Sony Is Working On AMD Ryzen LLVM Compiler Improvements - Possibly For The PlayStation 5

Nope, 7 nm process tech enables twice transistor density for a given chip size when compared to 14/16 nm process tech.

@ronvalencia: if it's the most powerful one, it'll be the most expensive one.

It depends on the chip size. 7nm enables Vega 64 to have RX-580 like chip area size geometry e.g. 470 mm2 / 2 = 235 mm2

X1X's GPU (~285 mm2) is larger than RX-580 (~232 mm2).

The problem is DRAM. Near monopoly by the Koreans.

As always this will be limited to production cost at the end of the day.

But yes, finally, this jump is needed.

Not that it will matter because the same low-risk game-types will be pumped out.

Higher CPU enables PC level Ashes of Singularity level NPC count.

Yea, I've heard this from the last 2 console generations, but something always stops it happening.

Last 2 gens Sony shows off tech demo's of millions of cubes and physics... bet they've managed to do with that this gen is add some nice shiny effects into God Of War.

Yeah because PC graphics haven't been held back by console for years, right? Besides no PC games have something called graphics settings which we can already push way ahead of consoles. Don't see how's that would be a problem but I'm glad you've come out of a closet and accepted you're a cow pretending to be a hermit.

Because 750Ti doesn't have tehe memory, still it fared way better in many games and as per Capcom's (Not ShitEnix) own words they considered PS4 to be an equivalent of 570 which was a mid range GPU when they did a demo of Deep Down. Funny that might be too high what they estimated as Deep Down isn't a reality nor any other games that came close to the graphics depicted in the demo on the PS4.

Yeah, thank you for agreeing that Jaguar is borderline cell/tablet level CPU.

Horizon Developer Guerrilla Games Recruiting for New Game

http://www.pushsquare.com/news/2018/05/horizon_developer_guerrilla_games_recruiting_for_new_game

Dutch developer Guerrilla Games has started a hiring spree, suggesting that its next project is decided and is about to enter full production. We wonder if it could be a sequel to the studio’s massively popular breakout hit, Horizon: Zero Dawn?

@pc_rocks: I guess the concept of textures is new to you. Yeah, moo.

Yeah obviously because a cow with 8GB of shared GDDR5would definitely know what good textures are over a PC gamer. You really got me there and I'll agree with KaliBurd there you do like to go sidetrack when you have been owned for being a cow. Indeed that sounds like moo. BTW when is the lem insult coming because otherwise you know you had no comeback.

@pc_rocks: Tall words from someone who admitted consoles hold pc gaming back then tries to "own" me by bringing up how games all of a sudden have tons of options that don't hold the games back.

We have a difference of opinion. I'm not really out to one up you but it is a fact that if literally every game on console had some mandate to run at 60fps that it would affect the lowest common denominator in terms of specs. Generation 7 was a thing. Why try so hard?

Of course it will be Ryzen based, there is literally no other option because we know they won't go Intel or drop X86 for some type of ARM variant.

so in simple english, how many TF is it? and what is the probability of the PS5 being at least 10 teraflops?

@ronvalencia: So by your estimates, at same clock speeds, a ryzen core would perform 2-3x better than a jaguar core in games? That’s what I would estimate it at myself.

@hawkeye44: @kali-b1rd:

CPU will be heavily bottlenecked by GPU next generation most likely. Specially if they aim for 4k resolutions, so there is a big chance those ryzens are going to run at really low clocks to push more GPU performance forwards.

I don't think we will see much change on what we see now on that department. 2-3x performance boost over jaguar is also not much to talk about on performance increase.

About the GPU performance.

AMD already sits at 14tflop gpu's and that generation is already old by now, let alone in 2020 which is a full generation after next generation of gpu's. I wouldnt'be shocked if next generation gpu's will sit at 16tflop+. So yea 10tflop isn't going to be much of a issue. They probably go balls deep on GPU's next generation anyway as they will need to have metric tons of gpu performance to keep up.

@appariti0n:

Totally depends on what ryzen you use, a 1200 ryzen sits at cinebench 475, while a 2700x sits at 1850. The fastest CPU however is the 7980xe oc'ed on PC that sits at 4448 score.

The xbox one X cpu should sit at around 170 from what i could find. which would put it on 3x slower then entry ryzen.

To give you a idea how shit jaguar is, a CPU released in 2008 ( 10 years ago ) sit at a easy overclock of 3,8ghz at 590 cinebench score.

@Gatygun: Right, I’m more interested in the IPC difference on a single core. I would estimate a single ryzen core as being at least 2, maybe 3x faster than a single jaguar core at the same clock speed.

IDSoftware's Doom 2016 is AMD Gaming Evolved title that used proper Async compute multi-threading model and AMD GPU's metal intrinsic functions.

Using Crapcom for as an authority on GPU comparison is a joke.

Desktop PC's Jaguar doesn't have X1X's CPU lower latency improvements. Ryzen includes improvements in lower latency and twice the math unit resource count.

X1X's CPU has some lower latency improvements from Ryzen, but it doesn't have full Ryzen improvements.

For example, X1X's quad core Jaguar CPU setup has gained Ryzen's quad core X layout

[C][C]

[C][C] X layout to improve latency with quad core CPU communication.

VS

[C][C][C][C] <------old quad core Jaguar layout.

My estimate for eight core Jaguar vs eight core Ryzen is more than 2X, since PS5's 7 nm Ryzen most likely from PC's 7 nm era Ryzen not 14 nm era Ryzen.

Using A4-5150 quad core CB R15 score is flawed since X1X's CPU is an 8 core setup.

2.3 Ghz version

My estimate for X1X's 8 core/8 threads CPU setup for Cinebench R15 is about 374 CB

8 core Ryzen at 2.3Ghz would be about 750 CB

Cinebench R15 has poor support for 256 bit AVX.

PS4 Pro has 1GB DDR3 memory to handle OS related workloads with 256 bit GDDR5-7000 memory setup.

If we carry this PCB design forward into 2020, it may have

DDR4 or DDR5 system memory e.g. 4GB

GDDR6 graphics/game/share memory e.g. 16 GB

OS workloads has very little business being in GDDR5/GDDR6 memory.

The Xbox One X is 6 teraflops and to be honest, they could've pushed it to 8tf if they really wanted to. But with Sony so far behind, it really wasn't necessary. Instead they went with a bump up in RAM to 12 GBs, a massive improvement over the 8GBs in every other console.

The PS5 and next gen Xbox will be easily over 10 Teraflops. I think a good estimate is 15 teraflops for next gen if they launch in 2021 which they will.

ARM Cortex A72 with quad core at 2.3 Ghz (Kirin 950) vs Intel Atom with quad core at 2.3 Ghz (Z3580)

CPU physics score

Intel Baytrail Atom Z3580 at 2.3 Ghz= 20,977

ARM Cortex A72 Kirin 950 at 2.3 Ghz = 14,784

AMD Athlon 5350 has quad core Jaguar at 2.05 Ghz. CPU physics score is 23,546 which is slightly superior to Intel Atom Z3580 at 2.3 Ghz's 20,977. AMD ZEN doubles Jaguar's IPC and math unit resources.

AMD Athlon 5350 quad core at 2 Ghz beats AMD A10-6700T 2 module Piledriver quad core/4 threads 2.5 Ghz.

SS3

Yeah, my bad for thinking you know anything about how PC games work. Because definitely console games really has the best AI, physics, simulation and world when they focus on graphics even at the cost of framerate. /sarcasm

So them sacrificing framerate won't solve the issue of the lowest common denominator since they still don't invest in AI or physics or simulation it self. All the invest their energies in the graphics department isn't something that can't be fixed on PC with sliders. But that is all besides the point, the point was you ponying up for graphics over framerate did sound like DC for consoles and I have never seen a hermit ever do that. And forget Gen 7 because Gen 8 is still a thing and I haven't seen any real progress in the technical side even with all the focus of graphics.

Yeah and PS4 was still a shitty low end closed PC when it was released in 2013.

Yeah my bad! I should have said 5GB shared memory since 3GB or so is reserved for OS. So no wonder the textures in console games are shit and covered with Vaseline.

PS4 is equivalent to R7-265 (212 mm2 size GPU) with two A4-5100 quad core CPU modules. PS4's APU used 348 mm2 size chip.

If Microsoft wasn't focusing on Kinect, XBO's 363 mm2 chip would have yield about 24-to-28 CU GPU. 32 MB ESRAM chip area would have yielded another 14 CUs.

PS4 Pro APU has ~321 mm2 size chip with ~232 mm2 GPU

X1X APU has ~359 mm2 size chip with ~285 mm2 GPU

Both MS and Sony are nearly consistent with their respective chip geometry budgets.

For PS5... Vega 64 class GPU at 235 mm2 (via 7 nm process tech) and 8 Ryzen embedded CPU cores. RX-480/RX-580 has 232 mm2 size chip.

To date, X1X is the largest GPU chip area size for a game console GPU. Which future game console to break 300 mm2 size GPU?

Ryzen 7 (8 core, 4MB L2 cache, 16 MB L3 cache) has 213 mm2 size chip via 14 nm process-tech, hence 7 nm version would be about 106.5 mm2

Vega 64 has 470 mm2 size chip via 14 nm process-tech hence 7 nm version would be about 235 mm2

Total APU chip area size estimate: 341.5 mm2 which is close to PS4's 348 mm2

Must be your eyes because they EASILY compete with PC......even surpass most stock PC games.

Yeah, they easily compete with PC games of the 90's and early 2000. No j/k, actually those early 2000's games has some awesome texture mods. So, no chance for these shitty textures hence a reason they are hidden behind Blur and all those Vaseline filters.

The GPU was the functional equivalent of a mid-range 570 that was released in 2010 and yes from all the benchmarks it was indeed very close to 7850/7870.

Sounds like ill be sticking with my pc.

Wow

@pc_rocks: see, There's a place we can definitely agree. The cpu was a potato due to being an early so I. There was honestly little hope of something new on the computational end simply because it was a win getting both ps and Xbox away from Power CPU architecture.

Still. I think the fact remains that having a baseline pushing 30fps in more graphically demanding scenarios is a lot better for us pc gamers. I'm hard pressed to find a reason why DragonQuest or Ni Oh or Chess needs to run at 60fps. It does nothing to enhance gameplay beyond a serious diminishing means of return. Conversely, games that struggle on console will reach 60fps or greater with a proper pc.

So, we have a difference of opinion. I don't think games running at 30fps are unplayable. I also think pc gets a better deal when a game is made to struggle on consoles. It has the upper hand. Just how I see it.

And that response is proof you're an idiot who's not to be taken seriously by anyone.

Nope.

Don't know about Dragon Quest or care, I think it's turn based but f**k me if any one forced me to play Nioh at less than 60FPS. It's not a game I would ever play at less than 60 FPS. I see where you're coming from but I would still argue that pushing for 60FPS would be a better deal for PC gamers since many Japanese games tie their business logic to frame rate and when they port it to PC, it's a mess. That situation has got somewhat better but still there's a lot to be improved in these games.

Yeah, we should take a person seriously who thinks laws of Physics can be changed on the whims of cows because definitely 5GB of shared RAM will net you better textures than PC games. j/k that was sarcasm, that idiot needs to guillotined.

Now THAT I can high five on. Can only hope that the improvement continues. Sega has been a damn champ this gen!

Consider this. The eight cores in the PS4 APU are single instruction. Ryzen is not only far superior in how cache is managed in the pipeline but they are dual-instruction as well. A quad-core Ryzen can outmath/process the cpu in the ps4 at the same clock-rate. Considering that 7nm chips will be able to run at higher frequencies to generate the same thermal dissipation, it will be a complete matchup to a lot of high end gaming PCs today if they opt for octo-core design. Transversely, they may opt for fewer physical cores due to dual instruction and use the die-space to balance things with a very powerful graphics chip.

Almost everyone in SW has been weighting the PS5 as if it is a given that it will be octo. That could very well be true, but if they opt to go hexacore, it gives die-space and budget back to making a more powerful GPU.

I think they will use 4 core 8 thread Ryzen so they have more room for a bigger GPU if they go the APU route

"was low end in 2013"?

Unhuh....

Released in "2018" and widely recognized as the best looking game this generation.

Dat GTX 570 equivalent tho!?

Must be rough on the PS4 if PS5 is on your mind especially when the next iteration of consoles is not going to offer anything note worthy.

Jaguar CPU has a peak dual instruction issue per cycle decoder with two instruction path from L1 and L2 cache. Ryzen doubles nearly everything from Jagaur from start to finish.

Ryzen has Trace cache (cache for decoded RISC instructions) and two post-decode instruction issue per cycle dispatcher in addition to quad instruction issue per cycle X86 decoders. Ryzen's peak instruction issue per cycle rate is six instructions per cycle. This section separates Intel Core I series and Ryzen from Core 2's quad instruction issue per cycle rate besides AVX v1/v2 256 bit (32 bytes per cycle bus) and AVX v3 512bit (64 bytes per cycle bus) internal bus improvements.

Future X86 CPUs may complete trace cache's instruction dispatchers into full blown X86 decoders, hence creating six X86 instruction decoders beast CPU. There's more IPC improvements for X86 CPU.

Both AMD and Intel CPU patent sharing agreements. Hyper-threading uses idle instruction decoder resources.

AMD GCN has WineFine (forward looking) technology i.e. both NV Kelper and Fermi are aging designs with modern games.

https://www.notebookcheck.net/Nvidia-drops-performance-optimisation-bug-fix-and-new-feature-support-for-Fermi-series-GPUs.296614.0.html

My laptop's 8870M (renamed R9-M270/R9-370, similar to desktop 7770/R7-250X, GCN 1.0) has up to date Shader Model 6_1 support. My laptop's GPU has no problems running Forza Motosport 6 DX12.

GTX 580 slower than R6-260X and on par with 7790. R7-265(PS4 PC GPU substitute) beats GTX 580.

R7-265 (33 fps) Mantle beats 580 (28 fps)

R7-265 = 1.89 TFLOPS ~= PS4 GPU = 1.84 TFLOPS

GTX 570 dead for Far Cry Primal.

Radeon HD 6950 result is 2X over GTX 570 LOL.

AMD GPUs are aging better than NVIDIA counterparts.

Vega 64's 470 mm2 size chip can be smaller when High Bandwidth Cache (Bollywood distraction) is removed.

X1X's GPU already has 44 CU GPU with about 285 mm2 size GPU at 16 nm process tech.. Scaled from X1X's 44 CU GPU for 64 CU, it yields 414 mm2.