Info PS5 die shots, rough floorplan, and analysis!

naukkis

Senior member
Jun 5, 2002
706
578
136
256 bit vectors are pretty useless for games and code that can make use of it can usually also executed faster in GPU. So cutting unnecessarily FPU and replacing it with something more useful in console SOC is pretty logical thing to do.
 

gdansk

Platinum Member
Feb 8, 2011
2,107
2,604
136
Seems like a waste of effort to remove those. Are they using some compiler option that mixes latencies for most znver2 instructions but with znver1 timings for AVX? Maybe it doesn't matter.

Lack of GPU L3 cache is not surprising but I wonder the impact this has on performance.
 
Last edited:

coercitiv

Diamond Member
Jan 24, 2014
6,203
11,909
136
We have new Xbox Series X SoC info on Anandtech, and some of it is somewhat linked with the discussion above:
For Scarlett, it is actually the CPU that becomes the limiting factor. Using AMD’s high-performance x86 Zen 2 cores, rather than the low power Jaguar cores from the previous generation, combined with how gaming workloads have evolved in the 7 years since, means that when a gaming workload starts to ramp up, the dual 256-bit floating point units on the CPU is where the highest thermal density point happens.

In this slide shown, although it doesn’t indicate what sort of workload is in play here, whether it is an active game or a power virus, Microsoft is showing 87.4ºC at the hotspot on the CPU side, while the GPU only has a 80.9ºC hotspot. Now this also comes down to the frequency choice and design point of the hardware, and finding the right balance between CPU power, GPU power, and overall thermal characteristics and acoustics.

1613414906323.png
 

Panino Manino

Senior member
Jan 28, 2017
821
1,022
136
It may have been a surprise, but it shouldn't. When I think back Cerny really talked about "removing bits that they don't need", right?
What a madman, I can't imagine anyone else going to a CPU maker and saying "I want this CPU you make, but I want these parts removed".
 
  • Like
Reactions: lightmanek

Carfax83

Diamond Member
Nov 1, 2010
6,841
1,536
136
256 bit vectors are pretty useless for games and code that can make use of it can usually also executed faster in GPU. So cutting unnecessarily FPU and replacing it with something more useful in console SOC is pretty logical thing to do.

I don't know about that. 256 bit vectors are used for physics calculations and particle effects in PC games by physics engines. Do you remember Ageia PhysX PPU? P4 and Athlon CPUs were so weak back then they couldn't do the calculations for cloth simulation at acceptable frame rates. Nowadays, modern CPUs can actually run cloth simulation faster than GPUs thanks to a new class of physics engines that have been built from the ground up to exploit wide vectors, SMT and multicore CPUs.

Epic's new Chaos physics engine will use up to AVX2 for instance if it's available. Physics effects has more room to drastically improve than graphics with the new generation of consoles, so if the XsX still retained its 256 bit FPUs and the PS5 did not, the former will certainly have an advantage.
 
  • Like
Reactions: Drazick

coercitiv

Diamond Member
Jan 24, 2014
6,203
11,909
136
From "The road to PS5", another tidbit of information remarked by a Twitter user:
CPU supports 256 bit native instructions that consume a lot of power
these are great here and there, but presumably only minimally used
if we plan for major 256 bit instruction usage, we need to set the CPU clock substantially lower or noticeably increase the size of the power supply and fan


Corroborate this with the Xbox heat-map showing that FPU temps are the ones dictating fan ramping, and you can see how FPU usage was seen as very important for power management by both MS and Sony engineering teams. The way these two companies built their custom SoC using essentially the same "LEGO" set only to arrive at such different destinations is indeed fascinating.
 

Valen

Junior Member
Apr 29, 2021
2
5
41
256 bit vectors are pretty useless for games and code that can make use of it can usually also executed faster in GPU. So cutting unnecessarily FPU and replacing it with something more useful in console SOC is pretty logical thing to do.
AVX 2 is not useless for games e.g. World of Tanks enCore's real-time BVH raytracing shadows are CPU AVX SIMD driven and conserves GPU's BVH raytracing hardware resource for other raytracing effects.

From "The road to PS5", another tidbit of information remarked by a Twitter user:

Corroborate this with the Xbox heat-map showing that FPU temps are the ones dictating fan ramping, and you can see how FPU usage was seen as very important for power management by both MS and Sony engineering teams. The way these two companies built their custom SoC using essentially the same "LEGO" set only to arrive at such different destinations is indeed fascinating.
PS5 CPU's AVX resource is less than desktop/laptop PC's Zen 2 e.g. less register file and FADD SIMD pipelines being missing.

PS5 CPU's AVX resource is between Zen 1.x and PC's Zen 2, and presumably Xbox Series X's "server-grade" Zen 2 CPU.

Removing FADD Vect/SIMD units, PS5 CPU's AVX resource FP operations are about half of normal Zen 2.

Intel Haswell has two AVX FP SIMD and Intel has a higher clock speed (close 5 Ghz) with Coffee Lake to mitigate its dual FP AVX Vect pipelines disadvantage relative to PC Zen 2(low 4 Ghz for all cores). Coffee Lake has three INT Vect pipelines and two of which overlap two FP Vect pipelines.


Cinebench R20 (raytracing) scores multithread
Intel Core i7 5960X = 2932, it's limited to dual AVX FP units with max 3.5Ghz clock speed turbo. Core i9-9900KS has 5 Ghz all cores

AMD Ryzen 7 3700X = 4760
AMD Ryzen 7 3800X = 4965

Intel Core i7 5960X is 61.6% of AMD Ryzen 7 3700X.
 
Last edited:
  • Love
Reactions: Carfax83

Valen

Junior Member
Apr 29, 2021
2
5
41
I don't know about that. 256 bit vectors are used for physics calculations and particle effects in PC games by physics engines. Do you remember Ageia PhysX PPU? P4 and Athlon CPUs were so weak back then they couldn't do the calculations for cloth simulation at acceptable frame rates. Nowadays, modern CPUs can actually run cloth simulation faster than GPUs thanks to a new class of physics engines that have been built from the ground up to exploit wide vectors, SMT and multicore CPUs.

Epic's new Chaos physics engine will use up to AVX2 for instance if it's available. Physics effects has more room to drastically improve than graphics with the new generation of consoles, so if the XsX still retained its 256 bit FPUs and the PS5 did not, the former will certainly have an advantage.
World of Tanks enCore's BVH shadows raytracing driven by CPU's AVX power.