21 September 2009, 02:12 | #41 | |
Registered User
Join Date: Mar 2008
Location: US
Posts: 315
|
I don't think it does revert to PIO -- (Ralph says it won't unless some rare condition exists where the accelerator has broken DMA to Chip access) hence the slowdown, the controller can only DMA to Chip RAM (since no accelerator RAM exists in the autoconfig space) and from there data is copied to accelerator RAM.
I'm tempted to pull my GVP sidecars out of the closet and run some benchmarks... I recall getting close to 3 MB/s with a Series II in my A2000, but can't remember what I measured with (might have been the RSCP program on the Guru ROM disk). Quote:
|
|
21 September 2009, 02:23 | #42 |
Moderator
Join Date: Nov 2004
Location: Eksjö / Sweden
Posts: 5,602
|
Oh! I thought Sysinfo was what everyone used. I guess I could be measuring Sysinfo's buffering here, but no... the "list directory" test I mentioned is plain enough.
I'm using Setpatch 37.34 (if that has anything to do with anything), will try 43.6b now. I'll have a look if the PP&S 040 likes the 040 libs. |
21 September 2009, 05:47 | #43 |
Moderator
Join Date: Nov 2004
Location: Eksjö / Sweden
Posts: 5,602
|
Buffers DO affect one thing at least: directory listing speed. Here's the sweet spot (for my 68040+GVP rig at least)
Test results: Buffers - speed (estimated) 11 - molasses 22 - molasses 30 - molasses 44 - molasses 55 - molasses 60 - Porsche 66 - Porsche 400 - Porsche 22 buffers were Porsche, with 32-bit ram disabled (IIRC). [<-- EDIT: NOPE, tested it properly just now.] It doesn't get gradually faster with increasing buffers, there's a clear boundary for when it goes VROOM! OK, I've now done some startup-testing and it resulted in the below paste. Seems binddrivers is not needed at all. (But maybe in Kick 1.3 mode?) I now have a zippy rig that autoboots to AsmOne in 2.5 seconds in 68040 mode and 4.2 seconds in 68000 mode. (From "white screen disappearing".) So now the floppy checks take more time than everything else (See my other thread here.) Code:
echo "LMB=LoadWB..." path sys:util sys:s add init040 addmem FASTROM >NIL: setpatch43_6b >NIL: ;---binddrivers205 >NIL: cpu DATACACHE NODATABURST INSTCACHE INSTBURST >NIL: cpu040 CBack_Off >NIL: MOUSECOND IF NOT WARN bpls resident c:assign pure assign t: GAMES:t assign ENV: sys:prefs/env-archive assign ENVARC: sys:prefs/env-archive assign prefs: sys:prefs assign KEYMAPS: devs:keymaps assign A: dh0:asmskool assign CO: DH0:CODE assign DS: DH0:CODE/DEVSYS assign st-00: dh0:snd resident assign remove system/Setmap Asm maxfont fastfonts >NIL: prompt %s> AsmOne \F\220\=\vA:work/d/stunner\r\ ;----------------------------------------------------- ELSE ;bpls 2 ;;;;; c:setpatch >NIL: c:version >NIL: addbuffers >NIL: df0: 15 Failat 21 resident >NIL: c:List pure add resident >NIL: c:Copy pure add resident >NIL: c:Assign pure add resident >NIL: c:Execute pure add makedir ram:T ram:Clipboards ram:env ram:env/sys copy >NIL: ENVARC: ram:env all quiet noreq assign ENV: ram:env assign T: ram:t ;set up T: directory for scripts assign CLIPS: ram:clipboards assign REXX: s: if exists sys:Monitors join >NIL: sys:monitors/~(#?.info) as t:mon-start execute t:mon-start delete >NIL: t:mon-start endif BindDrivers205 setenv Workbench $Workbench setenv Kickstart $Kickstart IPrefs ; echo "Amiga Release 2. Kickstart $Kickstart, Workbench $Workbench" conclip mount speak: mount aux: mount pipe: path ram: c: sys:utilities sys:rexxc sys:system s: sys:prefs sys:wbstartup add if exists sys:tools path sys:tools add if exists sys:tools/commodities path sys:tools/commodities add endif endif ; If this is the initial boot (i.e. keyboard env variable is not set) ; then execute PickMap which will query for a keymap and set the ; keyboard env variable. ; ; if keyboard env variable is set, set the keymap if ${sys/keyboard} NOT EQ "*${sys/keyboard}" setmap ${sys/keyboard} else PickMap.hd sys: endif if exists s:user-startup execute s:user-startup endif rexxmast >NIL: LoadWB endcli >NIL: ENDIF A bug that has been with me for years (misc expansions, misc WB versions) is that in Workbench, after some use, it hangs. I tested it out, and I can reproduce it simply by double-clicking on the harddisk and dragging a random folder icon. I can drag it for about 1 second (30 pixels), then it stops while the pointer moves, and I'm still holding the left mousebutton. I can't click on any icon from this point, and if I close the folder (click top left), the pointer freezes hard, every time. Now it's not the controller or the 68040, since it happens in 68000 mode too. The WBstartup is as you can see identical to the WB2.05 one. I've tried stacksize and misc stuff... I'm lost Is it an 'illegal insertion of Kick 2.0 in an OCS machine' conflict? ---------------{@}----------------- Anyway, now I'm ready to bench the A590 in hope for better speeds. My V6.6 roms should fail on my 2GB drive, well, let's see. (My partitions are 1 GB.) Maybe burning a Guru ROM (prediction: not possible...) will have other benefits, like faster speed? Is it possible, or should I burn A2091 ROM 7.0? I'm a bit skeptic to the A590setup.dms... it wants WB 1.3 disks on install, and the only pertinent file on the whole disk seems to be A590HD in the Expansion dir... Thoughts on a plan of attack on the A590, or is it easy peasy? Last edited by Photon; 21 September 2009 at 23:15. |
21 September 2009, 09:05 | #44 | |
Thalion Webshrine
Join Date: Jan 2004
Location: Oxford
Posts: 14,331
|
Quote:
Possible, but not practical. Burn the A590/A2091 ROM v7.0 instead. |
|
21 September 2009, 10:44 | #45 |
Registered User
Join Date: Mar 2008
Location: US
Posts: 315
|
Is the line turning off the copyback cache needed for something? Same with CPU DATABURST (don't think that tag does anything on 040/060 anyway). If using the MMULib libraries, you could delete those lines completely... the CPU lib should set everything optimally.
init040 ADDMEM >NIL: SetPatch QUIET MuFastROM ON PROTECT >NIL: ;kickstart mirrored and protected MuFastZero ON ;remaps zero page You could even use MuFastZero in combination with MuMove4k to try and get expansion.library in Fast RAM, but I'm not sure how that would work with your board needing an ADDMEM command... might have to surf the docs and experiment a little... but a quick glance sez it can work with some extra fiddling. |
21 September 2009, 16:16 | #46 |
Moderator
Join Date: Nov 2004
Location: Eksjö / Sweden
Posts: 5,602
|
It's for compatibility with apps. I turn it on for 3D apps tho.
Cool. alexh: Will burn ROM 7.0 then, if V6.6 is incompatible/slow. Any thoughts on the hang-bug in WB? Damaged icon.library? :P Or Kick 2.0 gfx.lib expects ECS chips? |
21 September 2009, 21:19 | #47 |
Moderator
Join Date: Nov 2004
Location: Eksjö / Sweden
Posts: 5,602
|
A590 first impressions: The controller, while looking brand new and awesome with an air of solid quality (pics are coming), it was a definite step-down.
It autobooted my 2GB *DISK* fine, booting time just a bit slower than GVP. However, it considered my second 1GB partition (at end of 2GB disk) as "Not a DOS disk". So it's confirmed that >1GB disks are fully possible with A590 ROM V6.6, and also that 0.98 GB partitions are okay (so far at least). Test results: (60 buffers etc... same settings as above) Mode - speed in Sysinfo 68000+A590 fastram - 1126 KB/s 68040+A590 fastram only - 1136 KB/s 68040+A590 fastram+"CPU FASTROM" - 1176 KB/s 68040+A590 fastram+32-bit ram - 395 KB/s It was fine to connect a PSU to the harddisk power cable. |
21 September 2009, 21:37 | #48 | |
WinUAE developer
Join Date: Aug 2001
Location: Hämeenlinna/Finland
Age: 49
Posts: 26,502
|
Quote:
You can can also use filesystem with direct-scsi like PFS3. |
|
21 September 2009, 23:12 | #49 |
Moderator
Join Date: Nov 2004
Location: Eksjö / Sweden
Posts: 5,602
|
OK, then it's 100% certain that V6.6 roms can handle partitions up to 1GB on >1GB disks, if they are within the first GB of the disk. So the A590 faq had an error in it. (Saying drives >1GB can't be used and that the simplest solution was to order a Guru Rom.)
If I can get my EPROM burner from my other place tomorrow I'll try V7.0. The speeds Ralph Babel mentioned suggest that with 32-bit RAM it falls back to PIO (~300 KB/s), and without it, it DMAs to chip, then copies to 24-bit fastram?? 890 ns=1123 KB/s. Silly way to do it with onboard fastram, says I. Unless the faq's "real use" speeds without a Guru ROM, 1.4 MB/s, is "my" 1.1 MB/s. Recommendations for tools like the rawscsispeed mentioned? Will examine the faq for some spiduptrix. Last edited by Photon; 22 September 2009 at 14:20. Reason: clarity |
22 September 2009, 00:53 | #50 |
Moderator
Join Date: Nov 2004
Location: Eksjö / Sweden
Posts: 5,602
|
Pictures:
A590 Anyone want a mint condition 20MB Seagate SCSI HDD!? GVP Impact Series II A500-HD8 revision 6 Anyone know the value of the C13 SMD cap? Not a biggie, I guess. Same as C14 in upper left!? (47 uF, right?) |
22 September 2009, 01:08 | #51 |
Global Moderator
Join Date: Aug 2008
Location: Sidcup, England
Posts: 10,300
|
Yes please!
Your A590 is exactly the same as one I've just bought from PowerPie5000. It hasn't arrived yet, but I understand it still has the original 20MB Seagate SCSI drive installed. Would you burn a V7.0 ROM for me too, please, if it turns out to be a worthwhile upgrade? Last edited by prowler; 22 September 2009 at 01:34. Reason: Typo |
22 September 2009, 01:20 | #52 |
Moderator
Join Date: Nov 2004
Location: Eksjö / Sweden
Posts: 5,602
|
You really want the harddisk!? *rubs eyes* Oh well, paypal postage... :P
And I know, I was about to buy it Well, if it's just like mine, you will be able to put multi-GB drive in it as long as you split it into <1GB partitions. Or maybe I misunderstand you... If I burn one for me, I can burn one for you for a few € ofc. |
22 September 2009, 01:32 | #53 | |
Global Moderator
Join Date: Aug 2008
Location: Sidcup, England
Posts: 10,300
|
Yes, please.
I'll be happy to. Oh yes, I remember now. Quote:
Thanks. I'm looking forward to the results of your experiments with this ROM version. |
|
22 September 2009, 01:56 | #54 |
Moderator
Join Date: Nov 2004
Location: Eksjö / Sweden
Posts: 5,602
|
I have an original Macintosh. But hmm... no, I will not attempt to stick it in there, even if there are some strange adapt0rs to do it. It whirrs so nicely
And I won't put a 20MB SCSI drive in anything else either, so sure Will test and weigh it after I try out the A590 some more. What, if any, software is newer than the "WB 1.3" driver disk on amiga.resource.cx !? Does it even look at a single file outside the ROM..? - can I put something in the startup-sequence that will "patch it to be accelerator-optimized"? |
22 September 2009, 04:42 | #55 |
Moderator
Join Date: Nov 2004
Location: Eksjö / Sweden
Posts: 5,602
|
OK, took a break from SCSI and compared straight IDE to IDE through IDEfix (not express) on my A1200-030.
With IDEfix: 1807 KB/s Without it: 1803 KB/s Someone said the tiny buffer gave slight speed improvements. Well, maybe on my A600? Will test, at least. I did get a measurement of 401 KB/s on the A590, a record for 68040+32-bit ram... also tried addbuffers 800, which did nothing in Sysinfo with/without 68040 or with/without 32-bit ram. Sysinfo showed 60 buffers instead of 860, but that might be that it just shows (SELECTED) RDB info. I didn't feel like changing the #buffers in the RDB, as the A590 doesn't recognize the second partition. GVP: Still unsure about the cap - it should be a normal anti-noise cap for the chips, but 47uF is huge... anyone know the value or has a service manual? I was unable to find info on how to decode the digits or find a manual. Now to cut the piece of shit that is Overdrive into pieces so I can re-use the 2.5" cable at least... plastic and superglue. Wow. |
22 September 2009, 09:07 | #56 | |||
Registered User
Join Date: Mar 2008
Location: US
Posts: 315
|
Quote:
Quote:
Also one more helpful tidbit from Ralph: Quote:
|
|||
22 September 2009, 09:13 | #57 |
WinUAE developer
Join Date: Aug 2001
Location: Hämeenlinna/Finland
Age: 49
Posts: 26,502
|
|
22 September 2009, 14:17 | #58 |
Moderator
Join Date: Nov 2004
Location: Eksjö / Sweden
Posts: 5,602
|
Well, hm yes, but it's still true. I meant partitions as in "partitions that are this big are possible", not as plural. Wording changed for clarity.
Damion: Weird, in most equipment such caps are ~22nF (lentil type ). But it has a + mark, so I'm convinced... I'd like to add one, and I have only electrolytics and drop-shaped ones which are likely tantalum electrolytic (middle one on the pic) Is yours the tantalum? Last edited by Photon; 22 September 2009 at 15:06. |
22 September 2009, 18:36 | #59 | |
Registered User
Join Date: Mar 2008
Location: US
Posts: 315
|
Mine are electrolytic, axial thru-hole. (Probably dried up at this point :P) The board looks about similar to yours, just a bit older design.
http://amiga.resource.cx/exp/impact5...t_impact500mk2 (mine is Rev 3) Quote:
|
|
22 September 2009, 19:08 | #60 |
Moderator
Join Date: Nov 2004
Location: Eksjö / Sweden
Posts: 5,602
|
Right-o.
OK, I have a Pocket Programmer 2A from Xtronics.com hooked up. Software says it needs a file called zntport.sys. I'll have a more thorough look for it later, but if someone happens to have it ... I got the impression it comes with Visual Basic, is that so? If someone has it, then it should be in the %windows%\system32\drivers folder. Last edited by Photon; 22 September 2009 at 20:03. |
Currently Active Users Viewing This Thread: 1 (0 members and 1 guests) | |
Thread Tools | |
Similar Threads | ||||
Thread | Thread Starter | Forum | Replies | Last Post |
CF card speed comparison | Photon | support.Hardware | 79 | 05 March 2014 19:36 |
The scsi interface metc vs squirrel scsi conflict | rampartsagain | support.Hardware | 29 | 24 June 2012 17:50 |
Cyberstorm PPC vs Blizzard PPC CPU Speed Comparison | CU_AMiGA | support.Hardware | 9 | 15 March 2012 14:25 |
Wanted: A500 SCSI interface with ram | Photon | MarketPlace | 0 | 03 February 2010 19:39 |
|
|