Blitz - Originally known, by the designer, as "Arcin32" - The 32Bit IDE Interface - NOW WORKING!

[rodders]

I already have an IDE adapter for the CF card, I just need something to connect it to the ARCIN32. Its the difference in pitch between the two that's the problem.

[IanJeffray]

I already have an IDE adapter for the CF card, I just need something to connect it to the ARCIN32. Its the difference in pitch between the two that's the problem.

Then you probably have the wrong adapter - guessing you have a CF to 44pin IDE, wheras what you need is a CF to 40pin one.

[rodders]

Well I can use it in my A3020 instead.
What is the purpose of the 44pin connector on the Arcin32 then?

[rodders]

Ah, just realised its probably to take the DOM adapter.

[daveejhitchins]

Ah, just realised its probably to take the DOM adapter.

It is . . . I have adaptor boards if you ever need one. You'll see how things fit together in the back of the User guide

Dave H.

[baz4096]

I also have a Kinetic

Dave H.

Hi Dave,

I don't suppose you have any test results from your Kinetic machine? Does your Blitz interface work with a Kinetic SA CPU?

All the best

Chris

[daveejhitchins]

I also have a Kinetic

Dave H.

Hi Dave,

I don't suppose you have any test results from your Kinetic machine? Does your Blitz interface work with a Kinetic SA CPU?

All the best

Chris

Hi, Chris . . .

There were problems to do with RAM when using the Kinetic and ARCIN32 - this was in a RISCPC which wouldn't let you use the Kinetic's on-board RAM so there wasn't any benefit to be had from the Kinetic's extra fast RAM!

As I'm not a 'user' of any particular machine e.g. I just use them for testing my accessories, so I let Ian Jeffray have the Kinetic. He's been very helpful with lots of other related issues I've had!

BRs - Dave H.

[sirbod]

There were problems to do with RAM when using the Kinetic and ARCIN32 - this was in a RISCPC which wouldn't let you use the Kinetic's on-board RAM so there wasn't any benefit to be had from the Kinetic's extra fast RAM!

Looking at the Kinetic PRM it would imply Podule DMA should work if the RiscPC has DRAM installed. I believe it was supposed to DMA to IOMD mapped RAM and then soft-copy it across to the Kinetic RAM. However, I'm not sure if DMAManager was modified to allocate IOMD based RAM for this to work, so suspect all Podule DMA will fail.

There's two potential workarounds so far as IDE goes, use PIO mode or ensure DMA capable Podules aren't fitted to DMA capable slots.

Either way, any DMA capable Podule would essentially be wasted in a Kinetic and likely be slower than a non-DMA Podule.

I have a Kinetic RiscPC and a selection of DMA capable IDE Podules (although not a Blitz) so could test this once I get the thing booting - its a bit temperamental.

[IanJeffray]

There's two potential workarounds so far as IDE goes, use PIO mode or ensure DMA capable Podules aren't fitted to DMA capable slots.

You may have misunderstood the issue. Original Blitz works fine with Kinetic (drops to PIO), the new one does does not - it _requires_ DMA or it doesn't operate at all - even, for example, in a standard StrongARM machien in a non-DMA slot. This is yet to be understood.

[sirbod]

You may have misunderstood the issue. Original Blitz works fine with Kinetic (drops to PIO), the new one does does not - it _requires_ DMA or it doesn't operate at all - even, for example, in a standard StrongARM machien in a non-DMA slot. This is yet to be understood.

I think we're talking about two differnt things. The posts above are related to this card working in a Kinetic.

Your issue is something I covered in this post with the rules it uses for DMA mode. You could also try *Unplug DMAManager or !(<private address>+&458)=-1 which force DMA off regardless - assuming we're talking about IDEFS v3.28.

It might be worth trying a "PIO only" drive to see if it works, that will confirm if its a problem specific to DMA capable drives switching to PIO, or a generic Podule/Module issue.

[baz4096]

I think we’re talking about the same thing. I recently acquired a kinetic card and had heard that they don’t work with DMA only Podules. I’d also heard that Dave’s Blitz card had issues with non-DMA machines, and was trying to establish if my Blitz card would work with my Kinetic SA. I’ll likely not try that particular combination in my main machine, but will try with a less important drive when time allows and see if I can contribute anything to help debug the Blitz DMA issue. Thanks for the idea regarding using a non-DMA capable drive too.

[sirbod]

I think we’re talking about the same thing.

It depends if your Podule is an "original" or the new one discussed in this thread. I wasn't sure which one you had.

If you have the original Podule, it "should" work as-is either in DMA or non-DMA mode in a Kinetic. If it doesn't work in a DMA capable slot, it will be due to the general Kinetic DMA compatibility issues I mentioned above. Moving the Podule to a non-DMA Podule slot will usually get Podules working, provided the Podule and its Module(s) all support non-DMA transfers.

If you have the new Podule, it's not going to work due to the issues noted by Ian where IDEFS is failing to fall back to non-DMA transfers.

I've provided various details about how IDEFS v3.28 decides to use DMA transfer in my posts above, which need some testing done to see if forcing the Module into non-DMA gets it working. It's not actually clear if the issue is hardware or software related, my guess is its probably hardware related as the Module hasn't changed - unless the latest version also introduces the issue to the original Blitz, I'm not sure if that's been tested.

The quick test that anyone could perform is to try *Unplug DMAManager, power-cycle and see what happens. It's also worth testing a "PIO only" drive to rule out issues on the drive interface side. Then there's trying the card in a non-DMA slot (which I assume has already been tried), trying a non-page-aligned DiscOp, we're then into the low-level stuff, like seeing if an ATA IDENTIFY command works when the DMA transfers are failing, checking the byte order of the returned data, modifying IDEFS etc.

Although I can't help out on the hardware side, I can help with testing/modifying IDEFS with the new Podule, if Dave/Ian would be willing to send one to me.

[IanJeffray]

Some hours of playing and a ridiculous amount of podule and ARM pulling later. I can state the following:

* With StrongARM & RISC OS 4.39:
- Original Blitz works well in DMA slots. 6MB/sec.
- Original Blitz 'works' in non-DMA slots, but at a rather pathetic 1.8MB/sec max read speed, with the same 2GB DOM. Worse than most/all other IDE cards, even ArcInV6C.
- New Blitz works well in DMA slots. 6MB/sec read speeds from 2GB DOM.
- New Blitz does not work AT ALL in non-DMA slots. Causes ALL podule slots to show up as "extended type" podules, causes data abort before "RISC OS" displayed, and unplugging DMAManager has NO effect.

* With Kinetic & RISC OS 4.39:
- Original Blitz 'works' in DMA slots, but at the rather pathetic 1.8MB/sec max read speed (NB; DMAManager is still active)
- Original Blitz 'works' in non-DMA slots, but at the rather pathetic 1.8MB/sec max read speed. (NB; DMAManager is still active)
- New Blitz also 'works' in DMA slots (much to my VERY great surprise!), but at the rather pathetic 1.8MB/sec max read speed. (NB; DMAManager is still active)
- New Blitz in a non-DMA slot PREVENTS RISC OS FROM STARTING AT ALL, until DMAManager is unplugged. With DMAManager unplugged, the machine takes "ages" to start, but then does start, with no IDEFS and *podules shows "extended type &0000" for all slots, as with StrongARM.

* With A7000+ & RISC OS 4.39:
- There are no DMA slots (Arm7500FE has no podule DMA) - New Blitz doesn't work at all here - didn't test original Blitz.

The results for the failing modes are the same without any IDE device connected.

Others can draw their conclusions from this.

[sirbod]

Sounds like the new Podule has a hardware issue to me.

The Kinetic speeds are what I'd expect as it's having to double-buffer via IOMD DRAM.

[daveejhitchins]

Sounds like the new Podule has a hardware issue to me.

I does . . . We've worked out that the ABLE code given to me is not the latest version. I'm still working on finding the latest.

Dave H.

[dpsharp]

* With A7000+ & RISC OS 4.39:[/b]
- There are no DMA slots (Arm7500FE has no podule DMA) - New Blitz doesn't work at all here - didn't test original Blitz.

Per emails, I've got two of the New Blitz cards here that both seem to work on my A7000+ RISC OS 3.7 using a Risc PC 2-port backplane but don't work on it with the new CJE Micro A7000+ backplane (https://www.cjemicros.co.uk/micros/indi ... JE-AMD10EQ).

Let me know if you want to borrow the podule/backplane or even the whole machine if it's helpful for diagnosis/comparison.

[IanJeffray]

* With A7000+ & RISC OS 4.39:[/b]
- There are no DMA slots (Arm7500FE has no podule DMA) - New Blitz doesn't work at all here - didn't test original Blitz.

Per emails, I've got two of the New Blitz cards here that both seem to work on my A7000+ RISC OS 3.7 using a Risc PC 2-port backplane but don't work on it with the new CJE Micro A7000+ backplane (https://www.cjemicros.co.uk/micros/indi ... JE-AMD10EQ).

Your experience is a massive anomoly as far as I'm concerned. Per the above you can see it doesn't even work on a RiscPC without DMA, so how it can work on an A7000, who knows? But you've got a slightly newer revision of the board than I have... though I expect that the FPGA/ROM code is identical.

You'll also note though that I've only tested with 4.39 -- I'll set up with 3.7 on A7000+ and give that a shot, but as it seems to be a hardware-level issue (FPGA bitstream, probably) it seems unlikely OS version would have an impact.

I have tested extensively with both the CJE and RiscPC backplane, in both RiscPC and A7000, as you're aware per email.

Let me know if you want to borrow the podule/backplane or even the whole machine if it's helpful for diagnosis/comparison.

Borrowing the podule would certainly be of some interest - to check on a wider number of machines here, and to compare against operation of other boards. Or I can send you one of my 'new' boards which doesn't work without DMA, as noted above. And/or 4.39 ROMs to try. For reference, I've tried on around 6 different machines here, with two samples each of the 'New' and original Blitz cards. I've tried a wide range of media too (CF, DOM, mSATA) but note that the issues I explained just above do appear to be wholly media-independent, so I think we can rule that out, at least.

Can you confirm it's IDEFS 3.28 you're running?

EDIT: I've tested RISC OS 4.03 and 3.7 and they both still exhibit the same behavior (or worse) than I described above.

[dpsharp]

Your experience is a massive anomoly as far as I'm concerned. Per the above you can see it doesn't even work on a RiscPC without DMA, so how it can work on an A7000, who knows? But you've got a slightly newer revision of the board than I have... though I expect that the FPGA/ROM code is identical.

Agreed on anomoly - this machine is a bit half-eaten by battery leak but I can't imagine how that would make it work when it shouldn't.

Borrowing the podule would certainly be of some interest - to check on a wider number of machines here, and to compare against operation of other boards. Or I can send you one of my 'new' boards which doesn't work without DMA, as noted above. And/or 4.39 ROMs to try.

I'm happy either way, I can experiment with anything you send or send my podule to you, just let me know what you prefer. I'd not tried my second New Blitz podule til this morning as had left it unopened so I was pleasantly surprised when it worked too.

Can you confirm it's IDEFS 3.28 you're running?

EDIT: I've tested RISC OS 4.03 and 3.7 and they both still exhibit the same behavior (or worse) than I described above.

Yes, IDEFS 3.28 Confirmed. Good to know the OS doesn't seem to make a difference on your end.

*ROMmodules in case useful...

[IanJeffray]

* With Kinetic & RISC OS 4.39:
- Original Blitz 'works' in DMA slots, but at the rather pathetic 1.8MB/sec max read speed (NB; DMAManager is still active)
- Original Blitz 'works' in non-DMA slots, but at the rather pathetic 1.8MB/sec max read speed. (NB; DMAManager is still active)
- New Blitz also 'works' in DMA slots (much to my VERY great surprise!), but at the rather pathetic 1.8MB/sec max read speed. (NB; DMAManager is still active)

Additional pain now uncovered - the above is only true if the NIC slot is also empty. With an EtherLan600 present, the machine will not start up in any of the above circumstances - just an abort and immediate lockup.

With my favoured 'fastest with Blitz' 2GB DOM (no smaller, no larger) I'm seeing 2.3MB/sec, which is still not great -- marginally slower even than an ArcInV6C in the same machine. (For comparison, Unipod in the same machine, same DOM, manages 3.4MB/sec)