keirf / amiga-stuff Goto Github PK

Hi,
I modified my A500 to have 2MB Chip mem, like it is done here:
http://www.rasterline.com/images/amiga/A500_REV6A_2MEG_CHIP.jpg

It works without issues, but ATK fails unpredictable during memory test with Kick3.1.4 with an exception (black screen).
With Kick1.3, the test runs fine. Other mem test tools (like MBRTest-2) do not have issues, regardless of Kick version.

Just wondering if this is a known issue or not. I've got an ACA500plus and SysTest seems to hang almost as soon as I launch it. It accepts input sometimes for about a second and then stops, the mouse still moves but no other input is recognized.

I should also mention that I put an 8372A Fat Agnus into it (an old v1 motherboard) so i could use 1MB chip ram. I've tried with the chip ram on and off with no luck. I've also played with the Aca500plus settings, disabling various things to try and get it to work to no avail.

Thanks

Hello,

Thanks for all the work done in this great piece of software.

On real Amiga 500, with ACA500plus extension, pressing the Help key always cause a crash instead of rebooting the computer. The crash screen is green and titled EXC1070330296 and showing register values. It seems to come from the internal exception handler of the ACA500plus. The 7-segment display of ACA show : E0

My configuration is :
Amiga 500 rev 6
equiped with 512 kb trapdoor extension
and ACA500plus !

I recorded a video to show you the problem : https://youtu.be/FR0Uk4Dvif4?t=39
You can rewind this video to see more details of my configuration.

I have also tried booting directly from the AmigaTestKit ADF, in plain kickstart 1.3, in ACA500 "cloaking" mode. Pressing Help also crash the machine with the same green screen.

Hi,

would it be possible to add the means for supporting HD disk drives (1760 kB) in testkit?

They assert a oscillating READY pattern instead of the constant 0xFF sent by DD drives, resulting in a drive id of 0x55555555 instead of 0xFFFFFFFF. Also, the drive's motor spins at only half the speed resulting in READY taking twice the time of being pulled low upon motor startup and INDEX pulses only about every 400 ms for HD disks.

While both the READY pattern is correctly read (while being interpreted as drive id ???) and the INDEX signal timing is showing as being correct, I get warnings about late or missing READY. Also, the read and write tests fail. I think they assume the flux of a 880kB disk coming from the drive regardless of id pattern and timing.

I know that real HD drives in Amigas are rather rare. But I've come up with a little board that mounts inside a Sony floppy to achieve just the right id sequence and ready timing as well as having the motor run at half speed, and I want to get it as accurate and similar to the unobtainable FZ357A as possible. Testkit would be a wonderful tool to test such converted drives, especially when combined with other floppy drives, Goteks, etc. and connected as DF1: in bigbox Amigas. I'd be very grateful if my suggestion could be put on the roadmap.

Thanks in advance,
Torsten

there is another glitch happening to the output in case a ready-error occurs (like the shown one: "signal is oscillating")

1. Lower/Upper-Text is shifted down a row
2. the Read-Matrix is painted over this

Originally posted by @fook42 in #27 (comment)

I think the only major feature currently missing in SysTest is the ability to set the RTC to an arbitrary date.

It would be nice if it could be added :).

I know, this is quite rare device, but with atari paddles connected, it says "paddles not detected". But analog joystick test works with them, so maybe name this test analog+paddles?

I have V1.16 that works without any issue on my modified Sony MPF 920 z/121 and z/131 drives. However, testing my A500 with V1.18 booting fails with a red screen! Initially I thought it's a disk problem but after using new disks, testing on another A500 and another modified drive I can confirm the issue; it does not boot on a modified PC drive (Sony MPF920 to be specific).

[ Firstly, thanks for this very useful test suite - I'm running it on all my Amigas to diagnose any faults. Only thing really missing is maybe more extensive gfx diagnostics (blitter/copper/Agnus etc.)! And a way to quit the utility when running via Workbench is also useful ]

However, I noticed a minor issue with the Mouse / Joystick / Gamepad check:

Both ports are detected as mice on my A1200 (w/ rev 1d4 mainboard) using a Tac-2 stick which was properly detected as such on my A500. The same also occurs when running SysTest via FS-UAE (emulating an A1200) and using a USB-joystick on OSX through the emulation layer.

This is using the (AFAIK) most recent SysTest release from Feb 18th, 2017.

I have a couple of memory expansions for the A500 with RTC based on the 72421 clock chip. The program reports incorrectly the M6242B chip. I suggest that the RTC chip number is removed unless there is a way to figure it out.

Some memory controllers (i.e. terriblefire cards) will implement burst mode for the MOVE16 command. At some point it would be nice to test ram with this instruction. I noticed that my ram passed systest but had MOVE16 issues.

Burned this to a bootable CD and it doesn't work on the CD32.
Also tested loading it from an IDE drive attached to my CD32 and it didn't work.

Loads to the menu screen, but shows no text and it's locked up there.

Find ISO attached. Behaves the same way in WinUAE with CD32 config.
systestCD32.zip

Sept 17 2017 20:00:25 build is the one that doesn't work. Seems like some people have another version that does? Still trying to determine this.

This is gonna be a short novel but please bear with me.

A while back I got an A4000D Rev B with some very mild corrosion from the leaking battery. It appeared to be in very nice shape overall, but looks can be deceiving :)
Machine was constantly in reset state - fixed, mouse input was erratic - fixed, audio channels had difference in amplitude - fixed, Zorro cards were not being recognized - fixed etc etc...

Anyway, during the fixing period I was using ATK 1.4, among other things, to check on the issues. 3 chips below the RTC (U177, U975 and U976) were replaced (RTC chip was left intact) together with some other stuff and everything appeared to be working fine, RTC included.

I had no time to play with the Amiga no more but resumed work a few days ago. Meanwhile ATK 1.5 was released and I started using it without much though, just to retest the motherboard quickly before i start putting everything back in the case. However, RTC test would constantly semi-hang the machine after a random period of time (from few seconds up to few minutes). So, completely baffled, as I was pretty sure everything was working fine a while back I went to replace the resistor pack and RTC chip just because I had spare anyway.
Ok, fine, I'm good to go now, for sure. Well... no, ATK was still semi-hanging on RTC test. WTF WTF Panic! So, as RTC seamed to be working fine from DOS I decided to give ATK 1.4 a go, and it works fine.

I'm not sure what has been changed from 1.4 to 1.5 but I've now left 1.4 running on 2 separate occasions for 24 hours on the RTC test screen and it survived, while 1.5 constantly semi-hangs after a short period of time.
What I mean by semi-hanging could best be described as time no longer updates, keyboard input is no longer registered, input from the mouse buttons is no longer registered but mouse cursor can be moved around and items do highlight when you mouse over them, just can't actually navigate anywhere any more.

Another great piece of software from you - have you ever released anything crap ?!!

I’ve been working on resurrecting an old trapdoor expansion with a failed (unpublished) GAL.

I’ve obviously been making heavy use of ATK as part of this.

What I can’t 100% test is the chip ram banks as seen from the Agnus Denise and Paula.

Could you add video tests in each of the chip ram banks based on what’s detected ?
Ie, an image at $00xxxx, $08xxxx, $10xxxx, $18xxxx

For bonus points borrow DiagROM scroll to test the blotter.

AmigaTestKit v1.10 incorrectly detects my Amiga 1000 as a PAL machine.

The machine is a unmodified NTSC model / Rev 6 motherboard:

• 28.3636 mhz crystal
• 8361R3 Agnus

The Chipset IDs shown on the CIA / Chipset menu are '0000' for Agnus and 'ffff' for Denise.

It seems that the Audio Test crashes on accelerated machines.

The time before the crash happens is quite random. Most times it will happen within 5-10 seconds after the mod starts playing, but sometimes the song might get to the end and loop once before it crashes.

I first noticed this on my A500+ with TF530@50MHz. I dismantled the machine piece by piece to find the cause, but I never managed to eliminate it. Then I noticed the machine is perfectly stable in WB and games, and the ATK RAM test is also OK.

Then I spoke to a few friends and found out that they also experience the same problem with TF534 and TF536.

Tested on v1.12. It does NOT seem to happen with the original 68000 and I haven't been able to reproduce it in UAE either.

I guess this might also be a bug in the TF line of cards, I can't rule that out as I know no-one with a different accelerator.

I am attaching a crash log from my machine and one from a friend's, who has a TF536@50MHz.

Would it be possible to make a ROM image of this? Something that can be burned on an EPROM and put on a real, driveless Amiga?

I had to memory test a machine which threw an error roughly twice a day. So I ran it with the monitor off, and occasionally checked on it.

I think it would be very helpful if the power LED toggled in brightness (i.e. toggle the audio filter status), to let the user know when to take a photo of the screen and then restart testing.

Hi, I have an Amiga 1200 that displays the moving checkerboard as a static one. I have another A1200 that displays it correctly moving, with all other hardware being the same. Would this point to a faulty Lisa chip in the first A1200 [how does the test derive the alternating image?] and would it be possible to devise/add a test for further diagnosis in your wonderful ATK?
(The strange thing is using the faulty A1200 normally I can't notice anything wrong?) - thanks!

Hi Keir, could you please add the following information about the older versions of atk and docker ...

diff --git a/README.md b/README.md
index 04dbed2..d8d814e 100644
--- a/README.md
+++ b/README.md
@@ -12,14 +12,16 @@ Hub)](https://hub.docker.com/r/rjnorthrow/atk). You can run it as
follows to generate the latest Amiga Test Kit zip file in your current
directory:

-docker run -v $(pwd):/output --rm -ti rjnorthrow/atk
+docker run -v $(pwd):/output --rm -ti rjnorthrow/atk:latest

-COMMIT and MAKE_OPTS may be set in the environment to build a particular version of Amiga Test Kit, and to specify extra build parameters:
+COMMIT, MAKE_OPTS and atk version (v1.7, v1.10, latest) may be set in the environment to build a particular version of Amiga Test Kit, and to specify extra build parameters:

-docker run -e COMMIT=testkit-v1.3 -e MAKE_OPTS=-j4 -v $(pwd):/output --rm -ti rjnorthrow/atk
+docker run -e COMMIT=testkit-v1.3 -e MAKE_OPTS=-j4 -v $(pwd):/output --rm -ti rjnorthrow/atk:v1.7

+For versions up to 1.7, use rjnorthrow/atk:v1.7. For versions 8, 9 and 10 use rjnorthrow/atk:v1.10
+

Build From Source (Manual Method)

Requires a GCC cross-compiler toolchain targetting

Trying to use :

docker run -e COMMIT=testkit-v1.8 -e MAKE_OPTS=-j4 -v $(pwd):/output --rm -ti rjnorthrow/atk

...(or later versions) results in:

...
CC video.o
VASM ptplayer/ptplayer.o
vasmm68k_mot -Felf ptplayer/ptplayer.asm -o ptplayer/ptplayer.o
make: vasmm68k_mot: Command not found
make: *** [../base/Rules.mk:65: ptplayer/ptplayer.o] Error 127
make: *** Waiting for unfinished jobs....

This is when running rjnorthrow/atk [554c6212d705] on Docker version 18.06.3-ce, build d7080c1

It looks like v1.3-1.7 do not use ptplayer which is why there is no need/dependecy on vasmm68k_mot - git history confirms the testkit/Makefile had ptplayer added 2020-06-03

Hello,

Grateful for having added the KS reading in the tool.

A small glitch. When using not standard roms (in one of my cases, a custom KS 1.3 with the scsi.device to boot on a A600) the ATK hangs on startup. I mean, it shows the red and blue screen without text.

Would it be possibile to bypass that check and just shows undetected or whatever you like in these corner cases?

Thank you and keep up the good work!

In an Amiga 4000 repair video (Adrians Digital Basement). He installed some 1Mb memory which was incorrectly configured to via a jumper to act as 4Mb memory.
The memory test however still passed as it seems the memory test was seeing the misconfigured RAM addresses as mirrors of the working RAM.
I feel like the random fill test could be setup differently to detect this situation. It seems like the random fill patern is too repetetive to catch this error.
So in his case a 16Mb ram test passed, when only 4Mb was actually in the computer.

Issue is described here from https://youtu.be/mJkiblw8Wl8?t=867 to 16:30

1. Configurable delay between fill and check, perhaps as a new top-level option. This can detect DRAM refresh problems.
2. Custom range as top-level option.

[From Chris Morley via FB]

I have an A2000 rev 6.2. with an Udo Neuroth Turbo-Karte, that is an accelerator with a 020 and 4 MB of Fast 32-bit RAM.

Contrary to that the above page says, the RAM does NOT autoconfig, it needs a tool from the install floppy to be run (Add32Bit), which will instantly "insert" the RAM into the system. I also have an Oktagon card with 4 MB of Fast RAM, which gets autoconfig'ed at $200000-$5fffff, so the RAM from the Udo Neuroth gets mapped at $600000-$9fffff.

The system appears to be working correctly, though I haven't done extensive testing on it. I have tried to run the ATK Memory Test and:

• If I boot it straight away, it won't see the ram from the Udo Neuroth Card (which is expected AFAIK).
• If I run the "manual" memory scan it will detect 5 MB of Fast RAM, that is the 4 MB from the Oktagon + 1 MB from the Udo Neuroth, and it will all test fine.
• If I run the WB version of ATK once WB has booted and the memory has been configured by the tool, it will see the whole 8 MB (4+4) but the test will fail as soon as it checks the $600000-$67ffff range.

I have tried to edit the start/end addresses to test 1 MB from the Udo Neuroth at a time and this way it will always succeed. So I guess what's happening is that the memory is wrapping at every 1 MB, but I don't understand why this is happening. Other memory tests I have tried either do not see the Udo Neuroth RAM or detect no problems (but they might be crappy and not detect "wrapovers"). Could this be a bug in ATK or more likely a problem with my card?

The card has a "1/4 MB" jumper, but I'm not sure which setting corresponds to the jumper open or closed and the manual does not help. All I can say is that it is currently closed and if I open it the RAM config tool will crash, so it looks like it's in the 4 MB position.

Hello,

I've run the latest release version (on an A1200 with a 8mb trapdoor extension) and I get a strange memory detection:

• Kickstart memory scan : chip 2Mb, fast 8Mb
• Direct memory scan: chip 2Mb, fast 4Mb

It's not really clear to me what is the difference between both memory scan types and what can drive to this different detection.

When I run the test, I get errors starting at address 0x60000, what corresponds to the second 4Mb of fast ram that is not detected by the 'direct' memory scan.

Any hint on what is going on would be very helpful :)

in floppy drive test, using head calibration, only 3 positions within the floppy disk are accessible ( track 0, 40, 79 ).
my suggestion: having a complete "map" of a disk (all 80 tracks) showing the misalignment of each track as a sum(?).
layout may be like in disk-copiers (as e.g. Xcopy).
That would give a better "overview", how the misalignment of heads or tracks are placed all over the disk.

sequence may be: access each track once, calculate the sectors (+,0,-), add them up to a number, write the number to the "grid-like" disk-map and continue with the next track...
tracks where all sectors are fine should show up as 0 .. tracks with sectors of higher number than expected should be positive numbers (1 to 9,A,B .. or 1 to 11 ? ) .. with negative numbers these may show up as inverted numbers..

Hi Keir,

I have a rev 1, EC60 which gets reported as a full 060.

I would suggest you displayed both parts of the PCR so you can identify the cpu's correctly instead of interpreting a certain revision as full or EC/LC.

Something like this:

68<EC/LC>060,

Example: 68EC060, rev1

Code can be improved further to determine EC or LC:

EC = no mmu, no fpu
LC = mmu, no fpu
RC = mmu, fpu

I can find the code to determine mmu and fpu if you like.

Sorry for another post :), I am sure you are busy with tons of other things but just some suggestions....

• Interlaced checkerboard

• Overscan test screen showing boundary edges/corners for various screen sizes

• Some sort of up/down & left/right scrolling test for ghosting/refresh
  e.g. Pinball Fantasies where the screen blurs badly going up and down if using an old slow refreshing LCD screen.
  Same game, top text/score scrolling e.g. looks a lot clearer on certain CRT's than others depending on long or short phosors.
  [perhaps a speed adjustable scrolling checkerboard?]

• PAL/NTSC software switch [feasable?]

More than happy to test & help in any way I can!

Currently the Testkit doesn't work with Amiga 1000! It would be great if you extend it to also cover this computer!

Im seeing the axis on my Voltmace joystick swapped in the test so that left/right show as up/down and vice-versa, and restricted to the top left of the movement box. i know the joystick works fine in Microprose F1 etc.

Would it be possible to add a calibration feature?

First, thank you very much for such a great utility! I use it very often, and it works very well. It has helped me in finding many hardware issues on many of my Amiga!

Nevertheless, AmigaTestKit crashes on my Amiga 1200 equipped with BlizzardPPC/060 (FPU+MMU) and a Kickstart 3.1.4.
The crash can be reproduced on a similar setup on FS-UAE (when using a full 68060 CPU with MMU, not the default one which comes without MMU).

After digging into the issue, it appears that it is not possible in this hardware configuration to copy anything in the memory range 0x400-0x3FFF. The copy loop in base/init.S ends up with only zeroes in this memory range, which makes the program crash at a later point.

I've read on EAB that Kickstart 3.1.4 reserves 16k of chip memory for emulators (to avoid tools like PrepareEmul for ShapeShifter) which may partly explain why ATK doesn't work, but the problem only exhibits when one also uses a BlizzardPPC with complete 68060.

I've compiled a version of ATK where I've changed the base address from 0x400 to 0x4000 in the file base/amiga.ld and now it works perfectly fine on FS-UAE and on my real computer.

Hi Keir,

Firstly, the Amiga Test Kit is absolutely fabulous and recently helped me diagnose an issue with one of my CIA chips, so thank you!

I was wondering if you would consider adding a black & white checkerboard pattern to the Video section of Amiga Test Kit?

I appreciate you probably don't want that menu to turn into a full video calibration system, but a checkerboard pattern (ideally with the option to alternate it on each frame) would be extremely useful for finding the optimal horizontal sample rate and sampling phase with video scalers such as the OSSC.

You can see a video of a checkerboard pattern being used to dial in pixel perfect calibration settings here - https://youtu.be/dEr3uFPs7gU?t=155

Happy to answer any questions about this request, I think it would be a really useful feature.

Many thanks,
Andy.

To better support calibrating floppy drives with an analog head calibration disk and oscilloscope, the following features in the "Calibration tool" would be great:

1. Allow to choose which side to calibrate (i.e. upper or lower head). If both sides are being read, switch the head aligned with the Index signal, otherwise the oscilloscope output switches in the middle of the track which is hard to decipher
2. offer to seek to the current track from inside and/or outside ca 40 times ("step out")
3. In addition to track 0,40 and 80, offer track 2, 69, 75, 77 and 79 as preselection (see https://www.apple.asimov.net/documentation/hardware/storage/disks/Dysan%20Alignment%20Disk%20info.pdf for a list of tracks )
4. display rotation speed (in ms) and offer motor on/off in the Calibration tool as well (just like in the Signals test).

When looking for a tool to byteswap/split/join Kickstart files for EPROM burning, I found a few. But Kickconv seems the most useful tool I've found so far. The other tools require 32-bit DOS and a lot of scripting, or don't do the job at all.

So, what do I need to compile Kickconv for Windows, on a Windows machine? I think the tools mentioned in the readme only apply to Linux.

Or does a precompiled Windows version exist somewhere?

Hi Keir,

First of all, thanks for the great ATK that I have been using it since its initial releases. Moreover, I ended up with a CD32 re-capped by you :-)

I know it is not essential, however some kind of kickstart tool would be handy especially for self-built (or homebrew) roms. I seem to recall some code to process roms in the repository (byteswapping? doubling?) however I might be wrong.

I am not sure which of them would be feasible, however as some ideas you may consider checksum validation, showing the KS modules versions, analysing multiple roms on 27C800 or 27C160s, etc.

This I believe is not easy when booting from the DF0, however when luanching from the DH0 a grabkick that would be capable to save elsewhere (DH0/DF0?) might help.

Finally, you may consider showing the kickstart version in the initial screen. I know this is readily is available in the initial amiga screen, again this is just for completeness of the screen where you can see the 680x0, AGA/ECS PAC/NTSC and 50/60Hz:-)

Once again, just some ideas. ATK is already very good as it is now!

Thank you for your consideration and keep up the good work.

Hello, i'm facing this problem; during compile of testkit i get this error:

CC memory.o
memory.c: In function ‘memcheck’:
memory.c:401:18: error: array subscript [131072, 6946816] is outside array bounds of ‘char[80]’ [-Werror=array-bounds]
401 | p[1<<17] = 0xaaaa;
| ~~~~~~~~~^~~~~~~~
memory.c:348:10: note: while referencing ‘s’
348 | char s[80];
| ^

latest git release (3433c4b*) - but i have to say, i'm using gcc version 9.2.0. maybe this issue is only due to this new compiler, but i seams that the code is somewhat "not correct", i guess?

Using V1.16 on my A500+ with custom made 68000 accelerator, one of the fast-ram slices starts at 0x200000 and ends
at 0xbfcfff (done to exploit as much memory as possible, added using addmem) but ATK says its upper range is 0xbfffff. If I do nothing, the program freezes when it tries to test past 0xbfcfff, but if I manually change the upper to correct value, everything is ok. Other non-standard ranges tests ok, but they have 64KB granularity. No big deal at all, but I thought I could mention it, as other
test-software (like analyzer) detects and tests this memory range correctly.

It would perhaps be useful to keep the memory test running even after errors are found.

For example I only saw bit 2 to be broken after running the test several times, swapped out the chip and bit 2 is not broken anymore. But now bit 5 gives errors after a longer period of running. I would have been nice to catch both of those on one go.

ATK v1.18 on A4000 with Chinon FZ-357A.

Under floppy tests, the ID is found as "AAAAAAAA (DS-HD 80)" when a HD disc is in the drive. Signal test all looks good, read test looks good, write test looks good. The drive appears completely working.

However the head calibration test fails in a way that strongly indicates a bug in ATK. With every revolution of the spindle, the number of found sectors increases, until it shows 11/11 okay, then missing sectors shift in from the left, until no good sectors are found. The pattern repeats over and over, with the found sectors shifting to the right.

Since it should be showing 22 sectors, not 11, I think this is clearly just wrong. I would guess that only half of the buffer is being inspected for sectors, and the shifting pattern is caused by a changing delay in where the read data begins within the buffer.

I have reproduced the issue under fs-uae. I will be happy to test the fix on real hardware.

the main splash screen reports ROM: unknown
it is a minor issue since all tests work normally

Amiga A500+, Gotek, KS ROM 3.1.4, Trapdoor 1Mb, TF534 Accelerator (68030/50Mhz, 4Mb RAM in Z3 space, FPU and IDE)

If I boot the test kit with a vanilla 68000 CPU it sees the Oki clock in the test and I can set it and reset it fine and groovy.

When I boot with the TF accelerator the test kit says ** No Clock Detected **, but in Workbench (3.1.4) I can see and set the time just fine.

The middle mouse-button is not detected with the "Mouse, Joystick, Gamepad" (F4) test on one of my A1200s. It works fine on my 1d4 board (also w/ previous versions of SysTest going back to October 2017), but neither of them detects the MMB on the 1d (rarer, older mainboard), despite the MMB working properly in WHDLoad, and with the same mouse and adapter (TOM+) being used on both A1200s.

There are no docs about what toolchain is needed to build this stuff. Could we get some.. i just tried GCC 6 and vbcc ... neither work.

I’ve noticed that 4Mb Zorro iii (but not zorro ii) cards seem to show the range

0x40000000-0x4040ffff

On systest. Should be
0x40000000-0x403fffff

Causes systest to fail when testing the ram.

When building with Docker, since v-1.17, the zip output file has the Git commit ID instead of the version number and it's the same when specifying -e COMMIT=testkit-v1.18:

docker run -ti --rm -v$(pwd):/output rjnorthrow/atk:latest

Expected output: AmigaTestKit-v1.18.zip, actual output: AmigaTestKit-803ec6a.zip

Up to v1.16, the zip file was named as expected

I got an Amiga 1200 rev. 1D1, kickstart 3.0 with a 8MB fastram card from amigakit. Just after a few minutes of running ram test I get an Exception.

In ATK, it would be nice to be able to test analog joysticks too. This would make sure that the rough ADCs in Paula work correctly.

And/or an actual analog joystics, of course :).

Hi,

There doesn't seem to be a .adf file in the testkit folder as described in the readme.

Regards

Chris

Hello,

thanks for this nice software. I think it'd be nice to have CPU and FPU burn tests like on popular PC test suites like OCCT, Prime95, etc.

I think software should ask for how many runs to do with the option of "forever" too so one can leave the Amiga "burning" for some hours and be sure it doesn't have anything bad.

An IDE test would also be very nice considering the IDE adapter on A600/A1200 is quite picky on what CF cards / adapters it can work with, ESPECIALLY on A600 with Furia accelerators.