwhich you keep all variables the same expect the GPU.
It still doesn't make your own conclusion (PS5 Pro close to a 7800 XT) any more valid though. We will see but I would bet on Digital Foundry conclusion much more than yours.
 
Last edited:

Ruru

S.T.A.R.S.
Joined
Dec 16, 2012
Messages
12,958 (2.96/day)
Location
Jyväskylä, Finland
System Name 4K-gaming / media-PC
Processor AMD Ryzen 7 5800X / Intel Core i7-6700K
Motherboard Asus ROG Crosshair VII Hero / Asus Z170-K
Cooling Arctic Freezer 50 / Alphacool Eisbaer 240
Memory 32GB DDR4-3466 / 16GB DDR4-3000
Video Card(s) Asus RTX 3080 TUF OC / Powercolor RX 6700 XT
Storage 3.3TB of SSDs / several small SSDs
Display(s) Acer 27" 4K120 IPS + Lenovo 32" 4K60 IPS
Case Corsair 4000D AF White / DeepCool CC560 WH
Audio Device(s) Sony WH-CN720N
Power Supply EVGA G2 750W / Fractal ION Gold 550W
Mouse Logitech MX518 / Logitech G400s
Keyboard Roccat Vulcan 121 AIMO / NOS C450 Mini Pro
VR HMD Oculus Rift CV1
Software Windows 11 Pro / Windows 11 Pro
Benchmark Scores They run Crysis
I'm disapointted that there's no GTA V re-re-remastered.
 
Joined
Oct 6, 2023
Messages
31 (0.07/day)
Location
Bahrain
Processor AMD Ryzen 9 7950X
Motherboard ASUS ROG Crosshair X670E Hero
Cooling NZXT Z73 RGB
Memory 64GB G.Skill Trident Z RGB DDR5 6200MT/s CL30
Video Card(s) Zotac NVIDIA RTX 4090 Trinity
Storage 2TB Samsung 990 Pro M.2 SSD + 15.36TB Micron 9300 Pro U.2 SSD
Display(s) LG C9 UHD 120Hz OLED TV
Case Lian Li O11D XL ROG
Audio Device(s) Denon AVR + 5.1 Setup
Power Supply Corsair HX1000i
Mouse Logitech G903
Keyboard ASUS ROG Strix Flare
Yes but Digital Foundry did base themselves on gaming tests and found out the RX 6700 was indeed very close to the original PS5. Besides, if the architecture wasn't the same then you'd be right to dissociate TFLOPs and gaming perfs, but since it was supposed to be based on a similar architecture, it was a strong hint that they would be close.
The architecture isn't entirely the same but its close, PS5 is based on RDNA2 but with modifications that can affect performance, and again compute performance is one part of the equation, not all of it.

Hmm, do you have the links ? I may be lazy but since those are quite old I couldn't find them easily. But arguably one of their latest video is the one I sent you and they do say the RX 6700 is "very very close", and the PS5 Pro should be close to a RX 6800 if not taking into account RT and super sampling, so I'm sorry to trust them more than you since they indeed put quite some effort into testing this.


It still doesn't make your own conclusion (PS5 Pro close to a 7800 XT) any more valid though. We will see but I would bet on Digital Foundry conclusion much more than yours.
These aren't my claims/conclusion, these are the facts we have right now, PS5 Pro = slightly downclocked 7800XT.

RX6700 = RDNA2 36CU @ 2450 MHz (11.29 TFLOPs) / 10GB dedicated @ 160 bit @ 16 Gbps (320 GB/s)
PS5 = RDNA2 based 36CU @ 2233 MHz (10.29 TFLOPs) / 16GB shared @ 256 bit @ 14 Gbps (448 GB/s)

7800XT = RDNA3 60CU @ 2430 MHz (37.32 TFLOPs) / 16GB dedicated @ 256bit @ 19.5 Gbps (624.1 GB/s)
PS5 Pro = RDNA3 60CU @ 2344 MHz (36 TFLOPs) / 16GB shared @ 256 bit @ 18 Gbps (576 GB/s)

PS5 has ~%10 less compute and %40 more bandwidth vs RX6700 ---> PS5 = RX6700?
PS5 Pro has ~%3.7 less compute and ~%8.4 less bandwidth vs 7800XT ---> PS5 Pro nowhere near 7800XT?

Zero logic.

We are running in circles here, you can believe whatever you want to believe, these won't change the facts.
Next time I'll put a 7900XTX on Linux on a slower system and when I see its not performing as well as a 7800XT on a faster Windows machine, I'll claim that 7800XT>7900XTX, backed up by "facts" lol.
 
Joined
Oct 12, 2021
Messages
41 (0.04/day)
System Name Deneb
Processor AMD Ryzen 7 7800X3D
Motherboard Asus ProArt X670E-Creator WiFi
Cooling CPU: Noctua D15. Case additional : ML140 Pro blue PWM (400-2000) x1, ML140 Pro RGB PWM (400-1200) x2
Memory G.Skill Flare X5 DDR5 6000Mbps EXPO 2x16Go
Video Card(s) Geforce RTX 4070 Asus Dual 12Go
Storage SSD Solidigm (SK Hynix) P44 Pro 2To (Gen4). HDD WD80EAZZ (8To, CMR) x2 in AMD RAID1
Display(s) Asus VG27AQ 27" 1440p 165Hz ELMB Sync, Freesync/Gsync compatible
Case Fractal Design R6 Tempered Glass Black
Audio Device(s) Motherboard soundcard. Logitech Z623 2.1 THX speakers.
Power Supply Corsair RMx 2018 850W
Mouse Logitech G502
Keyboard Filco Majestouch Convertible 2 (USB/BT) TKL Cherry blue
Software Windows 11 Pro
Benchmark Scores Cinebench R23 18440, Geekbench 6 CPU 2671/14823, GPU 181588, 3DMark Speedway 4723, Steel Nomad 4019
The architecture isn't entirely the same but its close, PS5 is based on RDNA2 but with modifications that can affect performance, and again compute performance is one part of the equation, not all of it.



I was talking about the Digital Foundry article comparing the original PS5 with GPUs.

These aren't my claims/conclusion, these are the facts we have right now, PS5 Pro = slightly downclocked 7800XT.

RX6700 = RDNA2 36CU @ 2450 MHz (11.29 TFLOPs) / 10GB dedicated @ 160 bit @ 16 Gbps (320 GB/s)
PS5 = RDNA2 based 36CU @ 2233 MHz (10.29 TFLOPs) / 16GB shared @ 256 bit @ 14 Gbps (448 GB/s)

7800XT = RDNA3 60CU @ 2430 MHz (37.32 TFLOPs) / 16GB dedicated @ 256bit @ 19.5 Gbps (624.1 GB/s)
PS5 Pro = RDNA3 60CU @ 2344 MHz (36 TFLOPs) / 16GB shared @ 256 bit @ 18 Gbps (576 GB/s)
The fact here is that the PS5 Pro is supposed to have 45% perfs more than the original one, and that's not the case with the 7800 XT, so it's not a slightly downclocked 7800 XT, THAT is a fact. I don't know where your numbers about the PS5 Pro come from.

PS5 has ~%10 less compute and %40 more bandwidth vs RX6700 ---> PS5 = RX6700?
I already explained about bandwidth, this is shared with the CPU, hence the additional bandwidth.

PS5 Pro has ~%3.7 less compute and ~%8.4 less bandwidth vs 7800XT ---> PS5 Pro nowhere near 7800XT?

Zero logic.
You are giving numbers about the PS5 Pro out of nowhere.

We are running in circles here, you can believe whatever you want to believe, these won't change the facts.
Next time I'll put a 7900XTX on Linux on a slower system and when I see its not performing as well as a 7800XT on a faster Windows machine, I'll claim that 7800XT>7900XTX, backed up by "facts" lol.
The thing is I'm not the only one saying the PS5 Pro won't be close to a 7800 XT. So you're right and Digital Foundry or Linus are wrong ?
 
Joined
Oct 6, 2023
Messages
31 (0.07/day)
Location
Bahrain
Processor AMD Ryzen 9 7950X
Motherboard ASUS ROG Crosshair X670E Hero
Cooling NZXT Z73 RGB
Memory 64GB G.Skill Trident Z RGB DDR5 6200MT/s CL30
Video Card(s) Zotac NVIDIA RTX 4090 Trinity
Storage 2TB Samsung 990 Pro M.2 SSD + 15.36TB Micron 9300 Pro U.2 SSD
Display(s) LG C9 UHD 120Hz OLED TV
Case Lian Li O11D XL ROG
Audio Device(s) Denon AVR + 5.1 Setup
Power Supply Corsair HX1000i
Mouse Logitech G903
Keyboard ASUS ROG Strix Flare
I was talking about the Digital Foundry article comparing the original PS5 with GPUs.
What you requested is in the two links I sent, read/watch them.
The fact here is that the PS5 Pro is supposed to have 45% perfs more than the original one, and that's not the case with the 7800 XT, so it's not a slightly downclocked 7800 XT, THAT is a fact. I don't know where your numbers about the PS5 Pro come from.

You are giving numbers about the PS5 Pro out of nowhere.
Numbers come from MooreLawIsDead leak ~6 months ago, where he posted detailed specs of the PS5 Pro which he got it from a developer who was briefed by Sony about the device, these specs turned out to be 100% correct.

BTW his YouTube video got deleted because of a complaint by Sony, so yeah these are facts.

His latest leaks shows a game gaining way over 45% performance, this is not proven yet though.
The thing is I'm not the only one saying the PS5 Pro won't be close to a 7800 XT. So you're right and Digital Foundry or Linus are wrong ?
Testing 101 states in order to test something properly you need to have all other variables the same, I proved to you that DigitalFoundry tests wasn't as it had too many other variables that can affect the result, so you can't take proper conclusion from it, its nothing more than a scientific estimation. Mine is based on facts, the specs don't lie.
 
Joined
Oct 12, 2021
Messages
41 (0.04/day)
System Name Deneb
Processor AMD Ryzen 7 7800X3D
Motherboard Asus ProArt X670E-Creator WiFi
Cooling CPU: Noctua D15. Case additional : ML140 Pro blue PWM (400-2000) x1, ML140 Pro RGB PWM (400-1200) x2
Memory G.Skill Flare X5 DDR5 6000Mbps EXPO 2x16Go
Video Card(s) Geforce RTX 4070 Asus Dual 12Go
Storage SSD Solidigm (SK Hynix) P44 Pro 2To (Gen4). HDD WD80EAZZ (8To, CMR) x2 in AMD RAID1
Display(s) Asus VG27AQ 27" 1440p 165Hz ELMB Sync, Freesync/Gsync compatible
Case Fractal Design R6 Tempered Glass Black
Audio Device(s) Motherboard soundcard. Logitech Z623 2.1 THX speakers.
Power Supply Corsair RMx 2018 850W
Mouse Logitech G502
Keyboard Filco Majestouch Convertible 2 (USB/BT) TKL Cherry blue
Software Windows 11 Pro
Benchmark Scores Cinebench R23 18440, Geekbench 6 CPU 2671/14823, GPU 181588, 3DMark Speedway 4723, Steel Nomad 4019
What you requested is in the two links I sent, read/watch them.

Numbers come from MooreLawIsDead leak ~6 months ago, where he posted detailed specs of the PS5 Pro which he got it from a developer who was briefed by Sony about the device, these specs turned out to be 100% correct.

BTW his YouTube video got deleted because of a complaint by Sony, so yeah these are facts.

Regarding the detailed specs you mention, I found is this :
As The Verge noted at the time, changes in AMD’s architecture means it’s difficult to directly compare teraflops directly between PS5 and PS5 Pro, and that in reality the comparison is more like 10.28 versus around 17 teraflops (indeed, the documentation leaked by Moore’s Law is Dead suggests “rendering is about 45% faster”).
The Verge also argues :
it looks like Sony might be moving from what’s effectively an AMD Radeon RX 6700 inside the PS5 to something close to the Radeon RX 7800 XT.
It's in terms of hardware and architecture, because they do also mention +45% in terms of perfs, and the 7800 XT again is not +45% compared to the RX 6700 but +70-75% (except in RT). Back to the start of the problem.
Everyone, including me, says it's close to a 7800 XT, especially because 7800 XT has currently the newest known AMD architecture and RT units, and it's the same number of CU, but it doesn't match in terms of perfs, hence Digital Foundry and others saying it would be a quite badly downclocked 7800 XT if it was close to be one. I do think that's what everyone's meaning, but you can still hope for Mark Cerny and his team to be depreciating their console while it's in fact better than what they said... I don't believe they did that to be frank. And as the Verge says (and I kinda implied it too), TFLOPs don't compare between architectures, so we can't conclude anything as long as we don't test it. For me what you show is not facts, the only fact we have is this +45% rendering perfs.

His latest leaks shows a game gaining way over 45% performance, this is not proven yet though.
Hmm this leak doesn't show that much FACT as you seem to insist on this point. At least it's not very elaborate... What game was tested, was it in RT ? etc etc. It's promising, but "up to 45%" seems like a more reliable number again here. I mean you kinda imply a leak is more reliable than the official statement by Sony at this point...

Testing 101 states in order to test something properly you need to have all other variables the same, I proved to you that DigitalFoundry tests wasn't as it had too many other variables that can affect the result, so you can't take proper conclusion from it, its nothing more than a scientific estimation. Mine is based on facts, the specs don't lie.
Well regarding MooreLawIsDead, Reddit automod doesn't agree with you :
 

Attachments

  • Capture d'écran 2024-09-17 013607.png
    Capture d'écran 2024-09-17 013607.png
    96.9 KB · Views: 27
Last edited:
Joined
Jul 5, 2013
Messages
28,208 (6.74/day)
No thank you Sony. The PS5 is STILL an ugly as hell overpriced unit. When you folks get back to practical designs at a reasonable price, like the PS4/PS4Pro, I'll buy one. Also, if you do away with the disc drive and physical media, I'll NEVER buy it. Make sure you continue physical media and that those games WILL play without the internet.
 
Last edited:
Joined
Feb 1, 2019
Messages
3,655 (1.70/day)
Location
UK, Midlands
System Name Main PC
Processor 13700k
Motherboard Asrock Z690 Steel Legend D4 - Bios 13.02
Cooling Noctua NH-D15S
Memory 32 Gig 3200CL14
Video Card(s) 4080 RTX SUPER FE 16G
Storage 1TB 980 PRO, 2TB SN850X, 2TB DC P4600, 1TB 860 EVO, 2x 3TB WD Red, 2x 4TB WD Red
Display(s) LG 27GL850
Case Fractal Define R4
Audio Device(s) Soundblaster AE-9
Power Supply Antec HCG 750 Gold
Software Windows 10 21H2 LTSC
No they didn't. People keep mentioning this, it's not true. The options are not very clearly marked but you can definitely backup your console without live services (this is not new, this is there from day one!). I think this confusion started because the PS5 didn't have the same direct option the PS4 had to allow for migration, but you can still backup your console to a USB drive including your saved games.



They should design this better but the options are clearly there!



Your math is way off, you don't need PS Plus premium for saves or to play online, PS Plus essential is enough and goes for 59£ a year (48£ if you buy during black friday or days of play in june). PS Plus Premium is the equivalent-ish to Xbox Game Pass with access to games on a subscription base.

The final long term price can vary wildly, if you're counting on buying a controller per year you should also count on buying a mouse per year and a keyboard every two years when comparing to PC :D I can tell you I still have my Dualsense from launch day, I don't feel any drift but if I did I'd take the thing apart and just fix it (I know the average won't but anyway...)
You actually tested you can backup to a USB stick (not drive) and then restore it from the stick without a factory reset on PS5 games?

When I searched it people are saying the same as my experience, ps4 BC saves no problem, native ps5 game saves, some kind of limitation.
 
Joined
Aug 23, 2024
Messages
34 (0.29/day)
The message is clear....consoles will be much expensive in the future(PS6).
It is neither good or bad....4k high fidelity visuals needs beefy hardware...costly ones.
It also means we will wait for PS6 much longer than we thought.
Sony won't release even higher end stationary console while £700 mashine will be on the market any time soon - +5 years.
PC graphics will also benefit from better playstation hardware but we will see how those ports will perform on personal computers.
I just truly hope PS5 Pro bring real stable 4k 60fps gaming without compromise on game quality settings...
 
Last edited:
Joined
Jun 18, 2021
Messages
2,567 (2.01/day)
You actually tested you can backup to a USB stick (not drive) and then restore it from the stick without a factory reset on PS5 games?

When I searched it people are saying the same as my experience, ps4 BC saves no problem, native ps5 game saves, some kind of limitation.

No and as far as I know that's indeed not possible, but that doesn't mean you need to pay to backup your data which you don't. It could be improved (it probably will before PS6 or now with PS5 Pro) but the statement "you have to pay to backup" is completely false.

The message is clear....consoles will be much expensive in the future(PS6).
It is neither good or bad....4k high fidelity visuals needs beefy hardware...costly ones.
It also means we will wait for PS6 much longer than we thought.
Sony won't release even higher end stationary console while £700 mashine will be on the market any time soon - +5 years.
PC graphics will also benefit from better playstation hardware but we will see how those ports will perform on personal computers.
I just truly hope PS5 Pro bring real stable 4k 60fps gaming without compromise on game quality settings...

Blame the competition (or lack there of), but either way I still think that this will be temporary. They'll probably see the sales numbers slow and progressively decrease the prices and the base PS6 will most likely sell for the same 500$ price tier every recent entry console.
 
Joined
Nov 26, 2021
Messages
1,702 (1.52/day)
Location
Mississauga, Canada
Processor Ryzen 7 5700X
Motherboard ASUS TUF Gaming X570-PRO (WiFi 6)
Cooling Noctua NH-C14S (two fans)
Memory 2x16GB DDR4 3200
Video Card(s) Reference Vega 64
Storage Intel 665p 1TB, WD Black SN850X 2TB, Crucial MX300 1TB SATA, Samsung 830 256 GB SATA
Display(s) Nixeus NX-EDG27, and Samsung S23A700
Case Fractal Design R5
Power Supply Seasonic PRIME TITANIUM 850W
Mouse Logitech
VR HMD Oculus Rift
Software Windows 11 Pro, and Ubuntu 20.04
The architecture isn't entirely the same but its close, PS5 is based on RDNA2 but with modifications that can affect performance, and again compute performance is one part of the equation, not all of it.





These aren't my claims/conclusion, these are the facts we have right now, PS5 Pro = slightly downclocked 7800XT.

RX6700 = RDNA2 36CU @ 2450 MHz (11.29 TFLOPs) / 10GB dedicated @ 160 bit @ 16 Gbps (320 GB/s)
PS5 = RDNA2 based 36CU @ 2233 MHz (10.29 TFLOPs) / 16GB shared @ 256 bit @ 14 Gbps (448 GB/s)

7800XT = RDNA3 60CU @ 2430 MHz (37.32 TFLOPs) / 16GB dedicated @ 256bit @ 19.5 Gbps (624.1 GB/s)
PS5 Pro = RDNA3 60CU @ 2344 MHz (36 TFLOPs) / 16GB shared @ 256 bit @ 18 Gbps (576 GB/s)

PS5 has ~%10 less compute and %40 more bandwidth vs RX6700 ---> PS5 = RX6700?
PS5 Pro has ~%3.7 less compute and ~%8.4 less bandwidth vs 7800XT ---> PS5 Pro nowhere near 7800XT?

Zero logic.

We are running in circles here, you can believe whatever you want to believe, these won't change the facts.
Next time I'll put a 7900XTX on Linux on a slower system and when I see its not performing as well as a 7800XT on a faster Windows machine, I'll claim that 7800XT>7900XTX, backed up by "facts" lol.
The 7800 XT has one thing that the PS5 Pro doesn't: a 64 MB last level cache which increases effective bandwidth far more than the 19.5 GT/s figure indicates. In addition, due to sharing a memory interface with the CPU, the PS5 Pro's GPU doesn't get to consistently use all 18 GT/s of its 256-bit memory interface.
 

f0ssile

New Member
Joined
Jan 5, 2023
Messages
24 (0.03/day)
Location
Italy
The 7800 XT has one thing that the PS5 Pro doesn't: a 64 MB last level cache which increases effective bandwidth far more than the 19.5 GT/s figure indicates. In addition, due to sharing a memory interface with the CPU, the PS5 Pro's GPU doesn't get to consistently use all 18 GT/s of its 256-bit memory interface.
If only that were all... He's forcing where he wants in an absurd way. Saying that the PS5 GPU bandwidth is not disturbed at all by the "work" of the processor is ridiculous.

Not only that, it is obvious that the latency of the GDDR is not suitable for the processor, which among other things already lacks 3/4 of L3 cache.

As has already been rightly said, shared memory is good for saving consumption and occupation. The rest are against, latency and bandwidth disturbed by the 2, unlike the PC where the bandwidth is sumable and the latency is clean, and optimal for the processor.

In the same video he posts, DF clearly says that usually the PS5 goes a little slower than the 6700 (4:22).
That 6700 was downclocked to the PS5 frequency, so the comparison he posts to say that the PS5 performs better is not even good...

The 6700 also has infinity cache, 80MB of cache comparable to a sort of L3.

The evidence of its alleged power does not exist. It is only a hope passed off as plausible.
Always from the video you can also see that they are pairing it with a 3070 Ti, which is at least 20% below the 7800 XT.

There is a video that refutes it 2 times, however it tries to pass it off in its favor.
The competence, objectivity and will to be so are missing. It is a losing challenge from the start... :)


Even the DF ones are ridiculous:
First they put it at the level of 3070 Ti, then in interviews he says (always him) that PS5-Pro is similar to a downclocked 4070, but the price of 550 puts it in full...

They are trying hard in another article trying not to contrast PS5-Pro to PC, a platform that in the tests that count is almost always penalized.

The latest news in hindsight is DLSS which gains value 6 years later, when something that resembles it is available on PS.
Same thing they did for years before SSDs and the more frequent 60fps on consoles.

They are so unbalanced that they broke the scale as soon as they touched it... Yet the masses follow them like a flock.

I don't know, maybe that's what they deserve...
 
Last edited:
Joined
Feb 1, 2019
Messages
3,655 (1.70/day)
Location
UK, Midlands
System Name Main PC
Processor 13700k
Motherboard Asrock Z690 Steel Legend D4 - Bios 13.02
Cooling Noctua NH-D15S
Memory 32 Gig 3200CL14
Video Card(s) 4080 RTX SUPER FE 16G
Storage 1TB 980 PRO, 2TB SN850X, 2TB DC P4600, 1TB 860 EVO, 2x 3TB WD Red, 2x 4TB WD Red
Display(s) LG 27GL850
Case Fractal Define R4
Audio Device(s) Soundblaster AE-9
Power Supply Antec HCG 750 Gold
Software Windows 10 21H2 LTSC
No and as far as I know that's indeed not possible, but that doesn't mean you need to pay to backup your data which you don't. It could be improved (it probably will before PS6 or now with PS5 Pro) but the statement "you have to pay to backup" is completely false.
Sure for you I will change it to you have to pay for convenient backups.
 
Joined
Oct 12, 2021
Messages
41 (0.04/day)
System Name Deneb
Processor AMD Ryzen 7 7800X3D
Motherboard Asus ProArt X670E-Creator WiFi
Cooling CPU: Noctua D15. Case additional : ML140 Pro blue PWM (400-2000) x1, ML140 Pro RGB PWM (400-1200) x2
Memory G.Skill Flare X5 DDR5 6000Mbps EXPO 2x16Go
Video Card(s) Geforce RTX 4070 Asus Dual 12Go
Storage SSD Solidigm (SK Hynix) P44 Pro 2To (Gen4). HDD WD80EAZZ (8To, CMR) x2 in AMD RAID1
Display(s) Asus VG27AQ 27" 1440p 165Hz ELMB Sync, Freesync/Gsync compatible
Case Fractal Design R6 Tempered Glass Black
Audio Device(s) Motherboard soundcard. Logitech Z623 2.1 THX speakers.
Power Supply Corsair RMx 2018 850W
Mouse Logitech G502
Keyboard Filco Majestouch Convertible 2 (USB/BT) TKL Cherry blue
Software Windows 11 Pro
Benchmark Scores Cinebench R23 18440, Geekbench 6 CPU 2671/14823, GPU 181588, 3DMark Speedway 4723, Steel Nomad 4019
If only that were all... He's forcing where he wants in an absurd way. Saying that the PS5 GPU bandwidth is not disturbed at all by the "work" of the processor is ridiculous.

Not only that, it is obvious that the latency of the GDDR is not suitable for the processor, which among other things already lacks 3/4 of L3 cache.

As has already been rightly said, shared memory is good for saving consumption and occupation. The rest are against, latency and bandwidth disturbed by the 2, unlike the PC where the bandwidth is sumable and the latency is clean, and optimal for the processor.

In the same video he posts, DF clearly says that usually the PS5 goes a little slower than the 6700 (4:22).
That 6700 was downclocked to the PS5 frequency, so the comparison he posts to say that the PS5 performs better is not even good...

The 6700 also has infinity cache, 80MB of cache comparable to a sort of L3.

The evidence of its alleged power does not exist. It is only a hope passed off as plausible.
Always from the video you can also see that they are pairing it with a 3070 Ti, which is at least 20% below the 7800 XT.

There is a video that refutes it 2 times, however it tries to pass it off in its favor.
The competence, objectivity and will to be so are missing. It is a losing challenge from the start... :)


Even the DF ones are ridiculous:
First they put it at the level of 3070 Ti, then in interviews he says (always him) that PS5-Pro is similar to a downclocked 4070, but the price of 550 puts it in full...

They are trying hard in another article trying not to contrast PS5-Pro to PC, a platform that in the tests that count is almost always penalized.

The latest news in hindsight is DLSS which gains value 6 years later, when something that resembles it is available on PS.
Same thing they did for years before SSDs and the more frequent 60fps on consoles.

They are so unbalanced that they broke the scale as soon as they touched it... Yet the masses follow them like a flock.

I don't know, maybe that's what they deserve...
That's interesting, I didn't notice about the L3, totally agree with you.
 
Joined
Nov 20, 2021
Messages
118 (0.10/day)
Flawed comparison, and funny enough even your incomplete PC is still more expensive than a PS5 Pro.

1) Ryzen 7 5700X3D is not $150, its ~$200, And the 5700X (Cheapest 8C Ryzen on sale for PCs) is ~$160.
2) The 5700X doesn't come with a cooler (+$15 A good non noisy cooler).
3) The $75 AM4 doesn't come with WIFI, Bluetooth and PCIE Gen4 combined, (+$20 for a WIFI6/BT5 card).
4) RX6800 is inferior, PS5 Pro has a custom GPU, the GPU part is a slightly downclocked 7800XT with RDNA4 RT hardware.
5) Blu-ray drive is not required to game on a PS5 Pro so remove from both.
6) Where are the input devices on the PC? LOL (+$13 for cheap keyboard and mouse) and (+$70? For a Dual sense).
7) Where is the OS on the PC? LOL.
8) Vertical stand is not required to play on a PS5 Pro, and even if you want one you don't have the buy the official one, you can buy a third party for like $10.
9) PS+ is not required to play on PS5 Pro, if you choose to get it then the extra benefits it provides offsets the cost you pay for the service.

Even when you buy the cheapest part that satisfy/close to satisfy PS5 Pro specs and get the Windows key for like $5 (Not legal?) you are still looking at $1000 minimum, $300 more than a PS5 Pro.
1) Micro Center has the 5700X3D on sale for $150 and at the time WAS the price. It’s usually under that by $10 or $20 and it’s more of the gamer CPU. Also thank you for making the cheaper with the 5700x suggestion…

2) you’re right and oh no had to add a $30 cooler…

3 So why can’t you use an Ethernet port? So that $20 can towards that cooler and it’s always better to play wired than wireless.

4) RT maybe but with FSR that’s argument is null and void. Rasterization is the only difference and that’s down to basically having MORE BANDWIDTH for the GPU. Doesn’t matter if the 7800XT is dowclocked if it’s getting choked by the width bus bandwidth JUST like the PS5. PS5 has the GPU of the 6700 non XT and doesn’t even get CLOSE to that performance. Not even close to a 6600XT. So I don’t expect to be anywhere near the 7800 XT and much closer to the RX6800.

5) only added that to make it fair and without makes the PS5 Pro MORE EXPENSIVE

6) PS5 Pro comes with a controller already but doesn’t have a stand so add $30(or $10 to save some money)to that and your right the PC doesn’t come with Keyboard and Mouse (for a monitor you can use your TV just like a console can) so add about $50(or replace the BLU-Ray Drive with K&B)

7) Windows can be found for less $20 and Linux is free lol(thanks Steam Deck bet..I mean users )

8) From the census from the PS community, it is required so sure

9) You’re absolute right…you don’t NEED to play online but if you want to play games online that’s NOT FREE TO PLAY AKA regular COD and not Warfare or Fortnite. But games like FFXIV DOES need a PS subscription. Also this one of of the argument that doesn’t make sense to me. Why buy a new device and NOT use it to its full ability? If you just want to play PS games, you could have saved yourself $500 and get a PS4(or Pro) and still have a great time(or you know build a PC that doesn’t need the latest and greatest tech).

The math I did is current and doesn’t involved used PC parts. I don’t know where you got a $1000 as I provided the math that it doesn’t.
 
Last edited:
Joined
Oct 6, 2023
Messages
31 (0.07/day)
Location
Bahrain
Processor AMD Ryzen 9 7950X
Motherboard ASUS ROG Crosshair X670E Hero
Cooling NZXT Z73 RGB
Memory 64GB G.Skill Trident Z RGB DDR5 6200MT/s CL30
Video Card(s) Zotac NVIDIA RTX 4090 Trinity
Storage 2TB Samsung 990 Pro M.2 SSD + 15.36TB Micron 9300 Pro U.2 SSD
Display(s) LG C9 UHD 120Hz OLED TV
Case Lian Li O11D XL ROG
Audio Device(s) Denon AVR + 5.1 Setup
Power Supply Corsair HX1000i
Mouse Logitech G903
Keyboard ASUS ROG Strix Flare
Regarding the detailed specs you mention, I found is this :

The Verge also argues :

It's in terms of hardware and architecture, because they do also mention +45% in terms of perfs, and the 7800 XT again is not +45% compared to the RX 6700 but +70-75% (except in RT). Back to the start of the problem.
Everyone, including me, says it's close to a 7800 XT, especially because 7800 XT has currently the newest known AMD architecture and RT units, and it's the same number of CU, but it doesn't match in terms of perfs, hence Digital Foundry and others saying it would be a quite badly downclocked 7800 XT if it was close to be one. I do think that's what everyone's meaning, but you can still hope for Mark Cerny and his team to be depreciating their console while it's in fact better than what they said... I don't believe they did that to be frank. And as the Verge says (and I kinda implied it too), TFLOPs don't compare between architectures, so we can't conclude anything as long as we don't test it. For me what you show is not facts, the only fact we have is this +45% rendering perfs.
RDNA2 to RDNA 3 TFLOPs comparison is more or less valid if you are talking about rasterization performance and take into account that RDNA3 has dual issue FP32 ALUs (TLDR divide RDNA3 TFLOPs by 2 and compare to RDNA2).

As I said before about performance, believe whatever you want to believe.
Hmm this leak doesn't show that much FACT as you seem to insist on this point. At least it's not very elaborate... What game was tested, was it in RT ? etc etc. It's promising, but "up to 45%" seems like a more reliable number again here. I mean you kinda imply a leak is more reliable than the official statement by Sony at this point...
Read the last few words of the sentence you quoted.
Well regarding MooreLawIsDead, Reddit automod doesn't agree with you :
Every detail Sony gave about the PS5 Pro was leaked by him correctly 6 months ago, whether he was reliable or not before doesn't matter to me, on the PS5 Pro he was dead correct and I would actually believe him if he leaked something else about it, up to you whether or not you believe him.

The 7800 XT has one thing that the PS5 Pro doesn't: a 64 MB last level cache which increases effective bandwidth far more than the 19.5 GT/s figure indicates. In addition, due to sharing a memory interface with the CPU, the PS5 Pro's GPU doesn't get to consistently use all 18 GT/s of its 256-bit memory interface.
Which is why I'm saying its based on Navi32/7800XT and not Navi32/7800XT.

1) Micro Center has the 5700X3D on sale for $150 and at the time WAS the price. It’s usually under that by $10 or $20 and it’s more of the gamer CPU. Also thank you for making the cheaper with the 5700x suggestion…

2) you’re right and oh no had to add a $30 cooler…

3 So why can’t you use an Ethernet port? So that $20 can towards that cooler and it’s always better to play wired than wireless.

4) RT maybe but with FSR that’s argument is null and void. Rasterization is the only difference and that’s down to basically having MORE BANDWIDTH for the GPU. Doesn’t matter if the 7800XT is dowclocked if it’s getting choked by the width bus bandwidth JUST like the PS5. PS5 has the GPU of the 6700 non XT and doesn’t even get CLOSE to that performance. Not even close to a 6600XT. So I don’t expect to be anywhere near the 7800 XT and much closer to the RX6800.

5) only added that to make it fair and without makes the PS5 Pro MORE EXPENSIVE

6) PS5 Pro comes with a controller already but doesn’t have a stand so add $30(or $10 to save some money)to that and your right the PC doesn’t come with Keyboard and Mouse (for a monitor you can use your TV just like a console can) so add about $50(or replace the BLU-Ray Drive with K&B)

7) Windows can be found for less $20 and Linux is free lol(thanks Steam Deck bet..I mean users )

8) From the census from the PS community, it is required so sure

9) You’re absolute right…you don’t NEED to play online but if you want to play games online that’s NOT FREE TO PLAY AKA regular COD and not Warfare or Fortnite. But games like FFXIV DOES need a PS subscription. Also this one of of the argument that doesn’t make sense to me. Why buy a new device and NOT use it to its full ability? If you just want to play PS games, you could have saved yourself $500 and get a PS4(or Pro) and still have a great time(or you know build a PC that doesn’t need the latest and greatest tech).

The math I did is current and doesn’t involved used PC parts. I don’t know where you got a $1000 as I provided the math that it doesn’t.
3) You trying to make a PC comparable to a PS5 Pro, since the PS5 has WIFI then you need to include one as well, WIFI is more convenient.
4) PS5 = less than 6600XT? LOL OK.
6) Stand for PS5/PS5 Pro is not required, its optional, I'm using mine without a stand, also you need both K&M and a controller to operate a Gaming PC, you can't play all games on a just a K&M.
7) Linux is free but unreliable for high end PC Gaming, cheap Windows keys are illegal.
8) Optional, not required, however its so valuable that you are likely to get it since it has so many games and extra benefits that its totally worth it.
9) Which is why its called optional, you have the option to get it or not, either way you'll enjoy the console whether you got it or not.

How the PC is $1000? Well add all the correct parts and you'll see how you can't get it below $1000 unless you go used or cut corners.
 
Joined
Oct 12, 2021
Messages
41 (0.04/day)
System Name Deneb
Processor AMD Ryzen 7 7800X3D
Motherboard Asus ProArt X670E-Creator WiFi
Cooling CPU: Noctua D15. Case additional : ML140 Pro blue PWM (400-2000) x1, ML140 Pro RGB PWM (400-1200) x2
Memory G.Skill Flare X5 DDR5 6000Mbps EXPO 2x16Go
Video Card(s) Geforce RTX 4070 Asus Dual 12Go
Storage SSD Solidigm (SK Hynix) P44 Pro 2To (Gen4). HDD WD80EAZZ (8To, CMR) x2 in AMD RAID1
Display(s) Asus VG27AQ 27" 1440p 165Hz ELMB Sync, Freesync/Gsync compatible
Case Fractal Design R6 Tempered Glass Black
Audio Device(s) Motherboard soundcard. Logitech Z623 2.1 THX speakers.
Power Supply Corsair RMx 2018 850W
Mouse Logitech G502
Keyboard Filco Majestouch Convertible 2 (USB/BT) TKL Cherry blue
Software Windows 11 Pro
Benchmark Scores Cinebench R23 18440, Geekbench 6 CPU 2671/14823, GPU 181588, 3DMark Speedway 4723, Steel Nomad 4019
RDNA2 to RDNA 3 TFLOPs comparison is more or less valid if you are talking about rasterization performance and take into account that RDNA3 has dual issue FP32 ALUs (TLDR divide RDNA3 TFLOPs by 2 and compare to RDNA2).

As I said before about performance, believe whatever you want to believe.
I believe one thing, it's the official statement of +45% rendering power for PS5 Pro over previous gen, the rest are leaks or speculations.

Read the last few words of the sentence you quoted.
I didn't quote a sentence so I had a hard time figuring out what u were meaning. You mean his leaks were removed by Sony right ? So what ? It doesn't mean leaks are right... Actually it's even worse for Sony if the leaks are wrong and they will want to have them removed even more.

Every detail Sony gave about the PS5 Pro was leaked by him correctly 6 months ago, whether he was reliable or not before doesn't matter to me, on the PS5 Pro he was dead correct and I would actually believe him if he leaked something else about it, up to you whether or not you believe him.
How can you be so sure if there is not yet any official confirmation of the leaked info regarding flops and such ??

Which is why I'm saying its based on Navi32/7800XT and not Navi32/7800XT.
??

3) You trying to make a PC comparable to a PS5 Pro, since the PS5 has WIFI then you need to include one as well, WIFI is more convenient.
4) PS5 = less than 6600XT? LOL OK.
6) Stand for PS5/PS5 Pro is not required, its optional, I'm using mine without a stand, also you need both K&M and a controller to operate a Gaming PC, you can't play all games on a just a K&M.
7) Linux is free but unreliable for high end PC Gaming, cheap Windows keys are illegal.
8) Optional, not required, however its so valuable that you are likely to get it since it has so many games and extra benefits that its totally worth it.
9) Which is why its called optional, you have the option to get it or not, either way you'll enjoy the console whether you got it or not.

How the PC is $1000? Well add all the correct parts and you'll see how you can't get it below $1000 unless you go used or cut corners.
While I do think the PS5 Pro will be slightly cheaper than a PC with equivalent hardware, it's getting closer than it has ever been. I did the maths with japanese prices, 120k yen for the PS5 Pro, and I was able to build a very similar machine with a 4060 Ti for around 140K yen. I agree the PS5 Pro is still decently positioned vs PCs in pure gaming use, and if you don't use services requiring monthly subscriptions, but it's pretty damn close now, and since with a PC you can do so much more than just gaming, I do think the PC is getting the advantage overall here.


By the way the PS5 Pro is now finally in the TPU database. It's not final data but it's also very different from the data you claim to be already facts :
At least the flops are very different to say the least. Not sure if it will end up being true or not, but what's sure is that your data aren't that much facts yet.
 
Last edited:
Top