Mr.Phoenix
Member
Better to install on the external drive. Speed is about the same. Or the difference is negligible.
100%
-
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.
PS5 Pro devkits arrive at third-party studios, Sony expects Pro specs to leak
- Thread starter Bernoulli
- Start date
rnlval
Member
RTX 3080's GPU design is not a problem when it has 10 GB, 12 GB, and 20 GB VRAM-equipped SKUs.
There's nothing wrong with consumers pressuring NVIDIA to increase VRAM at certain price points.
FYI, RDNA 3.5 is on the road map.
Minimising AMD's RT problem is still limited by the dual issue 33 TFLOPS shader scaled GPU.
rnlval
Member
The blocking behavior with all RDNA 2 GPUs is for non-hardware accelerated BVH transverse, NOT with hardware-accelerated BVH tests.
rnlval
Member
That presentation is very misleading, on the RT part.
The performance difference in games is almost negligible between RDNA2 and RDNA3.
Just look at the 6650XT and the 7600. Both have 32 CUs and similar clocks.
But in RT there is only a 2% performance improvement.
Higher resolution (higher pixels) can expose additional geometry workload.
Focus on 6900 XT (80 CU, 128 ROPS) and 7900 XT (84 CU, 192 ROPS) scaling.
There is a 34.5% RT improvement between similar 80-to-84 CU-scaled AMD GPUs.
PS5 Pro is claimed to have a 96 ROPS scaled RDNA-X-based GPU and RX 7700 XT has 96 ROPS with dual issue 35 TFLOPS. Unlike the game consoles, the PC GPU's TDP cooling budget is not shared with the CPU.
Hardware BVH transverse removes shader resource consumption for this workload, hence RX 7700 XT with hardware BVH transverse should land about RTX 4070 level to RTX 4070 Super.
PS; RX 7700 XT is factory overclocked beyond reference 7700 XT's 35 TFLOPS.
Based on TFLOPS and TMU scaling, RX 7900 XTX with hardware BVH transverse should be about RTX 4080 Super level.
Last edited:
rnlval
Member
AMD's copyright is on PS5 APU silicon.
rnlval
Member
Compute with 8 concurrent compute contexts has existed with Xbox 360. 1st gen GCN doubled this capability with two ACE (8 concurrent compute contexts for each ACE) units.
GCN's origin from Xbox 360 GPU.
CPU-GPU fusion with shared pointers existed in Xbox 360. This is an implied load design.
PS3's CELL-SPU is unable to perform the shared CPU pointers with RSX. CELL-SPU has an explicit load design.
Normal multi-core CPUs can share pointers.
Note why Xbox One has backward compatibility with Xbox 360.
Hawaii GCN has 8 ACE units and Sony's semi-custom Pitcairn GCN also has Hawaii 8 ACE units update.
Async compute(I/O via TMU path) is a workaround for GCN's 64 ROPS (I/O) forever problem when GTX 980 Ti has 96 ROPS (I/O). AMD made sure RDNA 2 and RDNA 3 had competitive ROPS I/O performance.
Implementing the full RT hardware acceleration is part of AMD's current "to-do" list.
AMD(ATI) GpGPU design won the game console battle against IBM SPU.
NVIDIA CUDA GpGPU design has a design win with Nintendo's game console.
Intel ARC GpGPU is the wild card.
Last edited:
I'd take $649. I want Cerny to make the most efficient configuration with APU and components reaching $600 to make.
If it only costs $20 more to manufacture in 4nm, then 5nm is a mistake.
300mm2, 4nm - $170
Mr.Phoenix
Member
Nothing stops them from switching to 4nm a year or two later when its price has come down. Same way the 6nm process they switched to for the PS5 in 2022 was available back in 2020 when the console launched.
Actually, what will really rock your knocker... 5nm process was already on the market by the time the PS5 launched, and 5nm made it to market before the 6nm process that the PS5 only started using in 2022. By that time, 5nm was almost 3 years old and 6nm was two years old. That should help give you guys some perspective on how quick to adopt price nodes sony is.
This is why I keep saying the idea that Sony would somehow jump onto the bandwagon of the latest and best chip process is completely unrealistic. Chances are, both the PS5pro and PS5slim will be on that 5nm node this year, so when you say "just" $20 more, extrapolate that to the like 30-40M chip contract Sony would be signing with AMD on that particular node.
You know, that's how they do it, sony doesn't just go and buy chips when they want them, they make an order for xxx million chips on a particular node, get a deal based on the amount and they are stuck with that until they exhaust that order.
This is the reason MS can't just jump onto the latest and greatest nodes, when MS placed their order for XSX and XSS chips on 7nm, they could have ordered like 30M chips (as an example, it could be more or less), the more they order, the less they pay for them. But if they have been unable to sell 30M XS consoles, then they can't go in and place an order for more chips on a newer node because they would still have to pay for the 7nm chips.
Killjoy-NL
Member
Isn't the Sony CES presentation today?
If there is no mention of a PS5 Pro, I'm going to continue pressing X to doubt.
If there is no mention of a PS5 Pro, I'm going to continue pressing X to doubt.
James Sawyer Ford
Gold Member
I wouldnt expect PS5 Pro to be announced until a few months before launch
Killjoy-NL
Member
PSVR2 was announced at CES.
James Sawyer Ford
Gold Member
It's a new device that doesn't compete with existing console sales.
Mr.Phoenix
Member
You should probably get a new controller, cause your x button would be worn out by September.
The only way they mention it now is if its coming in March.
Killjoy-NL
Member
Kind of irrelevant.
Besides, if rumors suggest that Sony expects leaks about the Pro devkits, announcing it today would make senss.
Killjoy-NL
Member
Why buy a new controller when one comes with my brandnew Pro day 1 if it does release?
Mr.Phoenix
Member
not irrelevant. PSVR2 is a standalone device that ONLY works with a PS5. Announcing it wouldn't make anyone go, "I will hold off on buying a PS5 until its released".
PS5ppro however can make people take a wait-and-see approach, even if not to buy it, but that they hope it coming means the PS5 gets a price drop. So it makes no sense to announce it months early.
And it leaking doesn't mean a damn thing. The people that would know of said leak are really small.
Killjoy-NL
Member
SmokSmog
Member
More like 50$~, last time VRAM was around 27$ per 8GB in bulk orders.
Keep in mind 8Gb = 1GB module
PS5 is using 16Gb modules
Last edited:
Rocco Schiavone
Member
Was the PS4 Pro announced at the CES?? Nope. So why do expect a PS5 pro announcement?? Console will be announced a few months before release.
Crayon
Member
GT7 wouldn't run off my expansion SSD. Black screen. Another person here had the same issue. I don't remember having that problem with anything else, but I guess that can happen.
Killjoy-NL
Member
Just saying I'll continue regarding it as nothing but rumors, since people are already 100% certain.Was the PS4 Pro announced at the CES?? Nope. So why do expect a PS5 pro announcement?? Console will be announced a few months before release.
I don't see any reason to treat this differently from any other rumor.
Sure, but a 300mm at 4nm will be a lot beefier and more power at only $20 more. Again, the base PS5 will cover the marketshare play, the pro doesn't have to conform to that. They should not be afraid to price it at $200-$300 more than the base model.
The PS5 base model has to be cheap. It has to be affordable for most in order to gain marketshare. Sony receives around 30% or less for every game sold, let alone PS+ subscription. That's how the console market operates. In other words, using the most reasonably affordable node for the PS5 that would cut the budget for perhaps $450 BOM was chosen. Again, the PS5 needs to be affordable to gain market saturation.
You are correct and I agree. That's why it is unrealistic to expect the PS6 to use the latest and best chip process. Unless you want it to be priced at $600 - $800 and Sony start to lose marketshare. That would be stupid. But we are not talking about new gen yet. This is mid-gen.
That's now how it works. There is this thing called value for money. If it's simply about the price, then PS5 should have been as weak as Series S selling at $299. Remember, it's you who gave the $20 dollar increase in value from 5nm to 4nm.
winjer
Gold Member
The issue is that N4 is being used a lot by NVidia, AMD and other companies for AI chips.
There might not be enough wafers left for a consumer product like a console.
PaintTinJr
Member
And this rare engineering sample of the the Cell BE processor, like the main one IIRC only SCEI copyright on it, and yet every man and his dog knows that all the PowerPC instruction set and PPE design belongs to IBM, and Toshiba own technologies within the Cell BE too, so that means absolutely nothing
Sony was part of STI with toshiba and ibm, that's very different from how the x86 license works with AMD and Intel and why Sony and MS has to have AMD make the chips and sell them to Sony.
Panajev2001a
GAF's Pleasant Genius
What disk did you get for it? What are the specs? Heatsink on or off?
Perrott
Member
The Pro won't stop regular console sales.
We knew that PS4 Pro was very real by April 2016 and it even was officially confirmed by Sony in June. It came out in November, but until then PS4 sales kept going on strong regardless of the rumors and then the fact that an enhanced console release was imminent. Hell, even after the Pro hit the shelves, the regular model continued to be the most sold of the two by a wide margin.
Killjoy-NL
Member
We also got another thread on Gaf stating that PS4 Pro sold 14.3M units, which is roughly <10% of total sales.
I'm willing to bet the majority of buyers was the core Playstation fanbase that double-dibbed, making the regular console sales argument insignificant.
PaintTinJr
Member
Are you being ridiculous? The shaders only have critical path blocking behaviour for their own compute shader, and only if written that way if not using acceleration. You have it back-to-front.
The blocking in vanilla RDNA2 like the PC cards and Series was shown in Xbox's technical breakdown of the series, and the blocking is caused by a CU being limited to either use a BVH accelerator or a texture unit at any one time, not both, and as shaders largely rely on texture samplers to get work done -and that capability is facilitated by the TMU - that is why it is either/or, BVH accelerator or TMU.
Mr.Phoenix
Member
In truth, they would likely be paying more than that anyways, the problem with being on the latest or most performant node is that at that point you are competing with everyone else and as such would have to pay a premium for fab allotment. There is a reason even AMD uses an older node for their MCDs and IO dies in their GPUs and CPUs respectively. Besides of course cache and mem PHY not shrinking as quickly as logic.
Actually happy you said this. "gain market saturation". I have said before that Sony is in the market share business. So My question for you is this, which of the following two options do you think Sony would go for?
- $400 PS5 + $700 PS5pro and sell 17M consoles per year? or
- $400 PS5 + $500 PS5pro and sell 20M consoles per year?
And if really in doubt as to why there is no need to be chasing the super high end, just look at the percentage of steam users in the $400-$600 GPU space and those in the $1000+ space, or look at the fact that even AMD is going to focus on mid-range GPUs $600ish cards and avoiding the high end with their next GPUs.
That is exactly how it works, especially if you are talking about value for money. Sony doesn't just add components looking for the best "value for money" and then total everything at the end of their shopping spree. Nope, they have a target, which could be $500 and they project that extrapolated into billions into their fiscal calendar spend YEARS in advance. Eg. in 2019 there were fiscal projections for 2020 showing how much they were expected to lose/make, which was reflective of the fact that there would be a hardware launch in that year. that's how FAR ahead these guys make their projections.
Once those projections are made, they shop out the best deals and tech to match those targets.
Last edited:
Why? Sony could have pre-hype for the core crowd and towards the release date go all out with marketing for casuals.
What good is it to keep the Pro a secret, assuming it's real now that there has been indirect confirmation of PS5+?
Mr.Phoenix
Member
Some reasons. The first is that they do not have to do any of that. They don't have to generate pre-hype for the "core" crowd, cause the core crowd doesn't need convincing. And the core crowd already probably has a PS5 anyway. The other main reason is that the last thing they want is for that core crowd to be advising causals to wait for the PS5pro and even being able to tell them all the ways it would be better than the normal PS5. The bottom line is that it could negatively impact sales right now. Info on it would leak eventually, but that's acceptable as long as that info stays within circles such as these.
PS5+? What indirect confirmation are you referring to?
The Immortals Of Aveum dev alluded to a higher graphics preset for game if there was Pro model. Now, that could be speculation but there's no reason Sony couldn't make an S model of the PS5 and called that a Pro.
Mark Maratea: On consoles only, it does an adaptive upscale - so we look at what you connected from a monitor/TV standpoint... and there's a slot in the logic that says if a PS5 Pro comes out, it'll actually upscale to different quality levels - it'll be FSR 2 quality rather than standard FSR 2 performance.
That seems like a weird thing to have for a theoretical console.
www.eurogamer.net
Mark Maratea: On consoles only, it does an adaptive upscale - so we look at what you connected from a monitor/TV standpoint... and there's a slot in the logic that says if a PS5 Pro comes out, it'll actually upscale to different quality levels - it'll be FSR 2 quality rather than standard FSR 2 performance.
That seems like a weird thing to have for a theoretical console.
Inside Immortals of Aveum: the Digital Foundry tech interview
Ahead of the launch of Immortals of Aveum, Alex Battaglia and Tom Morgan spoke with Ascendent Studios' Mark Maratea, Julia Lichtblau and Joe Hall about the game
www.eurogamer.net
Last edited:
PaintTinJr
Member
AFAIK it is more an issue of volume/price. IBM's PowerPC licensing is the same copyright situation,, just that they don't fabricate enough cheaply to compare to the amount PlayStation was going to make, so Sony took the lead at getting the chips made, just like Microsoft did with the Xenon, but Nintendo didn't with the Cube , Wii or WiiU AFAIK.
PaintTinJr
Member
I've had a little look at all the technical specs of those AMD cards and it is quite interesting how on the RT side, it looks like only the 6800XT is getting even semi optimised performance for that game.
IF you take the comparative teraflops of the bottom two cards, the 7700XT is wasting roughly 42% of its higher TFLOP(FP32) theoretical amount to get the same results in RT. But as it turns out, the RDNA2 card is doing two FP32 flops per StreamProcessor, compared to 4 FP32 flops for the RDNA3 card.
If you then take the 20TF/s of the RT 6800XT and divided that into the TFLOPs of the higher RDNA3 cards to get a frame-rate multiplier, to multiply the 26fps RT value, and then take 58% (42% inefficiency for RDNA3), it predicts each of those RT frame-rates within 1-3fps.
So I would say all those RDNA3 results are because the 64 streams per CU doing 4 FP32 instructions per clock are more wasteful with that game's RT implementation and probably not a hardware issue
Last edited:
Killjoy-NL
Member
At 14M sales, ~10% of the PS4 installbase made the jump to PS4 Pro.
Casuals don't care that much about Pro-consoles, minus that tiny minority of the minority with a PS4 Pro.
Edit:
Wrong percentage, it's a bit over 10%.
Last edited:
Mr.Phoenix
Member
Oh ok.. well, this can also just be good future-proofing.The Immortals Of Aveum dev alluded to a higher graphics preset for game if there was Pro model. Now, that could be speculation but there's no reason Sony couldn't make an S model of the PS5 and called that a Pro.
Mark Maratea: On consoles only, it does an adaptive upscale - so we look at what you connected from a monitor/TV standpoint... and there's a slot in the logic that says if a PS5 Pro comes out, it'll actually upscale to different quality levels - it'll be FSR 2 quality rather than standard FSR 2 performance.
That seems like a weird thing to have for a theoretical console.
Inside Immortals of Aveum: the Digital Foundry tech interview
Ahead of the launch of Immortals of Aveum, Alex Battaglia and Tom Morgan spoke with Ascendent Studios' Mark Maratea, Julia Lichtblau and Joe Hall about the game
It however does go in line with what I have been saying for ages now. Thet the PS5pro "Quality" mode would just take what was doing FSR2 with an internal rez of 900p-1080p ( OG PS5 performance mode) up to 1440p while hitting 60fps. All with the IQ presets of what was the quality mode on the OG console. And people's expectations should fall in line with what you need on the hardware side of things to accomplish that.
A good way to do this would be top just look at the GPUs that are averaging 65-75fps @1440p in games that were released in the last 12 months.
Ahhhh... the dual issue compute issue. A simple way to put this is that those FP32 TF numbers on the 7xxx series and Nvidia 4xxx cards, aren't quantifiably real. If you instead just half the TF numbers, to get their equivalent 6xxx series "type" TF, you begin to see that the numbers all line up.
AMD and Nvidia are claiming architectures that allows them to double the FP32 throughput, but that shit is nonexistent in any games thus far. If you ask me thus far they are just as good as stat padding, cause I can't understand how there is that much potential power sitting on the table and no dev is using it thus far.
PaintTinJr
Member
No, I don't think you are understanding what I wrote as this isn't a nvidia issue. This is about the instructions per (streamprocessor per) clock utilisation, and the switch from 2 to 4 FP32 instructions costs 42% in efficiency in the RT, which is either through sparse workloads per StreamProcessor on the RDNA3 software because of lack of optimisation or the cost of texture(sampler access in shaders) is possibly more costly. The consistent % cost across all the RDNA3 cards IMO definitely points to a purely software issue.
Mr.Phoenix
Member
You aren't wrong, with that it "points to a pure software issue", but you aren't right either.
This is VOPD, (vertex operations, dual), that's the what AMD introduced in RDNA3 that allows them double FP32 calculations per clock. Nvidia has something similar, albeit going about its implementation differently. Either way, that is why on both RDNA3 and anything from Nvidia 3xxx series and up, you have FP32 compute that is listed as the same as FP16 compute. They are listing the dual issue compute numbers, vs the actual logical core compute value. Hence why we now say things like RDNA2 TF vs RDNA3 TF.
If however, you ignore the dual issue stuff, that would mean you would be halving all those listed TF numbers they show for the affiliated GPUs, and all of a sudden, you see that the real-world performance we are getting, is in line with the percentage differences between them and previous GPUs. But to further prove this, this is not just an RT workload thing, take a look at any and every game benchmarked on these GPUs, those with and without RT, you will notice, every single time, that their listed compute numbers never translate to the percentage difference between them and the previous cards. Not even close.
And yes, an argument can be made that on the software side of things, devs are not taking advantage of this dual-issue compute stuff, which is why we re not seeing even a meaningful percentage boost to suggest something like that is being used, but that's got make you wonder, why is EVERY game, both Rt and non-RT based... not using it?
Crayon
Member
I wish I could remember rn but I can tell you it had an official ps5-ready endorsement from Sony. It even kinda matched the console. Heatsink, yes. Iirc, Sony has some kind of stake in the company.
PaintTinJr
Member
No, that's still not the issue IMO, as I'm not talking about FP16 as FP32, that's another thing entirely where precision is different. With FP16 you double the RDNA FLOPS(FP32) to get the half FLOPS(FP16), this is the instruction count and how you actually calculate native full precision FLOPs(FP32) on the cards. StreamProcessor count X instructions per (Processor) clock x boost clock
Think of the newer 4 instruction FP32 per processor as a vector unit in each processor, doubling the size of the vector unit does actual increase the compute, but if you are on average only (just over) half filling the vector unit (wasting a FP32 or more per processor per clock) then your utilisation is going to be down 40%, compared to a card that had more processors and has half the vector width, as the overspill cost is just once at the last row.( as seen by the 6800XT vs 7700XT)
When offline renderers use this silicon they do a prepass to align, bin pack and batch the workloads to give superior utilisation, and in Epic's follow up technical videos on UE5 early access the lead programmer did explain how the nanite and software lumen work as very long (two pass) compute shaders to maximise utilisation using FP16 - although the FP16 isn't the point. So there is examples of the work scaling on these TFLOPs number IMO..
In the non-rt workloads the inefficiency - as show in Alan Wake - goes from being slightly worse at 46% inefficient to 58% with the 7900XTX, which makes sense because as the frame-rate increases on a semi-complex workload (with lots of context switching) frame drops increase because of increasing stress on memory latency, which lowers utilisation.
Last edited:
playsaves3
Member
That dude is off the highest would be 699
schaft0620
Member
There should be no expectation of PS5 Pro at CES. If its launching in the fall, there is no benefit to announcing it early, in fact it could hurt sales. Sony would do a 3-6 month marketing beat. If they announce it at CES, I would say its coming out in March or April which according to rumors is not the case.
SIE itself is in the middle of a leadership change, I don't see massive announcement coming before a new CEO is announced.
I personally don't expect PlayStation to be at CES in any meaningful way (I hope I am wrong).
Last edited:
Killjoy-NL
Member
Fair enough, but I'm going by this thread talking about how Sony expects leaks.
Might as well beat them to it.
playsaves3
Member
I think your informing there would be a decent amount of high end gamers like pc gamers who would be far more interested in a 700 pro than a 500 one
Mr.Phoenix
Member
A $500 console, would likely have a BOM of $480-$520. Would be equivalent to a ~$800-$1000 PC. Or a PC gamer spending around $400 -$600 for a GPU upgrade. In such a situation, you can entice them to pick up a PS5pro instead.
A $700 Pro will likely have a BOM of $680-$750. Will be equivalent to ~$1300 -$1500 PC hardware. In this scenario, you are still selling to the people in the scenario above, as the people in this scenario who can afford to spend $1500 on a PC, wouldn't be interested in a PS5 for its specs. They would rather just spend on the PC.
Regardless of all that though, there really is no basis for Sony to make a $700 console. No matter what, there will be more people to sell to at $500 or even $600 than there would be to sell to at anything above that. We have seen how poorly a $600 console sells. I don't know where you are getting this impression that there are so many people willing to buy such a console.
hinch7
Member
They wouldn't show it off in CES anyways. They just launched the PS5 refresh a few months ago and the Pro will (most likely) be coming in the Holidays.
Last edited:
Crispy Gamer
Member
It's a bit irrational to expect Sony to announce it this early but I do expect some type of leak and even if the leak is substantial Sony still wouldn't acknowledge a Pro console until they're ready to sell it.