assurdum
Banned
I'm not sure to understand your continuous ping pong in different fronts. Who ever talks about issue?
-
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.
Imo MS locking at sustained speeds is a mistake to be corrected
- Thread starter longdi
- Start date
rnlval
Member
PS5 has a 40 CU scale RDNA 2 based GPU. NAVI 10 is a 40 CU scale RDNA.
assurdum
Banned
Navi 10 again hasn't the cache scrubs inside the ps5. Furthermore CUs if I'm not wrong are tiny in navi10 because RDNA1.
Last edited:
rnlval
Member
Sony's PS4's customized L2 cache was largely a non-issue when my GCN 1.0 7970 1Ghz murdered it. You're repeating PS4 style argument and I'm NOT going to fall for it, again.
rnlval
Member
Hint, PS4 GPU's L2 cache line write protection during a flush.
assurdum
Banned
Wut. Jesus you are chaotic as hell with such many argument mixed in the bug. The hell it has to do PS4 with the ps5. It's even tough to follow you. Are you in purpose just to create confusion or whatelse? Go to inform yourself about the ps5 cache subsystem if you really to care and stop to mess all those useless stuff. It seems clear to me you don't even care of the matter. Man even Road to the ps5 is fine at this point.
Last edited:
rnlval
Member
1. Reminder, this is an XSX topic.
2. Sony doesn't have visibility over XSX's semi-custom changes.
3. PS5 GPU's cache scrubbers handle GPU's cache coherency issues with the CPU or DMA sources e.g. from SSD. If GPU's cache line referenced a memory address that was affected by SSD's DMA memory write, it shouldn't flush the entire GPU L2 cache.
Last edited:
assurdum
Banned
And again what's your point? You don't believe to Cerny or the other engineerd claims who talks about how fundamental is custom cache system in the ps5 to save the bandwidth/RAM usage in the GPU? Fine. But it's hilarious by the same person who find more legitime to spread the worst perfomance on series X is there is a software glitch without any concrete prove. Almost paradoxical.
Last edited:
Mr Moose
Member
Isn't that a 3.7TF card?
rnlval
Member
PS5 GPU cache scrubber is familiar, Hint: CPU L2 cache line behavior vs other DMA access. I'm a C++ business software programmer.
Last edited:
assurdum
Banned
You put a lot of grasp here, isn't it? How did you know such stuff? What are the result on the ps5 performance for example? You mixed a lot of tech stuff and sometimes with some gross approximation and to finish to land to the nothing.
Last edited:
rnlval
Member
You're a hypocrite.
FlyyGOD
Member
I should listen to a youtuber why? These guys dont work for Sony or amd. They work on a couch from home.
rnlval
Member
I don't recall NAVI 10 has Vega 10's high bandwidth cache (HBC) coupled with NVMEs.
assurdum
Banned
Oh no you are. You barely know what's inside the ps5 and you discredit everything about it but you are perfectly fine to claim in the same time series X perform worse on multiplat because there is a software glitch, something is hardly demonstrable but it's a fact because who can't provide otherwise? Very smart
Last edited:
assurdum
Banned
I don't understand if parroting the specs details of everything has a real sense, especially when the conclusions are incredibly approximative and mostly personal.
rnlval
Member
You're a hypocrite while I describe PS5 GPU's cache scrubber function. You latched on PS5's "cache scrubber" PR just like yesteryear's PS4's L2 cache customization fanboys.
Tomeru
Member
Man, this isjust proof that forum experts are only good at writing long posts, and game devs are only good ("good") at making games. All that "difference in power is quite staggering" and "the power is there, its just reality" is such bullshit.
Everyone should stop talking out of their asses. And I mean everyone.
Though that would make gaf a ghost town, so let the fighting resume.
Everyone should stop talking out of their asses. And I mean everyone.
Though that would make gaf a ghost town, so let the fighting resume.
assurdum
Banned
And define the better performance on ps5multiplat caused to a software glitch on series X, is less fanboy? Really?
Last edited:
rnlval
Member
You made this topic personal i.e. "But I think an engineer is a bit more reliable.in such things than you with all respect." You have started a flamewar based on a personality war.
rnlval
Member
It didn't stop PS4 GPU L2 cache customization fanboys bashing GCN 1.0 7970s.
Papacheeks
Banned
Redtech gaming literally leaked Infinity cache. No one knew anything about that he did, and it was true. SO who ever he knows who either works or has close ties with AMD.
The thing about how custom PS5 is compared to PC counterpart is how early they had devkits and showing games with ray tracing. AMD's software package for ray tracing wasn't even done yet when Sony was developing games.
assurdum
Banned
Oh. A flamewar because I find your opinion partial and less reliable of the people who works on it? It's you who don't believe to the ps5 cache system is a thing, it's all fanboy chats blabla when even a chief engineer talk about it repeatedly. But meanwhile you substained absurdity like the software glitch on series X. The coherency.
Last edited:
rnlval
Member
Again, you latched on PS5's "cache scrubber" PR just like yesteryear's PS4's L2 cache customization fanboys. For the record, I watched Mark's PS5 presentation on "cache scrubber", it sounds similar and you claimed I didn't watch it? LOL
assurdum
Banned
Again how you can't see that is just your personal take? Geometry engine is used from generation to generation, now it means GE work the same on ps5? Let me admit for a second the method of the cache system on ps5 it's the same of the past (doubt it's exactly the same anyway) , nothing of revolutionary , but it lands to give better usage of RAM and bandwidth and squeeze more from the gpu. What would changes in the whole perspective of the things? It can't provide better performance because is nothing of new?
Again if a chief engineer bet so much on it, I give him the benefit of the doubt.
To end here, if performance on ps5 multipiat has surprised everyone it means maybe Sony engineers are not wrong at all? Or it's just acceptable to think the API not works on series X?
Last edited:
Coulomb_Barrier
Member
I think PS5-style variable clocks will be the new standard for consoles in the future.
It's an innovation that has rendered 12 tflops looking a bit silly after Xbox has lost all the multiplat performance face offs.
Whether XsX Pro will have variable clocks is debatable though, it might be too early for MS to copy their rival's design. But keeping 52CUs but allowing it to clock past the measly 1.8Ghz of the Series X, say to 2.3Ghz, would offer huge performance increases IF they removed other system bottlenecks.
It's an innovation that has rendered 12 tflops looking a bit silly after Xbox has lost all the multiplat performance face offs.
Whether XsX Pro will have variable clocks is debatable though, it might be too early for MS to copy their rival's design. But keeping 52CUs but allowing it to clock past the measly 1.8Ghz of the Series X, say to 2.3Ghz, would offer huge performance increases IF they removed other system bottlenecks.
alabtrosMyster
Banned
People used to say this for the PS4 as well.
Also, you seem to forget the ps5's cache... That diminishes (if not alleviate, or completely trounces) the memory speed difference.
alabtrosMyster
Banned
The PS4 is more like a 7850 - 7870 (pretty sure it's closer to the 7850).
This is pretty amazing what they are able to do with the hardware!
TheThreadsThatBindUs
Member
This.
What cache? I' ve recently watched road to ps5 and have not heard anything that suggests it, more so that it can be used with rasterization.
The only mentions are cache scrubbers in IO
Care to point out where? I might have missed it
Last edited:
rnlval
Member
NAVI 10 has a 252 mm2 size chip.
309 -252 = 57 mm2
That's 57 mm2 remaining for CPU and another semi-custom request changes e.g. CU audio, RT cores, cache scrubbers, hardware checkerboard, and 'etc'. For hardware backwards compatibility, semi-custom request changes from PS4 and PS4 Pro need to be ported into PS5 e.g. hardware checkerboard.
You can't fit BiG NAVI's ~100 mm2 128 MB L3 cache in a PS5 APU.
Last edited:
rnlval
Member
1. There's no court of law "discovery process" with NDA laden PS5 and XSX and this forum. Sony is less transparent with their hardware platforms e.g. it took a social media storm to extract hardware RT information from Sony.
2. Your "You barely know what's inside the ps5" claim is hypocritical. The operational word is semi-custom. Unlike you, I state cache scrubber's function NOT latching on PR technobabble.
3. Your "you are perfectly fine to claim in the same time series X perform worse on multiplat because there is a software glitch", It would be funny with just a cache scrubber feature resulted in XSX being half the power of PS5. Are you claiming PS5 being RX 6800 XT level GPU?
RX 6800 murdered PS5's results.
Last edited:
assurdum
Banned
Who ever said that? Has been MS really that more transparent about the series X limitations as the splitted bandwidth? He claims they can have 390 Gb and 560Gb available in the same time their paper which is completely false. So because they show to you the SOC they are really that more transparent of Sony? As substained 12 TF performance bullshit spread everyday ? Do you really want to engage this discussion? You don't believe custom cache system on ps5 give a true benefit, good. But let's not pretend to be more objective to everyone else when the same argument you use in defence of series X is the same bullshit PR but with the difference you believe to MS but Sony is full of shit because you haven't seen in the hot chip presentation. But of course if multiplat perfom really close doesn't counts as data because don't fit your narrative, isn't it?
Last edited:
Oh found some of Matts principal engineer of sony tweets. He uses a zen 2 cpu as an example and then a coherency engine (not cache?) to eliminate some bandwidth bottlenecks.
It is quite a bit too random to make any logical conclusion of how it works,.more so suggesting that rasterizarion benefits greatly from it.
Further clarification is needed
rnlval
Member
Hint: RX 5600 XT 36 CU at the same clock speed as RX 5700 36 CU, 5600 XT 36 CU OC's performance penalty is minor. RX 5600 XT has a multi-MB L2 cache to mitigate this issue.
5600 XT 36 CU has 336 GB/s memory bandwidth just like XSX's 6GB memory address range. Your "390 Gbs" value is not even correct. LOL
Last edited:
assurdum
Banned
Lol. Continue to hide your fanboism with the data spelling. You don't know a shit about the ps5 hardware but of course Sony lies about it, meanwhile you are totally engaged in the MS propaganda. And what address the range? You can't use 336 GB and 560 GB contemporary if you have a total of 560 GB, that's something is not clear in what MS provided as many other stuff. Now go to research on Wikipedia, seems the whole discussion is based on your search on it and the wait for. For me can end here, I'm not pretending to change your mind but doesn't act as you are the more impartial because eh.
Last edited:
rnlval
Member
With your screenshot that shows PS5 being nearly two times over XSX result, are you claiming PS5 has RX 6800 level GPU? The real fanboy comes from you. Hypocrite.
assurdum
Banned
What screenshot? Whenever I said such stupidity? And don't call him hypocrite please. You are the last person who can accuse someone about it .
Last edited:
rnlval
Member
rnlval
Member
You are supporting this argument https://www.neogaf.com/threads/imo-...istake-to-be-corrected.1577809/post-261215811
assurdum
Banned
Lol. No, it doesn't clarify how the bandwidth can stay at 560 GBs when it uses 336 GBs. It's literally impossibile in the same bus that's the problem of use separate speed in an unified bandwidth. Again you parrot whatever you seen posted by MS but you don't use the logic
32-bit datatype can NOT be strip across the entire 320-bit bus
Try again.
Last edited:
rnlval
Member
Are you claiming PS5 has RX 6800 level GPU?"What happened to your Rasterizer debate?"
No one's denying XSX's advantages. PS5 has its advantages too due to higher clock speed that's evident in more than one game's performance. Like I've said:
What happened to your memory bandwidth debate?
I can do this too.
assurdum
Banned
I'm not supporting anything. It's you who believe ps5 uses cache system is like the PS4 because you don't believe on Sony presentation, it's only propaganda blablabla but meanwhile you can't see how there are bottlenecks on series X when the first multiplat prove it. Now I don't say it can't beat ps5 in the future, indeed I fully expect it, but the difference will be so ridiculous than people will remember more their propaganda of the most powerful hardware the world and how it fails a the beginning of the generation than the better performance in the future multiplat. Imo. So I don't know even it's that worthy continue to discuss about it.
Last edited:
rnlval
Member
Note that BiG NAVI has a very fast 128 MB L3 cache at~ 1.11 TB/s with GDDR6-16000 16 GB at 512 GB/s. L3 cache hit rate is ~58%, hence BiG NAVI is faking a ~1.1TB/s VRAM (638 GB/s effective from L3 cache + 512GB/s from GDDR6)
Last edited:
assurdum
Banned
The hell it has to do with series X. You make a lot of confusion . Isn't it linked to the infinity cache ?
Last edited:
rnlval
Member
Apply the same principle when both XSX GPU and RX 6800 has fast and slow speed memory zones. For XSX, larger high-speed memory storage = higher hit rates.
Last edited:
assurdum
Banned
Lol. No. Absolutely. Are you crazy? Series X mock more the separate RAM/bandwidth architecture on pc. Even MS claimed it
Last edited:
Md Ray
Member
Nope. It has 7900 XT, if you ask me.