Click Here
home features news forums classifieds faqs links search
6069 members 
Amiga Q&A /  Free for All /  Emulation /  Gaming / (Latest Posts)
Login

Nickname

Password

Lost Password?

Don't have an account yet?
Register now!

Support Amigaworld.net
Your support is needed and is appreciated as Amigaworld.net is primarily dependent upon the support of its users.
Donate

Menu
Main sections
» Home
» Features
» News
» Forums
» Classifieds
» Links
» Downloads
Extras
» OS4 Zone
» IRC Network
» AmigaWorld Radio
» Newsfeed
» Top Members
» Amiga Dealers
Information
» About Us
» FAQs
» Advertise
» Polls
» Terms of Service
» Search

IRC Channel
Server: irc.amigaworld.net
Ports: 1024,5555, 6665-6669
SSL port: 6697
Channel: #Amigaworld
Channel Policy and Guidelines

Who's Online
62 crawler(s) on-line.
 16 guest(s) on-line.
 1 member(s) on-line.


 kamelito

You are an anonymous user.
Register Now!
 kamelito:  1 min ago
 Rob:  6 mins ago
 eliyahu:  17 mins ago
 duga:  22 mins ago
 DiscreetFX:  22 mins ago
 pixie:  26 mins ago
 Vidar:  44 mins ago
 emeck:  57 mins ago
 amiwell:  1 hr 3 mins ago
 kolla:  1 hr 10 mins ago

/  Forum Index
   /  Amiga News & Events
      /  RISC V Laptop announced... Could this be the ultimate AMIGA hardware?
Register To Post

Goto page ( Previous Page 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 )
PosterThread
cdimauro 
Re: RISC V Laptop announced... Could this be the ultimate AMIGA hardware?
Posted on 17-Sep-2022 3:55:34
#161 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3084
From: Germany

@agami

Quote:

agami wrote:
@Threading

Nice solution for allowing older single-threaded apps to run on a multi-threaded OS.

https://www.osnews.com/story/135304/multi-threading-and-globals-on-pumpkin-os/

Well, it's an obvious solution: the TLS is for keeping each thread's specific data set.

However you need proper coding for it and change several parts, as you can see. Plus, it's NOT that much efficient.

Usually a compiler should do a good work by collecting all globals and put them on the data section (of the executable). Then the o.s. loader allocates the data section on proper pages, so that they could be duplicated with the usual copy-on-write method when multiple threads are created and the new thread is changing the data section.

I don't understand why that wasn't possible with PumpkinOS.


@matthey

Quote:

matthey wrote:
agami Quote:

Nice solution for allowing older single-threaded apps to run on a multi-threaded OS.

https://www.osnews.com/story/135304/multi-threading-and-globals-on-pumpkin-os/


Duplicate everything global? The Amiga can do that too. It gives an instance of AmigaOS for each task/process (I believe this is what AxRunTime does on Linux). It works fine and increases security but eliminates code sharing which is one of the biggest advantages of the AmigaOS. The benefit of shared libraries, pure programs and the tiny footprint of the AmigaOS goes away.

You don't need it with modern o.ses: see above my reply. At zero development costs (no need to create proper libraries).
Quote:
We can have hundreds if not thousands of separate AmigaOS instances on modern high performance PCs so why not while we emulate the 68k CPU and Amiga chipset. It would be a virtual Amiga or virtually not an Amiga.

The problem with the Amiga emulation is that the more accurate you do it and the more performances are needed due to the very tight integration / interaction / synchronization of CPU and chipset.

For o.s.-friendly applications it's different, however, and they can be made blazing-fast (because the chipset doesn't need to be emulated and/or only a part of it could be kept).

 Status: Offline
Profile     Report this post  
Hammer 
Re: RISC V Laptop announced... Could this be the ultimate AMIGA hardware?
Posted on 19-Sep-2022 9:04:18
#162 ]
Elite Member
Joined: 9-Mar-2003
Posts: 4598
From: Australia

@bhabbott

Quote:

I have been interested in using the Pi Zero in various projects several years, but every time I look at buying one they are out of stock! Now people are saying they won't be available until some time next year.

Very frustrating, particularly with the performance that PiStorm is getting now. Guess I will just have to be satisfied with the Vampire for now...

After testing PiStorm with my Amiga game selection set. Emu68 JIT has some odd behavior differences from WinUAE-JIT.

My Amiga games selection is grouped into multi-parallax, arcade ports, smooth, long quests, short, sports, and PC ports.

Examples,
WHDLoad-Startup script has EmuControl ICNT= 2 NOCACHE
(this is for JIT instruction count depth set to 2 instructions and no cache).


WHDLoad-Cleanup script has EmuControl ICNT= 256 CACHE
I used CMake Nightly Builds https://github.com/michalsc/Emu68/actions
Using Nightly Build #603:


"Teenage Mutant Hero Turtles: The Coin-Op" has graphics render issues. TF1260 doesn't have this issue.
Golden Axe, sound glitch with the player's sword action.

The following games have at least a parallax background.
Jim Power doesn't start
Zool, graphics rendered red colors and has major graphics render issues.
Kid Chaos, start menu issues e.g. can't select.


Games that seem to run okay (1st level run) e.g.
Lionheart,
Zool 2,
Saint Dragon,
Shadow Fighter,
Fire and Ice,
Riskywood,
Brian the Lion (very minor copper effect glitch after map scene, doesn't affect gameplay),
Elf Mania,
Bubble and Squeak
Agony
FlimboQuest (a very minor glitch in the intro screen)
Turrican 3 (intro music timing issues)
Toki
Silk Worm
Mr Nutz

Last edited by Hammer on 19-Sep-2022 at 11:31 AM.
Last edited by Hammer on 19-Sep-2022 at 11:22 AM.
Last edited by Hammer on 19-Sep-2022 at 11:19 AM.
Last edited by Hammer on 19-Sep-2022 at 11:03 AM.
Last edited by Hammer on 19-Sep-2022 at 09:09 AM.
Last edited by Hammer on 19-Sep-2022 at 09:08 AM.
Last edited by Hammer on 19-Sep-2022 at 09:05 AM.

_________________
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 3080 Ti
Ryzen 9 7900X, DDR5-5600 32 GB RAM, GeForce RTX 3080 Ti
Amiga 1200 (rev 1D1, KS 3.2, TF1260, 68060 @ 63 Mhz, 128 MB)
Amiga 500 (rev 6A, KS 3.2, PiStorm/RPi3a/Emu68)

 Status: Offline
Profile     Report this post  
Hammer 
Re: RISC V Laptop announced... Could this be the ultimate AMIGA hardware?
Posted on 19-Sep-2022 11:27:54
#163 ]
Elite Member
Joined: 9-Mar-2003
Posts: 4598
From: Australia

@cdimauro

Quote:

The problem with the Amiga emulation is that the more accurate you do it and the more performances are needed due to the very tight integration / interaction / synchronization of CPU and chipset.

WinUAE is a narrow-threaded software, hence multiple WinUAE instances are not a major issue.

_________________
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 3080 Ti
Ryzen 9 7900X, DDR5-5600 32 GB RAM, GeForce RTX 3080 Ti
Amiga 1200 (rev 1D1, KS 3.2, TF1260, 68060 @ 63 Mhz, 128 MB)
Amiga 500 (rev 6A, KS 3.2, PiStorm/RPi3a/Emu68)

 Status: Offline
Profile     Report this post  
cdimauro 
Re: RISC V Laptop announced... Could this be the ultimate AMIGA hardware?
Posted on 19-Sep-2022 20:04:30
#164 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3084
From: Germany

@Hammer

Quote:

Hammer wrote:

After testing PiStorm with my Amiga game selection set. Emu68 JIT has some odd behavior differences from WinUAE-JIT.

[...]

"Teenage Mutant Hero Turtles: The Coin-Op" has graphics render issues. TF1260 doesn't have this issue.
Golden Axe, sound glitch with the player's sword action.

The following games have at least a parallax background.
Jim Power doesn't start
Zool, graphics rendered red colors and has major graphics render issues.
Kid Chaos, start menu issues e.g. can't select.

Games that seem to run okay (1st level run) e.g.
[...]
Brian the Lion (very minor copper effect glitch after map scene, doesn't affect gameplay),
[...]
FlimboQuest (a very minor glitch in the intro screen)
Turrican 3 (intro music timing issues)

Why don't you submit those bugs? Michal could fix Emu68 with those inputs.
Quote:

Hammer wrote:
@cdimauro

Quote:

The problem with the Amiga emulation is that the more accurate you do it and the more performances are needed due to the very tight integration / interaction / synchronization of CPU and chipset.

WinUAE is a narrow-threaded software, hence multiple WinUAE instances are not a major issue.

The problem is that running WinUAE multiple times is useless: WinUAE is used with a single instance.

That's why what counts is single thread/process maximum performance. The number of cores doesn't count (besides WinUAE uses some extra thread for offloading some stuff to the "host"; but they are only small things) and aren't useful for running other things.

 Status: Offline
Profile     Report this post  
Hammer 
Re: RISC V Laptop announced... Could this be the ultimate AMIGA hardware?
Posted on 20-Sep-2022 1:56:13
#165 ]
Elite Member
Joined: 9-Mar-2003
Posts: 4598
From: Australia

@cdimauro

Quote:
Why don't you submit those bugs? Michal could fix Emu68 with those inputs.

Did you assume I didn't post an alternative test set in Discord that includes mschulz and clever?

I'm building test cases.

Quote:

The problem is that running WinUAE multiple times is useless: WinUAE is used with a single instance.

That's why what counts is single thread/process maximum performance. The number of cores doesn't count (besides WinUAE uses some extra thread for offloading some stuff to the "host"; but they are only small things) and aren't useful for running other things.

Modern X86 CPUs have SMT, hence less compute intense threads can be easily handled.

I run two WinUAE instances that relates to my A500 rev 6a/PiStorm and A1200 rev 1d1/TF1260 microSD installations.

_________________
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 3080 Ti
Ryzen 9 7900X, DDR5-5600 32 GB RAM, GeForce RTX 3080 Ti
Amiga 1200 (rev 1D1, KS 3.2, TF1260, 68060 @ 63 Mhz, 128 MB)
Amiga 500 (rev 6A, KS 3.2, PiStorm/RPi3a/Emu68)

 Status: Offline
Profile     Report this post  
cdimauro 
Re: RISC V Laptop announced... Could this be the ultimate AMIGA hardware?
Posted on 20-Sep-2022 5:13:38
#166 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3084
From: Germany

@Hammer

Quote:

Hammer wrote:
@cdimauro

Quote:
Why don't you submit those bugs? Michal could fix Emu68 with those inputs.

Did you assume I didn't post an alternative test set in Discord that includes mschulz and clever?

I'm building test cases.

OK. There was no mention so I was supposing not.
Quote:
Quote:

The problem is that running WinUAE multiple times is useless: WinUAE is used with a single instance.

That's why what counts is single thread/process maximum performance. The number of cores doesn't count (besides WinUAE uses some extra thread for offloading some stuff to the "host"; but they are only small things) and aren't useful for running other things.

Modern X86 CPUs have SMT, hence less compute intense threads can be easily handled.

SMT doesn't help for emulation. Rather, it could easily hurt performances of the main thread.
Quote:
I run two WinUAE instances that relates to my A500 rev 6a/PiStorm and A1200 rev 1d1/TF1260 microSD installations.

OK, so you run two installations in parallel and let them go for their way.

I don't know how long it could take, since the Amiga o.s. is very small.

Anyway, it's a single use case, and it's something which isn't that important. Normally people doesn't run multiple WinUAE instances, because they have a single set of hands for using them...

 Status: Offline
Profile     Report this post  
Hammer 
Re: RISC V Laptop announced... Could this be the ultimate AMIGA hardware?
Posted on 21-Sep-2022 3:52:31
#167 ]
Elite Member
Joined: 9-Mar-2003
Posts: 4598
From: Australia

@cdimauro

Quote:

SMT doesn't help for emulation. Rather, it could easily hurt performances of the main thread.

It depends on SMT implementation e.g. Zen 3's enabled SMT or disabled SMT has near zero performance difference.

Zen 2's enabled SMT or disabled SMT has a performance difference. Context switching with software multithreading has a higher overhead cost e.g. Core i5-2500 (4C/4T) vs Core i7-2600 (4C/8T).

_________________
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 3080 Ti
Ryzen 9 7900X, DDR5-5600 32 GB RAM, GeForce RTX 3080 Ti
Amiga 1200 (rev 1D1, KS 3.2, TF1260, 68060 @ 63 Mhz, 128 MB)
Amiga 500 (rev 6A, KS 3.2, PiStorm/RPi3a/Emu68)

 Status: Offline
Profile     Report this post  
cdimauro 
Re: RISC V Laptop announced... Could this be the ultimate AMIGA hardware?
Posted on 21-Sep-2022 4:56:50
#168 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3084
From: Germany

@Hammer

Quote:

Hammer wrote:
@cdimauro

Quote:

SMT doesn't help for emulation. Rather, it could easily hurt performances of the main thread.

It depends on SMT implementation e.g. Zen 3's enabled SMT or disabled SMT has near zero performance difference.

Zen 2's enabled SMT or disabled SMT has a performance difference. Context switching with software multithreading has a higher overhead cost e.g. Core i5-2500 (4C/4T) vs Core i7-2600 (4C/8T).

The SMT implementation doesn't matter: if you have another thread which is running on the same core, it'll hurt the performance of the other thread, because both threads are contending the resources of the core.

That's why many times for games it's much better to completely disable the SMT: to be sure that a single thread is using / has exclusive access to all resources of the core where it's running.

On emulation this is even more important, because usually there's only one (main) thread which is running and doing all the work and it's much better to don't have other threads running on the same core.

I hope that it's clear now.

 Status: Offline
Profile     Report this post  
Hammer 
Re: RISC V Laptop announced... Could this be the ultimate AMIGA hardware?
Posted on 28-Sep-2022 6:19:11
#169 ]
Elite Member
Joined: 9-Mar-2003
Posts: 4598
From: Australia

@xe54
https://www.hpcwire.com/2022/09/23/nvidia-shuts-out-risc-v-software-support-for-gpus/
A viewpoint from NVIDIA about RISC-V


Nvidia Shuts Out RISC-V Software Support for GPUs
Nvidia was an early adopter of RISC-V for controllers in its GPUs, but that’s the best use of the architecture for now, said Jensen Huang, CEO at Nvidia, during a press briefing with the Asia-Pacific press.

“We like RISC-V because it’s open source… but more importantly, it’s adaptable. We can use it for all kinds of interesting configurations of CPUs. However, RISC-V is not appropriate yet and not for some time for external third-party software,” Huang said.

By comparison, the x86 and Arm architectures have a large software ecosystem that isn’t fragmented and is stable, regardless of the supplier it comes from, Huang said.

The benefit of RISC-V being open source and adaptable could also have its disadvantages, Huang said.

The RISC-V architecture is more like a chip version of Linux, and is free to license and modify. The goal is for companies to make and manufacture their own chips at a low cost, while cutting reliance on proprietary x86 and ARM architectures, which have to be bought or licensed.

The RISC-V architecture has a base instruction set on which companies can customize by putting their proprietary extensions. For example, Nvidia competitor Imagination has made its own RISC-V CPU called Catapult, on which it can bundle its compatible GPU for graphics and AI. Imagination offers full software and debug support. Similarly, others offer RISC-V AI chips with vector extensions with their own software stacks.


There-in lies the problem. Huang views that incoherent software ecosystem, with different software offerings tuned to different chips, as the disadvantage for RISC-V. He indicated that contributing to a fragmented ecosystem won’t be healthy for the development of RISC-V.

“We’ll see how the world evolves in the long term. But building an ecosystem that is software compatible, that is architecturally compatible, it’s very, very hard to do,” Huang said, adding “can you make a RISC-V that is like an ecosystem, like Arm and x86? Of course, but it will probably take a decade or two.”


Huang’s view could be a reflection of how Apple views RISC-V. Apple is replacing Arm controllers with RISC-V cores on non-user facing parts, semiconductor analyst Dylan Patel said in a newsletter post earlier this month. Those parts typically rely less on system software.

Last edited by Hammer on 28-Sep-2022 at 06:20 AM.
Last edited by Hammer on 28-Sep-2022 at 06:19 AM.

_________________
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 3080 Ti
Ryzen 9 7900X, DDR5-5600 32 GB RAM, GeForce RTX 3080 Ti
Amiga 1200 (rev 1D1, KS 3.2, TF1260, 68060 @ 63 Mhz, 128 MB)
Amiga 500 (rev 6A, KS 3.2, PiStorm/RPi3a/Emu68)

 Status: Offline
Profile     Report this post  
Hammer 
Re: RISC V Laptop announced... Could this be the ultimate AMIGA hardware?
Posted on 28-Sep-2022 6:31:07
#170 ]
Elite Member
Joined: 9-Mar-2003
Posts: 4598
From: Australia

@cdimauro

Quote:

cdimauro wrote:

The SMT implementation doesn't matter: if you have another thread which is running on the same core, it'll hurt the performance of the other thread, because both threads are contending the resources of the core.

That's why many times for games it's much better to completely disable the SMT: to be sure that a single thread is using / has exclusive access to all resources of the core where it's running.

On emulation this is even more important, because usually there's only one (main) thread which is running and doing all the work and it's much better to don't have other threads running on the same core.

I hope that it's clear now.

For Zen 3

https://www.anandtech.com/show/16261/investigating-performance-of-multithreading-on-zen-3-and-amd-ryzen-5000/2
Dolphin 5.0 emulator benchmark
SMT Off = 100%
SMT ON =99.1%

the fact that we see no difference means that AMD’s implementation of giving a single thread access to all the resources even in SMT mode is quite good


https://www.anandtech.com/show/16261/investigating-performance-of-multithreading-on-zen-3-and-amd-ryzen-5000/3
Actual gaming benchmarks for AMD Ryzen 9 5950X, SMT On vs SMT Off

Conclusions: SMT On

The debate about enabling SMT is dependent on the specific microarchitecture design.


Zen 3's single thread out-of-order depth to extract higher IPC didn't scale with the additional resources being added to the CPU core.

The majority of Zen 4's improvement is with the front end e.g. larger reorder buffers for deeper IPC extraction, lower cache latencies, wider cache I/O, improved Load/Store and 'etc'. Zen 4 core is still similar to Zen 3 core, but improved front end.

The SMT implementation does matter when Zen 3 is like two CPU cores superglued together and the installment for significant front-end improvements is on Zen 4.



Last edited by Hammer on 28-Sep-2022 at 07:01 AM.
Last edited by Hammer on 28-Sep-2022 at 07:00 AM.
Last edited by Hammer on 28-Sep-2022 at 06:49 AM.
Last edited by Hammer on 28-Sep-2022 at 06:46 AM.
Last edited by Hammer on 28-Sep-2022 at 06:41 AM.
Last edited by Hammer on 28-Sep-2022 at 06:34 AM.

_________________
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 3080 Ti
Ryzen 9 7900X, DDR5-5600 32 GB RAM, GeForce RTX 3080 Ti
Amiga 1200 (rev 1D1, KS 3.2, TF1260, 68060 @ 63 Mhz, 128 MB)
Amiga 500 (rev 6A, KS 3.2, PiStorm/RPi3a/Emu68)

 Status: Offline
Profile     Report this post  
matthey 
Re: RISC V Laptop announced... Could this be the ultimate AMIGA hardware?
Posted on 28-Sep-2022 18:04:40
#171 ]
Super Member
Joined: 14-Mar-2007
Posts: 1684
From: Kansas

Hammer Quote:



There-in lies the problem. Huang views that incoherent software ecosystem, with different software offerings tuned to different chips, as the disadvantage for RISC-V. He indicated that contributing to a fragmented ecosystem won’t be healthy for the development of RISC-V.

“We’ll see how the world evolves in the long term. But building an ecosystem that is software compatible, that is architecturally compatible, it’s very, very hard to do,” Huang said, adding “can you make a RISC-V that is like an ecosystem, like Arm and x86? Of course, but it will probably take a decade or two.”


Huang’s view could be a reflection of how Apple views RISC-V. Apple is replacing Arm controllers with RISC-V cores on non-user facing parts, semiconductor analyst Dylan Patel said in a newsletter post earlier this month. Those parts typically rely less on system software.


RISC-V is accepted as a simple customizable low power architecture for deeply embedded uses. That is practically the opposite of a powerful standardized architecture with popular/mass ecosystem support though. There can be some crossover but it is limited. It is possible to turn off RISC-V optional enhancements like a SIMD/vector unit and compression in the compiler to run on a simpler base hardware but then it will run poorly on high performance hardware. The RISC-V ISA already made choices to keep it simple instead of high performance which will likely limit max integer performance. This is why ARM AArch64 decided to add more powerful instructions and addressing modes borrowed from CISC and standard FPU/SIMD enhancements. They needed to do this to compete in performance with x86-64 but the resource usage is high for embedded use and they didn't gain all the CISC advantages most notably variable length instruction code density and scaled immediate and displacement advantages. If just examining x86-64, it is easy to miss these advantages as they have been mostly lost. ARM AArch64 hardware will not scale down to Thumb2 hardware territory but it is not a problem as ARM uses two ISAs to scale from low end embedded markets to high end desktop and server markets. x86(-64) couldn't scale very low and was too power hungry for embedded markets while the 68k once scaled from low end 16/32 bit microprocessor embedded markets to workstation markets which it pioneered, revolutionized and dominated.

32b embedded market: 68k->SuperH->ARM Thumb->ARM Thumb2
32b workstation market: 68k->x86->x86-64

PPC was supposed to replace the 68k in these markets and Motorola killed off the 68k prematurely while it was still much more popular than the PPC replacement. Intel, which gained the desktop market out of sheer luck, made an Itanic mistake away from x86 but kept x86 development alive which they were able to return to. The 68k was a very competitive 2nd choice for the desktop market at IBM for the PC so the 68k very nearly dominated the desktop market which we would likely be using today in some form instead of x86.

Hammer Quote:

Conclusions: SMT On


Cdimauro is still right. The best case single thread performance is with SMT off. The best performance is to use another core rather than another thread on the same core. Once no more cores are available for a new thread, then SMT improves total performance but threads may take longer to execute. Shared resources of a single core by 2 threads is not as good as having full resources. It's nice that the SMT implementation is efficient enough that it doesn't usually need to be turned off. There still likely are performance impacts from SMT. Threads sharing a core while other cores are asleep may use less power. Supporting SMT likely always slows down your cores as larger shared resources like caches with more ways to avoid increased address aliasing from SMT sharing have longer access times. Related threads could be sharing some of the same cached data improving performance but it is more likely that threads are contending for cache resources which may even cause cache thrashing in the worst case. SMT is not free by any means. It is a good fit for x86-64 cores as performance/thread is good but performance/W is not so performance/thread can be traded for performance/W. ARM CPUs on the desktop are usually trying to boost weak performance/thread and are willing to trade performance/W where it is better so ARM has often avoided SMT. ARM cores are smaller than x86-64 cores so it is cheaper to add more cores too.

Last edited by matthey on 28-Sep-2022 at 06:31 PM.

 Status: Offline
Profile     Report this post  
cdimauro 
Re: RISC V Laptop announced... Could this be the ultimate AMIGA hardware?
Posted on 28-Sep-2022 19:52:39
#172 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3084
From: Germany

@matthey: exactly.

@Hammer: the tests were made with a Ryzen 5950X, which has 16 cores. So, it's quite difficult for single threaded jobs to be disturbed by other threads. Also games usually use a maximum 8 core.

If you want to verify how the SMT influences the performance then the test is very simple:

- take Dolphin (or WinUAE, or another emulator which stresses the main process/thread) and using the affinity "park" it to a specific core. Then do the same with another application which is doing something similar (or a number-cruncher, like SuperPi or yCrunch) but which should be using just a single thread, but put it on a different core. Finally, start the Dolphin's benchmark (or do some heavy computation on WinUAE which is NOT using I/O: just pure computation) and the other heavy application, and take note of Dolphin's performance.

- now do exactly the same, but parking both applications on the same core.

I'm pretty sure that you'll see a HUGE difference on the results of both tests (which is logical & expected).

 Status: Offline
Profile     Report this post  
Hammer 
Re: RISC V Laptop announced... Could this be the ultimate AMIGA hardware?
Posted on 29-Sep-2022 6:39:45
#173 ]
Elite Member
Joined: 9-Mar-2003
Posts: 4598
From: Australia

@cdimauro

Quote:

cdimauro wrote:
@matthey: exactly.

@Hammer: the tests were made with a Ryzen 5950X, which has 16 cores. So, it's quite difficult for single threaded jobs to be disturbed by other threads. Also games usually use a maximum 8 core.

If you want to verify how the SMT influences the performance then the test is very simple:

- take Dolphin (or WinUAE, or another emulator which stresses the main process/thread) and using the affinity "park" it to a specific core. Then do the same with another application which is doing something similar (or a number-cruncher, like SuperPi or yCrunch) but which should be using just a single thread, but put it on a different core. Finally, start the Dolphin's benchmark (or do some heavy computation on WinUAE which is NOT using I/O: just pure computation) and the other heavy application, and take note of Dolphin's performance.

- now do exactly the same, but parking both applications on the same core.

I'm pretty sure that you'll see a HUGE difference on the results of both tests (which is logical & expected).

Reminder, Ryzen 7 5800X has 16 threads and its overkill for WinUAE usage i.e. lower cost Ryzen 5 5600X would deliver a similar performance.

For WinUAE and Dolphin 5.0 usage, there is very little difference between Ryzen 7 5800X and Ryzen 9 5950X. Ryzen 7 5800X has extra TDP headroom for higher OC since it's missing the additional 8 cores from Ryzen 9 5950X.

https://www.anandtech.com/show/16261/investigating-performance-of-multithreading-on-zen-3-and-amd-ryzen-5000/2
Using 3D Particle Movement with AVX2 benchmark, SMT On yielded 177.5% boost.
3D Particle Movement is very light on the cache and memory requirements.

Ryzen 9's 12 and 16 cores are HEDT CPU core counts with mainstream memory bandwidth.

WinUAE is mostly a single-thread application with minor multithreading.


From https://www.youtube.com/watch?v=ZxA-6iNp304
For Ryzen 7 5800X vs Ryzen 7 5800X3D, secondary threads are minimally used as shown in this Red Dead Redemption on RPCS3 Emulator.




Last edited by Hammer on 29-Sep-2022 at 06:57 AM.
Last edited by Hammer on 29-Sep-2022 at 06:52 AM.
Last edited by Hammer on 29-Sep-2022 at 06:50 AM.
Last edited by Hammer on 29-Sep-2022 at 06:40 AM.

_________________
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 3080 Ti
Ryzen 9 7900X, DDR5-5600 32 GB RAM, GeForce RTX 3080 Ti
Amiga 1200 (rev 1D1, KS 3.2, TF1260, 68060 @ 63 Mhz, 128 MB)
Amiga 500 (rev 6A, KS 3.2, PiStorm/RPi3a/Emu68)

 Status: Offline
Profile     Report this post  
Hammer 
Re: RISC V Laptop announced... Could this be the ultimate AMIGA hardware?
Posted on 29-Sep-2022 7:01:30
#174 ]
Elite Member
Joined: 9-Mar-2003
Posts: 4598
From: Australia

@matthey

Quote:

Cdimauro is still right. The best case single thread performance is with SMT off. The best performance is to use another core rather than another thread on the same core. Once no more cores are available for a new thread, then SMT improves total performance but threads may take longer to execute. Shared resources of a single core by 2 threads is not as good as having full resources. It's nice that the SMT implementation is efficient enough that it doesn't usually need to be turned off. There still likely are performance impacts from SMT. Threads sharing a core while other cores are asleep may use less power. Supporting SMT likely always slows down your cores as larger shared resources like caches with more ways to avoid increased address aliasing from SMT sharing have longer access times. Related threads could be sharing some of the same cached data improving performance but it is more likely that threads are contending for cache resources which may even cause cache thrashing in the worst case. SMT is not free by any means. It is a good fit for x86-64 cores as performance/thread is good but performance/W is not so performance/thread can be traded for performance/W. ARM CPUs on the desktop are usually trying to boost weak performance/thread and are willing to trade performance/W where it is better so ARM has often avoided SMT. ARM cores are smaller than x86-64 cores so it is cheaper to add more cores too.

Cdimauro is not correct e.g. Single thread Dolphin 5.0 doesn't need SMT to be turned off. LOL

For Zen 3, disabling SMT feature is a pointless action for WinUAE and Dolphin context.

Windows platform has 3rd party tools like Process Lasso that helps end users to disable SMT on a per-process basis and store the CPU thread affinity profile for each program without system reboot.

Disabling SMT outright is a piece of old fashion advice when there's a better method for doing both worlds without a system reboot.

Again: Keep SMT ON and use Process Lasso when it's needed.


PS; I also use Process Lasso to limit CPU thread count usage on mobile APU, hence allowing the iGPU to obtain higher TDP headroom. I don't need to reboot Windows for 4 CPU threads configuration.



Last edited by Hammer on 29-Sep-2022 at 07:17 AM.
Last edited by Hammer on 29-Sep-2022 at 07:14 AM.
Last edited by Hammer on 29-Sep-2022 at 07:10 AM.

_________________
Ryzen 9 7950X, DDR5-6000 64 GB RAM, GeForce RTX 3080 Ti
Ryzen 9 7900X, DDR5-5600 32 GB RAM, GeForce RTX 3080 Ti
Amiga 1200 (rev 1D1, KS 3.2, TF1260, 68060 @ 63 Mhz, 128 MB)
Amiga 500 (rev 6A, KS 3.2, PiStorm/RPi3a/Emu68)

 Status: Offline
Profile     Report this post  
cdimauro 
Re: RISC V Laptop announced... Could this be the ultimate AMIGA hardware?
Posted on 2-Oct-2022 12:37:09
#175 ]
Elite Member
Joined: 29-Oct-2012
Posts: 3084
From: Germany

@Hammer

Quote:

Hammer wrote:
@cdimauro

Quote:

cdimauro wrote:
@matthey: exactly.

@Hammer: the tests were made with a Ryzen 5950X, which has 16 cores. So, it's quite difficult for single threaded jobs to be disturbed by other threads. Also games usually use a maximum 8 core.

If you want to verify how the SMT influences the performance then the test is very simple:

- take Dolphin (or WinUAE, or another emulator which stresses the main process/thread) and using the affinity "park" it to a specific core. Then do the same with another application which is doing something similar (or a number-cruncher, like SuperPi or yCrunch) but which should be using just a single thread, but put it on a different core. Finally, start the Dolphin's benchmark (or do some heavy computation on WinUAE which is NOT using I/O: just pure computation) and the other heavy application, and take note of Dolphin's performance.

- now do exactly the same, but parking both applications on the same core.

I'm pretty sure that you'll see a HUGE difference on the results of both tests (which is logical & expected).

Reminder, Ryzen 7 5800X has 16 threads and its overkill for WinUAE usage i.e. lower cost Ryzen 5 5600X would deliver a similar performance.

For WinUAE and Dolphin 5.0 usage, there is very little difference between Ryzen 7 5800X and Ryzen 9 5950X. Ryzen 7 5800X has extra TDP headroom for higher OC since it's missing the additional 8 cores from Ryzen 9 5950X.
[...]WinUAE is mostly a single-thread application with minor multithreading.

Correct and already stated: if you have plenty of cores then applications which are using only a few threads/processes are usually not affected.
Quote:
From https://www.youtube.com/watch?v=ZxA-6iNp304
For Ryzen 7 5800X vs Ryzen 7 5800X3D, secondary threads are minimally used as shown in this Red Dead Redemption on RPCS3 Emulator.

In fact there are more than enough cores.

Anyway the game is has stuttering and is sluggish: those processors don't look enough for such emulated beast.
@Hammer

Quote:

Hammer wrote:
@matthey

Quote:

Cdimauro is still right. The best case single thread performance is with SMT off. The best performance is to use another core rather than another thread on the same core. Once no more cores are available for a new thread, then SMT improves total performance but threads may take longer to execute. Shared resources of a single core by 2 threads is not as good as having full resources. It's nice that the SMT implementation is efficient enough that it doesn't usually need to be turned off. There still likely are performance impacts from SMT. Threads sharing a core while other cores are asleep may use less power. Supporting SMT likely always slows down your cores as larger shared resources like caches with more ways to avoid increased address aliasing from SMT sharing have longer access times. Related threads could be sharing some of the same cached data improving performance but it is more likely that threads are contending for cache resources which may even cause cache thrashing in the worst case. SMT is not free by any means. It is a good fit for x86-64 cores as performance/thread is good but performance/W is not so performance/thread can be traded for performance/W. ARM CPUs on the desktop are usually trying to boost weak performance/thread and are willing to trade performance/W where it is better so ARM has often avoided SMT. ARM cores are smaller than x86-64 cores so it is cheaper to add more cores too.

Cdimauro is not correct e.g. Single thread Dolphin 5.0 doesn't need SMT to be turned off. LOL

For Zen 3, disabling SMT feature is a pointless action for WinUAE and Dolphin context.

Windows platform has 3rd party tools like Process Lasso that helps end users to disable SMT on a per-process basis and store the CPU thread affinity profile for each program without system reboot.

Disabling SMT outright is a piece of old fashion advice when there's a better method for doing both worlds without a system reboot.

Again: Keep SMT ON and use Process Lasso when it's needed.


PS; I also use Process Lasso to limit CPU thread count usage on mobile APU, hence allowing the iGPU to obtain higher TDP headroom. I don't need to reboot Windows for 4 CPU threads configuration.

So, let me recap: disabling the SMT should not be required, according to your sources, but an application is used to provide a similar feature at runtime.

Which looks a contraddiction...

Anyway, previously I've provided you a way to verify the effects of the SMT.

 Status: Offline
Profile     Report this post  
Goto page ( Previous Page 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 )

[ home ][ about us ][ privacy ] [ forums ][ classifieds ] [ links ][ news archive ] [ link to us ][ user account ]
Copyright (C) 2000 - 2019 Amigaworld.net.
Amigaworld.net was originally founded by David Doyle