Poster | Thread |
wawa
| |
Re: Oh, boyz, it's seems that the new baby can beat them all Posted on 13-Jun-2011 17:50:22
| | [ #61 ] |
|
|
|
Elite Member |
Joined: 21-Jan-2008 Posts: 6259
From: Unknown | | |
|
| @pavlor
Quote:
68060 is not much faster per clock than original Pentium (I even assume 68060 is slower). Pentium 90 MHz scores 2.88 SpecInt95 (single core integer benchmark), 603e 300 MHz 7.5 SpecInt95 and 604e 233 MHz 9 SpecInt95.
|
if that is true then 060 would be quite close to compete clock by clock with 603/604. is the superscalarity taken itno account in this benchmark? |
|
Status: Offline |
|
|
Samurai_Crow
| |
Re: Oh, boyz, it's seems that the new baby can beat them all Posted on 13-Jun-2011 17:52:51
| | [ #62 ] |
|
|
|
Elite Member |
Joined: 18-Jan-2003 Posts: 2320
From: Minnesota, USA | | |
|
| @NutsAboutAmiga
One common misconception about piracy is that pirates never pay for stuff. Usually if they find something they like, they'll go to some extra effort to buy a legit copy of what they've been running bootleg copies of if for no other reason than to get a printed manual or tech support or the other intangibles of pride in ownership.
Say all you want about pirates killing the market. From what I've seen, it's the publishing companies driving people away with draconian DRM schemes and into the hands of shareware, freeware, and open-source developers.
It's anti-piracy tirades and DRM schemes of Microsoft and other companies as well that keep me coming back to Linux and AROS. It's the generous freebies that keep me coming back. Sure that can be taken to wild extremes, but I think piracy is a tempest in a teapot. |
|
Status: Offline |
|
|
wawa
| |
Re: Oh, boyz, it's seems that the new baby can beat them all Posted on 13-Jun-2011 17:55:46
| | [ #63 ] |
|
|
|
Elite Member |
Joined: 21-Jan-2008 Posts: 6259
From: Unknown | | |
|
| @NutsAboutAmiga Quote:
Having AmigaOS4 for 68k makes senses, because it gives 68k programers the possibility to use modern API's like AmigaInput and Reaction, closing the gab between the old and the new.
|
i dont know what amigainput is but what essentials does os4 reaction offer beyond the version available on 3.9? i have seen that thesource of os4 netsurf differs quite from what would be needed on 3.x, i dont exactly know though what functionality is gained by that. personally i would rather prefer to have updated mui/zune as there is much more valuable software using it. |
|
Status: Offline |
|
|
NutsAboutAmiga
| |
Re: Oh, boyz, it's seems that the new baby can beat them all Posted on 13-Jun-2011 18:11:01
| | [ #64 ] |
|
|
|
Elite Member |
Joined: 9-Jun-2004 Posts: 12796
From: Norway | | |
|
| @wawa
AmigaOS4 applications often use Reaction (because we don't see a future in MUI if its not updated for AmigaOS only MorphOS)
Yes we are getting some fixes for MUI in update 3, and yes MUI is faster, but Reaction is coming along great.
Rigo has done lots of stuff on Reaction, and lots more is planned.
_________________ http://lifeofliveforit.blogspot.no/ Facebook::LiveForIt Software for AmigaOS |
|
Status: Offline |
|
|
NutsAboutAmiga
| |
Re: Oh, boyz, it's seems that the new baby can beat them all Posted on 13-Jun-2011 18:16:52
| | [ #65 ] |
|
|
|
Elite Member |
Joined: 9-Jun-2004 Posts: 12796
From: Norway | | |
|
| |
Status: Offline |
|
|
AmigaBlitter
| |
Re: Oh, boyz, it's seems that the new baby can beat them all Posted on 13-Jun-2011 18:29:23
| | [ #66 ] |
|
|
|
Elite Member |
Joined: 26-Sep-2005 Posts: 3512
From: Unknown | | |
|
| @Samurai_Crow
Quote:
In the meantime both the MiniMig and NatAmi teams are working to improve Classic compatibility. It hasn't escaped me that ACube produces MiniMigs as well as SAM4x0 series computers. Perhaps we can work something out. |
Not to mention that Acube may decide a program upgrade for their Lattice XP2 FPGA. 40k fpga (already wired to memory) is a quite bigger FPGA. In that case the 460 PPC is already on board, the memory and all other devices are on board too.
Last edited by AmigaBlitter on 14-Jun-2011 at 12:58 PM.
_________________ retired |
|
Status: Offline |
|
|
wawa
| |
Re: Oh, boyz, it's seems that the new baby can beat them all Posted on 13-Jun-2011 19:06:36
| | [ #67 ] |
|
|
|
Elite Member |
Joined: 21-Jan-2008 Posts: 6259
From: Unknown | | |
|
| @NutsAboutAmiga
no offence, but i couldnt think of a reaction application i would like to have backported to 3.x these days. except maybe for chris netsurf gui, but in context of alternatives (mos owb) i would hestiate even about that.
instead reaction or mui, zune was proposed as common standard for all amiga systems since it is open source. as reaction and mui are proprietary we will never reach an agreement for common denominator based on them. besides reaction is too limited and the progress appears too slow. alas one could currently claim that about zune too. |
|
Status: Offline |
|
|
wawa
| |
Re: Oh, boyz, it's seems that the new baby can beat them all Posted on 13-Jun-2011 19:09:18
| | [ #68 ] |
|
|
|
Elite Member |
Joined: 21-Jan-2008 Posts: 6259
From: Unknown | | |
|
| @NutsAboutAmiga
so amigainput is some sort of replacement for lowlevel.library? seriously ive never heard about it, except it is open source it doesnt pay to establish this as standard, i think. |
|
Status: Offline |
|
|
pavlor
| |
Re: Oh, boyz, it's seems that the new baby can beat them all Posted on 13-Jun-2011 20:13:35
| | [ #69 ] |
|
|
|
Elite Member |
Joined: 10-Jul-2005 Posts: 9578
From: Unknown | | |
|
| @wawa
Quote:
if that is true then 060 would be quite close to compete clock by clock with 603/604. |
With 603 yes, with 604e no.
Quote:
is the superscalarity taken itno account in this benchmark? |
I assume yes - SPEC CPU is de facto standard among workstation benchmarks. |
|
Status: Offline |
|
|
wawa
| |
Re: Oh, boyz, it's seems that the new baby can beat them all Posted on 13-Jun-2011 22:32:35
| | [ #70 ] |
|
|
|
Elite Member |
Joined: 21-Jan-2008 Posts: 6259
From: Unknown | | |
|
| @pavlor
i recall 604 being regarded one of the best ppc designs clock by clock. |
|
Status: Offline |
|
|
Mechanic
| |
Re: Oh, boyz, it's seems that the new baby can beat them all Posted on 14-Jun-2011 1:05:58
| | [ #71 ] |
|
|
|
Elite Member |
Joined: 27-Jul-2003 Posts: 2007
From: Unknown | | |
|
| @thread
Although I'm a OS4 fan I really would like to see NatAmi take it's own direction.
It's been said here that including a PPC in the FPGA my be possible, and that's cool, but perhaps it might be worthwhile to investigate adding other processor type functions in the FPGA that are more closely related to the other hardware on the board instead of to an API or software. If people go out and buy a OS then they would certainly want it supported in the future, which may put a limit on some nifty hardware possibilities. I know, too soon to think about such things.
Brain went to sleep awhile ago, time for body to follow.
|
|
Status: Offline |
|
|
NutsAboutAmiga
| |
Re: Oh, boyz, it's seems that the new baby can beat them all Posted on 14-Jun-2011 12:07:31
| | [ #72 ] |
|
|
|
Elite Member |
Joined: 9-Jun-2004 Posts: 12796
From: Norway | | |
|
| |
Status: Offline |
|
|
Mechanic
| |
Re: Oh, boyz, it's seems that the new baby can beat them all Posted on 14-Jun-2011 12:57:12
| | [ #73 ] |
|
|
|
Elite Member |
Joined: 27-Jul-2003 Posts: 2007
From: Unknown | | |
|
| |
Status: Offline |
|
|
minator
| |
Re: Oh, boyz, it's seems that the new baby can beat them all Posted on 14-Jun-2011 23:33:44
| | [ #74 ] |
|
|
|
Cult Member |
Joined: 23-Mar-2004 Posts: 989
From: Cambridge | | |
|
| @AmigaBlitter
Quote:
at least in certain test and conditions. Both classic and some new generation hardware |
There's lies, damned lies, and benchmarks.
He's comparing standard memcopy routines (which are probably not that great) against a benchmark he's written himself.
When we see 3rd parties running tests properly (i.e. the same code on all machines) I expect to see rather different results.
It's an interesting project and claims like this just make the entire thing look dubious and dodgy, which is a pity because I get the impression it's not.
They really don't need to do this. Anyone with even the slightest technical inclination will be very put off by stunts like this.
_________________ Whyzzat? |
|
Status: Offline |
|
|
AmigaBlitter
| |
Re: Oh, boyz, it's seems that the new baby can beat them all Posted on 15-Jun-2011 8:02:15
| | [ #75 ] |
|
|
|
Elite Member |
Joined: 26-Sep-2005 Posts: 3512
From: Unknown | | |
|
| @minator
I don't think so. Specs are public, you can do the math by yourself. _________________ retired |
|
Status: Offline |
|
|
ErikBauer
| |
Re: Oh, boyz, it's seems that the new baby can beat them all Posted on 15-Jun-2011 12:32:46
| | [ #76 ] |
|
|
|
Super Member |
Joined: 25-Feb-2004 Posts: 1141
From: Italy | | |
|
| |
Status: Offline |
|
|
umisef
| |
Re: Oh, boyz, it's seems that the new baby can beat them all Posted on 15-Jun-2011 13:13:04
| | [ #77 ] |
|
|
|
Super Member |
Joined: 19-Jun-2005 Posts: 1714
From: Melbourne, Australia | | |
|
| @AmigaBlitter
Quote:
I don't think so. Specs are public, you can do the math by yourself.
|
Indeed. And doing so (as well as reading the related threads, and looking at the screenshots) reveals:
(a) These benchmarks were not accessing Natami's DDR2 main memory at all. They were accessing a smallish amount of fast SRAM which is located on the CPU card, and which even Gunnar tends to refer to as "L2 cache".
(b) The 68060 benchmarks were taken with the CPU cache in write-through mode, whereas the "comparison values" come from PPCs configured for write-back mode. Now, in actual use, write-back tends to give much better performance than write-through; However, it does mean that copying 10MB is actually reading 20MB and writing 10MB, unless one takes care to avoid the read-cacheline-one-write-miss-allocation[1] --- meaning that write-through is better for benchmarks involving large streaming writes.
(c) The combination of (a) and (b) is particularly deceptive; The 4MB asynchronous SRAM on the CPU card are nice and fast for lots of individual accesses, and don't have any concept of bursting. The DDR2 main memory, on the other hand, is very, very much optimized for burst accesses, to fill or write-back a cache line worth of data. In fact, the licensable DDR controllers for the FPGA used treat any write of something smaller than a whole burst by reading a whole burst into the local buffer, updating parts of that buffer, and writing it back --- which is positively awful for performance. Write-through on the 68060 means no write bursts at all...
So it will definitely be interesting to see how actual performance in real world tasks on a real-world setup turns out. And if we are real lucky, we might even get some performance info from actually running the actual 68050 on actual FPGA hardware, running actual real world tasks. Seeing how "500MHz 030" is pretty exactly what the 68060 card does right now, on real hardware.
[1]: Gunnar is definitely aware of this --- see his posts on the very subject.[2]
[2]: Interestingly enough, that thread also has Gunnar report 178MB/s from glibc's memcpy on the Efika (and 347MB/s with tuned code), yet when recently comparing the Natami to PPC, he "remembers" a value of 100MB/s |
|
Status: Offline |
|
|
Hammer
| |
Re: Oh, boyz, it's seems that the new baby can beat them all Posted on 15-Jun-2011 13:50:42
| | [ #78 ] |
|
|
|
Elite Member |
Joined: 9-Mar-2003 Posts: 5246
From: Australia | | |
|
| @Samurai_Crow
Quote:
It's not just OS 4's version of E-UAE, I'm bashing. I'm also bashing WinUAE and any other Amiga emulator out there. The problem with JIT-based solutions is that they spend a lot of time compiling old software into the new format. Add to that that the Copper-list functionality isn't yet mapped to the shader functionality of the underlying graphics chips and you have a problem. Finally, no version of UAE is multithreaded. |
As for multi-threading issue, my Intel Core i7-740 Mobile Quad Core overclocks to 2.93Ghz and Intel Core i7-2600 overclocks to 3.8Ghz during narrow thread processing.
AMD Bulldozer has 1Ghz overclock (turbo mode) i.e. can go as high as 4.2Ghz.
Modern X86 CPUs matches the AmigaOS/AROS's lack of support for SMP/SMT processing.
PS; DX9/DX10 has single CPU thread for GPU command generation, hence AMD/Intel CPU turbo overclocking features with narrow thread situations. DX11 overcomes this middleware issue.
Quote:
I've got an Intel Mac Mini running the version of E-UAE that already has a JIT (Intel). When I run my favorite game, Total Chaos AGA, it struggles to maintain chipset timing. The main flow of the program gets JITed but the chipset emulation strains the system immensely. If the chipset emulation could run in another thread or something, then at least the second core of my Core 2 Duo would get in on the action . |
Try WinUAE 2.3.2? Anyway, it's too bad the old desktop Intel Core 2 doesn't have turbo overclocking modes.
I tested WinUAE 2.3.2 on my ACER Iconia W500 tablet (AMD C-50 APU with 1.0Ghz Dual Core Bobcat and DDR3-1066 memory) with my ADF games and they played fine..
I don't recall having WinUAE/ADF games speed issues with my old Sony Vaio FW45 (Intel Core 2 Duo P8700 @ 2.53Ghz, dual DDR2-800 memory) laptop.
Last edited by Hammer on 15-Jun-2011 at 02:48 PM. Last edited by Hammer on 15-Jun-2011 at 02:42 PM. Last edited by Hammer on 15-Jun-2011 at 02:32 PM. Last edited by Hammer on 15-Jun-2011 at 02:29 PM. Last edited by Hammer on 15-Jun-2011 at 02:29 PM. Last edited by Hammer on 15-Jun-2011 at 02:26 PM.
_________________ Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68) Amiga 500 (Rev 6A, KS 3.2, PiStorm/RPi 3a/Emu68) |
|
Status: Offline |
|
|
Samurai_Crow
| |
Re: Oh, boyz, it's seems that the new baby can beat them all Posted on 15-Jun-2011 16:00:08
| | [ #79 ] |
|
|
|
Elite Member |
Joined: 18-Jan-2003 Posts: 2320
From: Minnesota, USA | | |
|
| @Hammer
Quote:
Hammer wrote: @Samurai_Crow Quote:
I've got an Intel Mac Mini running the version of E-UAE that already has a JIT (Intel). When I run my favorite game, Total Chaos AGA, it struggles to maintain chipset timing. The main flow of the program gets JITed but the chipset emulation strains the system immensely. If the chipset emulation could run in another thread or something, then at least the second core of my Core 2 Duo would get in on the action . |
Try WinUAE 2.3.2? Anyway, it's too bad the old desktop Intel Core 2 doesn't have turbo overclocking modes.
I tested WinUAE 2.3.2 on my ACER Iconia W500 tablet (AMD C-50 APU with 1.0Ghz Dual Core Bobcat and DDR3-1066 memory) with my ADF games and they played fine..
I don't recall having WinUAE/ADF games speed issues with my old Sony Vaio FW45 (Intel Core 2 Duo P8700 @ 2.53Ghz, dual DDR2-800 memory) laptop.
|
I don't have a Windows machine to try WinUAE on. And the fact that as a large program, Total Chaos AGA overflows the JIT cache making it run slowly on every version of UAE that I've tried. It's rare for an Amiga program to be developed to this point of memory consumption and still use memory efficiently as Total Chaos AGA does. It's reasons like this that Team Chaos Leader has joined the NatAmi team because emulators don't always cut the muster. Look for a Total Chaos version enhanced for the NatAmi's capabilities later on. |
|
Status: Offline |
|
|
ChrisH
| |
Re: Oh, boyz, it's seems that the new baby can beat them all Posted on 15-Jun-2011 19:18:19
| | [ #80 ] |
|
|
|
Elite Member |
Joined: 30-Jan-2005 Posts: 6679
From: Unknown | | |
|
| @umisef That's a great technical analysis/deconstruction .
@Samurai_Crow Quote:
One common misconception about piracy is that pirates never pay for stuff. |
Because it is TRUE in the majority of cases. Kids who grow up being able to easily download/stream stuff for free end-up believing that this is their right, and that they should not have to pay.
On the other hand, I agree that draconian DRM schemes cause more harm than good (when paying customers get a worse experience than pirates, that actually encourages piracy!). The police should go after the people making money from piracy, and governments should educate people, rather than criminalising children & families who can least afford to pay.Last edited by ChrisH on 15-Jun-2011 at 07:24 PM.
_________________ Author of the PortablE programming language. It is pitch black. You are likely to be eaten by a grue... |
|
Status: Offline |
|
|