Good old Colonel James it's nice to meet you again after more than 10 years and see you still fighting for your team .
Proven wrong by what? just to get me a culture about it
Last edited:
-
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 Die Shot has been revealed
- Thread starter LivingD3AD
- Start date
Proven wrong by what? just to get me a culture about it
Loxus
Member
So way is the PS5 performing and sometimes even out performing the XBSX?
You guys are who makes this Forum look bad with your fan boy nonsense.
Tools alone isn't enough to explain why the PS5 is performing the way it is.
Engineering also plays a role.
James Sawyer Ford
Gold Member
can’t go higher because of thermal issues.
a 400Ghz increase on a wider chip will consume a ton more power that’s not realistic for a console form factor
longdi
Banned
Eh no. Im just curious and concern why one of the more definition feature upgrade from zen->zen2 had been dropped. Extra fpu muscle help for 3d games iirc.FUD machine.
The work arounds suggested here seems weak.
-Do it on gpu? Why not both?
-It's too hot! I thought part of the excitement moving from jaguar to zen2 is that we finally have a fuller fatter cpu now. But the liquid meta and continuous boost capabilities..
-Do it as per zen, split it between units and run another cycle..but we are zen2 now!
-Probably not as useful because ps5 need not run windows or something...but this is avx, i doubt my pc win10 requires or use avx256..
-mark cerny said it! ..but he just said 256bit, not avx256? He even hinted this shit gets hot and affect their variable boost design
Last edited:
Loxus
Member
Read my post again properly.
So your telling me RDNA 2 GPU's doesn't have Infinity Cache?
DaleinCalgary
Member
Can't we all just hold hands and agree PC sucks at HDR
Last edited:
because all games on which a comparison was made do not use a single hardware advantage features of the xsx. Hivebusters and gear tactics are the only one (and in any case taking advantage of all the features).
And still we having something like hitman 3 that reach to render 44% of more pixels on the screen.
We don't make the forum look bad, but users who think they have the truth in their pocket without knowing what's going on. The ps5 will improve its performance until reach + or - it's ceiling of 10.2tf . The xsx Will improve until it's perfomance reach closer to it's too at 12.1tf
Last edited:
sorry I was thinking you was taking about the ps5 )Read my post again properly.
So your telling me RDNA 2 GPU's doesn't have Infinity Cache?
the infinite cache it is due from the bus of only 256bit .. the xsx have a 320 bit bus (still the IC is a monster bw amplifier)
Last edited:
Md Ray
Member
Let me guess... Teraflops advantage?
Here's a reminder to the 'teraflops' crowd by Unity's principal engineer:
Basically means "most of the time GPUs aren't 100% ALU (TFLOPs) bound". He is also the developer of Claybook, and is showing evidence of the ALU usage using profiling tools.
This "it'll show even more down the road" isn't true because games/scenes can be bound by different parts of the GPU, it's not teraflops always. PS5 GPU has 22% more geometry, triangle rasterization throughout, and faster caches, etc. And when scenes are bound by these, PS5 will pull ahead or at least be on par XSX and we've seen this in multiple titles now.
Last edited:
Bryank75
Banned
Have they figured out why the PS5 GPU is slapping the XSX one yet?
I mean we all thought there would be some type of difference...like I thought it would be small (about 10%) and others thought it would be much bigger. But most didn't expect PS5 to come out ahead in so many head to heads or almost even when all is considered.
I mean we all thought there would be some type of difference...like I thought it would be small (about 10%) and others thought it would be much bigger. But most didn't expect PS5 to come out ahead in so many head to heads or almost even when all is considered.
The Hamburglar
Banned
Your concern is touching. The ps5s of the world are appreciative of your concern for their well-being and if they’re full rdna2 or not.
Why don’t you just type lol rdna1 and be done with it?
absolutely true.
now tell all the advantages of a wider gpu and you will understand why all the market and gpu maker are going that route )
Last edited:
alabtrosMyster
Banned
There is a point of diminishing returns on your yields, 400mhz can make a big difference in volyage (heat) on a giant APU like this, especially when everything has to run at full throttle all the time-for a relatively small gain in performance.
Somewhere an engineer made the call, it's probably the best they could muster given all the other factors that they had to account for.
DaleinCalgary
Member
SenjutsuSage
Banned
I feel they're close enough also, but maybe my definition of different performance tier is being looked at the wrong way. Nvidia and AMD whenever they release GPUs they have performance tiers. They have mid-range, and then they have their enthusiast options that are a bit better at higher resolutions/visual settings.
In Nvidia GPU terms I think Series X is the RTX 2080 and potentially better approaching or exceeding 2080 Super at full potential.
In Nvidia GPU terms I think PS5 is the RTX 2070 and potentially better approaching or matching 2070 Super at full potential.
Many would consider those cards fairly close also in practical terms, but we just know the 2080 cards are a cut above the 2070 cards. That's what I believe the difference to be between the two machines.
Why do I use the word exceeding for series x in relation to 2080 Super, but not for PS5 in relation to 2070 Super? Memory Bandwidth. There's a key memory bandwidth element that I think is being largely overlooked between the two consoles. Series X has 560GB/s for the GPU at its very best (yes I know, shared CPU and asymmetric design) PS5 has 448GB/s (also shared CPU and a much more ideal fully symmetrical design)
The key drawback I feel for Series X right now is in its ability to fully benefit from all that compute and use all that memory bandwidth in a satisfying and consistent enough fashion to really show what it can do. With 10GB as a hard limit for 560GB/s that isn't quite so easy, which is why Sampler Feedback Streaming is so vital, and also why Mesh Shaders is an important factor as well. But the faster short term solution is to use SFS I think.
If these two consoles were runners PS5 would be Tyson Gay in his proper track gear, Series X would be Usain bolt, but wearing these pants and shoes. Series X needs to get on its track gear
No PS5 only owner will ever feel like they got shafted is my take. Just because I feel it's ultimately going to prove to be a bit weaker compared to Series X doesn't mean I think the PS5 is weak or lacking in real capability. I don't.
Bryank75
Banned
I heard there were some bugs too...that guy Alex Bugaga mentioned on DF.
alabtrosMyster
Banned
All chips eventually run 'too hot'.
The rest is (on your part)
Same reason you don't walk with both your hands and legs... We have a better way to perform 'fpu' tasks now.
SenjutsuSage
Banned
You want the simplest answer?
Series X:
10GB of GPU Optimal Memory at 560GB/s (Series X GPU is using this, its performance is great)
6GB of Standard Memory at 336GB/s (If the GPU touches any of this RAM performance WILL suffer)
Meanwhile on PS5:
16GB at 448GB/s
That is the primary reason we see PS5 winning ANY performance battles so far. It's developers having games that are organized in their memory access patterns in a certain way, and it's just plain simpler to work with the PS5's fully symmetric memory design. There's less to plan for or think about. Series X's design requires more careful planning, or use of a feature like Sampler Feedback Streaming to greatly mitigate the just 10GB of faster RAM for the GPU. If the Series X GPU touches that other 3.5GB reserved for games past that 10GB, the GPU's performance will dip because now you have a 52 Compute Unit 1825MHz GPU now trying to survive on just 336GB/s of memory bandwidth. It becomes bandwidth starved.
Some developers handle this better than others. In cases where series x and ps5 perf are identical a dev has just opted for parity (don't have an issue here so long as the game is solid on both), or the additional headroom offered by Series X allows it to JUST match what PS5 is outputting or brings it slightly below. Control is clear evidence of there being more headroom for higher performance on Series X. That additional performance headroom on Series X existing is how you get cases like in Hitman 3.
That is the reason PS5 sometimes performs better. It's down to the memory design causing the Series X GPU to be bandwidth starved in momentary blips. There are solutions to this.
Lysandros
Member
He will mention that whenever convenient until the end of the generation.
Fredrik
Member
Overclocking 400MHz would be overkill but if I was MS I would absolutely try going up to 1.9 or 2GHz just to see what would happen. Iirc they increased the cpu clock on Xbox One.
thicc_girls_are_teh_best
Member
IIRC unified cache was only in reference to the CPU's L3$, not a unified last-level cache between CPU and GPU. The Infinity Cache was something separate and referencing the GPU specifically.
Maybe I've heard wrong, though.
And for all intents and purposes, right now they're practically performing on par with each other and that's the main takeaway in terms of analysis so far.
However Microsoft should be announcing some more features at their event later this month. I think DirectML is supposed to be focused on among a few other technologies (not all of them pertaining to Series consoles tbf). I'm curious if anything they discuss will elaborate on things they've mentioned in the past but haven't gone super in-depth on.
Some of this is true for sure: PS5 has a higher culling rate (17.84 gtri/s) and triangle rasterization rate (8.92 gtri/s). The cache part is more tricky though. Assuming 1 ns latency for the L0$, PS5's L0$ bandwidth should be about 10.172 TB/s. The thing is that while the GPU does run 405 MHz faster, that mainly affects per-CU cache bandwidth on a cyclic basis.
AKA if you break down L0$ bandwidth on a per-frame basis that's 169.53 GB/s per frame in a 60 FPS target on PS5 vs. 200.425 GB/s per frame for the same on Series X. However, like I was saying earlier this throughput on Series X only really comes with game engines that are designed to scale higher. Assuming a game is running PS5's GPU at peak, for similar performance on Series X you'd need to saturate 44 of the CUs to account for the slower GPU clock.
But that's just mainly in terms of achieving similar cache bandwidth rates; obviously once you saturate more of the Series X's GPU beyond 36 CUs that starts to give it some other advantages particularly with texture fillrate and BVH traversal intersections for RT. So I guess what I'm getting at is yes, it's always been more than TF, but there's also some idea Microsoft designed their system solely around TF and that actually isn't true, either.
Both Sony and Microsoft took a lot of other things into consideration besides raw TF peaks; they just prioritized many of those things somewhat differently.
Last edited:
FlyyGOD
Member
Or could it be that these are 1st generation ports of old games on New hardware? I don't think any developers have scratched the surface of what these machines are capable of.Why don't we wait until the RDNA 2 features are utilized before we make judgements on the power narrative.
IntentionalPun
Ask me about my wife's perfect butthole
DERP, I think you are right.
MasterCornholio
Member
I don't want to sound like an asshole but less talk and more delivering is what I want to see.
If they promise massive performance gains that will be what I will expect.
FatKingBallman
Banned
But Hitman 3 is performing better on PS5. What the hell you are babbling about.
assurdum
Banned
So basically it seems you haven't a single clue of what this thread talking about...but sure series X rulez
assurdum
Banned
Listen a single game is enough to prove the others are just rushed port, bad tools or payed by Sony secretly to screw the image of series X. You have to accept the reality of the fact, don't be a child.
Last edited:
Microsoft met their goals they are taking zero chances after the rrod cost them billion or 2. They have over cooled shit since and left plenty of headroom for a billion reasons. The original Xbox one was laughingly huge for this reason. The one x was a cooling beast like the series x. The series s is conservative clocked laughingly to avoid any cooling issues in a small form factor.
Clear
CliffyB's Cock Holster
Well, an awful lot of pc code isn't especially optimal. So if you are aiming for a broad-spectrum of applications the easiest route is just more of everything. There are plenty of people (and corporate clients especially) that will happily pay through the nose for performance gains, so there's less impetus to try the sort of radical rethink that we see with PS5's design.
As I've been saying, Cerny's team was free to approach the problem of price/performance without nearly so much legacy baggage as the Xbox team, who will have had to keep in mind that the software ecosystem was always going to include PC as a target.
Free will knight
Neo Member
Devs not having enough time to optimise the game for xbox is the most likely answer to that. Dev kits for ps5 came out early 2019 where as xbox dev kits came out late 2019. With a lot of countries going into lockdown early 2020 this restricted access to the dev kits since they had to stay at the studios . Which also explains why MS didn't have gameplay to show us last year and why halo was shown on a similar spec PC. Devs likely had maybe 14 to 16 months with the PS5 dev kits but 8 to 10 with xbox.
DESTROYA
Member
You really think SONY paid off a company to make Hitman 3 worse fo a competitor? LMAO
If you think any company would agree to make something they produce to purposely look bad for another console your pretty delusional.
Omega Supreme Holopsicon
Member
I think part of the problem of the series X is that it doesn't have dram cache for its ssd unlike the ps5, which I think does have additional dram for its ssd.
Also I think games could very well start using over 10GB of graphics memory at a time, thanks to the massive I/O asics and dram ps5 can not only use 11-12GB of graphics memory but it can do so per scene, as it is fast enough to stream 12+GB into memory for the next scene in about a second.
James Sawyer Ford
Gold Member
their power supply and cooling likely won’t allow for it, even small gains like that have a huge impact on power
we should ask to red gaming something for this insider info
ToTTenTranz
Banned
Glad to see sebbbi stepping in to clear some nonsense.
The PS5 also has the same amount of shader engines as the SeriesX, so less WGPs/CUs per work distributors, meaning ALU utilization should be higher.
The only clear-cut advantage I saw in the SeriesX was theoretical memory bandwidth, but it seems that has been somewhat mitigated by the memory contention issues due to the memory pools with different bandwidths.
I wonder if Microsoft didn't lose an opportunity here to just launch the console with all 10 channels populated with 2GB chips for a total of 20GB GDDR6, even if they had to charge an extra $50 for the console. They'd have both the bandwidth and memory amount advantages with no caveats other than price.
Regardless, IMO the SeriesX's biggest problem will be developers in general concluding that they need to make the game to the SeriesS first, as common denominator with lower RAM amounts, and then scale up. If they do the opposite then the result will be the SeriesS without high-FPS modes and/or no ray tracing effects.
The list of developers complaining about that has been steadily piling up.
If Infinity Cache becomes a crucial component of Navi 3x's chiplet architecture (e.g. for faster coherency between GPU chiplets?), then I think it's very likely yes.
Sony already has patents on multi-chip (multi-GPU or even multi-SoC) solutions, probably for the PS5 Pro.
Last edited:
yes, I am also sure that the I / O will have a fundamental role in the course of the generation but as someone has already said, it is not that Ms has not also thought about these situations. the SFS the Mesh shader the VRS they all come into play to substantially decrease the required bandwidth. but until devkits, engines, tools and ESPECIALLY devs will go in mutual agreement towards the use of these the console will not express its potential. To have some glimpse of his performances in the short term we can only hope in turn10, the coalition and playground ... usually it is the teams that first experiment with new technologies
thicc_girls_are_teh_best
Member
You're certainly not alone in that; Microsoft IMHO dropped the ball both by not having a 1P showcase ready by launch (they could've even delayed games like Bleeding Edge to spruce up with optimizations for Series X and S), and by virtually not hyping the Gears 5 Hivebusters DLC whatsoever (felt like a stealth release basically).
Relying too much on 3P timed exclusives nets you with games like The Medium which, well, it has great-looking backgrounds and the main character model looks nice enough (plus she has a nice butt
), but heavily uncanny valley and janky animations among other things. Microsoft'll need more impressive visual showcases to come through in order to start bringing a performance narrative to their way (or at least mitigate 3P performance delta if current trend continues). It's practically guaranteed. Only question is how large will the cache be. 32 MB - 64 MB would be my guess. If there's a Series X upgrade in the pipeline I guess that will also have some form of IC at similar size; by that point (2023), it should be relatively commonplace.
Seems both systems are topping out at around 215 watt TDPs, which is already a good bit for consoles. A GPU clock increase of even probably 75 MHz would probably add another 5-6 watts on top of that, which could probably exponentially increase thermal output (power and heat don't share a linear relationship).
If Microsoft do any clock adjustments or upgrades, it'd most definitely be CPU related. The GPU would cause too much an issue there and even if they could they are still bound by the memory bandwidth so that's more peak GPU power with less bandwidth per TF (and with the GPU clocked even higher they would need to probably increase the CPU clock by a dependent amount to make sure it doesn't bottleneck the GPU).
So basically, from greatest to least chance of clock adjustments for Series X would be CPU >>> GDDR6 memory clock >>>>>>>>> GPU clock
Clear
CliffyB's Cock Holster
Once again, this saves rendering bandwidth not data bandwidth. Everything you describe helps to discard redundant data, but what it doesn't address is feeding the data to the GPU. If you feed it a 10,000 vertex mesh, that is then culled and tesellated down to 5,000 for rendering, rasterization speed increases but the i/o load remains untouched.
MasterCornholio
Member
You're certainly not alone in that; Microsoft IMHO dropped the ball both by not having a 1P showcase ready by launch (they could've even delayed games like Bleeding Edge to spruce up with optimizations for Series X and S), and by virtually not hyping the Gears 5 Hivebusters DLC whatsoever (felt like a stealth release basically).
Relying too much on 3P timed exclusives nets you with games like The Medium which, well, it has great-looking backgrounds and the main character model looks nice enough (plus she has a nice butt
They certainly did screw up by not proving the systems power right up to launch. Since they bragged about the XSX power I was expecting a clear difference at launch. I'm listening to what people have to say about the XSX but I prefer if they actually deliver instead of just talking about what might be possible. It kind of reminds me of Sony and the PS3s Cell in a way.
What Microsoft needs more than ever is their own Demon Souls. Something that shows what the system is really capable of. Even Digital Foundry was extremely impressed with what Bluepoint achieved in that game. Now it's Microsofts turn to do something similar.
P.S Oh and no more Craigs please.
Tickle My Tendies
Member
Timestamped
FatKingBallman
Banned
Actually, it's more easy to scale down than scale up
MasterCornholio
Member
Epic Sax CEO
Banned
I'm the only one that don't like using Nvidia GPUs to compare?
Turing/Ampere and RDNA2 are completely different architectures making comparisons always inconsistent because each will deliver different results depending on the game.
i think that there will not be any problem regarding that with both architectures customization and the nvme ssd imho both console are over specced in that regard. I do not see how there will be a leap forward in the use of data in a way where a nvme with hw decompressor and other stuff will not be enough
Omega Supreme Holopsicon
Member
I think the lack of dram could cause issues with relying on fast streaming. Aren't both ps5 and Series X TLC ssds? TLC tends to have bad performance, sometimes a portion of the TLC ssd is converted into SLC to serve as cache and to improve performance. Not sure if either of the consoles are doing that.
Clear
CliffyB's Cock Holster
My point was mainly that what we are seeing is a fundamental philosophical difference in approach in how best to manage workload. To my mind a far more interesting thing to observe than just comparing spec-sheets.
Especially because these things are not mutually exclusive; Sony have gone fairly conservative in terms of core specs but are hoping to make up the shortfall with efficiency gains along the entire pipeline, however there's no reason these same alterations couldn't be applied to a much larger, faster, more state-of-the-art chip.
I just think we should want Sony's stuff to be as effective as advertised because it could have massive benefits for all AMD devices going forwards.
Proelite
Member
.
Last edited:
Starlight Lotice
Member
I wish I could understand the images but all I think of is this.
Guessing it is missing something?
Guessing it is missing something?