12 December 2018, 09:53 | #1 |
Registered User
Join Date: May 2014
Location: Norway
Posts: 379
|
Real-world difference between bus speeds
Hi there,
I've been wondering what are the actual difference in performance between the different zorro buses (inkluding fast Zorro II, which is supposedly ZorroIIx2). I've never had a machine with Zorro III but I understand it's bandwith is about the same as VLB, while Zorro II's bandwith is a little less then ISA (which I find surprising). Does anyone know, or have measured, how the difference in bandwith effect the performance of cards (such as Cybervision 64/3D)? It might be difficult to create an "all-things-being-equal" situation when going from one bus to another as a Zorro II and Zorro III machine will be so different in so many ways, but lets say an A1200/60 with a Zorro II busboard, then in fast-zorro II mode, then an A4000/60 on the Zorro III bus. Does a Cybervision saturate the Zorro II bus? Would it saturate it in fast-mode as well? Is there really any point in buying an RTG-card for the zorro bus in a machine that could potentially work with a PCI card? That's just for graphics. I understand that few SCSI cards were made for Zorro III and that the PCI SCSI cards that work with mediator busboards are for some reason capped at about 4mb transfer-rates, but what about other expansion cards like sound/music, videoediting and ethernet? |
12 December 2018, 10:02 | #2 |
-
Join Date: Jul 2003
Location: Helsinki / Finland
Age: 43
Posts: 9,911
|
It is noticeable for example when swapping around between more large/colourful RTG screens than can fit into your video card's memory.
There is only one Zorro III network card, it doesn't give much improvement over Zorro III. The best speed improvements can be had with CPU card local expansions. Last edited by Jope; 12 December 2018 at 10:09. |
12 December 2018, 10:31 | #3 |
Registered User
Join Date: Feb 2008
Location: RNO
Posts: 1,007
|
Wasn't "real life" Zorro II throughput more like 3.5MB/s rather than 5.3MB/s said in that page? At least in practise chunky demos ran more smoothly over AGA than on CV64/3D in Zorro II on my 060 setup.
And the Zorro II bandwidth limitation really shows with games and demos that push raw chunky data on screen and don't use any 2D accelerated functions. For example, playing Napalm in 640x512 mode is quite impossible with Zorro II graphics card, but it flies without any problems over Zorro III or PCI. Zorro II is somehow acceptable when you use system friendly utilities, and definitely better than AGA for that use, but generally Zorro III or PCI are so much more enjoyable setups. Sound cards are fine in Zorro II and you get DSP options which is really nice compared to PCI ones. I guess network cards should be in theory too. Although I don't know if it was Zorro II or just old crappy network card designs, but for me it was like day and night when going from Hydra AmigaNet in Zorro II to Realtek cards in Mediator. Hydra was always a bit unreliable and max transfer speeds were 600kB/s in good weather. 10Mpbs PCI NIC gave steady 1MB/s, and 100Mbps card went over that (but was restricted by CPU then soon). So, in my experience PCI has been much better solution for networking. |
12 December 2018, 12:23 | #4 |
Registered User
Join Date: May 2014
Location: Norway
Posts: 379
|
That's interesting, so for Zorro II graphics are really just productivity-enhancers, and not much use for gaming? Do any of you have any experience with fast-zorroII? I interested to learn if it really results in twice the performance.
|
12 December 2018, 13:51 | #5 | |
Registered User
Join Date: Feb 2008
Location: RNO
Posts: 1,007
|
Quote:
I have ZorroIV board (with MediatorZIV), so I'd basically have those double speed Zorro II slots, but I only have the standard CV64/3D which can't take advantage of the better speed. But if it'd work at all how it was advertised, I guess 2x speed for Zorro II would make it noticeably better experience... because the raw speed is what it lacks and if it'd really be close to double speed, I guess I could be quite happy with it. CV64/3D in Zorro II was mostly "if this could be even sligthly faster it wouldn't annoy me that much" experience... it was all the time at the worse side, but wouldn't have needed that much to get on the acceptable side |
|
12 December 2018, 14:34 | #6 |
Puttymoon inhabitant
|
Sorry, couldn't help myself
[ Show youtube player ] |
12 December 2018, 16:05 | #7 | |
10MARC
Join Date: Jul 2018
Location: Tucson, AZ, USA
Posts: 214
|
Quote:
I had a Picasso 2 in my A2000 (when it was alive), and it worked fine for day to day use. RTG is worth it 100%, just plan on using the standard RBG port for gaming, |
|
13 December 2018, 10:12 | #8 |
Registered User
Join Date: Jun 2009
Location: Dublin, then Glasgow
Posts: 6,379
|
The thing is, most of the OS-friendly drawing commands are accelerated by passing the instruction directly to the graphics card itself. This takes far less bandwidth than transferring image data, so the limited bandwidth of Zorro-II isn't quite as big an issue as the numbers might make it appear. Using a Picasso-IV in an Amiga 2000 is a far, far better OS experience than AGA, despite AGA having higher bandwidth, purely because the actual pixel generation in many cases doesn't have to traverse the bus at all.
If a game uses the OS routines to draw its graphics (and doesn't simply use it as a frame buffer), then it will still benefit greatly from a Zorro-II RTG card compared to the native chipset. |
13 December 2018, 11:25 | #9 |
Registered User
Join Date: Feb 2008
Location: RNO
Posts: 1,007
|
But can you name any games doing that? I can't... or maybe some small freeware ones running in an OS window... but "real" commercial games are just using it as a dummy frame buffer, and they are affected big time.
|
13 December 2018, 11:32 | #10 |
Registered User
Join Date: Jun 2009
Location: Dublin, then Glasgow
Posts: 6,379
|
Yep, true. I'm not familiar with the specific mechanisms of most games. Anything 3D that's not using Warp3D will of course be software rendered, but perhaps there are some 2D games that use the OS blitting routines... Foundation, maybe? I might install that on my 2000 the next time I have some spare time.
|
13 December 2018, 19:08 | #11 | |
Registered User
Join Date: Nov 2009
Location: Top of the world
Posts: 161
|
Quote:
I would guess colonization. It surely looks like it is 100% system routines. |
|
16 December 2018, 20:29 | #12 |
Registered User
Join Date: May 2014
Location: Norway
Posts: 379
|
I've bought a CV3D, so I'll test to see what the benefits of the fastZII mode is at least. I'll post my findings here. I'll also see about overclocking, just for fun, to see if there are benefits also in ZII/fastZII-modes.
|
17 December 2018, 07:35 | #13 | |
Registered User
Join Date: Feb 2008
Location: RNO
Posts: 1,007
|
Quote:
I also have overclocked my CV64/3D to work marginally faster on a Zorro II bus... |
|
17 December 2018, 07:38 | #14 |
Registered User
Join Date: Feb 2008
Location: RNO
Posts: 1,007
|
|
17 December 2018, 13:13 | #15 | |
Registered User
Join Date: May 2014
Location: Norway
Posts: 379
|
Quote:
How much did you overclock it? Do you know how fast your ram is? I understand 70ns is normal for Phase5 cards, but DCE-cards often have 50ns ram it seems. Do you cool the card? I'll try colonization while I'm at it, but it will be difficult to gauge any speed improvement as it already runs VERY fast on a stock A1200. WHen I first asked about game-performance I was thinking about RTG-games. I had a PicassoIV back in the day but quickly upgraded to a Voodoo3, and the PIV was left doing scandoubling. :P I don't remember how well it performed on the ZII bus, and how great the improvement was over stock AGA. What I wanted to know was: Will SCUMMVM, Doom, Shapeshifter etc. run noticably faster on an RTG card on the ZII bus then on AGA, or will the ZII bottleneck be so dramatic that it's niglectible? My guess would be that even on ZII RTG will be dramatically faster. Correct me if I'm wrong, but mode promotion doesn't result in improved performance. 15 years ago I spent many days trying to get the best possible performance in Sim City 2000. One of the things i tried was mode promotion, but there was no performance increase. I tried it both with my Voodoo 3 and with my Picasso IV. |
|
17 December 2018, 13:44 | #16 |
Registered User
Join Date: Jun 2015
Location: Germany
Posts: 1,924
|
Except for systems with very low CPU power ZII will be slower than properly programmed AGA software for programs working with a chunky buffer.
|
17 December 2018, 13:58 | #17 | |
Registered User
Join Date: May 2014
Location: Norway
Posts: 379
|
Quote:
EDIT: It will be interesting to see if that is true for Fast ZII as well. Last edited by Overmann; 17 December 2018 at 14:14. |
|
17 December 2018, 14:49 | #18 |
Registered User
Join Date: Jun 2015
Location: Germany
Posts: 1,924
|
Well, it is an oversimplification because some programs may have a stupid approach and set pixels individually which may lead to working directly with planar graphics which would be very slow. This would be faster with ZII than AGA. However, the important factors here are that AGA has twice the bus speed as ZII due to the 32 bit bus (if I'm not very much mistaken) and that chunky2planar conversion can be hidden pretty much in the copying of the buffer to chip memory (even more so if AGA had as slow a bandwidth as ZII).
|
17 December 2018, 15:19 | #19 | |||
Registered User
Join Date: Feb 2008
Location: RNO
Posts: 1,007
|
Quote:
Quote:
Quote:
|
|||
31 January 2020, 14:09 | #20 |
Registered User
Join Date: May 2014
Location: Norway
Posts: 379
|
I've done some tests now. I have my A1200 set up with a DCE Cybervision 64/3D and have run sysspeed graphics tests with both ZII and fast ZIII, I have also compared the results to my A3000 030@25 with a Phase5 CV63/3D and the results are a little surprising. ZII and fastZII test exactly the same. So there must be something wrong somewhere, but my ZIV busboard was recently recapped and appears to be working perfectly.
The card in my A1200 test about 5 times faster then the CV64/3D in my A3000. I expect the CPU is bottlenecking the A3000 scores. My A1200 also test significantly slower then the A2k_060 module in sysspeed (about half the performance), which is strange.. Any thoughts ? :P EDIT: I am using the previous version of P96 on both my machines. Is it possible that I need to use a specific driver to use the CV64/3D in fast-ZII mode? Last edited by Overmann; 31 January 2020 at 14:31. |
Currently Active Users Viewing This Thread: 1 (0 members and 1 guests) | |
Thread Tools | |
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
Ram speeds with Accelerators | Toryglen-boy | support.Hardware | 9 | 29 April 2017 13:35 |
More cycle-exact speeds | Leandro Jardim | request.UAE Wishlist | 5 | 28 June 2013 08:44 |
Prince of Persia vs the Real World | s2325 | Nostalgia & memories | 4 | 18 April 2012 22:47 |
G-Rex4000D: any REAL speed difference between voodoo3 2000 and 3000 ??? | keropi | support.Hardware | 10 | 06 February 2007 02:21 |
WinUAE Speeds? | Amigaboy | support.WinUAE | 6 | 18 August 2003 11:00 |
|
|