Poster | Thread |
cdimauro
|  |
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon Posted on 1-Nov-2015 9:25:05
| | [ #821 ] |
|
|
 |
Elite Member  |
Joined: 29-Oct-2012 Posts: 3097
From: Germany | | |
|
| @pavlor
Quote:
pavlor wrote: @cdimauro
Quote:
Is this supported (or will be in "two more weeks")by the o.s.? |
Do you even know topic you are talking about? |
Do you even know the meaning of a question mark in a sentence? Quote:
Two PCIe slots were blessing for further developement of RadeonHD driver. So yes, you can use two PCIe GFX cards in OS4. |
How they are supported? Are you able to use both cards at the same time?
Like playing two different 3D games at the same time, for example. Or one 3D game which splits the display processing to the two cards, albeit using only one to display the final framebuffer. Quote:
Quote:
And an accelerator unit for the FFT, but you need proper code to use it. |
From benchmarks we saw, no specific PA6T compiled binaries were needed. |
Yes, nevermind. I saw that the offload unit isn't for FFT acceleration, but only for encryption. Quote:
Quote:
there were already G4s (even dual core) available at the time, that were less expensive. |
PA6T has much better expansion capability than MPC8640D. |
Like what? MPC8640D supports DDR2, PCI-express, and Gigabit Ethernet too. Is there something missing? Quote:
Quote:
If you don't plan to use floating point-intensive applications, then YES. Otherwise it's a NO. |
As someone other in this thread wrote: Let's wait for real-world benchmarks.  |
Someone already posted them, and I re-quote "someone else" comment about it:
FPU in e500v2 is weak even without discussed compatibility issue (benchmarks show e500 1.2 GHz is comparable to G3 600 MHz in FPU depending tasks - that is with native SPE compiled code). We don´t know performance penalty for emulated FPU instructions, but don´t expect miracles.
It seems that historians have an unbelievable tendency to forget...  |
|
Status: Offline |
|
|
Yasu
|  |
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon Posted on 1-Nov-2015 9:36:39
| | [ #822 ] |
|
|
 |
Regular Member  |
Joined: 13-Oct-2015 Posts: 224
From: Stockholm, Sweden | | |
|
| I also have a hard time seeing how the X1000 "killed of the Amiga". It's not like the Amiga market would be in a better state now without it. It's simply not possible to make a PPC motherboard for a, in the eyes of the general public, competative price.
Both A-Cube and A-Eon is trying to do the best of a very bad situation. Without their hardware, I think AmigaOS 4 would be in a much worse state now. Unless Hyperion had done the same thing as MorphOS Team and started to port to PPC Macs, but that's purely speculative. _________________ Amiga Forum - Sweden's best Amiga Magazine
My MorphOS Blog
"Free speech includes other peoples right to offend you." |
|
Status: Offline |
|
|
pavlor
|  |
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon Posted on 1-Nov-2015 9:53:25
| | [ #823 ] |
|
|
 |
Elite Member  |
Joined: 10-Jul-2005 Posts: 9441
From: Unknown | | |
|
| @cdimauro
Quote:
How they are supported? Are you able to use both cards at the same time? |
I recommend reading RadeonHD developement blog for better info...
Quote:
Like what? MPC8640D supports DDR2, PCI-express, and Gigabit Ethernet too. Is there something missing? |
(Much) less SerDes lanes: only up to 1x PCIe x8.
Quote:
FPU in e500v2 is weak even without discussed compatibility issue |
Well, your point is it will be unusably slow, my point is better to wait for actual benchmarks. I must base my statements on evidence, not (un)wishful thinking. I´m historian, not oracle.  |
|
Status: Offline |
|
|
OlafS25
|  |
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon Posted on 1-Nov-2015 10:23:10
| | [ #824 ] |
|
|
 |
Elite Member  |
Joined: 12-May-2010 Posts: 6213
From: Unknown | | |
|
| @pavlor
the developer on the german site admitted that FPU emulation will be slow and first boards will sold as "beta" or "dev" boards or similar (if I remember right). They want to improve speed later. It is principle of hoping. The rest is speculation. |
|
Status: Offline |
|
|
cdimauro
|  |
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon Posted on 1-Nov-2015 10:25:09
| | [ #825 ] |
|
|
 |
Elite Member  |
Joined: 29-Oct-2012 Posts: 3097
From: Germany | | |
|
| @pavlor
Quote:
pavlor wrote: @cdimauro
Quote:
How they are supported? Are you able to use both cards at the same time? |
I recommend reading RadeonHD developement blog for better info... |
I did it, up to the first 2012 post, and I haven't found any information about multiple graphic cards support. Can you point me to something? Thanks. Quote:
Quote:
Like what? MPC8640D supports DDR2, PCI-express, and Gigabit Ethernet too. Is there something missing? |
(Much) less SerDes lanes: only up to 1x PCIe x8. |
Directly from MPC8640D facts sheet, you have 8+8 SerDer lines, which can be used as 2x8 PCIe, or (much more useful) 1x8 PCIe + 1x4 PCie + 1x4 Serial RapidIO. Quote:
Quote:
FPU in e500v2 is weak even without discussed compatibility issue |
Well, your point is it will be unusably slow, my point is better to wait for actual benchmarks. I must base my statements on evidence, not (un)wishful thinking. I´m historian, not oracle.  |
Then you should pay attention to the fact that what I've written is related to the regular FPU emulation (so, using the trap mechanism). And I've reported some data about it, directly from the e500v2 reference manual (which isn't published by me, but by Freescale, and they should know what they are talking about, right? ). But, again, you're free to confute them, eh! 
Last but not least, I've already stated that using binaries compiled for native SPE unit utilization will provide performances comparable to the existing low-end end machines.  |
|
Status: Offline |
|
|
OlafS25
|  |
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon Posted on 1-Nov-2015 10:25:25
| | [ #826 ] |
|
|
 |
Elite Member  |
Joined: 12-May-2010 Posts: 6213
From: Unknown | | |
|
| @pavlor
it will be slow when needing FPU. This will affect applications and games to a different degree. How much is speculation right now. |
|
Status: Offline |
|
|
Overflow
 |  |
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon Posted on 1-Nov-2015 10:26:35
| | [ #827 ] |
|
|
 |
Super Member  |
Joined: 12-Jun-2012 Posts: 1628
From: Norway | | |
|
| @OlafS25
It partly depends on pricelevel at this point.
If the price is "entrylevel", then even issues with FPU can encourage more people to join the AOS4 club. At this point people are aware of the possible FPU performance issue, so benchmarks should be available for you to evaluate before you purchase. |
|
Status: Offline |
|
|
cdimauro
|  |
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon Posted on 1-Nov-2015 10:27:53
| | [ #828 ] |
|
|
 |
Elite Member  |
Joined: 29-Oct-2012 Posts: 3097
From: Germany | | |
|
| @OlafS25
Quote:
OlafS25 wrote: @pavlor
the developer on the german site admitted that FPU emulation will be slow and first boards will sold as "beta" or "dev" boards or similar (if I remember right). They want to improve speed later. It is principle of hoping. The rest is speculation. |
FPU emulation speed can be improved (compared to the Linux's FPU trap handler, which is written in C), but the limit itself is represented by the trap mechanism, which is slow by itself.
What you can expect for the Tabor board is that there'll be released ad-hoc binaries, in order to directly use its SPE unit, and get acceptable speeds. |
|
Status: Offline |
|
|
OlafS25
|  |
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon Posted on 1-Nov-2015 10:33:28
| | [ #829 ] |
|
|
 |
Elite Member  |
Joined: 12-May-2010 Posts: 6213
From: Unknown | | |
|
| @Overflow
if the price is very low then people might oversee that the hardware is slow and even slower when needing FPU (I compare it to common hardware standard of 2015/2016 and not 2005). As you wrote price will decide how many people might be interested in it. |
|
Status: Offline |
|
|
pavlor
|  |
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon Posted on 1-Nov-2015 10:51:21
| | [ #830 ] |
|
|
 |
Elite Member  |
Joined: 10-Jul-2005 Posts: 9441
From: Unknown | | |
|
| @cdimauro
Quote:
Can you point me to something? Thanks. |
Not exactly developers log, but these user reports may give you some clue: http://www.amiga.org/forums/showthread.php?t=69522 http://amigax1000.blogspot.cz/2014/12/looking-back-at-2014-on-amiga.html
Quote:
Directly from MPC8640D facts sheet, you have 8+8 SerDer lines, which can be used as 2x8 PCIe, or (much more useful) 1x8 PCIe + 1x4 PCie + 1x4 Serial RapidIO. |
Looking at X1000, you have: 1x PCIe x8 (or x16 when not using the other x16 slot) 1x PCIe x8 2x PCIe x1
Looks much better. 
Quote:
Then you should pay attention to the fact that what I've written is related to the regular FPU emulation |
We don´t know what method will be used by Hyperion... |
|
Status: Offline |
|
|
cdimauro
|  |
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon Posted on 1-Nov-2015 11:15:38
| | [ #831 ] |
|
|
 |
Elite Member  |
Joined: 29-Oct-2012 Posts: 3097
From: Germany | | |
|
| @pavlor
Quote:
In the first link the only thing which can be seen is that the guy is using two monitors, but they can be driven by a single card.
The second link reports only that the guy is selling two cards, but there's no mention about using both them.
So, there's absolutely nothing which can help answer my previous two questions about multiple cards (not monitors) support on OS4. Quote:
Quote:
Directly from MPC8640D facts sheet, you have 8+8 SerDer lines, which can be used as 2x8 PCIe, or (much more useful) 1x8 PCIe + 1x4 PCie + 1x4 Serial RapidIO. |
Looking at X1000, you have: 1x PCIe x8 (or x16 when not using the other x16 slot) 1x PCIe x8 2x PCIe x1
Looks much better.  |
I don't think that only 2 SerDer lines can be classified as "much better".
The MPC8640D allows you to use 2x8 PCIe, or a better 1x8 PCIe + 1x4 PCIe + 2x1 PCIe + 2 Serial RapidIO configuration, which are similar to what the X1000 offers, with the second config which is more useful for a user, since it doesn't penalize much the graphic performance. Quote:
Quote:
Then you should pay attention to the fact that what I've written is related to the regular FPU emulation |
We don´t know what method will be used by Hyperion... |
For regular FPU emulation there's nothing other than the already discussed trap & emulate mechanism. NOTHING. And it's very slow BY ITSELF. This is set on the stone and you cannot change it, even invoking spirits with some Voodoo ritual.
pavlor, you haven't read the e500v2 manual, and you have no low-level coding experience (I don't know if you have coding experience in general, to be honest), and even an OS4 coder said that the FPU emulation is slow. So, why do you continue to insist on it? |
|
Status: Offline |
|
|
OlafS25
|  |
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon Posted on 1-Nov-2015 11:33:16
| | [ #832 ] |
|
|
 |
Elite Member  |
Joined: 12-May-2010 Posts: 6213
From: Unknown | | |
|
| |
Status: Offline |
|
|
pavlor
|  |
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon Posted on 1-Nov-2015 11:47:58
| | [ #833 ] |
|
|
 |
Elite Member  |
Joined: 10-Jul-2005 Posts: 9441
From: Unknown | | |
|
| @OlafS25
That says nothing about actual performance of such solution. Exactly as I wrote... |
|
Status: Offline |
|
|
pavlor
|  |
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon Posted on 1-Nov-2015 11:54:50
| | [ #834 ] |
|
|
 |
Elite Member  |
Joined: 10-Jul-2005 Posts: 9441
From: Unknown | | |
|
| @cdimauro
Quote:
but they can be driven by a single card. |
It is the opposite: OS4 driver supports more RadeonHD card, but not more displays from the same card...
Quote:
with the second config which is more useful for a user, since it doesn't penalize much the graphic performance. |
You need some space also for bridge chip...
Quote:
This is set on the stone and you cannot change it, even invoking spirits with some Voodoo ritual. |
As I wrote, I will wait for real implementation.
Quote:
and even an OS4 coder said that the FPU emulation is slow. |
Remember my proposed bet? I wrote 1/10 of native performance would be success. 1/10 - that is "slow", but still much better than your expectations (whatever these are). |
|
Status: Offline |
|
|
cdimauro
|  |
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon Posted on 1-Nov-2015 12:12:03
| | [ #835 ] |
|
|
 |
Elite Member  |
Joined: 29-Oct-2012 Posts: 3097
From: Germany | | |
|
| @pavlor
Quote:
pavlor wrote: @cdimauro
Quote:
but they can be driven by a single card. |
It is the opposite: OS4 driver supports more RadeonHD card, but not more displays from the same card... |
OK. But that doesn't answer my previous questions about cards utilization by the o.s.. Quote:
Quote:
with the second config which is more useful for a user, since it doesn't penalize much the graphic performance. |
You need some space also for bridge chip... |
For what purpose? Quote:
Quote:
This is set on the stone and you cannot change it, even invoking spirits with some Voodoo ritual. |
As I wrote, I will wait for real implementation. |
Sure, but it'll be slow. Quote:
Quote:
and even an OS4 coder said that the FPU emulation is slow. |
Remember my proposed bet? I wrote 1/10 of native performance would be success. 1/10 - that is "slow", but still much better than your expectations (whatever these are). |
As I stated previously, it strictly depends on the mixture of FPU and regular ("integer") instructions that are used by an application. Usually you don't execute ONLY FPU instructions, but you need to execute some "integer ones" also.
But if you have an FPU intensive code (I mean: the FPU instructions are much more than the integer ones), then your 1/10 bet is impossible to be satisfied. As I already reported, to just execute the FIRST instruction of the trap handler, you need 10 cycles (in the BEST case: see the e500v2 manual), and in such amount of time the SPE unit can execute up to 20 instructions, giving a much worse ratio of 1/20.
And I repeat again: that's just for the first instruction of the handler, which is NOT the normal case (because you have to execute more instructions to complete the emulation of the "offending" FPU instruction).
That are technical facts, pavlor: no bets neither guesses. |
|
Status: Offline |
|
|
fjudde
|  |
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon Posted on 1-Nov-2015 12:26:47
| | [ #836 ] |
|
|
 |
Regular Member  |
Joined: 12-Mar-2003 Posts: 146
From: Stockholm/Sweden | | |
|
| @cdimauro
Quote:
In the first link the only thing which can be seen is that the guy is using two monitors, but they can be driven by a single card. |
I found this in the first link: Quote:
For sale is a pristine condition AmigaOne X1000 Computer. Mine is presently running dual displays so has two video cards inside. Really cool when you connect to large displays to it. Includes: ... Original Radeon HD 4650 Video Card Radeon R9 270 Video Card ... |
cdimauro also wrote: Quote:
The second link reports only that the guy is selling two cards, but there's no mention about using both them. |
And I found this in the second link:
Quote:
My AmigaOne X1000 with Ducky Zero custom Amiga keyboard and two Radeon HD cards running AmigaOS4.1.6 and DOpus 5.9 |
cdimauro wrote: Quote:
So, there's absolutely nothing which can help answer my previous two questions about multiple cards (not monitors) support on OS4. |
You lost me there, or have I totally misunderstood your discussion with pavlor??????
By the way, I think you also crisscrossed the links._________________ "Great spirits have always encountered violent opposition from mediocre minds" - Albert Einstein |
|
Status: Offline |
|
|
pavlor
|  |
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon Posted on 1-Nov-2015 12:33:05
| | [ #837 ] |
|
|
 |
Elite Member  |
Joined: 10-Jul-2005 Posts: 9441
From: Unknown | | |
|
| @cdimauro
Quote:
USB, SATA?  |
|
Status: Offline |
|
|
cdimauro
|  |
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon Posted on 1-Nov-2015 12:51:52
| | [ #838 ] |
|
|
 |
Elite Member  |
Joined: 29-Oct-2012 Posts: 3097
From: Germany | | |
|
| @fjudde: I've exchanged the links (first took a look at the second, and then the first), but doesn't change the picture, because I haven't found the information which I was looking for.
I quote myself:
"Like playing two different 3D games at the same time, for example. Or one 3D game which splits the display processing to the two cards, albeit using only one to display the final framebuffer."
@pavlor: for USB and SATA you can use the two spare lines for the Serial RapidIO in the second configuration. But, in general, you can use 8 SerDers for the graphic card (PCIe), and the other 8 for some 1x PCIe slots and the rest for USBs and SATAs. |
|
Status: Offline |
|
|
pavlor
|  |
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon Posted on 1-Nov-2015 12:59:29
| | [ #839 ] |
|
|
 |
Elite Member  |
Joined: 10-Jul-2005 Posts: 9441
From: Unknown | | |
|
| @cdimauro
Quote:
"Like playing two different 3D games at the same time, for example. Or one 3D game which splits the display processing to the two cards, albeit using only one to display the final framebuffer." |
This is certainly not possible.  |
|
Status: Offline |
|
|
OlafS25
|  |
Re: New hardware: PPC-Motherboard A1222 "Tabor" by Acube/A-eon Posted on 1-Nov-2015 13:12:32
| | [ #840 ] |
|
|
 |
Elite Member  |
Joined: 12-May-2010 Posts: 6213
From: Unknown | | |
|
| @pavlor
I think we could stop here. But propably it will continue another 10 pages 
We know that FPU emulation is slow and that will certainly affect software. We do not know how much because AmigaOS does not yet run so no real-world tests possible. We do not know price and availability either. X5000 has priority and it is scheduled this year so development on AmigaOS for Tabor will not really start before next year (my personal guess).
If the hardware is very very cheap some people might be tempted to buy it despite shortcomings. But no price or date yet. But of course we can continue for the next 10 months propably  |
|
Status: Offline |
|
|