Click Here
home features news forums classifieds faqs links search
6025 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
64 crawler(s) on-line.
 18 guest(s) on-line.
 0 member(s) on-line.



You are an anonymous user.
Register Now!
 Fl@sh:  7 mins ago
 Georg:  18 mins ago
 matthey:  43 mins ago
 amigang:  1 hr 22 mins ago
 hardwaretech:  1 hr 25 mins ago
 michalsc:  1 hr 40 mins ago
 RobertB:  1 hr 44 mins ago
 mac6:  1 hr 56 mins ago
 bison:  2 hrs 13 mins ago
 RobertJDohnert:  2 hrs 24 mins ago

/  Forum Index
   /  Amiga General Chat
      /  3.3
Register To Post

Goto page ( Previous Page 1 | 2 | 3 Next Page )
PosterThread
number6 
Re: 3.3
Posted on 9-Jun-2021 15:07:06
#21 ]
Elite Member
Joined: 25-Mar-2005
Posts: 11251
From: In the village

@AmigaBlitter

There is a phrase "the devil you know".

Quote:
meaning that, when forced to decide, it's better to choose a difficult or undesirable option that one is used to or familiar with rather than an unknown person or thing.


I believe that plays a part in thinking around here.

#6

_________________
This posting, in its entirety, represents solely the perspective of the author.
*Secrecy has served us so well*

 Status: Offline
Profile     Report this post  
bison 
Re: 3.3
Posted on 9-Jun-2021 15:08:24
#22 ]
Super Member
Joined: 18-Dec-2007
Posts: 1881
From: N-Space

@NutsAboutAmiga

How did you get this impressively big axe to grind with all things related to Cloanto, Mike Battilana, etc?

_________________
"Unix is supposed to fix that." -- Jay Miner

 Status: Offline
Profile     Report this post  
bison 
Re: 3.3
Posted on 9-Jun-2021 15:13:18
#23 ]
Super Member
Joined: 18-Dec-2007
Posts: 1881
From: N-Space

@number6

My last post timed out after two minutes, probably because we were trying to post at about the same time. Apparently someone posting temporarily locks the entire thread. Interesting.

_________________
"Unix is supposed to fix that." -- Jay Miner

 Status: Offline
Profile     Report this post  
Hammer 
Re: 3.3
Posted on 9-Jun-2021 15:24:59
#24 ]
Elite Member
Joined: 9-Mar-2003
Posts: 4270
From: Australia

@matthey

Quote:

matthey wrote:
NutsAboutAmiga Quote:

Amiga tried to be different, and different was not always a good thing.
Zorro III, made it hard to use the best graphic chips on the market, and sound cards.


Let's compare Zorro III to other 32 bit buses.

VMEbus32 - 1981 release
NuBus - 1987 release
EISA - 1988 release
Zorro III - 1990 release in Amiga 3000 (should have been earlier and with AGA)
PCI 1.0 - 1992 release but not popular in PCs until 1994
PCI 2.1 - 1995 release allowing 66 MHz operation

VMEbus32 theoretical bandwidth is 40 MB/s
NuBus maximum burst throughput of 40 MB/s with average throughput of 10 to 20 MB/s
EISA theoretical bandwidth is 33 MB/s
Zorro III theoretical bandwidth is 150 MB/s
PCI 1.0 33MHz 32b theoretical bandwidth is 133 MB/s
PCI 2.1 66MHz 32b theoretical bandwidth is 266 MB/s

There were bugs in Busters which limited throughput but CBM cut R&D budgets and didn't have a clear development roadmap. There are other important factors besides throughput but Zorro III stands up pretty well as being flexible and relatively trouble free. The winners write the history books and define the industry standards though.


PCI's theoretical 132 MB/s I/O bandwidth is backed by Pentium's 64 bit FSB at 50/60/66 Mhz clock speeds. 68040 and 68060 have 32 bit FSB.

486's bandwidth is dependant on the CPU's bus speed: It started at 100 MB/s for CPUs with a 25 MHz bus, increased to 133 MB/s at 33 MHz and 160 MB/s at 40 MHz, and reached 200 MB/s at 50 MHz.

486's VL-bus for the dGPU idea was recycled for AGP which is located at the North Bridge. AGP was created in 1997.

PCI was later mass-produced in numbers while ZIII remained as niche workstation Amiga boxes.

My 1996 era Pentium 150 classic motherboard + S3 Trio 64U card has a lower large chip count when compared to A1200/A3000/A4000.

Last edited by Hammer on 09-Jun-2021 at 03:37 PM.
Last edited by Hammer on 09-Jun-2021 at 03:36 PM.
Last edited by Hammer on 09-Jun-2021 at 03:34 PM.
Last edited by Hammer on 09-Jun-2021 at 03:29 PM.
Last edited by Hammer on 09-Jun-2021 at 03:28 PM.
Last edited by Hammer on 09-Jun-2021 at 03:26 PM.

_________________
Core i9-9900K, DDR4-3800 32 GB RAM, GeForce RTX 3080 Ti
Ryzen 9 3900X, DDR4-3200 32 GB RAM, GeForce RTX 2080 Ti
Amiga 1200 (rev 1D1, KS 3.2, TF1260, 68060 @ 63 Mhz, 128 MB)
Amiga 500 (rev 6A, KS 3.2, 68K 50Mhz, 12 MB RAM)

 Status: Offline
Profile     Report this post  
NutsAboutAmiga 
Re: 3.3
Posted on 9-Jun-2021 17:22:09
#25 ]
Elite Member
Joined: 9-Jun-2004
Posts: 11881
From: Norway

@AmigaBlitter

I’m only a spectator, maybe a fortune teller, most defiantly not a news anchorman man, my opinions are my own.

Last edited by NutsAboutAmiga on 09-Jun-2021 at 05:52 PM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

 Status: Offline
Profile     Report this post  
NutsAboutAmiga 
Re: 3.3
Posted on 9-Jun-2021 17:23:22
#26 ]
Elite Member
Joined: 9-Jun-2004
Posts: 11881
From: Norway

@number6

OMG, SMILIFACE.

its like watching an episode of the circle / Netflix, where envy one is trying to get some money at end of the show, Its a bit like that, is it not.

Fake people, conspirator, catfishes, and backstabbers. The nice people are the people voted out first or taken advantage of. Well you know what remains.

Sure, pick the devil you know. Or at least know is delivers something.
the grass is not always greener on the other side. If you are trying get to greener grass by tumbling down the mountain, who knows if ever get to ate the grass.

Last edited by NutsAboutAmiga on 09-Jun-2021 at 05:41 PM.
Last edited by NutsAboutAmiga on 09-Jun-2021 at 05:34 PM.
Last edited by NutsAboutAmiga on 09-Jun-2021 at 05:33 PM.
Last edited by NutsAboutAmiga on 09-Jun-2021 at 05:30 PM.
Last edited by NutsAboutAmiga on 09-Jun-2021 at 05:28 PM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

 Status: Offline
Profile     Report this post  
number6 
Re: 3.3
Posted on 9-Jun-2021 17:38:41
#27 ]
Elite Member
Joined: 25-Mar-2005
Posts: 11251
From: In the village

@NutsAboutAmiga

Quote:
The nice people are the people voted out first or taken advantage of.


Finally...
A 1 line summary of how things work around here. heh.

#6

_________________
This posting, in its entirety, represents solely the perspective of the author.
*Secrecy has served us so well*

 Status: Offline
Profile     Report this post  
matthey 
Re: 3.3
Posted on 9-Jun-2021 20:58:43
#28 ]
Super Member
Joined: 14-Mar-2007
Posts: 1151
From: Kansas

Hammer Quote:

PCI's theoretical 132 MB/s I/O bandwidth is backed by Pentium's 64 bit FSB at 50/60/66 Mhz clock speeds. 68040 and 68060 have 32 bit FSB.


The 68060 had better cache efficiency than the Pentium, due to better code density and more cache ways, allowing similar performance with the 32b data bus while allowing cheaper 32b memory to be used, saving lines on boards allowing them to be cheaper and saving pins on the CPU allowing it to be cheaper. There is likely a 68060 energy savings as well from the narrower but well matched bus width and certainly from the better cache efficiency. The Pentium had higher memory bandwidth which it needed to keep up with more efficient designs like the 68060.

While DDR memory transfers are 64b wide, 68k replacement hardware today does not necessarily use it. As I recall, some memory is only 16b like on the Buffee (TI 3358 SoC is a decade old), original Vampire, FleaFPGA Ohm and likely others. It shouldn't be too much of a bottleneck on very low performance hardware but may be for the Buffee.

Hammer Quote:

486's bandwidth is dependant on the CPU's bus speed: It started at 100 MB/s for CPUs with a 25 MHz bus, increased to 133 MB/s at 33 MHz and 160 MB/s at 40 MHz, and reached 200 MB/s at 50 MHz.

486's VL-bus for the dGPU idea was recycled for AGP which is located at the North Bridge. AGP was created in 1997.

PCI was later mass-produced in numbers while ZIII remained as niche workstation Amiga boxes.

My 1996 era Pentium 150 classic motherboard + S3 Trio 64U card has a lower large chip count when compared to A1200/A3000/A4000.


The 68k and Amiga fell behind going into the mid-90's mostly because of improved chip process die sizes they were not taking advantage of. Moore's Law really kicked in hard at that time. The Amiga was taking years to enhance and integrate chips where leading technology companies were updating hardware at least yearly and mass production was necessary for survival. With Moore's Law slowing today, we now have well designed chips on the market for over a decade like the TI 3358 SoC used in the Buffee. The integrated memory controller is the part which has aged the most due to it being difficult to support future memory types when developed. One way around this would be to use on chip eDRAM in an SoC ASIC which also eliminate pins, saves board space and provides better performance than external memory but costs more.

 Status: Offline
Profile     Report this post  
Hammer 
Re: 3.3
Posted on 10-Jun-2021 5:44:17
#29 ]
Elite Member
Joined: 9-Mar-2003
Posts: 4270
From: Australia

@matthey

Quote:

The 68060 had better cache efficiency than the Pentium, due to better code density and more cache ways, allowing similar performance with the 32b data bus while allowing cheaper 32b memory to be used, saving lines on boards allowing them to be cheaper and saving pins on the CPU allowing it to be cheaper. There is likely a 68060 energy savings as well from the narrower but well matched bus width and certainly from the better cache efficiency. The Pentium had higher memory bandwidth which it needed to keep up with more efficient designs like the 68060.

While DDR memory transfers are 64b wide, 68k replacement hardware today does not necessarily use it. As I recall, some memory is only 16b like on the Buffee (TI 3358 SoC is a decade old), original Vampire, FleaFPGA Ohm and likely others. It shouldn't be too much of a bottleneck on very low performance hardware but may be for the Buffee.

FYI, my classic Pentium 150's motherboard 512 KB L2 cache.

When compared to classic Pentium FPU, 68060's FPU is not pipelined. This factor is important for Quake-type games which pushed Pentium competitors with weak FPUs into the niche market segment.

Pentium's 64-bit bus is needed for FPU workloads since they exceeded the 32-bit bus width. Modern desktop PCs have dual-channel DIMM which is a 128-bit bus.

168-pin DIMM SDR has 64 bits wide channel without DDR.

LPDDR permits 16- or 32-bit wide channels.

AMD 4700S APU has Xbox Series X's 320-bit GDDR6-14000 memory bus.

Lower latency yields better performance, hence AMD's 192 MB L3 cache for Ryzen 9 5950 improves gaming performance by a further 12 to 15 percent when compared to normal Ryzen 9 5950.

For Quake benchmarks, Pentium 166 (with 430VX chipset) can do 37.30 fps. https://thandor.net/benchmark/33


From terriblefire
Buffee to TF bus runs at 100Mhz. TF doubles bus cycles up to 32bit. So no slowdown from buffee being 16bit (well there is a 10ns penalty over a TF1260)


Quote:

The 68k and Amiga fell behind going into the mid-90's mostly because of improved chip process die sizes they were not taking advantage of. Moore's Law really kicked in hard at that time. The Amiga was taking years to enhance and integrate chips where leading technology companies were updating hardware at least yearly and mass production was necessary for survival. With Moore's Law slowing today, we now have well designed chips on the market for over a decade like the TI 3358 SoC used in the Buffee. The integrated memory controller is the part which has aged the most due to it being difficult to support future memory types when developed. One way around this would be to use on chip eDRAM in an SoC ASIC which also eliminate pins, saves board space and provides better performance than external memory but costs more.

68K was killed by Motorola/Freescale and there was no "AMD" cloner for the 68K market.

68K licensees have weak CPU R&D.

Intel has plans to kill X86 with Itanium and it was rendered useless with AMD's K8 Sledgehammer/Clawhammer which forced Intel to develop project Yamhill (aka Prescott) i.e. Pentium IV with X86-64 compatibility.

Intel Itanium and IBM PowerPC 970 attempted to take desktop PC crown and they were beaten down by AMD's K8 Athlon 64 and K8 Opteron.

IBM forcing Intel for second source insurance has worked for keeping X86 PC alive.

-----------------

Amiga 3000's complexity didn't lead to low-cost desktop PC manufacturing.


PC gaming doomed the Amiga when games such as Doom were released.


From https://webcache.googleusercontent.com/search?q=cache:NfZ_G7tzVw0J:https://www.latimes.com/archives/la-xpm-1993-12-23-fi-4940-story.html+&cd=11&hl=en&ct=clnk"
The year 1993, 486 at 33Mhz PC price in California, USA
Article date: DEC. 23, 1993

A year ago, a San Francisco-area PC clone dealer known for its low prices was advertising a fully equipped 33 Mhz 486 PC for $1,388. Today, that same machine costs about $1,000



A500's October 1987 introductory price is $699 USD. Note that $1000 USD 486 33Mhz based PC in December 1993 is approaching A500's October 1987 introductory price range.

The year 1993 marks the major rise in PC gaming.



No 3rd party Amiga CPU accelerator will match Commodore's economics of scale.

For year 1993, the uncompetitive nature with 68K can also stem from Motorola not just from Commodore. Commodore is just of many 68K platform vendors who jumped ship away from 68K e.g. PA-RISC 7150 based Amiga Hombre.

None of the 68 K-based PC vendors was able to make 68040 in 68000's and 68020's higher unit numbers.



68000 based A500's $699 USD October 1987 introductory price is about USD $1,600 in 2020 equivalent.

--

From USA's Amiga World Magazine (November 1993), page 58 of 100,
Price listed in USD in November 1993

A1200/020, 2MB, price $379
A3000/030 at 25Mhz, 5MB, 105HD, price $899
A3000T/030 at 25Mhz, 5MB, 200MB HDD, price $1199
A3000T/040 at 25Mhz, 5MB, 200MB HDD, price $1599
A3000s are missing AGA chipset.



Cost estimate for 68040 card, $1599 - $1199, cost for 040 card = $400

A1200's $379 + 040 card's $400 = $779.



Commodore could have out-of-the-box configured A1200 with 68040 at 25Mhz for slightly above $779 (i.e. add 4MB fast ram, small HDD) which could compete against $1000 out-of-the-box 486 33Mhz based PC. Amiga 1200 with 68LC040 + 4MB fast ram at around $779 cost would be targeting Doom type PC gaming.

Amiga ECS upgrade is like the near-useless Commodore 128 upgrade i.e. both products focus on low color display high resolution with aging gaming hardware.

I owned an Amiga 3000/030 at 25Mhz in early 1992 and my major criticism against Commodore's management is ECS is not good enough against similar priced entry-level SVGA-equipped PCs.

Commodore runs down the Amiga platform like C64-to-C128 move!

My Dad could have purchased 486 based PC instead of the Amiga 3000, but my Dad plays sports games on the Amiga 500. My Dad already has 386SX with 387 FPU for His office work at the same time.

Last edited by Hammer on 10-Jun-2021 at 05:49 AM.

_________________
Core i9-9900K, DDR4-3800 32 GB RAM, GeForce RTX 3080 Ti
Ryzen 9 3900X, DDR4-3200 32 GB RAM, GeForce RTX 2080 Ti
Amiga 1200 (rev 1D1, KS 3.2, TF1260, 68060 @ 63 Mhz, 128 MB)
Amiga 500 (rev 6A, KS 3.2, 68K 50Mhz, 12 MB RAM)

 Status: Offline
Profile     Report this post  
matthey 
Re: 3.3
Posted on 11-Jun-2021 2:30:20
#30 ]
Super Member
Joined: 14-Mar-2007
Posts: 1151
From: Kansas

Hammer Quote:

FYI, my classic Pentium 150's motherboard 512 KB L2 cache.


I found 50ns SIMMs for my 68060 CSMK3, clocked up the memory bus by 50% and still was able to use minimum wait states which gave a nice performance boost. This and external L2 are nowhere near as good as on chip L2 though. Funny thing is the Amiga or an embedded system can run from memory the size of L2 caches today and no L2 is needed in this case. An on chip 2MiB cache could be used as L2 for external memory or reconfigured as very fast main memory with no L2 for example.

Hammer Quote:

When compared to classic Pentium FPU, 68060's FPU is not pipelined. This factor is important for Quake-type games which pushed Pentium competitors with weak FPUs into the niche market segment.

Pentium's 64-bit bus is needed for FPU workloads since they exceeded the 32-bit bus width. Modern desktop PCs have dual-channel DIMM which is a 128-bit bus.


68060 FPU instructions are partially processed in the pipelined integer units so it is not as bad as it seems. The FPU operates in parallel with integer units so it does pretty good on mixed integer and floating point workloads which are scheduled correctly. A fully pipelined FPU likely would have helped floating point intensive code and may have made scheduling easier.

I suspected the 68060 suffered from too small of DCache when doing heavy floating point workloads. A 64 bit data bus likely would have improved FPU performance but an on chip L2 likely would give better performance today. Certainly both an L2 and 64 bit data bus are possible in higher performance and higher cost hardware.

Hammer Quote:

168-pin DIMM SDR has 64 bits wide channel without DDR.

LPDDR permits 16- or 32-bit wide channels.


Yes. LPDDR is targeted at mobile and embedded systems where power and cost may be more important than performance.

Hammer Quote:

AMD 4700S APU has Xbox Series X's 320-bit GDDR6-14000 memory bus.

Lower latency yields better performance, hence AMD's 192 MB L3 cache for Ryzen 9 5950 improves gaming performance by a further 12 to 15 percent when compared to normal Ryzen 9 5950.


High end GPUs need crazy amounts of memory bandwidth and data is more commonly not cached than for CPU cores. It is amazing that mass production allows performance features like this for an APU in a relatively inexpensive console. I wouldn't want to compete in the high performance console market but there is probably room for cheap mini consoles.

Hammer Quote:

From terriblefire
Buffee to TF bus runs at 100Mhz. TF doubles bus cycles up to 32bit. So no slowdown from buffee being 16bit (well there is a 10ns penalty over a TF1260)


Clocking up the bus isn't energy efficient but gives bandwidth and may be cheaper than a wider bus. The nice thing about DDR memory is that it doubles the bandwidth without clocking up the bus.

Hammer Quote:

68K was killed by Motorola/Freescale and there was no "AMD" cloner for the 68K market.

68K licensees have weak CPU R&D.


The trademarks are expired now so no license is needed. The only 68k design with higher performance than the 68060 is the Apollo Core using modern memory performance and larger caches.

Hammer Quote:

Intel has plans to kill X86 with Itanium and it was rendered useless with AMD's K8 Sledgehammer/Clawhammer which forced Intel to develop project Yamhill (aka Prescott) i.e. Pentium IV with X86-64 compatibility.

Intel Itanium and IBM PowerPC 970 attempted to take desktop PC crown and they were beaten down by AMD's K8 Athlon 64 and K8 Opteron.

IBM forcing Intel for second source insurance has worked for keeping X86 PC alive.


The 68k had a 2nd source producer too in Hitachi. Unfortunately, they developed an incompatible SuperH from the 68000 design, for which they were sued, and then licensed that to ARM for the Thumb encodings. It was easier to change ISAs for the 68k than for x86 on the desktop. Itanium at least allowed to retain x86 compatibility but VLIW proved more expensive to develop and less general purpose than CISC. PPC was difficult to adopt because it was a new ISA. IBMs failure at creating a high performance PPC design in the G5 and the bad reputation of low performance PPC designs like the 603, partially due to poor code density, meant PPC only had mid performance reasonably energy efficient processors that were practical but not exciting.

Hammer Quote:

For year 1993, the uncompetitive nature with 68K can also stem from Motorola not just from Commodore. Commodore is just of many 68K platform vendors who jumped ship away from 68K e.g. PA-RISC 7150 based Amiga Hombre.


Yes, even Motorola was not able to keep up with Intel for the desktop market. Unfortunately, they panicked, fell for the RISC hype and gave up the embedded market which would have given them the economies of scale ARM has used, with 68k technology, to challenge Intel today.

Hammer Quote:

Amiga ECS upgrade is like the near-useless Commodore 128 upgrade i.e. both products focus on low color display high resolution with aging gaming hardware.

I owned an Amiga 3000/030 at 25Mhz in early 1992 and my major criticism against Commodore's management is ECS is not good enough against similar priced entry-level SVGA-equipped PCs.

Commodore runs down the Amiga platform like C64-to-C128 move!


I agree. CBM treated the Amiga like another C64. They were good at reducing production costs but failed to integrate the technology fast enough and forgot about value. They should have had an Amiga SoC with 68020+AGA+RTG with enhancements in the CD32 for half the price or with 4MiB of fast memory. Amateurs did a better job of developing SAGA than CBM who led the world in PC sales for several years. Inconceivable!

 Status: Offline
Profile     Report this post  
Hammer 
Re: 3.3
Posted on 11-Jun-2021 3:46:56
#31 ]
Elite Member
Joined: 9-Mar-2003
Posts: 4270
From: Australia

@matthey
Quote:

I found 50ns SIMMs for my 68060 CSMK3, clocked up the memory bus by 50% and still was able to use minimum wait states which gave a nice performance boost. This and external L2 are nowhere near as good as on chip L2 though. Funny thing is the Amiga or an embedded system can run from memory the size of L2 caches today and no L2 is needed in this case. An on chip 2MiB cache could be used as L2 for external memory or reconfigured as very fast main memory with no L2 for example.

Sometime in 1996, it's either Cyberstorm 68060 at 50Mhz + CyberGraphics 64 (S3 Trio 64V VLB variant) upgrades for my Amiga 3000 vs new build Pentium 150 + S3 Trio 64UV PCI + Yamaha OPL3 16bit soundcard. The main target use case: playing Quake.

Pentium's Socket 7 430VX motherboard has the 512 KB L2 cache.

New build Pentium PC was cheaper and faster. I overclocked Pentium 150 into 166 Mhz with 60 to 66 Mhz FSB jumper.


Quote:

68060 FPU instructions are partially processed in the pipelined integer units so it is not as bad as it seems. The FPU operates in parallel with integer units so it does pretty good on mixed integer and floating point workloads which are scheduled correctly. A fully pipelined FPU likely would have helped floating point intensive code and may have made scheduling easier.

I suspected the 68060 suffered from too small of DCache when doing heavy floating point workloads. A 64 bit data bus likely would have improved FPU performance but an on chip L2 likely would give better performance today. Certainly both an L2 and 64 bit data bus are possible in higher performance and higher cost hardware.

Motorola's R&D attention was diverted from 68K i.e. 88000 and PowerPC R&D. It's most likely there was not enough time to complete 68060 FPU.

Intel mostly focused on Pentium R&D.

Cyberstorm has additional chips to support Amiga 3000/4000 Zorro III additional complexity.

Amiga 4000 design wasn't native for 68040/68060 CPUs since it has a 68030 bus.

The mentioned new build Pentium PC was cheaper and faster. Pentium 430VX chipset-based motherboard has less large package chip count when compared to A1200/A3000/4000, that's not including Cyberstorm 68060 accelerator card.

Quote:

High end GPUs need crazy amounts of memory bandwidth and data is more commonly not cached than for CPU cores. It is amazing that mass production allows performance features like this for an APU in a relatively inexpensive console. I wouldn't want to compete in the high performance console market but there is probably room for cheap mini consoles.

NAVI 21 RX 6800/6800XT/6900XT's 128 MB Infinity Cache + real-time Delta Color Compression/Decompression works great for traditional raster 3D PC games (non-raytracing).

128 MB Infinity Cache acts like an L3 cache for the NAVI 21 GPU.

RX 6800XT has external memory bandwidth similar to RTX 3070 (256 bit GDDR6-14000) with traditional raster 3D performance similar to RTX 3080 (with 320 bit GDDR6X-18000+) and RTX 3080 Ti/3090(with 384-bit GDDR6X-18000/19000).

Recent game console APUs do not include PC NAVI 2x's Infinity Cache (L3 cache).

128 MB Infinity Cache on NAVI 21 was interesting.

Quote:

Clocking up the bus isn't energy efficient but gives bandwidth and may be cheaper than a wider bus. The nice thing about DDR memory is that it doubles the bandwidth without clocking up the bus.

Desktop gaming (also desktop graphics apps) is about maximum performance for the money.


Quote:

The 68k had a 2nd source producer too in Hitachi. Unfortunately, they developed an incompatible SuperH from the 68000 design, for which they were sued, and then licensed that to ARM for the Thumb encodings. It was easier to change ISAs for the 68k than for x86 on the desktop. Itanium at least allowed to retain x86 compatibility but VLIW proved more expensive to develop and less general purpose than CISC. PPC was difficult to adopt because it was a new ISA. IBMs failure at creating a high performance PPC design in the G5 and the bad reputation of low performance PPC designs like the 603, partially due to poor code density, meant PPC only had mid performance reasonably energy efficient processors that were practical but not exciting.

IBM's X86 second source insurance requirement has worked.

Hitachi didn't attach 68K front end with SuperH RISC core like AMD's K5 which has 29K RISC core.

SuperH is closer to a Japan Inc CPU project with attempts to be instruction set independent from the US-dominated CPU instruction sets.

SH1 and SH2 were used in Sega Saturn, Sega 32X, and Capcom CPS-3.

https://www.youtube.com/watch?v=LUZ436FXB4U
Digital Foundry's Sega Saturn's Quake analysis. Unstable frame rates with a range from 10 to 27 fps. Mostly in 15 to 20 fps.

Amiga 1200 with 68060+fast ram or Amiga Hombre with PA-RISC/OpenGL would have countered Sega Saturn.

For gaming PC, Pentium CPU family was important to counter 3DO, Sega Saturn, and Sony PS1 game console generation.


Quote:

I agree. CBM treated the Amiga like another C64. They were good at reducing production costs but failed to integrate the technology fast enough and forgot about value. They should have had an Amiga SoC with 68020+AGA+RTG with enhancements in the CD32 for half the price or with 4MiB of fast memory. Amateurs did a better job of developing SAGA than CBM who led the world in PC sales for several years. Inconceivable!

The missing approach was a balance between low-cost A1200 and high-cost A4000.

Pentium 75/90/100 in 1994 has pushed older Pentium 60/66 into a lower price category.

Pentium 120 in 1995 has pushed older Pentium 75/90/100 into a lower price category.

Pentium 133/150/166/200 in 1996 has pushed older Pentium 75/90/100/120 into a lower price category. Pentium Pro has pushed Pentium 133/150/166/200 into a lower desktop category.

When I purchased Pentium 150 in 1996, it wasn't the top-tier SKU from Intel. I overclocked Pentium 150.

The Apollo team did a good job with "What IF" i.e. the missing 68K cloner like AMD for X86.

Anyway, CBM's MOS 65xx team was supposed to be the Motorola 6800 cloner team. Good management is important.


Last edited by Hammer on 11-Jun-2021 at 03:53 AM.
Last edited by Hammer on 11-Jun-2021 at 03:50 AM.

_________________
Core i9-9900K, DDR4-3800 32 GB RAM, GeForce RTX 3080 Ti
Ryzen 9 3900X, DDR4-3200 32 GB RAM, GeForce RTX 2080 Ti
Amiga 1200 (rev 1D1, KS 3.2, TF1260, 68060 @ 63 Mhz, 128 MB)
Amiga 500 (rev 6A, KS 3.2, 68K 50Mhz, 12 MB RAM)

 Status: Offline
Profile     Report this post  
matthey 
Re: 3.3
Posted on 11-Jun-2021 22:46:56
#32 ]
Super Member
Joined: 14-Mar-2007
Posts: 1151
From: Kansas

Hammer Quote:

Motorola's R&D attention was diverted from 68K i.e. 88000 and PowerPC R&D. It's most likely there was not enough time to complete 68060 FPU.


The first PPC development started in 1992 and was half funded and staffed by IBM. The 88k development likely seemed like more of a duplication of resources with the PPC development especially after PPC adopted the 88k bus and considering the 88k was struggling to gain market share. The 68k market was still healthy and there was reason to believe the 68060 would be profitable which I have heard it was. There were parts of the 68060 which seem unfinished and unpolished but then it was the first of a new generation (microarchitecture) of 68k processors with a well design foundation to build on. The lack of a fully pipelined FPU could have been because of a transistor budget or cost benefit analysis revealing it was not worthwhile considering the performance goals which appear to be focused on integer performance.

Hammer Quote:

Intel mostly focused on Pentium R&D.


Let's not forget about the Intel i860. The 2nd gen came out in 1991 but they may have been working on a 3rd generation before cancellation and switching to ARM XScale. Intel did seem to be more persistent with the x86 but the PC boom had turned it into a cash cow for them. All they had to do was make a moderately higher performance x86 compatible processor and it was pretty much guaranteed of market success even if Intel management had doubts about the architecture in the long run.

Hammer Quote:

Cyberstorm has additional chips to support Amiga 3000/4000 Zorro III additional complexity.

Amiga 4000 design wasn't native for 68040/68060 CPUs since it has a 68030 bus.

The mentioned new build Pentium PC was cheaper and faster. Pentium 430VX chipset-based motherboard has less large package chip count when compared to A1200/A3000/4000, that's not including Cyberstorm 68060 accelerator card.


Some of the early PC standards were cheap and inferior but competition replaced them with better and better standards. CBM seemed to standstill watching it all happen.

Hammer Quote:

Desktop gaming (also desktop graphics apps) is about maximum performance for the money.


GPUs are practically limited by power requirements. Most computer users don't have 1000+ W power supplies and don't want their computer turning into a space heater. Consoles have even smaller power budgets.

Hammer Quote:

IBM's X86 second source insurance requirement has worked.

Hitachi didn't attach 68K front end with SuperH RISC core like AMD's K5 which has 29K RISC core.

SuperH is closer to a Japan Inc CPU project with attempts to be instruction set independent from the US-dominated CPU instruction sets.


The big mistake of SuperH was the fixed length 16 bit encoding. Lack of space for immediate and displacement bits in a 16 bit length instruction means many more instructions are required which are often dependent so can't be executed in parallel (superscalar). Switching to RISC already requires more instructions. SuperH needed nearly 50% more instructions than the 68k in some code. This is likely to kill single thread performance which is important for games.

Hammer Quote:

SH1 and SH2 were used in Sega Saturn, Sega 32X, and Capcom CPS-3.

https://www.youtube.com/watch?v=LUZ436FXB4U
Digital Foundry's Sega Saturn's Quake analysis. Unstable frame rates with a range from 10 to 27 fps. Mostly in 15 to 20 fps.


The Sega is not very impressive. My 68060@75MHz Amiga with Voodoo gfx has a higher frame rate and looks much better at 512x384x16. Consoles are generally not the highest performance hardware available and the Sega Saturn had some rather poor 3D hardware and has low memory for 3D games.

Hammer Quote:

Amiga 1200 with 68060+fast ram or Amiga Hombre with PA-RISC/OpenGL would have countered Sega Saturn.


Maybe if the price could have dropped with mass production and if the hardware could have beat the Sony Playstation to market. The CD32 aimed for a lower console price with just 2D. Actually, the 2D Sega Genesis was still selling well when the Saturn came out and Sega ran out of Genesis units. Most Sega Genesis games look better than those early pseudo 3D games too.

Hammer Quote:

For gaming PC, Pentium CPU family was important to counter 3DO, Sega Saturn, and Sony PS1 game console generation.


True although GPU performance was becoming as important or more important than CPU performance.

Hammer Quote:

The Apollo team did a good job with "What IF" i.e. the missing 68K cloner like AMD for X86.


The Apollo core performance in FPGA is disappointing and I don't like the ISA changes. SAGA has been very successful in FPGA and not just as a "what if" but a "what is".

 Status: Offline
Profile     Report this post  
BigD 
Re: 3.3
Posted on 12-Jun-2021 0:55:39
#33 ]
Elite Member
Joined: 11-Aug-2005
Posts: 5910
From: UK

Started new thread instead

Last edited by BigD on 12-Jun-2021 at 12:56 AM.

_________________
"Art challenges technology. Technology inspires the art."
John Lasseter, Co-Founder of Pixar Animation Studios

 Status: Offline
Profile     Report this post  
pixie 
Re: 3.3
Posted on 12-Jun-2021 13:14:31
#34 ]
Elite Member
Joined: 10-Mar-2003
Posts: 2675
From: Figueira da Foz - Portugal

@NutsAboutAmiga

Hyperion is in itself in the hole they managed to dig, why people still praise them is beyond belief

_________________
Indigo 3D Lounge, my second home.
The Illusion of Choice | Am*ga

 Status: Offline
Profile     Report this post  
NutsAboutAmiga 
Re: 3.3
Posted on 12-Jun-2021 20:36:20
#35 ]
Elite Member
Joined: 9-Jun-2004
Posts: 11881
From: Norway

@pixie

As I see it Amiga Inc, KMOS, ITEC, Hyperion, Cloanto are all invested making that hole bigger, every day they spend in the legal system, is waste of energy, time and money. There are also personal tragedies, Ben lost his marriage, Bill lost someone close to him, developers where not paid, people and developers lost interest. I see Cloanto as some who like to prolong the suffering, Hyperion had a settlement with Amiga Inc, it looked like the end of the tunnel, but no then came Cloanto.

Last edited by NutsAboutAmiga on 12-Jun-2021 at 10:03 PM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

 Status: Offline
Profile     Report this post  
matthey 
Re: 3.3
Posted on 12-Jun-2021 21:44:18
#36 ]
Super Member
Joined: 14-Mar-2007
Posts: 1151
From: Kansas

NutsAboutAmiga Quote:

As I see it Amiga Inc, KMOS, ITEC, Hyperion, Cloanto are all invested making that hole bigger, every day they spend in the legal system, is waste of energy, time and money. There are also personal tragedies, Ben lost his marriage, Bill lost someone close to him, developers where not paid, people and developers lost interest. I see Cloanto as some who like to prolong the suffering, Hyperion had a settlement with Amiga Inc, it looked like the end of tunnel, but no then came Cloanto.


Michele Battilana looks to me like a man on a mission to consolidate the Amiga IP and restore the Amiga brand. This requires being persistent and firm with Amiga IP squatters. The 2009 settlement agreement between Hyperion and Amiga Inc. does not look like it was made in good faith but rather under financial duress and likely with coercion. The agreement was heavily biased toward Hyperion with very few requirements for Hyperion yet Hyperion appears to have violated the contract by challenging ownership of Amiga IP. This is not an isolated incident of questionable ethical behavior for Hyperion but rather appears to be their modus operandi. There is little hope for Hyperion to do anything with the Amiga anyway as selling an old niche market OS is barely profitable and their checkered past limits other Amiga businesses working with and investing with them.

 Status: Offline
Profile     Report this post  
kolla 
Re: 3.3
Posted on 13-Jun-2021 9:05:52
#37 ]
Super Member
Joined: 21-Aug-2003
Posts: 1813
From: Trondheim, Norway

@matthey

SAGA successfull? Using what metrics?

_________________
B5D6A1D019D5D45BCC56F4782AC220D8B3E2A6CC

 Status: Offline
Profile     Report this post  
NutsAboutAmiga 
Re: 3.3
Posted on 13-Jun-2021 11:23:45
#38 ]
Elite Member
Joined: 9-Jun-2004
Posts: 11881
From: Norway

@matthey

Don’t think Michele Battilana should be the person talk about ethics, he built is business around cracked and abandoned software.

“Hyperion appears to have violated the contract by challenging ownership of Amiga IP”

By makeing updates to a old OS, so users can buy new versions of it? or are talking about being pressured into trying to protect their business? Because Michele Battilana was grabbing brand names left and right.

Sorry I don’t see Michele Battilana interference as something good, not seeing any plan forward, or talks about how to finance innovation forward. I don’t think Michele Battilana cares about software development or innovation, I believe he is only interested in protecting his emulation business.

"but rather under financial duress"

Amiga Inc put them self-there by not selling product people wanted to buy AA/AmigaDE was failure, there mobile venture was a mistake. They scammed people with coupons and t-shirts. Amiga Inc invested in BlackBerry I guess something based on UAE, of because it was old Amiga games, sadly for Amiga Inc, blackberry was not a hugely popular phone. They were trying to capitalize on other people’s work, or out sourcing, or trying to flip something like Intent/AmigaDE, they went for quick bucks, of course you can get UAE for android, and that product is free, the silly thing here is original Amiga games where better than Amiga DE games. They failed to create value, and unique experience with their new platform, and they were quickly outcompeted by google. Another mistake was development PC for Amiga DE, hardware dongle, instead of just selling the Amiga DE as development platform, they trying to sell old pc running at high price to the developers, to get develop write code for a unknown platform, It was naive. They were also known for snow man maker, guess what it was huge failure.

Last edited by NutsAboutAmiga on 13-Jun-2021 at 04:36 PM.
Last edited by NutsAboutAmiga on 13-Jun-2021 at 02:53 PM.
Last edited by NutsAboutAmiga on 13-Jun-2021 at 02:23 PM.
Last edited by NutsAboutAmiga on 13-Jun-2021 at 02:20 PM.
Last edited by NutsAboutAmiga on 13-Jun-2021 at 02:16 PM.
Last edited by NutsAboutAmiga on 13-Jun-2021 at 02:11 PM.
Last edited by NutsAboutAmiga on 13-Jun-2021 at 11:32 AM.
Last edited by NutsAboutAmiga on 13-Jun-2021 at 11:25 AM.
Last edited by NutsAboutAmiga on 13-Jun-2021 at 11:24 AM.

_________________
http://lifeofliveforit.blogspot.no/
Facebook::LiveForIt Software for AmigaOS

 Status: Offline
Profile     Report this post  
matthey 
Re: 3.3
Posted on 13-Jun-2021 19:17:31
#39 ]
Super Member
Joined: 14-Mar-2007
Posts: 1151
From: Kansas

Quote:

SAGA successfull? Using what metrics?


While maintaining Amiga custom chip compatibility, it was possible to upgrade SAGA (and the AmigaOS) to support modern hardware and standards. SAGA supports digital output standards, high resolutions practically limited by memory bandwidth, chunky RTG which can be displayed with and layered under existing ECS/AGA gfx, much improved sprites and much improved 16 bit sound. The logic should fit in a $10 FPGA or cost less than $1 in an ASIC. It is likely that 3D can be added to the chunky gfx layer. Where do you think SAGA could do better?

While SAGA is adequate for modern use, the Apollo core in a several times more expensive FPGA has barely enough performance for a modern browser. The Apollo core has good performance for a FPGA core but can't compete with most inexpensive 500MHz cores in ASICs. That reality is disappointing.

 Status: Offline
Profile     Report this post  
matthey 
Re: 3.3
Posted on 13-Jun-2021 20:32:31
#40 ]
Super Member
Joined: 14-Mar-2007
Posts: 1151
From: Kansas

NutsAboutAmiga Quote:

Don’t think Michele Battilana should be the person talk about ethics, he built is business around cracked and abandoned software.


I have no ethical problem with old cracked and abandoned software. Nobody is losing money when nobody is trying to make money from this software. Give me a break.

NutsAboutAmiga Quote:

By makeing updates to a old OS, so users can buy new versions of it? or are talking about being pressured into trying to protect their business? Because Michele Battilana was grabbing brand names left and right.


It looks like Hyperion tried to trademark Amiga related marks which they were forbidden from doing in the 2009 Settlement Agreement with Amiga Inc. Michele Battilana may have had restrictions on what he could trademark due to Amiga licenses before buying out Amiga Inc. but there are no such restrictions after the purchase and he should own all "Amiga" derived and related trademarks whether registered or not.

NutsAboutAmiga Quote:

Sorry I don’t see Michele Battilana interference as something good, not seeing any plan forward, or talks about how to finance innovation forward. I don’t think Michele Battilana cares about software development or innovation, I believe he is only interested in protecting his emulation business.


We haven't even heard Michele's plan. What is Hyperion's plan? Suck the last remaining dollar out of a dying niche market selling an old 68k AmigaOS? How do you think Hyperion plans to make AmigaOS successful when they have practically stopped development of their flagship AmigaOS 4 and do not have the resources to port it to another architecture?

NutsAboutAmiga Quote:

Amiga Inc put them self-there by not selling product people wanted to buy AA/AmigaDE was failure, there mobile venture was a mistake. They scammed people with coupons and t-shirts. Amiga Inc invested in BlackBerry I guess something based on UAE, of because it was old Amiga games, sadly for Amiga Inc, blackberry was not a hugely popular phone. They were trying to capitalize on other people’s work, or out sourcing, or trying to flip something like Intent/AmigaDE, they went for quick bucks, of course you can get UAE for android, and that product is free, the silly thing here is original Amiga games where better than Amiga DE games. They failed to create value, and unique experience with their new platform, and they were quickly outcompeted by google. Another mistake was development PC for Amiga DE, hardware dongle, instead of just selling the Amiga DE as development platform, they trying to sell old pc running at high price to the developers, to get develop write code for a unknown platform, It was naive. They were also known for snow man maker, guess what it was huge failure.


Amiga Inc. had a bad game plan and questionable ethics too. Hyperion stealing AmigaOS 4 from them was generally supported by the Amiga community as a sort of Robin Hood syndrome. Hyperion's game plan and business practices weren't much if any better though. They have burnt so many bridges with business partners and developers that they may not have many options left. It is time to sell out of the AmigaOS business and see what someone else can do with it.

 Status: Offline
Profile     Report this post  
Goto page ( Previous Page 1 | 2 | 3 Next Page )

[ 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