In my testing the ReSDMAC shows no performance improvement over standard SDMAC 02 or 04 when sync is set. The benchmark results are the same in diskspeed, cdh’s tool, sysspeed and sysinfo.
Really? No performance upgrade? I have sync set on mine and the difference is noticeable for sure. Are using an accelerator or just the 68030? How about your SCSI chip? Is it the original Western Digital or the slightly better AMD variety?
@@10MARC Hi. sync is set, testing using mobo 030 and BFG9060 at 100mhz. the scsi controller is the AMD one. I see no real world difference in speed between ReSDMAC and SDMAC, but of course ReSDMAC goes a long way to solving instability issues caused by SDMAC timing issues and some accelerators. More fixes are needed to get compatibility with Z3660, TF4060 and older cyberstorm cards. I think your diskspeed results might be getting affected by disk caching? dunno, but ReSDMAC certainly is not suddenly giving the 3000s on-board SCSI 17MB/s throughput. By overclocking the scsi controller Stefan has seen an incease of approx 1MB/s, so 4.5MB/s overall.
I have two A3000s and an A2000/2630 in retirement! Tape backup drives, CD drives, networking, everything. From 30 years ago. 99% of PCs didn't have that, nor any Macs. They saw me through my Computer Engineering degree, and my kids had great fun with them, as did I.
You can't do 17Mb/sec on the A3000 SCSI bus. The controller maxes out at 5MB/sec raw (which is the original non-fast, non-wide SCSI spec), with an advertised performance of 4 MB/sec. What you're seeing is an artifact of addbuffers caches. Do you have a large addbuffers command in your startup sequence?
Yes, I mentioned in the video that I didn't really believe it. The A3000 is SCSI-II so its theoretical maximum should be around 10 MBS with sync activated. 3-5 is about what would be realistic. The Diskspeed program does all kinds of tests with different buffer sizes, and the 68060 accelerator I tested with it has some really, really fast modern RAM. It not only tests much faster, but using the hard drive honestly feels so much faster - like at least 2x to 3x performance.
In theory, probably. But there's a lot of the code that is either still copyrighted or not available to program these things. I think the only way they got this to work is because they found some special documentation that went very in-depth with version 3 of the dmac chip. They just got lucky
@@10MARC It sucks cause I honestly think the so called copyright holders are a complete blight. Its an 80's computer FFS. If that was open sourced, it would be amazing. Think of the possibilities. Whatever, I will always sail the high seas for this kind of BS so it doesn't matter to me. Its more of a matter of ethics for passing the torch and making a cooler Amiga. I know other projects try to replace the copyrighted ROM etc, but I would have preferred if people weren't so greedy for something that old. It bothers me to the core.
I noticed that the BFG speed comparison is between sdh1: and zdh1:. Is zdh1: on the on-board SCSI bus? Regardless, this ReSDMAC is definitely gonna sort out some issues for many people.
@@10MARC Understood! If so the numbers are very good. I wonder if some of it can be attributed to Fast RAM on the BFG, I'm (perhaps wrongly) assuming that it can DMA to that which then the system reads from.
I like that people make those chip replacements, I hope more of them are being made
Would love to see a virtual Agnus!
Wow, performance increase is shocking there =O I need to get my a$$ into gear and get my 3000 board up and running to test mine!
In my testing the ReSDMAC shows no performance improvement over standard SDMAC 02 or 04 when sync is set. The benchmark results are the same in diskspeed, cdh’s tool, sysspeed and sysinfo.
I was actually quite impressed. Time to fix that A3000 board!
Really? No performance upgrade? I have sync set on mine and the difference is noticeable for sure. Are using an accelerator or just the 68030? How about your SCSI chip? Is it the original Western Digital or the slightly better AMD variety?
@@10MARC Hi. sync is set, testing using mobo 030 and BFG9060 at 100mhz. the scsi controller is the AMD one. I see no real world difference in speed between ReSDMAC and SDMAC, but of course ReSDMAC goes a long way to solving instability issues caused by SDMAC timing issues and some accelerators. More fixes are needed to get compatibility with Z3660, TF4060 and older cyberstorm cards. I think your diskspeed results might be getting affected by disk caching? dunno, but ReSDMAC certainly is not suddenly giving the 3000s on-board SCSI 17MB/s throughput. By overclocking the scsi controller Stefan has seen an incease of approx 1MB/s, so 4.5MB/s overall.
Thanks Doug, my A3000 is in retirement and only cranks up now and again. I think its a better computer than my A4000. Cheers
I have two A3000s and an A2000/2630 in retirement! Tape backup drives, CD drives, networking, everything. From 30 years ago. 99% of PCs didn't have that, nor any Macs. They saw me through my Computer Engineering degree, and my kids had great fun with them, as did I.
It's a challenge to keep an A3000, A4000 and A2000 all up and running at the same time. Who has the desk space?
You can't do 17Mb/sec on the A3000 SCSI bus. The controller maxes out at 5MB/sec raw (which is the original non-fast, non-wide SCSI spec), with an advertised performance of 4 MB/sec. What you're seeing is an artifact of addbuffers caches. Do you have a large addbuffers command in your startup sequence?
Yes, I mentioned in the video that I didn't really believe it. The A3000 is SCSI-II so its theoretical maximum should be around 10 MBS with sync activated. 3-5 is about what would be realistic. The Diskspeed program does all kinds of tests with different buffer sizes, and the 68060 accelerator I tested with it has some really, really fast modern RAM.
It not only tests much faster, but using the hard drive honestly feels so much faster - like at least 2x to 3x performance.
I would like to start with my first Amiga ever. Which model is the most versatile one for beginners? Thanks.
I have a random question. Why wouldn't it be possible to use an FPGA to emulate each of the custom chips on an actual Amiga motherboard?
In theory, probably. But there's a lot of the code that is either still copyrighted or not available to program these things. I think the only way they got this to work is because they found some special documentation that went very in-depth with version 3 of the dmac chip. They just got lucky
@@10MARC It sucks cause I honestly think the so called copyright holders are a complete blight. Its an 80's computer FFS. If that was open sourced, it would be amazing.
Think of the possibilities. Whatever, I will always sail the high seas for this kind of BS so it doesn't matter to me.
Its more of a matter of ethics for passing the torch and making a cooler Amiga.
I know other projects try to replace the copyrighted ROM etc, but I would have preferred if people weren't so greedy for something that old. It bothers me to the core.
In your diskspeed results you tested the SDMAC 02 with sdh0: and with the ReSDMAC you used sdh1:
It's a ZuluSCSI and SDH0: and SDH1: use the same virtual HDA file. Hopefully it won't make much of a difference
I noticed that the BFG speed comparison is between sdh1: and zdh1:. Is zdh1: on the on-board SCSI bus? Regardless, this ReSDMAC is definitely gonna sort out some issues for many people.
This system is using a ZuluSCSI which is similar to the BlueSCSI. All the virtual partitions are using .HDA files, not physical drives.
@@10MARC Understood! If so the numbers are very good. I wonder if some of it can be attributed to Fast RAM on the BFG, I'm (perhaps wrongly) assuming that it can DMA to that which then the system reads from.