Thirty7ven
Banned
Funny OP, you didn’t mind jumping to conclusions without material evidence for 11 months straight. Is this some sort of born again Christian phase you’re going through?
-
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.
The DiRT 5 Update Shows Exactly Why Making Outlandish Conclusions Of Hardware Is Dumb
- Thread starter thicc_girls_are_teh_best
- Start date
- Opinion Drama Game Dev
ethomaz
Banned
I don’t believe it is possible to build that PC like PS5 or even Series X.
So you are comparing Apple to Oranges.
Blue Spring
Read my tears about xbox here --->
You are one of the xbox fanboys that went into hiding after the comparisions started rolling in. Suprised Dirt 5 not looking like trash on xbox anymore gave you enough confidence to crawl out of your hole.
P.s. fanboys that hide behind neutrality are the worst kind. Even Riky is superior to you.
Last edited:
Leyasu
Banned
It wasn’t in any way shape or form a reasonable conclusion.
The game was clearly not ready for release on the series console. The same people that jumped up and down for months about cross gen games holding things back, were the ones trying to say that those very same cross gen games were too potent for the XsX to run.
Its laughable, and they are totally transparent.
I stated a fact. The difference is software. There are a few minor differences like 18 active CUs per SE vs 26 but that is not going to make the series X get lower FPS than the PS5 from a raw hardware standpoint.
Totally doable. Not a 100% replacement but close enough to know that the primary factor for the framerate differences is software.
DeepEnigma
Gold Member
It's not "just software."
ethomaz
Banned
That is not true.
There is big hardware differences.
+ Unified memory pool
+ Decompress and DMAs units
+ I/O specific co-processors
+ Cache scrubbers
These are from top of my head... we can find more if start to get deep in the arch.
Last edited:
Fake
Member
He spend most of his time on the Next Gen thread doing walltexts. He should edit his old posts before made this thread.
DeepEnigma
Gold Member
This thread a late lazy load. Trying to take the
Last edited:
Fake
Member
The dude is so out of the touch that is quite depressing. I was laughting all the day from those comparisons thread, but I don't even care about Dirt 5 comparison because IMO the game is still very very last gen no matter what plataform or what DF say. I played bunch of time Dirt 3 on Xbox 360 and Dirt 5 looks like nothing drastic really change to the vision of the game.
Those people really take console wars too far.
Last edited:
DeepEnigma
Gold Member
I remember when Dirt 5 was being treated like an Xbox exclusive for the month or two prior with all the hooting and hollering at PR language comparisons between the two versions and the like.
Then the comparisons came out and that train was jumped off like bad breakup rebound.
ethomaz
Banned
The issue I see with this thread was that it was premature.
There is no data about the new patch.
If the patch have dropped the performance on 120Hz mode then the thread is pretty much a lie.
If the patch doesn't drop the performance on 120Hz mode then maybe he has a point.
Sure so a PC will perform worse than a Console with the exact same spec if both could run the same software (OS, Firmware, drivers etc). Since both consoles have those features you can almost cancel out on both sides and your delta on the PC will be similar to the raw hardware delta give or take a few %.
We could even test this by running a 4750g locked at 3.6Ghz + 6800 at 12Tflops and compare games like Gears 5 / FH4 with the PC set to as close as we can to the console IQ settings to see what the FPS comes out as.
Raekwon26
Member
Glad more people recognise those fake BS.
OP is completely full of shit and anybody whose encountered his BS (usually a BS of wall of text) know what he truly is.
Fake
Member
Unfortunately just who used to post on Next Gen Thread will know that.
oldergamer
Member
+ Unified memory pool
+ Decompress and DMAs units
+ I/O specific co-processors
Are also on Series X. Not sure if they have cache scrubbers, or even need them.
DeepEnigma
Gold Member
He's been saying that BOTH consoles are not the same as slapping PC parts together.
ethomaz
Banned
Last edited:
assurdum
Banned
Aside unified memory pool (but it's virtually splitted so) such unit who talking about are quite standard and nothing special to squeeze out more by the hardware. And forgive me but series X beg for cache scrubbers with such higher number of CUs. It no need for what reason exactly? Bandwidth is quite low for such CUs counts and it's even splitted.
Last edited:
Sure but the main differences between the two when it comes to framerates are GPU, a tiny CPU difference and software. The hardware difference you can simulate with PC parts and while the raw performance numbers will not match the delta will be close.
ethomaz
Banned
CPU is in most cases the big offence in framerate... not the GPU (that is true for GPU bound games that are less than CPU bound).
The consoles have hardware units that free up the CPU while PC not.
You can't simulate any console with PC parts.
Because you can't get the same hardware with PC parts.
Last edited:
thicc_girls_are_teh_best
Gold Member
I went into hiding? You sure about that? I'm an Xbox fanboy? Are you sure about that?
Because my post history says otherwise. But it's okay; you're allowed to be wrong, angry, hostile, and stupid all at the same time. It's what you seem to be good at
Well technically speaking, cache scrubbers are 100% unique to PS5 and there's no denying that. Unified memory pool requires clarification because there are three types of "split" pools we could infer from that. There's "split memory" like in the SNES/Genesis/PS1/Saturn etc. eras where each processor block has its own dedicated block of physical memory, often of different memory types and speeds, but the developer able to allocate specific resources to each specific block without need of shadow-copying assets from one to the other (i.e remove need of an asset being duplicated in two memory pools). The same is also true of semi-recent systems like the PS3.
Then there's "split memory" like on PC where the CPU has its own system RAM (DDR derivative) and the GPU has its own VRAM (usually GDDR derivative). The two processors talk to each other (and therefore share data between each other's memories) over PCIe interconnect but all data intended for the GPU's memory has to first go in system RAM and is then shadow copied to the GPU over PCIe and then written to GPU's VRAM. This is starting to change a lot though thanks to features like GPUDirectStorage and Smart Access Memory, though by and large PC is still a nUMA architecture.
And then there's "split memory" like in the Series X which is technically all the exact same memory technology (GDDR6), and still all interfaced to the same memory controller. The only thing that causes the "split" is the capacity of the modules; since GDDR (and all volatile memory actually) writes/reads in parallel channels, then channel access has to be even across the modules. That's why if a system like Series X wants to access, say, 11 GB of data, it has to access 10 GB first across the 10 GDDR6 modules of the top 1 GB of each chip, then, access 6 GB across the lower 1 GB of the 6 2 GB modules.
That can create its own complications and I think some of those MIGHT be in play with some 3P game performance at this time, but is that really an issue that can't be readily overcome? No. If you want a system with a REAL memory bottleneck, look at the Atari Jaguar with its hardware bug that basically prevented accessing the RAM directly. Or even the PS3, which had two physically different pools of RAM that prevented OS features like party chat from ever being implemented. The Xbox One can also arguably fall into that designation, since the ESRAM just wasn't enough to overcome the bandwidth limitations of the DDR3 (FWIW Infinity Cache is very similar in intent and implementation for RDNA 2 as the ESRAM was on XBO, except the RDNA 2 GPUs are nowhere near bandwidth-starved on their larger GDDR6 memory capacity, and the IC is 4x the size of XBO's ESRAM).
Compared to those systems (some of which like XBO and especially PS3, found ways to overcome those challenges), the Series X is a relative breeze to work with in terms of absolute memory management, if less elegantly than PlayStation 5. A lot of any complications would mainly come from the sheer increase in game complexity and scope we've been seeing gen over gen, which can compound with a given hardware setup and create nonlinear scaling of factors for trickier needed optimizations.
Last edited:
Not trying to do that though, perhaps try reading and then comprehending. Let's try again.
The primary hardware difference between the two consoles outside of the SSD is the GPU with a tiny amount of CPU but probably barely worth mentioning.
The primary difference between the two PCs I compared is again the GPU.
So what I am saying is that if you have two GPUs with a similar on paper delta to the consoles and compare them the higher specified one is faster when bot CPU bound.
This should also be the case for the consoles yet it evidently is not so lets try and think of the reason for this.
It could be the PS5 has secret sauce that allows the shader units to have better utilisation.
An alternative is that the PS5 software stack is better leading to more optimised code and or lower OS/driver overhead than on Series X.
To me the software stack answer seems far more likely.
Devonshire
Banned
Development since last generation often is done on PC hardware, so I’m not sure how this can be true.
Last edited:
ethomaz
Banned
Wut?
What the development being often on PC has to do with what I said? lol
You can't build a PC like PS5/Xbox because there is no hardware parts like it is on these consoles.
PC had to brute force with CPU a lot of tasks executed by specific hardware units on consoles.
Last edited:
Devonshire
Banned
Since the PS4/Xbox 1 generation, Sony and Microsoft consoles are built from mostly off the shelf parts...nothing there that can’t be approximated with PC hardware.
They’re basically Steam Boxes.
assurdum
Banned
About that, I can be totally wrong but I noticed a curious thing about the CUs on ps5 compared the series X. MS has clearly reported on series X, CUs perform the 20% better compared the CUs of the past generation.
Cerny to the other side claimed the 36 CUs on ps5 are like to have 52 CUs of the old PS4 on it which not seems exactly a 20% of more perfomance. Something doesn't up to my math but again I could be easily wrong and misunderstood something.
Last edited:
MasterCornholio
Member
Well there is the speculation that the CUs are different between the two. However without a die shot we really don't know if that's true.
assurdum
Banned
Die shot are already leaked but no one has reported anything about the CUs unfortunately.
The link there
Last edited:
ethomaz
Banned
There is no x-ray die shot for PS5 yet.
MS claim the CUs have a 25% perf/clock compared with Xbox One X’s CUs.
That is the same increase claimed by AMD from GCN to RDNA.
RDNA to RDNA 2 has more ~15% increase in perf/clock on CUs... I’m saying ~15% but it is actually double digit... it can be a bit lower than 15 or a bit more.
Overall by AMD GCN’s CUs to RDNA 2’s CUs has a bigger than 38% (using 11% RDNA to RDNA 2) increase in perf/clock.
Last edited:
MasterCornholio
Member
I meant an x-ray so we can see how everything is organized on that die.
JackMcGunns
Member
6800 is ~52% faster than radeon 7 at 4k and both are 60 CUs. The 6800 also clocks 24% higher on average.
For the Vega -> RDNA2 IPC gain to be 38% it would mean the 24% clockspeed increase only accounts for 10% of the performance uplift (100 x 1.38 x 1.1 = 152 incase you are wondering why 38% + 10% does not equal 52%).
We know that RDNA2 clockspeed scaling is better than that so it cannot be that vega -> RDNA2 is a 38% IPC uplift and is likely closer to 25%.
This also does not account for the fact that 6800 has 96 ROPs vs the 64 in R7.
ethomaz
Banned
It is from AMD docs.
GCN to RDNA: 25% perf/clock increase on CU
RDNA to RDNA 2: double digit perf/clock increase on CU
Let’s say GCN is 100.
GCN = 100
RDNA = 100 + 25% = 125
RDNA 2 = 125 + xx%
For 11% = 138.75
For 15% = 143.75
For 19% = 148.75
I don’t believe they said double digit if it was 20%... that is the range I believe it is.
That basically matches with your Vega to RDNA 2 factoring the clock difference... clock scaling has does not affect perf/clock calc because both cards are running below the limit where you start to see issues with clock scaling.
Last edited:
It is multiplactive not additive.
Perf/clock/CU is also a nonsense stat because it will not factor bottlenecks so is very unlikely to be seen in a real use case.
I also do not think we will see a huge IPC gain going from RDNA to RDNA2 because the 6900XT with a higher clockspeed and 2x the CUs is only a bit more than 2x faster at 4k.
Anyway as shown the actual real world performance gain going from VEGA to RDNA 2 does not match the perf/cu/clock claims because those numbers won't be seen in a real workload.
ethomaz
Banned
I agree but both my comparisons were not real world performance... it is the actual IPC of the chips.
Like you said real world performance has a lot of variables and so bottlenecks that affect these theoretical numbers.
fran.fernandez
Member
Absurd.
let’s wait for PlayStation patch them. Later the Xbox , later the PlayStation, later the Xbox...
better, let’s wait 7 years. Lols
let’s wait for PlayStation patch them. Later the Xbox , later the PlayStation, later the Xbox...
better, let’s wait 7 years. Lols
Fredrik
Member
Yeah I think this gen will be an exhausting roller-coaster for those highly invested in having a gap, more patches will likely come for a long time, for all games.
And how many here has even bothered to start the game after the new patch? Raise your hand, please.
Instead of starting the game and check if there are issues and most importantly if they even matter, while actually playing it and having some fun, people just sit around grumping about downgrades or what not while waiting for a DF analysis to stare at a fps counter and play the console war game instead...
The game is actually fun. I wish more people talked more about this than used it as dumb console war fuel. It’s not like it’s the best racer ever but it’s pure arcade rally and we really don’t get many of these games anymore so I think we should make sure this is not the last one. Just go play it, even with some issues (still there occasionally on XSX which you should know if you’ve started it) it’s still solid fun.
Last edited:
Game is splendid .Yeah I think this gen will be an exhausting roller-coaster for those highly invested in having a gap, more patches will likely come for a long time, for all games.
And how many here has even bothered to start the game after the new patch? Raise your hand, please.
Instead of starting the game and check if there are issues and most importantly if they even matter, while actually playing it and having some fun, people just sit around grumping about downgrades or what not while waiting for a DF analysis to stare at a fps counter and play the console war game instead...
The game is actually fun. I wish more people talked more about this than used it as dumb console war fuel. It’s not like it’s the best racer ever but it’s pure arcade rally and we really don’t get many of these games anymore so I think we should make sure this is not the last one. Just go play it, even with some issues (still there occasionally on XSX which you should know if you’ve started it) it’s still solid fun.
II_JumPeR_I
Member
It is lol....
GHG
Gold Member
Don't be too hard on him, he only wrote a couple of hundred paragraphs jumping to conclusions in the run up to release.
Great Hair
Banned
Didnt they in the process ruin the PS5 performance? with the newest patch.
MiyazakiHatesKojima
Banned
Phil won.
Similar threads
- Poll
- Poll