Turbo board ponderings

[carls]

Hi all, not really a cry for help, but rather a few musings/observations recently made.

I've been running my A1200:s either on plain fastmem or on my Blizzard 1260 exclusively for quite some time. Disregarding that some things just won't work with a 060, both boards are rock solid. The 1260 will even run on a MoBo that otherwise seems to require timing fixes.

Now, after a long easter weekend I finally sorted my hardware out, which leaves me with three working A1200:s. So I decided to try out some different turbo boards. Here are my findings. All figures are from Sysinfo, so they're at least all equally bad

Phase 5 Blizzard 1220: Impressive piece of kit! Noticeably increases HDD speed (CF->IDE), also reported accordingly by Sysinfo. Overall this is a nice board, especially considering the speed to compatibility ratio. Although with the board I got, I experience some strange random crashes. Commodore's "reset" program (as included on the 3.x install disks) seems to do something that's not compatible with this particular board. It will guru every single time.

DKB Cobra: I've got the 28 MHz version of this 030 board, which is the first A1200 turbo I ever had. Sentimental values aside, it's kind of crappy but the speed increase is around that of the B1220. It's worth noting that the 28 MHz version has a full 030 CPU including the MMU, which could be useful for developers. As a bonus it will take SIMMs and is PCMCIA friendly above the 4 meg limit. However, disk speed increase is negligible compared to the B1220. I wonder why? The B1220 makes my CF card load at roughly 1.8 megs/second. With the Cobra, loading speeds stay at around 1.2-1.3 megs/second. Also, SysInfo reports a sligthly higher MIPS rating than the B1220. Except if I'm running the borderblank patch, then the rating is lower. Odd indeed. If anyone can tell me why it behaves like this I'd be very interested

DCE Blizzard 1260: Well, with this board everything is insanely fast, as long as the software will run on it. It runs cool and stable and the board seems to have a very high build quality. The ability to turn it off by pressing the "2" key when booting adds compatibility but for me this board is still a huge problem. I mainly watch demos, and a lot of nice midschool demos require fastmem or accelleration but won't run on the 060. However, all productivity software and all "modern" demos zip along like a hot knife in butter on this beauty.

Microbotics MBX 1200z: A fastmem board with a 68881 FPU. With memory fitted, my A1200 is reportedly more than two times faster (2.39 times, I believe) at integer math compared to an unexpanded A1200, according to Sysinfo. I really like this board. Rock solid performance, decent speed increase and high compatibility. I can also run Photogenics in an FPU optimized version, which appears to do fuck all to the overall performance

Magnum/Hawk RAM: Another fastmem board sold under a lot of different names back in the day. The performance is solid and the board design is nice but the memory handling seems to be sub par compared to the MBX board. SysInfo reports a 1.96 time speed increase compared to a vanilla A1200.

Well, there you have it. I'm really curious about the Cobra's behaviour. Other than that, I hope you'll appreciate this as a buyer's guide if you should ever stumble across one of these boards when you're looking to upgrade. Cheers!

[Zetr0]

Hello Carls

A very nice detailed list, thanks for sharing.

Perhaps I can offer some insight in regards to your Cobra's performance.

DKB Cobra: I've got the 28 MHz version of this 030 board, which is the first A1200 turbo I ever had. Sentimental values aside, it's kind of crappy but the speed increase is around that of the B1220.

There is very little difference between the 030 and the 020, it is essentially a revised chip with a few additional instructions and depending on the CPU an integrated MMU.

an 030 and 020 running at the same frequency will yield near identical results in regards to common benchmarks on the Amiga

As a bonus it will take SIMMs and is PCMCIA friendly above the 4 meg limit. However, disk speed increase is negligible compared to the B1220. I wonder why? The B1220 makes my CF card load at roughly 1.8 megs/second. With the Cobra, loading speeds stay at around 1.2-1.3 megs/second.

Interestingly this actually comes down to the design and timings of the adaptor.

Some adaptors are notably slower when it comes to accessing the motherboard resources, for example -

An Apollo 060@50Mhz is about 10% faster accessing CHIP ram than an Blizzard 060@50.
The Falcon A1240 040@25 can barely get more than 1.1MB per second via native IDE
The Apollo 030@40Mhz manages 1.7MB
A Marpet 1208 (8MB FAST RAM ONLY) achieves nearly 1.6MB.

In a nut shell, different cards with differing wait states will provide different performances under the same benchmarking software. Also revision of A1200 Motherboard will also have an impact on performance when using resources (IDE / CHIP RAM)

Interestingly you can increase performance on a PAL A1200 by setting in to 60Hz mode in the early start up ( aka NTSC mode - although its actually 60Hz PAL) - this will improve Adaptor to Motherboard resource performance by as much as 22%


Personally if you are looking for faster disk access I would humbly suggest a SCSI solution on your Blizzard 060, in ASYNC mode you should achieve about 8MB per second. Considering that the use of Blizzard SCSI is also DMA it will drastically reduce the CPU load in accessing / retrieving and writing data to the Hard Disk Device - as a bonus you can also add an additional 128MB of ram for a total of 256MB!

SysInfo reports a sligthly higher MIPS rating than the B1220. Except if I'm running the borderblank patch, then the rating is lower. Odd indeed. If anyone can tell me why it behaves like this I'd be very interested

While I could be wrong, but if I am not mistaken the borderblank program/patch actually uses CPU cycles to do what it does, as such performance under benchmarking software will appear lower. This is because the the CPU is also doing other job(s) alongside the benchmark.


I hope that helps.

[carls]

Interesting, thanks! I'm not really bothered by disk speeds, just thought it was an odd difference in performance.
Although I don't think the borderblank patch does anything with the CPU. IIRC it only sets a flag in a register checked by the copper which then blackens the screen borders. I believe this is a feature of the AGA chipset and OS 3.x.

Edit: In fact, I think I'm using the wrong border blanking software. Currently running BorderBlank, which is for OS 2.x and up and patches function calls. I think I should be using BBlank instead, which is for OS 3.x only, which works as described above. I'll try this out tonight and see what happens.