English Amiga Board


Go Back   English Amiga Board > News

 
 
Thread Tools
Old 23 June 2020, 11:54   #41
Daedalus
Registered User

Daedalus's Avatar
 
Join Date: Jun 2009
Location: Dublin, then Glasgow
Posts: 4,765
Look, rather than trying to comfort yourself by labelling things as opinions rather than facts, why don't you simply try it? There are dozens of things that can be affecting your setup, but try changing your current setup to a 68020 or 68030 and see how the benchmarking runs. Don't change anything else. It's not opinion that emulated CPU speed doesn't change significantly between CPU models, it's fact. Unless you're artificially slowing down other CPU models (e.g. using cycle exact settings or similar), you'll see no difference, whether that's benchmarking tools or playing MP3s.

It's great that you're happy with your current emulation setup, but blaming issues in your previous setup purely on your choice of CPU without really knowing what else has changed or could be affecting it is misleading to yourself.

As for FS-UAE / WinUAE differences, I use both and haven't seen any differences in this regard. The CPU emulation core is essentially the same for both, and performance is very similar.
Daedalus is online now  
Old 23 June 2020, 12:45   #42
AMIGASYSTEM
Registered User
AMIGASYSTEM's Avatar
 
Join Date: Aug 2014
Location: Brindisi (Italy)
Posts: 5,815
Si Daedalus is right, AfA OS for example with an 020/882 JIT and faster than a 060/FPU.

Even Picasso IV graphics card with only 4MB is faster than UAE Zorro III 256MB.

But if you want to use AfA OS with 060 Software and Libraries you can't use the 020 because some applications won't work.

Same thing for Picasso IV, where some programs with the 32Bit of UAE Zorro III have a better look.
AMIGASYSTEM is offline  
Old 23 June 2020, 16:18   #43
Weaselrama
Registered User

Weaselrama's Avatar
 
Join Date: Oct 2019
Location: Neunkirchen aP, DE
Age: 58
Posts: 328
Just to make sure I wasn't mis-remembering, I tried running a 68020, 68EC030, and a 68030, JIT disabled (I don't believe it could be enabled on the 68030), AfA libs disabled w/ both commands in startup-sequence commented out. Software failure before any of them even got to the WB and why would the WB start anyway? Most of my libraries are 68040, 040/060 or 060.

68040-NOMMU with both JIT enabled and disabled. Almost nothing worked but the WB started. iBrowse was slow as molasses on http sites even with JIT enabled (ok, compared to what I'm used to) and wouldn't bring up AmiSSL on https sites because the library is 68060 specific. What is it that I'm supposed to run a benchmark on? The only benchmark software I have is Sys Info and I've read more than once that it was worthless in emulation. Suggest one and I'll try it on the 68040-NOMMU vs the 68060-NOMMU. You wrote above not to change anything else.

Please explain the point of offering different CPU configurations in UAE if everything runs just fine on a 68030, as fast or even faster? This has not been my experience with either iBrowse, Netsurf, AMPlifier, or AmigaAMP. Let's not pretend there's a whole lot I could have screwed up in the process of upgrading over the past year. I keep a close eye on my startup-sequence and user-startup and keep a log of the libraries and shell commands I switch out with one exception, PeterK explained how the AfA libs replace the icon.library and picture.datatype. I had the idea they ran on top of them before he explained it, that, and using RemLib to be able to use his icon.library again.

You've run the benchmarks yourself and you've built configs with 040/060 libraries and run an emulated A4000 on a 68040 or 060 configuration and compared them?

I'm not trying to compare this to actual hardware running AfA. I'm happy to benchmark between the 040 and 060 if you tell me:

1. The benchmarking software you want me to use.
2. What you want me to benchmark, i.e, an idle system or specific software running? When I benchmark on the Mac, its usually a specific task or software such as running benchmarks on video rendering.
Weaselrama is offline  
Old 23 June 2020, 17:20   #44
Weaselrama
Registered User

Weaselrama's Avatar
 
Join Date: Oct 2019
Location: Neunkirchen aP, DE
Age: 58
Posts: 328
A few other notes:

There are few threads on EAB dealing with this but this is one I remember. "Fastest Possible" was removed from FS-UAE by Frode Solheim a few versions ago. I tried the fake 060 back when I was trying to run 060 demos from Pouet. Most require the 060 to be running.

http://eab.abime.net/showthread.php?...on%2C+68040%3F

Back to the 040 configuration, FFPlay with the 040 version trying to play the movies I have stuttered more and there were issues with audio latency. This clears up back in the 060 config with the 060 version of FFPlay. Again, this could be instruction set-specific.

I get the impression that few Amigans build an emulated machine for much more than playing older games. I try to push the Amiga for graphics, sound, video, and to see if I can do tasks I normally do on the Mac - for no better reason than I can and it's fun. I have tons of OCS/ECS demos that ran fine when I configured the A1200 originally w/o RTG. The 060-specific demos pretty much need the 060 enabled. I only have one game that "whizzes by" and that's Defender II. The rest of them work but they're WHDLoad games so they've been tweaked to run on anything. The games I actually play and have current saves are Duke, Doom, and Quake.

It probably does no good to even have conversations such as this. Many on EAB get extremely frustrated when you're not as knowledgeable, share their experience, or question what they say. A good example was watching AMIGASYSTEM and PeterK go 'round and 'round regarding conflicts between AfA's picture datatype and icon library. They talked past one another, PeterK gave up in frustration trying to explain it and yet AMIGASYSTEM got the conflict resolved in his own way and it worked.

I'm still happy to run a benchmark but does it matter? I'm not going back to a pre-AfA config because I can both see and experience the real-world improvements - My Mac is happier. My WB CPU is lower. I'm paranoid about how the Mac handles cooling. I'm not going back to a pre-68060 config because I don't see the point. My Mac ran hotter, I suffered from known RTG/AHI audio latency issues even on the new Mac in 68040 even with JIT enabled. I accept what you're saying about the emulated CPU and what I'm experiencing is likely compatiblity.

If it ain't broke, don't fix it. Plus, I'm having too much fun. I know I'm able to run webpages in iBrowse that would probably choke on a real A4000. I'm tickled pink to have font-smoothing and 4.1 style frames. When I run an Amiga, I'm in my happy place.

Here's an interesting thread from 2008, mostly outdated now I suppose. Everyone with an expert opinion and few agree with one another. Sort of like EAB. The one statement that sticks with me in light of this discussion is that a 68020 in UAE is faster than a real-world 060. The rest is a matter of compatibility, not to mention flaws in any flavor of UAE. It has come a long, long way.

I know you will come back and tell me I still don't get it. Ok, I probably don't, not completely anyway but perhaps next month or next year I will. I know I'll have a lot of fun getting there.

https://amigaworld.net//modules/newb...975&refresh=Go
Weaselrama is offline  
Old 23 June 2020, 18:58   #45
Weaselrama
Registered User

Weaselrama's Avatar
 
Join Date: Oct 2019
Location: Neunkirchen aP, DE
Age: 58
Posts: 328
I've been doing further reading. Where as I don't think a 68020 or 68030 would be a good match for some of the software I run in terms of compatibility, if I were to completely overhaul my configuration, replacing all libs and shell commands which were 060 with 040 or lower, re-installed iBrowse for the 68040 (this is the issue I ran into above when I switched FS-UAE to a 68040), I would have an Amiga configuration that was perceptually as fast as the 68060 config with some caveats:

I would likely run into the RTG/Audio latency bug in FS-UAE and I would have to make adjustments to AmigaAMP. This would also likely affect the 040 build of FFPlay. Lowering my screen resolution might overcome this without the need to change AmigaAMP, FFPlay would stay the same.

I would have to hunt down in my archives the software I run as there was quite a few of them that had 040 and 060 builds. It's possible I wouldn't have any conflict at all - I could take it on a case by case basis.

I would wait to install AfA to note the difference.

I'll let you know if I rebuild this to these specifications. I'm very curious now as to what results would be.
Weaselrama is offline  
Old 23 June 2020, 20:48   #46
Daedalus
Registered User

Daedalus's Avatar
 
Join Date: Jun 2009
Location: Dublin, then Glasgow
Posts: 4,765
Quote:
Originally Posted by Weaselrama View Post
It probably does no good to even have conversations such as this.
Indeed, especially when it's way off-topic, yet here you are, still trying to argue against easily verifiable facts.

Quote:
I'm still happy to run a benchmark but does it matter?
[...]
If it ain't broke, don't fix it. Plus, I'm having too much fun. I know I'm able to run webpages in iBrowse that would probably choke on a real A4000. I'm tickled pink to have font-smoothing and 4.1 style frames. When I run an Amiga, I'm in my happy place.
Once you're happy. The only reason I'm suggesting that you run some benchmarks is because your beliefs around the emulated CPU are misplaced, and benchmarking is the best way to objectively demonstrate that.

Quote:
I know you will come back and tell me I still don't get it. Ok, I probably don't, not completely anyway but perhaps next month or next year I will. I know I'll have a lot of fun getting there.
Yep But it's good that you're willing to explore it. And at the end of the day, enjoying it is the most important part.

Quote:
Originally Posted by Weaselrama View Post
Just to make sure I wasn't mis-remembering, I tried running a 68020, 68EC030, and a 68030, JIT disabled (I don't believe it could be enabled on the 68030), AfA libs disabled w/ both commands in startup-sequence commented out. Software failure before any of them even got to the WB and why would the WB start anyway? Most of my libraries are 68040, 040/060 or 060.
If you're not going to run some generic code, there's no point. A standard OS setup is a much better starting point, because once you start changing libraries, you're introducing other variables that might not necessarily be directly related to the CPU choice.

Quote:
iBrowse was slow as molasses on http sites even with JIT enabled (ok, compared to what I'm used to) and wouldn't bring up AmiSSL on https sites because the library is 68060 specific.
Again, sounds like a setup issue - maybe a CPU-specific executable is causing lots of traps because it's intended for a higher model CPU than the emulation provides. IBrowse loading simple pages is near-instantaneous for me, whether it's an '020 or an '040.

Quote:
What is it that I'm supposed to run a benchmark on? The only benchmark software I have is Sys Info and I've read more than once that it was worthless in emulation. Suggest one and I'll try it on the 68040-NOMMU vs the 68060-NOMMU. You wrote above not to change anything else.
Indeed, it's a little bit silly to compare such benchmarks to real hardware, because, as I've been trying to point out, emulation is a very different beast to real hardware and so the CPUs behave differently. But as a relative comparison, it's a good indication and demonstration of what is going on, since all it does is execute a tight loop of standardised instructions over and over. If you don't believe it, try something like AIBB, or even a real-world example like lha compressing an entire partition and timing it, as I suggested below.

Quote:
Please explain the point of offering different CPU configurations in UAE if everything runs just fine on a 68030, as fast or even faster?
I already answered your question on what's the point of emulating a 68060, but I'll do it again in case you missed it: For completeness, for testing or building specific software setups and for testing CPU-specific software, like '060-only demos, as you pointed out yourself. Just like the point in emulating an old network card when the bsdsocket.library emulation is so much better, or a Picasso IV when UAEGFX is so much more capable, or the specific IDE or RTC implementation of any machine, or the specific boot ROM of a specific accelerator.

Quote:
This has not been my experience with either iBrowse, Netsurf, AMPlifier, or AmigaAMP.
Which immediately points at compatibility or setup issues. I could find no perceptible difference using iBrowse or AMPlifier between an '020 and an '040 CPU setting.

Quote:
Let's not pretend there's a whole lot I could have screwed up in the process of upgrading over the past year. I keep a close eye on my startup-sequence and user-startup and keep a log of the libraries and shell commands I switch out with one exception, PeterK explained how the AfA libs replace the icon.library and picture.datatype. I had the idea they ran on top of them before he explained it, that, and using RemLib to be able to use his icon.library again.
I'm sure your setup is a beautifully hand-crafted work of CPU-specific art...

Quote:
You've run the benchmarks yourself and you've built configs with 040/060 libraries and run an emulated A4000 on a 68040 or 060 configuration and compared them?
Sometimes, yes, but most of the time it's generic 68020 code.

Quote:
I'm not trying to compare this to actual hardware running AfA. I'm happy to benchmark between the 040 and 060 if you tell me:
Quote:
1. The benchmarking software you want me to use.
Well, SysInfo is one, but for a real-world example, try compressing a partition using lha. There are '020 and '040-specific versions of the executable, and my testing hasn't found any difference while compressing my SYS: partition to RAM:, around 31 seconds, regardless of CPU chosen or lha executable.

Quote:
2. What you want me to benchmark, i.e, an idle system or specific software running? When I benchmark on the Mac, its usually a specific task or software such as running benchmarks on video rendering.
If you're running a benchmarker like SysInfo, it will disable multitasking to get the entire focus of the emulated CPU. If it's something else like lha, just do the very same thing - same sized Shell window after a fresh boot, for example. You could also try something like encoding an MP3, or any number of other CPU-intensive tasks that are otherwise stand-alone.

For pig iron, I ran it on FS-UAE, and guess what. Same results on 020, 030 and 040 - all around 1100-1200 MIPS. I also tried playing a 320kbps MP3 at 44.1kHz under each CPU - all played with no delay and no problems.

Bottom line, if using other CPU versions causes massive differences in speed, it's down to your setup, not the emulated CPU. All my tests, both real and artifical benchmarks, showed no significant difference.
Daedalus is online now  
Old 23 June 2020, 23:49   #47
Weaselrama
Registered User

Weaselrama's Avatar
 
Join Date: Oct 2019
Location: Neunkirchen aP, DE
Age: 58
Posts: 328
Quote:
Originally Posted by Daedalus View Post
Indeed, especially when it's way off-topic, yet here you are, still trying to argue against easily verifiable facts.
I didn't pull this thread this far off topic - another EAB'r pulled it that way with yet another conversation regarding his issues with icon libraries and Eastern and background corruption in AfA. I shouldn't have asked you about running AfA but I'm glad I did as it led to understanding CPU under emulation, covered in my remarks below.


Quote:
Originally Posted by Daedalus View Post
Once you're happy. The only reason I'm suggesting that you run some benchmarks is because your beliefs around the emulated CPU are misplaced, and benchmarking is the best way to objectively demonstrate that.
You missed my follow-up remarks. I read a lot earlier this afternoon, including conversations that went further into detail and helped me understand UAE emulation much better. I get it now, thanks. I will likely reconfigure to 68040 instead of 68030 and for no better reason than it will take me less time to switch out some of the libraries and shell commands, not to mention apps which were compiled for 060 such as FFPlay. I recognize that a 68030 base would be preferable but it's based on how much time I want to invest doing it. My slow-down in iBrowse when I switched FS-UAE over to 68040 was likely because I installed it with its PROGDIR 060 libs. Easily fixed. I'm glad I archive all my downloads.


Quote:
Originally Posted by Daedalus View Post
For pig iron, I ran it on FS-UAE, and guess what. Same results on 020, 030 and 040 - all around 1100-1200 MIPS. I also tried playing a 320kbps MP3 at 44.1kHz under each CPU - all played with no delay and no problems.
Off-topic - I can do true 44,100 in Songplayer in stereo but I don't like Songplayer except you can switch off AHI and thus, 44,100. I can live with AHI @ 22,050 futzing around on the Amiga and easily switch over to Internet Radio using AmigaAMP which is tied to AHI but otherwise set to stereo/44,100 max, no division. I've not spent much time on it but I've yet to coax anything higher than 22,050 out of the Paula AHI Music Unit but I know it can be done. I've downloaded the Toccata libs but I haven't installed them yet.

Last edited by Weaselrama; 24 June 2020 at 13:53.
Weaselrama is offline  
Old 24 June 2020, 11:26   #48
AMIGASYSTEM
Registered User
AMIGASYSTEM's Avatar
 
Join Date: Aug 2014
Location: Brindisi (Italy)
Posts: 5,815
Quote:
Originally Posted by Weaselrama View Post



Off-topic - I can do true 44,100 in Songplayer in stereo but I don't like Songplayer except you can switch off AHI and thus, 44,100. I can live with AHI @ 22,050 futzing around on the Amiga and easily switch over to Internet Radio using AmigaAMP which is tied to AHI but otherwise set to stereo/44,100 max, no division. I've not spent much time on it but I've yet to coax anything higher than 22,050 out of the Paula AHI Music Unit but I know it can be done. I've downloaded the Toccata libs but I haven't installed them yet.

In my opinion the best choice is the Device UAE 16Bit HiFi (uae.audio)

https://drive.google.com/file/d/18uj...XGelUyhfh/view
AMIGASYSTEM is offline  
Old 24 June 2020, 13:53   #49
Weaselrama
Registered User

Weaselrama's Avatar
 
Join Date: Oct 2019
Location: Neunkirchen aP, DE
Age: 58
Posts: 328
Quote:
Originally Posted by AMIGASYSTEM View Post
In my opinion the best choice is the Device UAE 16Bit HiFi (uae.audio)

https://drive.google.com/file/d/18uj...XGelUyhfh/view
(off-topic) I'm not completely certain but I don't think the uae.audio device is available in FS-UAE. I'm probably wrong but I've never seen a reference to it other than for WinUAE. I haven't spent a lot of time on audio configuration but every reference to AHI in FS-UAE mentions either Paula or Toccata.

EDIT: Toccata setup is a breeze in FS-UAE as of v3.x. I don't know why I never played with it before. For AmigaAMP, music unit 0 *must* be set to Toccata as that's the only music unit AmigaAMP uses.

Last edited by Weaselrama; 24 June 2020 at 16:22. Reason: add info
Weaselrama is offline  
Old 24 June 2020, 18:49   #50
AMIGASYSTEM
Registered User
AMIGASYSTEM's Avatar
 
Join Date: Aug 2014
Location: Brindisi (Italy)
Posts: 5,815
Quote:
Originally Posted by Weaselrama View Post
(off-topic) I'm not completely certain but I don't think the uae.audio device is available in FS-UAE.
Yes on FS-UAE is not supported uae.audio i asked FrodeSolheim some time ago HERE

On FS-UAE there are also other things not supported shown in this video.

[ Show youtube player ]

Last edited by AMIGASYSTEM; 24 June 2020 at 19:28.
AMIGASYSTEM is offline  
Old 24 June 2020, 19:32   #51
Weaselrama
Registered User

Weaselrama's Avatar
 
Join Date: Oct 2019
Location: Neunkirchen aP, DE
Age: 58
Posts: 328
Quote:
Originally Posted by AMIGASYSTEM View Post
Yes on FS-UAE is not supported uae.audio i asked FrodeSolheim some time ago HERE

On FS-UAE there are also other things not supported shown in this video.

[ Show youtube player ]
Not surprising. WinUAE has a long legacy and pedigree. Frode has been working hard on importing WinUAE features for version 4. Toccata works perfectly well for right now.
Weaselrama is offline  
Old 25 June 2020, 16:46   #52
Weaselrama
Registered User

Weaselrama's Avatar
 
Join Date: Oct 2019
Location: Neunkirchen aP, DE
Age: 58
Posts: 328
@Daedalus. Well, I won't waste anymore space on the subject. I had to reverse course and go back to the 68060. Not for speed - I compared my 040 and 060 setup as you suggested and they were precisely the same speed idle and while playing 320kbps mp3s @ 44,100 although I didn't get anything like the speed in MIPS you reported. Could be my version of Sys Info - Toni Wilen has stated that benchmarks in emulation were worthless. It did, however demonstrate that emulated CPUs were all the same: Faster than a real world 68k CPU and only different in instruction set. However, in my case, I had too many 060 AGA/RTG demos that wouldn't play with the fake 060 thingie. Mostly just the newer ones from about 2016 onward. So I went back to the 060 for that alone. No biggie - since there's no difference in speed, it doesn't matter which processor I use anyway. I would say the inability to play mp3s last year above 11025 in stereo was a combination of the known FS-UAE RTG audio latency and trying to run (at the time) too high a resolution in light of that bug and blaming it on the processor.

On to bigger and better things. This had one positive outcome: Taking the time to get Toccata running. Piece of cake since Toccata is fully supported now in FS-UAE...
Weaselrama is offline  
 


Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools

Similar Threads
Thread Thread Starter Forum Replies Last Post
Is there an app to change the workbench menu imagery? gulliver request.Apps 4 14 August 2017 18:44
ARexx help for a workbench menu action gulliver support.Other 7 15 February 2016 09:22
Looking for some basic IP tools en SDK tools Roland007 support.WinUAE 0 28 November 2012 22:34
Useful tools for Workbench ? Frazor New to Emulation or Amiga scene 12 30 January 2003 11:14
Best free Workbench 1.3 enhancement tools Paul request.Apps 0 29 January 2003 10:26

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump


All times are GMT +2. The time now is 10:19.


Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, vBulletin Solutions Inc.
Page generated in 0.07957 seconds with 15 queries